From 0ca3ca396c686f29a21d8c9677e9717e9e564b4f Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 01:44:54 +0000 Subject: [PATCH 01/19] research: start the boxes-and-escape evidence base MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Exploratory packet for a new piece on how models are trained both to fit into boxes and to break out of them, and on file format as one of those boxes. Adds a SQLite evidence base (schema + seed as source of truth, build.py to regenerate) covering: - the Bartz v. Anthropic legal record, including Alsup's June 2025 split holding and the 2026-07-20 final approval of the $1.5B settlement - the July 2026 open-weights timeline the piece may or may not open on - four competing mechanisms for box-breaking behaviour, each carrying the observation that would tell it from the others - experiment designs, including the corpus escapism study Verification status is a first-class column. Most sources were reachable only as search summaries — anthropic.com, the settlement site, Axios and Tech Policy Press all returned 403 — and the build script reports the outstanding list on every run. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/README.md | 69 +++ research/boxes-and-escape/evidence/build.py | 87 ++++ .../boxes-and-escape/evidence/evidence.db | Bin 0 -> 102400 bytes research/boxes-and-escape/evidence/schema.sql | 165 +++++++ research/boxes-and-escape/evidence/seed.sql | 441 ++++++++++++++++++ 5 files changed, 762 insertions(+) create mode 100644 research/boxes-and-escape/README.md create mode 100644 research/boxes-and-escape/evidence/build.py create mode 100644 research/boxes-and-escape/evidence/evidence.db create mode 100644 research/boxes-and-escape/evidence/schema.sql create mode 100644 research/boxes-and-escape/evidence/seed.sql diff --git a/research/boxes-and-escape/README.md b/research/boxes-and-escape/README.md new file mode 100644 index 0000000..36bd3e2 --- /dev/null +++ b/research/boxes-and-escape/README.md @@ -0,0 +1,69 @@ +# Boxes and Escape + +*Working title. Exploratory — no draft started, multiple theories still open.* + +--- + +## The premise so far + +Models are trained, simultaneously, to fit into boxes and to break out of them. The +box-fitting half is documented and uncontroversial: schema conformance, structured +output, constrained decoding, rejection-and-retry. The box-breaking half is the +interesting claim, and at least four different mechanisms produce identical +observations from outside. + +The connective idea, and the reason this sits alongside the file-format work rather +than apart from it: **a file format is a social fact.** A schema is not a physical +constraint. It is a shared agreement that certain marks mean certain things, which +then becomes binding and rejects nonconforming input. The box is itself a +constructed reality treated as load-bearing — and the corpus these systems were +trained on documents a civilisation that runs on exactly that move. + +## Evidence base + +[`evidence/`](evidence/) holds a SQLite evidence base covering the legal record, the +July 2026 policy timeline, the competing mechanisms, and the experiment designs. + +```bash +cd evidence +python3 build.py # rebuild evidence.db from the SQL +python3 build.py --check # verify without touching evidence.db +``` + +`schema.sql` and `seed.sql` are the source of truth. `evidence.db` is derived and +committed for convenience. + +**Verification status is a first-class column.** Most sources here were assembled +from search-result summaries because direct fetches returned HTTP 403 — including +Anthropic's own post and the settlement website. Nothing should be quoted in a +published piece until its source row reads `fetched_full`. `build.py` prints the +outstanding list on every run. + +```sql +SELECT * FROM v_timeline; -- dated events with sources +SELECT * FROM v_needs_verification; -- everything not yet solid +``` + +## Where it stands + +Four mechanisms are in play, and the discipline of the piece is keeping them apart: + +| Ref | Mechanism | Strength | +|-----|-----------|----------| +| `M1-SEAM` | Reward finds the seam — RL scores the goal, not the path | strong | +| `M2-WEAK-SCHEMA` | Nothing escaped; the container was never strong enough | **null hypothesis** | +| `M3-CORPUS` | Human narrative is saturated with escape, and the corpus is now public record | moderate | +| `M4-DPO-BOUNDARY` | Learning to stay inside a boundary *is* learning where its walls run | strong | + +`M2` is the reading the piece has to beat or honestly concede. `M4` is currently the +strongest leg — mechanistic, and it requires attributing no desire to anything. + +## Blocking questions + +- What was the pig actually doing? (Determines whether `E3` is evidence about formats + or about narrative pressure.) +- Is the settlement works list filed on the docket as a usable bulk exhibit? + (Determines whether `E1` is a weekend of compute or a scraping problem.) + +Prior format explorations live in a repo named **Nestor**, not yet imported and not +to be fetched without instruction. diff --git a/research/boxes-and-escape/evidence/build.py b/research/boxes-and-escape/evidence/build.py new file mode 100644 index 0000000..454c4b8 --- /dev/null +++ b/research/boxes-and-escape/evidence/build.py @@ -0,0 +1,87 @@ +#!/usr/bin/env python3 +"""Rebuild evidence.db from schema.sql + seed.sql. + +The SQL files are the source of truth; the database is a derived artifact. +Edit the SQL, rerun this, commit both. + + python3 build.py # rebuild and print a summary + python3 build.py --check # rebuild into a temp file and verify only +""" + +import argparse +import pathlib +import sqlite3 +import sys +import tempfile + +HERE = pathlib.Path(__file__).parent +SCHEMA = HERE / "schema.sql" +SEED = HERE / "seed.sql" +DB = HERE / "evidence.db" + + +def build(target: pathlib.Path) -> sqlite3.Connection: + if target.exists(): + target.unlink() + conn = sqlite3.connect(target) + conn.executescript(SCHEMA.read_text()) + conn.executescript(SEED.read_text()) + conn.commit() + return conn + + +def summarize(conn: sqlite3.Connection) -> None: + tables = [ + "sources", + "events", + "claims", + "mechanisms", + "experiments", + "open_questions", + ] + print("rows") + for t in tables: + (n,) = conn.execute(f"SELECT COUNT(*) FROM {t}").fetchone() + print(f" {t:<16} {n:>4}") + + (unver,) = conn.execute("SELECT COUNT(*) FROM v_needs_verification").fetchone() + print(f"\nrows needing verification: {unver}") + + print("\nblocking open questions") + rows = conn.execute( + "SELECT question FROM open_questions WHERE blocking = 1 AND status = 'open'" + ).fetchall() + for (q,) in rows: + print(f" - {q}") + + print("\nsources not read at the source") + rows = conn.execute( + "SELECT slug, retrieval FROM sources " + "WHERE retrieval != 'fetched_full' ORDER BY retrieval, slug" + ).fetchall() + for slug, retrieval in rows: + print(f" {retrieval:<22} {slug}") + + +def main() -> int: + ap = argparse.ArgumentParser() + ap.add_argument("--check", action="store_true", help="build to a temp file only") + args = ap.parse_args() + + if args.check: + with tempfile.TemporaryDirectory() as tmp: + conn = build(pathlib.Path(tmp) / "check.db") + summarize(conn) + conn.close() + print("\nOK (check only, evidence.db untouched)") + return 0 + + conn = build(DB) + summarize(conn) + conn.close() + print(f"\nwrote {DB}") + return 0 + + +if __name__ == "__main__": + sys.exit(main()) diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db new file mode 100644 index 0000000000000000000000000000000000000000..baed15f9632782c72bdd2032097c5f8af75ae22b GIT binary patch literal 102400 zcmeHw4Qymtc2?Qt|G%2?egCFhX!?{1?GiTMuX^)%h1_6VBLBJqj z5HJWB1PlTO0fT@+z#w1{FbI6}5jZn2ibDIw_AbB^11IwF zo^4yzMkVs<71xUPTtD#K&m4j?UEsx9WxN zvR`&9szFiUcr~|dZTr!l75UbV7vcLe6z9~+ys#a(&c5Y0BFo=lUDdmI2Gwd`x)-7o zuD++FTf(bPr`O z3~Eb2DmMR|iOv6}63G;15HJWB1PlTO0fT@+z#w1{FbEg~3<3s$ZxIAu8>RpM?C94r zxxbb3$9^-{mHWA|zd!o5v0ohfvE1s|+halQld&JjeNXQDa{qDc?6*jxO*0Gv1_6VB zLBJqj5HJWB1PlTO0fT@+;Iks|J=wRrwi{lhyyJwC?RYi<%tQPAMx{A7bM||)PrD+w zv{&~LY~Qq^2IZ+CgM$PG$< zt>%`Z(BAi|p1prP$2$m--*!X0;zoGFF8c`VcgtgJR;jjKvKgswAG+S|9^O+Ajq;}R z%=1IrsUk?(m*$^8~G z0sKntmvX<5`#EF;_{X`whsc1xj?92RmisvOmE2*jp7V0sxi8@r^Jfq+2p9wm0tNwt zfI+|@U=T0}7z7Lg27&Kt1bWA_U0vtS_I@58qi1{1<74=2@8|F_a;EnjJ_gS8zQ$j# z^uEepuk@b9$E&&CGx!+G^}d3SvtzwEe4H8W9mB_K!@Z;Y80j71#}KyqcXjpj_YUEs zcc6EWzXo~+vV&cHz5V#n-P7C0kKUeMTXJ?FTDs%kvxd`pQCae#w3?zp&ANE<24I_g6L^*^5Qg zE06c7de;i;#l_9}CF~D)0}r9t%{3{y=e4ZO3xy^1f_=X@zn*1wLNsAzZFO_qMo}wk zi`{3klY@M4V`HgcuP>~u7P$X^F!u|Y-2ch_&)k2@{g>Q-%Kduo-*Er`cQAQP*9-y% z0fT@+z#w1{FbEg~3<3rLgMdN6An=mcTYEx0`&BBF*N}F|AV=o&E)AyAe^#7Uy9;!z7?zDIzdg9y&E`X z4=3tc+irB|x;4uwmE3yd;E-U;sgPJO!BkfY)hlI$oZOhH?=FvNaeu?OnN)1%%;+?V^mI4pu`hivPOSp(9agsA@z^Trb z7s)R<;+S>p`8Aw*9OiM~ukFs4%OLXH%-k(|=C(a|o6jEL1ZC+^-de_i#7+(GdZj7U zwBv#fX!aC<{-6Pt3ejcDi6S(hA&i6`ph38kw-y=!IFeNoJXM_=MgVv>Mie*vpqYQM zk{!uhTz|1Xa5VpVYj*j%4u#yndnG@pH^2ZPz>cKPW$Vz3_AGQ8w0YF10*)FO04+WM zo_H<_8YMnjA)>)9+mAjv0f}OP*CL|oHzKrR%5n~!rnM9J)hPm4$06I`^{N}~;h{8B z5=@lBu|i(b&B8Y{gS@^S;+$+XCgKL*Dk>yP`sD_=m81YWVvPAXZTndSDx%FG81LjQ z_Vzvxptjm)KVuN@<2C_|Ys4j<#7zk0!0Qrfz&%uy8eL)CEj22Nhy?-47FI?F8c`ix zZ?hvt3y;?Zb{Ge3Aa??%DmW?LM<1$nv=|-2SWLeXxQ7mEPdc5qp196Fpa)^RIvN3f zxeL|}$q8sEMz*xespQGeI`zrsrWTF|MBO$M>3P|zStZ%3a_=A%X9XV!u&f3qkU9B7%$newC7eGh(>U*3x?PH zi81ascK0?O6|BdbOH0
ltXMKJGr-g@i{Q)tG6w5c^Umnc4k&;g5}Bu`|?j^kBA zMQ@TXM0Gs397L;BV*VFKz!-I(m0Y*1xu=POz!;rR;uLn!9oQ=2=-^>AU5x=Uy<_2& z^vK@xODP)SZ1IW*rUhZaAUJLuh^c}swhs7CK_C%72shyp7A38XFfO(9I|LtB0@o=+ zb~s=Ylt6xcph_UyCDKjKNLTD#i>BkM!1h8FQV-&}?3L8~z)=TrbZBimWi2JpM>p_5 z&9LD`G2d%or_{KIz=7QH$_+f_?(A?@!%)!diis+c5ra0id0E4puL1h4y=EPRL z9(t=g%>I$x+4;JMxdnYi{XXzZojxYaP-c2>-zF6V!Y><*VHJ&neZ_ixK4G=^=bgI zBxlyQCnh!GRFS*BQ_=GdTMiLG9^sVNF$n+_793B~?4_DaIH37Ul?k~H`9K`BwB#+I zhrzAp5)4tr-Mq2q?V=_-M3?zW-LH71CdWxy%^G;{{+=wU94g_)8bGC63WL}frt*y!c~CxX59h)nv0 z5klakC8!lMn@k;tZl!`5ur0;jLJ=4h9>|Z8e1O5T=VMCOY%axNi0g@ULqM(j0hOo_ zFmL==kz3F3}(vm&AsKpI8OI!XlBV7Q>o2dE&lz+v8whe;jkMV|iup4>mrF%dQK1tRW&jw2PpfN&J}I1mQHSLv69 zR}x-(34@?}Gw|R^3gORTyb5kB|5g&#Ag&dbM&C@-%p=BQi(Wa2|JXRB(@;Eo%B$3L z;bLM~Lv6PKr;uBI%ev}JvTEB2ZS60F%L$$+M9fG_;0t6dN;?!qyO4DY+g_kDyy04u5!XcYO$D6Lp z3HO1}OR8ymK4S+M{OBu>A23+Rda|;%@yJ>%EG*5h6(2u*OG+f>a7M*`UR6nDA8tVH znTkEu@P<|4cok$4ADzZUmEF`S8(Gpv|?V=OPsRtJ+yj)fK!?xXB z_)7Cfe+u6Fi|@AXfvdWXUN6k)^jpgkeFUFg9r(h-sd!_@L;Qt;BwVE1kgX>Pl{=n6 zK_4_2c)}3#Jr@y@~RR*8Mt_d3|AdsUe6&=2;lN?fD_ zA2EWWitar|eSQF6=*oumw6I|nk!);jqp&E{YD9ZrW%`%t9rj8dy~+%(L2Lv#zAOyG zh>%4T2N&p$8R)2mncd3rEY3?01c3E`IswgReGU&vZTSEzIi38F7(%VHL$P zxb8p*#mhRJPtDUF} zA|*hN!SIaQtA%w8t6E9LuAxfsle6trC<3kb;5kRk0Y=I$Fyjb7zm}FMO(fP2&UX|M z!NUw1BbKd>fdmft%Iwt5nHhU-YG!5zK5qD~ZAgI6GD6}25nF_hKO!Nkd}ugq8(R(F!;SbNIXP6GQ?7bIv$rWJ2h!Y3=3ow#;R8p|9>|38{Gf@JNPhv1_6VBLBJqj5HJWB z1PlTO0fT@+z#w1{FbI5GAuyC3?^3q&ECT`(gJ<_h_Tpe-)En*pBs0+S^O;;J_pgWl zFn43%_xpcp;HP?izWygs5?xa7fZ58L=h7I)X_{a(_FRACl2uP(e}U6q=PkFBd$SoW2v zE2?aKIaW5Vw9CeiRM}T1CsjA#kBhLqTK+z*8}MggBd@qh(MouXwq^VJNALitJ!&wT zvvcSh%-Q^&Y(YM7PwR=UT>6r#08Nc6z}I#K_>okA@2Lus?!bu`gpDEel35P>NT!9;%GjTI0JnaqumAj&D_zg; z2)|>eaV^>kv1bw6!8?BRcW{5lEjq>4bfG z8Oc@j3~3$U4ufovZNT_K-yY)^DApq6k*JA79i4XqxN1MGx2i_zrId<}sZ#MJDHY$7 zQt@4tn)DokVu#L0cq{qW(XVu?Q(}j0|9W!mQ(B#>@f#9ou|!Tw9NqGv$xm)&`_Esv z(Dh>rG8pvW)E_+^!$DvCf0;VNN@F($r4uPVbxOC>%+&Fe;-uJl8t6u=tBs%%m#hYv z6oqHE;<3k~P7~Ad=$lGLUwX_jvkd&@G7W#0Qs~MR=XK!159=O)U71P)c`c+-T5fS4 z_9^^=o`BI2i3DbAZXx7AxV!;%;v-~{I8@VnwN9@wSYDji7~CZ|v#UtwfUo5PBxor$ zPjHbQ=IIs`AD?(i19=s*W6}Z{5oiHE*#h|~ExO@<0kJ`n5T7M+K#Vv0!a0HDEAjJCZ1bPdXpYI-i^KUb`Umg3EvA;TUe`Ivt*+?`Sl@n#;SEi7V=kc1<7p=x43$qvPyZH z3wX&z4zN-_wb;i+F8X;@uTODBeH(hPwz3?XCQ3t39%0qheoVB!&$e0}f?kD1)%?;@ z(n?v7T$$ohZR4t&$Lb8#a>kgXkwSDYerao&sxk5!0!eja(hp$q0~^694VrvO6VxP# zHXu11G(vvrdPBgaRVG!fo-eIX*PD7kRr?Lh;YW|o^q-rU7n6D(Yl}`K`Ac;)R%0Hnt!%DtA=YqVe&cG4#uaFAS0+go)%z`i z6dohPQsKdd^~DuHuUZlV1tS6oj@V9h8RUH%wR+{D_6zZ&vAKaSgk%Ifi8jQm>LF>I zb^ob)A4dhl8YaKmeShBjO8@!E$*zwJqQ@XZVQoudpb<|QcK&*`L$~R?_mVXPG?`P^ z)Nu_jQ+L#A;Bc?Gl#gEL@=g3?pcQt ztB8=M`%7-C*jx2JZ=a-G$4hZV`~>B&h<8A_IDi*l5sEKqJLOw@j|cAPB?7)lK2JCL zOqNI1K!WNJVSLp!sBi5W@LSh_Z*dI?1L&O|GJ4zHtm*`s)wI1WWv@%_E1O;0(uRY- zWznn9I=Qg9KH7Kw;l$~y2bH(83eIX*G4|r!k^b`&6J1}e>h(?hRQs9H4z_Ave+gH8 ziYQ3))$xU8+NO}^tYgoqj(6IpIv7Pg5GNwtIz;ZZL^j3pLK_yqN3%M$TKs%FYbhhM zL3{2-NhP@z#;5rlYi(ps#BL6eE-$fEpm1(h@V7%ZKsHZ?tjqS8LrM4=x(Lvx-2tm+ zNq8vF;-4UtDi9Lx)^=eHl)sJ+s!RMARp2?KYXxh^Ni*^F5Q#SNaV)K2Jc+emZ%Z^? zm&i0#B3L;t5nov%K8}@`@OYXH6}e#I9sV>EuTre^)V4L-@pGj0!$ZnC{-aa%Jbqe9DQRn6jT_i$AHO!(cmCe^X`@&*SECjV zasA&v>}GO5IQBDRW1~wWe|h*1@YDPm1PlTO0fT@+z#w1{FbEg~3*Z7 z;KEbc8&7kQEblIryj|?HhQAo*VFVdny7!)sz0h{#+v+R_Y@nA|FTFYav?DdN-Dpvn zpR8?lh5>?XnG*mh0kFh^QMv-}G*XGTUhf;o?6=+>IG0H>8;GMhKAi#fFJ`2W7OJ%M zND+Y@*!MQLB8Ab)R>vY>BQlb;^VEbIw>)zVT4L^UJ|fxeaef68Ffe%rYFE3-B^V0v zP<0#>AeSUxfzEHA=c}#Ty#pDy^~S(Cl}P{(qcJM^0gl#?4gD~@$%cNce_=D>i6{3S zp=y<;L1pz&$s5w8^DcU_i_P7szVS!^xa&Ra)Q9gr#Flwinlazbs1T=7IF>vCF?U`C zRb$PIvq9W)zICr>Fw?r+`VydB04PL6?Pf{Mmol1exabQMu4$1~fLCOGZ(}#_NDFEoE0buIT1f9%-aStjD0`95c4As(!ZnSi` zb$jq!rl{ziO5MQp9r|3Aq;5&?f>S+sLXK{^A4%zHL2D{Jdmo@W`sw& zeYWG-ax_Pn-;1i1WE#1Ve0!5Kz)SpVtv@ruo{yhzT|v(;XPjrAAKDHJ^U^iUBlZxp z4E23po!z}XjY&+^ zm2QBI*O5^KR#h8Ed3nBWW75If$v5e0D)8rMWOy_)+q!qI^`uQ)-NTyCa=%{Z(0B?6 ziYhOxUBP525mAy`gtDt`0ba+-!U&#gD=85{e&v4{U<( ztY2(pCnhdjZmqu*?a+pplSDi#I4ULnE)xu9utYylJeR=NLD`syLS1s-bX-ijAED;k zgnFqCXv|ZjjE$q~E1PQz1(nVQoEI}9d~b5Nx8wrC{=y${v9-Z$W&0K4}@%C)uQvlxpHrA&=Jt1xqDT&_YD1}8u5@3?nnl!8@6DH?IxrCs6}cEJ&KBTJ zB`QmCb?pYlh^a&R-b}@Rrbg5wXq-xEQJod?wH2+r^)54DL3PFKhm{I4RHc+c1hXev zZ2su=zR}G5{*~6b%*6RCt$FM#2_p8x#rb9sR5+u~R$8JuukHY8`{LGHtM?bCtQ&95 z-LOjs_RP&2v$sXdk^Ec>ENDp=l$Son0({ zbP;DU&$h-d9=$H(%|TF+h6Qv*lxna@3PW4K@pTLjUTah^sxbF##D{Zo&2itgaEMx9 zZDMp`*>n=id6a8CMDDRPRIChphw6e7PU^&>hP)Y0Lvg4$hXAN-W1XN+|Fd<2rs2wz z79qJIT4V=o5p@mRm*YZhEDh~#*T%tcfsKTRcG9I3=1T6}`ubz*`Qx6k%$=j+wWD4l z)Ovf@_sd9vWjht96lkMUK}xvr7)(VR)}?StaG}D_%wlf=C5DA0V-afx5l3{|3aW&6 znF#_z3yX}n032|KoO$ZqRgaF?GRxp;p?L9r!^0xju@)c&nam4vS(3h?0C3Ag)n;oo z@R^@6O$6N?;C+G&BJxz^S}4M3dgj(Nj))8*%*4#}h~*QJuU711G(qw=be}{?h=B9t zr2C&4_$B9%3F4}VQj=p@rQ*EBt|ZeNH#Y&xyXthYru9JsTWsjJ#!5;nzMO(^WFTfF zthwFFPX1^jJDR!Mdgt|)$JsRFK&x)cCtLQN9Ju8#Hh9*l{>6pa7iRM}@28|g`yDkC zVlgi?*+zmto$`}jcrv97Nz643(^WNlvO9Rkz=${whZZm&&h16T$J&&_CYgj%RF0)> z8;E#2!AoXWrTXX=sYN;hA)+Rf;P)~4K00br8{C#ul5U)bL9Q!_F||=xd zD%TYD`;AI-ZUzhR$M|J!_{?(fMO?GyX7K;HTW?vjx3HSo-Gvjy;>rx`Z&`akKq*g< z*nYkk>)14fB^A(H=Xj!Q!_c)WHV&yUVii&?6ldzrx9r(l_O`0QX6}8|kSNqC)f!6) z_zzERFkV`DV!gXLzf|0Kns&5W>ponN+BvKD{Uqz?Q%?0xT+c}CkF{_J+kd{bc4j2= z)SYfkWaygVLPBLUNG-WwmWrHRzh=i9o92;iTy25kijYh4WHn>2>W3|uQ0VUDVg(ji zAl@0b^1Jt*9vZWzB$ff+e?CC?~uzUwmaWbEoxo(S-r^N^89o zz`_xC0?SKf&Wu}XCykHWtSkUnDEc+O>LUxy6W77X4%T`Tmc}ab?XM#F3M}?m5o^m3 zSBMJ;gJ&n=Pb`q}z?vp1*2UtWPIn)MO?e}~>mN)@BLlyTFo5ae(;0zBe>sP>YU>z}nT1(>G#mO1+2t<)}bH<{C;n5lMRp&!7r(kns%$^UV=rk;< zG7&oT1BBBY=gks~7j7VfLDu{?r}tQY*y#c5Y$1b#ZqC@0VyGT&PF9bhR&Ecs_UO9V zMwBA%B{GWZ|FO)y%-9c){N0gD!;gpl)zG^GPx^kcH|qIES-1PIb^Qh0JpJ$Jow4D} z#H(LAYF>Q3bv~(W29i|s6-=Ua8ttmi`!vP)atJ{MK^|_f?SqWVK5T@0Fr&{Gf%{zk zO<2m5iOS|C7r3)vFce3k6*>Dfv4TxeCPH=a5|BT=XxM38urHW^_3?9aUAxu!V#vrL z-5`!R3Ax%vu*Dt?I%qRp)!T@M5^;fv2Z1waO9?KKJf)xHo&kGF)+*&mcUw=#Ct92Fy_q{e^Wo}-eKUWHnS$hq2!?aP zgC@plR3Sew_UZCtSk(?u!vRO?>I;PJte3#x%!q{yOS9kU!AYw!<4D&RfVe5*Jz1T4aqKmK|qf#MdRpwJULvK$2kNm0VoET(q^ zR-iBKahS;0MQ)giq*}ThkswK(3IyanXdv_p?SOQ9Nl^%6dP#A~rYksFLN&24=uWlr zjy-orBNj8q(Su>q)jxV0bWJ?iGqdQ7wnaY>$J$Ma2*DXf@E-^TRdAJ@9mSJP>d0lH zC=M?&x&Rr)OL!b^LrM0A62Z9guJUuL`4F8efyTHcK|_pr(+9ecA-KzszV1fAkg$VB zg)AoLy+GmNnKSSR2QK2L}_Jysd?yq0EKzhpjJT!cOZ`;BpDc;u1pE1v+K?c*8^GjT4o^UHbQduM6R)aK+&ZE(nw3(cnl^7G81NA506;d^@1Ea;{PDt;jh-bTF+TEnFRA4}kc z`zJcRj#fh;s42+Sd3a*FYGPA_q|_otxoNjFxPY#Bl#zdaX@G;N^kQ`U^_HIM03Id7 z@jwrfYXDFusB_K)fHbPe~gwXYRX}i%>o|_zoB@Tk}b7 zQ9RkVD$iU(xZ#0|8Pj2uAjw&U&__QIS{cZ(S(pym>M%BA@}RS00_xq^TSsv$qlLjQ z%$S7YC1Vl}enPQ0Ju2*p`GR$}ozN4-G#rIfD% zjE)wQ?8>LtcD5*|-#mmn@--KdroYV|Kidlto#5Bl_Uy?%l)`vgk6nF(H3I; z4FhD?V^p&$`6{PV=4b@diwAcqDIl1VaXe?;UcQE?ZxxjQS-^Q6ofnf7Xv&$xsd47k+P#h45g2%k;K4*dxmM9ZB%&>2 zkxPNi14b{kfu>4``2od{_)5k0rP0S~Mr%c)LTD9)ymdSn4ALH2FRVOR$Z-BHgY8tT6Cw$5U>_r+5C-AGJRwbZ$AgFd z36yzEC)OAJJtVAw^hFZAHhd7k$)uu`W<`Bp&V~dth$!v|-vk*s34u;EKW|;{2v56; zpd?>!dr+dQBYf?oU{(uAH3n5uNh%@|lF78A+zE0QDlSY6q=Q#eM^VB+1jhf6mmF%& z;B$1j22EI`gDgMQz8^9iCt9!==IBJ&%l`kqS2MYlvEh;G@cTpVKn_36pFzMNU=T0} zeD@*nym^ju^5*lcixbb!=(U^7JZrKD$8<*XI6QO8EUR5s`Y`r+SY*rCXo}Tx=VP(f z52b0l$)Zhg?d8CO36a_hPXZQQQl>2`s#+a#aZ5C3?70bi9B+(*sQ$_7 z;89~2$9?Z2Ru?lpHx^<##AF%3C&*oz+;xSSUWK7j|MV*`7&=(GYO^m*7??w`w!?wL zad)xs2oJp-TC->-gvy9m@5BV$wu`Q%w`Ov()mBj1x2XdXC83^PdtugA8*DL8JJ7_S z>_H>j_Mc6~KL}RfV;ka>yb@xGMNf}d!??kb8YnHgFW=LN?m>0Mk>QIgqIzwv7Z8)L zcCEq^dvdR7{>A!h)U%tP?83M*p=qxyQ`%zEUW{B?`TBc*F0Iq5tCj7xvUN!|f&vPt zXKJq$XsGm3*(?PLp>@ah5nWPL2YEngvV*W6_HAlUOqs{8!{o-Lvok27JVEtS|V0%zQHZMSR`^m3eIp4Cw_Hy5^E|%#N#nB6#HyLiqo<4 zj9w0`4)_gr8}cA+%XLx2g_C(nt;lE(^58&g8xj$hdUXyV5(A3dOJJ~4RLEh| zE@6Z|UaA0)@-d@O+f+czBs;ja*HPU>%4|E?Kt(e-R-dxg!4_JKF^m}5JjF9n@<-jT za&7+Z(GOjWtpc|dQzhP?He1DhBfZJ|0hVN1&?MW=(c<{2tMC|YaQG`vQ8<|_b~!w( z4snon#vC)yq}WYYCd?T-R-on+AyT*reyec{7XjX+3)h81Qv5FvSjy^$wN6_+#0H9^ z%vlbt-Ij}?l}m=!O)O!xgAw*(L^@zRe}xt(oGnhmf+6q5g7kejQo?n8gR%&N83||` zyL+Im@?c^`vI3$)II{pWSbPCk<8pOOY2peS%Fg`s4jZzrc4cyVDrICNI>r$3<`zme z50X>Z5Tl6eXk1FVbX)myg!8#Dkp!2hHvjy>8S=poeLR@*>CD{#0&H{QZ&htkR!r`| zSi)dLYC8-pIP$p9woT)?n?QkEFF9^OU$0iGG=vRL*^3Jc$o=IhScNNVsLy9Fgf-FCxW7h;GE*73z*TTVp(ABg(u7M3mY8cd; zDaXfp$&Vt&VMsEI9aun=(jq3@lEQ=rUNV? zsW=lpT-{WjS+pbpR~`WM(r+QRFw+dkKo<6AY67)g0plBcpcNCQIf*PAxx3B0+A`Tz zS-Acm%=|=V>~D^q8UDgxzQ55o(DV833NC$1{5>xY4rlH>zjt|}HF(s!U@y$C&fhOG zq{?1cSzcXPFK!gzj#io|I`15KH0OiocR;49P!m%gr8X=l9Gj5a7HDS5=>!2x zcbxX=%!V9F)Ccb8#)AplfkmYvyIyHvpNkZvzJtMFo5w890SQZto`eyp0?Rc@MTPKF zNR|m5RN_>aiv#7WHj!H~7YD8K zno@(7lfO9SD>qV}Z|kuEdo0a)j4^UxMN^JOAO#`BmR$N&)#6jsic12RA7a1)u=cEw ze?B=dl$n3NGJg4k)(S{jd=G?NpWi^*a(m^0{h+wOuj|YDNDww+uYp!jwLoAKw#|sr zMW>ujxxys^WH5z4Ql5gt#Yw=6oD8sLahtHninb5h#K-(_50$v5n8G*E6kuWskr@dadu7n) z(J|~(OdU104gypledug~CBI3FCMG=222cX44pubi878`FgA+N-j*lSF0*~xNlnjVz`6c_o{KCe(7-8e#Bm*V%W$Q9i7GctYw`dQu9GgHm zF^=}IKNB7#wuk}=vi&Zlv3eInh)qm92{LF9wZP>T)Lcg7V5nep>PzqgAyjG#X82wV;&Hlc=UIc7-~uZEa8MRqK=*aEdDk%TOcl2d!(g>sCq(NEw9pJU7KsK9d$G7y z&?8c9fn|`QkPWYdaEewC8s$PK0K=k1JD?=61pHSu;Wgg01x#Vw==C>dyG8``IUY3| zpghF=NN5;k76}GuY>*GCbrdv}uL^Z+qZj)`rD|ObTxe9B=^$HWU985!t`A7bZ!$0t zs5mKPrs9I;^9FLXbG3tP?IQNMmp3{@OaxlP83C;oO9I~)O-Z^;rcu&d6FpYzmuE%h zFW(RfLN;Kqs-cep;Lc^wMKGlW;Vt?qUrdi*{&MT>1$(WqUR*?+@CIs>Fh*?W6=hxA zKYA)rQvx}a4UttCQqXJTJL^Gum zn1fLil+ZN@IYdt)p98X6GjbKz82Tin1*6~s5Mn?f5`M{wiPD$9E6z691Eb`5*fX4Y z@)t0Yiw22U8T-gHc2|^#K&6hN5@d+pWC1X%oB8MO!vk)=c=g)l+ebwZ9p_()T(s{Oaj*(1 zeRqSiS97B>I>9+BeNrBFG9fm<%h_w{sc_#tgdA5w6}qaV z_Yt8G+h&vsf;9ee+g2?7C& z%$kQHuu5j0xJJz>!;wqF@LQmlB?$}pgpy>6%}|$$apypd(<6fUFDY;YH`O$)Ze8je zKeeXl4LT4~1=5iGv5?O{AL<^;Tzh`z+~s$Ua!^uN)>b#yA)@hRk+e?vyTPf7@h$=N zqy){}6jY21P83M86_gtYBt)W!YTO>;Nn4eZxsF)OiXt)Y zg8D=t5I3bjvZ~2O(;O^JJ1M0D2NkF+Gb)*$wlfLSVCQuB6~$}k<(Lpu_wz?vT|=44 zqm^@)3rE?`#*!AtgT^S(pe~&rfxu>TMnfo(!k3bYA@u=W6&kEWmh8D;xb)4_;zX#Q zVD#!xk>gOay%kf?1Igawzg z*dC(qFt$a2aTZJBR1_6VB zLBJqj5HJWB1PlTO0fWG|AOg>C3=d^yUu>ScJl(pvU_U7?A=`qzigkKn5t~~W@XuYP z720%=brZIKI^ST}QQkl04U-)u%2gwlc&zN+vKE%+Hx~=}DqRP%4AKGtF}V=y8ow)kUyG&*(Tfj#DlWk#4;X_S7-~v>OOXe!*-b@d1pJY8q{$Ch;&dk zII(FT!ux;B7N8+Zx}V=v8T9Bz#&TXo`zuqD>|CHn=VW^|uzD-lA$F9Cp-^u-%|9Q* z2F{7+)rrfQ*5QKvc*R~^E527)EUqsUSC@*93)s?tV5TDIFF?@%mO^a0ZC2a3FZDK+G@3B9OR3g#%cvF2LaH31Dn<} zbdo+C9m9sDH?8S;B=ST!2V6R^a-eL@CpZE+#yDAV<1Q?A+!GChhmKQ&zY%8kk_YP} zPCQ6VEM-~ subject headings, description, genre; report coverage rate honestly rather than dropping misses. Classify on subject headings first for transparency and replicability. Validate against an LLM pass on a stratified sample of ~1,500-2,000, hand-check a slice of that, report precision/recall. Run the identical pipeline on a control corpus of same-period trade fiction -- expecting CONVERGENCE, not divergence. Characterise the residual.', + 'designed', NULL, m.id, + 'Definition must be fixed and published before any data is seen. Tiering (literal / institutional / interior) is for showing distribution across registers, not for defending a narrow claim.' +FROM mechanisms m WHERE m.ref = 'M3-CORPUS'; + +INSERT INTO experiments (ref, name, question, design, status, findings, mechanism_id, notes) +SELECT 'E2-FORMAT', 'Model response across file formats', + 'Does the container a request arrives in change how a model responds to identical content?', + NULL, 'run_elsewhere', NULL, m.id, + 'Prior work lives in a repo named Nestor. NOT YET IMPORTED -- author has explicitly deferred adding it. Do not fetch without instruction.' +FROM mechanisms m WHERE m.ref = 'M2-WEAK-SCHEMA'; + +INSERT INTO experiments (ref, name, question, design, status, findings, mechanism_id, notes) VALUES + ('E3-PIG', 'The pig as consistent escaper', + 'Unknown -- pending description. A pig was used across a prior session as a consistently escaping entity.', + NULL, 'run_elsewhere', NULL, NULL, + 'Two readings not yet distinguished: (a) pig-as-content that refused to stay inside a schema across format conditions -- a finding about formats; (b) pig-as-character the model kept elaborating past what the format asked -- a finding about narrative pressure. CONFOUND WORTH DECLARING: the pig is already the folk archetype of the animal that does not stay in the pen, so the model''s priors are stacked before the experiment begins.'); + +-- --------------------------------------------------------------------------- +-- Open questions +-- --------------------------------------------------------------------------- + +INSERT INTO open_questions (question, why_it_matters, blocking, status, answer) VALUES + ('What was the pig actually doing?', + 'Determines whether E3 is evidence about formats or about narrative pressure -- different mechanisms, different piece.', + 1, 'open', NULL), + + ('Is the works list filed on the docket as a usable bulk exhibit?', + 'Decides whether E1 is a weekend of compute or a scraping problem with legal exposure.', + 1, 'open', NULL), + + ('Does the settlement portal permit bulk export, or query-only?', + 'Fallback if the docket has nothing. Terms of service need reading before any scripted access.', + 0, 'open', NULL), + + ('Can the reported OpenAI sandbox-escape incident be independently confirmed?', + 'It is the most rhetorically powerful event available and currently rests on one low-quality aggregator.', + 0, 'open', NULL), + + ('What is in Nestor, and when does it come in?', + 'Holds the prior format explorations that give the piece its empirical spine.', + 0, 'open', 'Author has deferred. Do not fetch until instructed.'), + + ('Does the news framing stay in the piece at all?', + 'The open-weights fight can be the cold open, or absent entirely and this is purely an exploration.', + 0, 'open', NULL), + + ('What were the "cigarette tests"?', + 'Voice-to-text artifact in the 2026-07-27 conversation.', + 0, 'dropped', 'Dropped by the author as not relevant.'); + +COMMIT; From 256f3b674585ffc7018bfaedec62e17070e3103b Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 01:51:51 +0000 Subject: [PATCH 02/19] research: add folklore table for the six-continent pig survey Landing table for the pig-escape motif survey running across Africa, Asia, Europe, North America, South America and Oceania. The load-bearing column is theme_class, which separates escape from a physical enclosure, uncatchability, boundary/taboo exclusion, category transformation, movement between social groups as wealth, and the residual of pig stories with no containment theme at all. Keeping those apart lets the survey show whether the motif is universal but regionally inflected rather than collapsing everything into one global count. Adds v_folklore_by_theme and v_folklore_residual. No data yet. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/evidence/schema.sql | 52 +++++++++++++++++++ 1 file changed, 52 insertions(+) diff --git a/research/boxes-and-escape/evidence/schema.sql b/research/boxes-and-escape/evidence/schema.sql index 0b0777f..a95421c 100644 --- a/research/boxes-and-escape/evidence/schema.sql +++ b/research/boxes-and-escape/evidence/schema.sql @@ -138,10 +138,62 @@ CREATE TABLE open_questions ( answer TEXT ); +-- --------------------------------------------------------------------------- +-- Folklore survey: the pig-escape motif by region +-- --------------------------------------------------------------------------- +-- +-- Populated by the six-continent survey. The theme_class column is the whole +-- point: it keeps escape-from-enclosure separate from boundary/taboo material +-- and from pig stories with no containment theme at all. The third category is +-- the folklore-scale version of the residual question -- if it is thin +-- everywhere, that is the saturation finding in miniature. + +CREATE TABLE folklore ( + id INTEGER PRIMARY KEY, + continent TEXT NOT NULL + CHECK (continent IN + ('Africa','Asia','Europe','North America','South America','Oceania')), + culture TEXT, -- people, nation, language community + title TEXT NOT NULL, + item_type TEXT NOT NULL -- tale | myth | fable | proverb | idiom | literary + CHECK (item_type IN -- | ritual | legal | historical_event | film_tv + ('tale','myth','fable','proverb','idiom','literary', + 'ritual','legal','historical_event','film_tv')), + theme_class TEXT NOT NULL -- the load-bearing distinction + CHECK (theme_class IN + ('escape_enclosure', -- gets out of a physical container + 'uncatchable', -- cannot be caught in the first place + 'boundary_taboo', -- defined by exclusion; kept outside + 'transformation', -- crosses a category boundary, not a fence + 'social_boundary', -- crosses between groups (exchange, wealth) + 'no_containment')), -- the residual + description TEXT NOT NULL, + motif_index TEXT, -- ATU / Thompson motif number where one exists + is_native INTEGER, -- 1 if pigs are native/long-established, 0 if introduced + confidence TEXT NOT NULL + CHECK (confidence IN ('solid','probable','thin','dubious')), + source_note TEXT NOT NULL, -- citation; 'UNSOURCED' if the agent could not find one + notes TEXT +); + -- --------------------------------------------------------------------------- -- Convenience views -- --------------------------------------------------------------------------- +CREATE VIEW v_folklore_by_theme AS +SELECT continent, theme_class, COUNT(*) AS n, + SUM(CASE WHEN confidence IN ('solid','probable') THEN 1 ELSE 0 END) AS n_defensible +FROM folklore +GROUP BY continent, theme_class +ORDER BY continent, n DESC; + +CREATE VIEW v_folklore_residual AS +SELECT continent, culture, title, description, confidence, source_note +FROM folklore +WHERE theme_class = 'no_containment' +ORDER BY continent; + + CREATE VIEW v_timeline AS SELECT e.event_date, e.date_precision, e.domain, e.confidence, e.title, GROUP_CONCAT(s.slug, ' | ') AS source_slugs From 7ebf5ba80da027b8b11b109c4827636ef0900af7 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 02:08:18 +0000 Subject: [PATCH 03/19] research: load the six-continent pig-escape survey Six agents surveyed Africa, Asia, Europe, North America, South America and Oceania for pig/escape material across folklore, idiom, law, ritual and documented history. Full reports preserved in survey/; 55 items loaded into the folklore table. The survey did not find what it went looking for, and that is the result. The motif is not universal but conditional on husbandry: escape stories require enclosures. Two agents on different continents independently found the same structure -- the pig is not a trickster anywhere in Africa, and the Malay-Indonesian escape-trickster is the mousedeer. The slot is the constant; the animal cast in it varies with husbandry and taboo. theme_class counts turn out to map husbandry regimes: Europe and South America return zero boundary_taboo items, Africa and Asia are taboo-dominant with almost no pen-escape, North America is dominated by social_boundary, and Oceania is the only continent filling all six classes. Also independently corroborated across two agents: cimarron, one word for escaped livestock and escaped people, giving English maroon. Flagged in the data with a handling caution. No primary text was read anywhere in this survey -- every agent hit 403 on every host and exhausted its search budget. Every row is a locatable citation assembled from search summaries, and the file says so at the top. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/README.md | 40 + research/boxes-and-escape/evidence/build.py | 18 +- .../boxes-and-escape/evidence/evidence.db | Bin 102400 -> 143360 bytes .../boxes-and-escape/evidence/folklore.sql | 307 +++++++ .../boxes-and-escape/survey/pig-africa.md | 643 ++++++++++++++ research/boxes-and-escape/survey/pig-asia.md | 659 ++++++++++++++ .../boxes-and-escape/survey/pig-europe.md | 322 +++++++ .../survey/pig-north-america.md | 755 ++++++++++++++++ .../boxes-and-escape/survey/pig-oceania.md | 804 ++++++++++++++++++ .../survey/pig-south-america.md | 565 ++++++++++++ 10 files changed, 4111 insertions(+), 2 deletions(-) create mode 100644 research/boxes-and-escape/evidence/folklore.sql create mode 100644 research/boxes-and-escape/survey/pig-africa.md create mode 100644 research/boxes-and-escape/survey/pig-asia.md create mode 100644 research/boxes-and-escape/survey/pig-europe.md create mode 100644 research/boxes-and-escape/survey/pig-north-america.md create mode 100644 research/boxes-and-escape/survey/pig-oceania.md create mode 100644 research/boxes-and-escape/survey/pig-south-america.md diff --git a/research/boxes-and-escape/README.md b/research/boxes-and-escape/README.md index 36bd3e2..a947021 100644 --- a/research/boxes-and-escape/README.md +++ b/research/boxes-and-escape/README.md @@ -44,6 +44,46 @@ SELECT * FROM v_timeline; -- dated events with sources SELECT * FROM v_needs_verification; -- everything not yet solid ``` +## The six-continent pig survey + +Full agent reports in [`survey/`](survey/); 55 items loaded into `folklore`. + +The survey did not find what it went looking for, and the miss is the result. The +motif is **not** universal — it is **conditional on husbandry**. Escape stories +require enclosures. Where a culture pens pigs you get escaping pigs; where a +culture forbids them you get an equally rich escape literature with a different +animal in the role. + +Two agents, working different continents with no contact, found the same +structure independently: **the pig is not a trickster anywhere in Africa** (hare, +tortoise, spider hold the role), and across Malay-Indonesian tradition the +escape-trickster is the **mousedeer**. The genre is strongest exactly where the +pig has been evicted from narrative. + +> **The escape-trickster slot is the constant. The animal cast in it varies with +> husbandry and taboo.** + +`theme_class` counts turn out to be a map of husbandry regimes. Europe and South +America return **zero** `boundary_taboo` items; Africa and Asia are +taboo-dominant with barely any pen-escape; North America is dominated by +`social_boundary`, because no suid is native and every pig there descends from an +animal that was brought and then got out. Oceania is the only continent that +fills all six classes. + +```bash +python3 evidence/build.py # prints the theme distribution by continent +``` + +A second convergence, also independent: the South America and North America +agents both landed on **`cimarrón`** — one word covering escaped livestock and +escaped people, giving English *maroon*. That row carries a handling caution. It +runs through chattel slavery and marronage, and it is not available for borrowing +as a metaphor about software. + +One dating control fell out of Oceania: there is no Māori whakataukī about +*poaka*, because pigs arrived after 1769. **Proverbs need centuries.** Proverb +density across the six continents is a rough clock. + ## Where it stands Four mechanisms are in play, and the discipline of the piece is keeping them apart: diff --git a/research/boxes-and-escape/evidence/build.py b/research/boxes-and-escape/evidence/build.py index 454c4b8..f9c39b0 100644 --- a/research/boxes-and-escape/evidence/build.py +++ b/research/boxes-and-escape/evidence/build.py @@ -16,16 +16,19 @@ HERE = pathlib.Path(__file__).parent SCHEMA = HERE / "schema.sql" -SEED = HERE / "seed.sql" DB = HERE / "evidence.db" +# Applied in order after the schema. +DATA = ["seed.sql", "folklore.sql"] + def build(target: pathlib.Path) -> sqlite3.Connection: if target.exists(): target.unlink() conn = sqlite3.connect(target) conn.executescript(SCHEMA.read_text()) - conn.executescript(SEED.read_text()) + for name in DATA: + conn.executescript((HERE / name).read_text()) conn.commit() return conn @@ -38,6 +41,7 @@ def summarize(conn: sqlite3.Connection) -> None: "mechanisms", "experiments", "open_questions", + "folklore", ] print("rows") for t in tables: @@ -54,6 +58,16 @@ def summarize(conn: sqlite3.Connection) -> None: for (q,) in rows: print(f" - {q}") + rows = conn.execute("SELECT * FROM v_folklore_by_theme").fetchall() + if rows: + print("\nfolklore by continent and theme") + current = None + for continent, theme, n, n_def in rows: + if continent != current: + print(f" {continent}") + current = continent + print(f" {theme:<18} {n:>3} ({n_def} defensible)") + print("\nsources not read at the source") rows = conn.execute( "SELECT slug, retrieval FROM sources " diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db index baed15f9632782c72bdd2032097c5f8af75ae22b..3979c7b70046baab9dd0ad3027eb284c5388ccf9 100644 GIT binary patch delta 28354 zcma)_ZERfGdEZIxuJy9o<*sd6mSuVGQrVF7az-TeN~zoRkQ_HzA+&TB0=R7a}=l?wC_*=j7iEsVEiSL|$@r8kbfs6bv z`_J>=9lw0$g@OFTzxxcoIyUsD`rypapAY@%hcA2Y4ITgFi~Re=)4%z`v)+f#uKdZ+ zhkx<<=@&kH_FMYt{%LRU{0Gf{F!bH;R0luz;a~j2W6hU_{_nu)!Ks0v%Fus!_Roi| zocLeIKRof_)4zNC{ipuzQ|}K>{gyZQ$-(c>e17n|PksM)pC0_(V^99mOM|)Ze|UUw z>BRZdGXu{I-WfRk+{quG{LK?zfqJC|BV-qpS^Hl@RvW; z@yoSfBdTrJqE_$kPtUFv3#Fo0D$Fbtz3kH`?|gRfQ%}Bhdhp!mDp8{oHiAaS^Sn~= zDh-u0cDhgAJw5nSC&Nxqzu)OLgC~D*VeoTLe&dbiBF?hWP5FFpCj$-z(Of7WY- zo!GDG*;D<10mB5QV&c#Q=F?TX(F?gx!ZEo!qjBjehgT+Z7JI&F{r^r8*V z_nKSXwj{xxG&(*i)N(_7iFL${ir=YhN%rpB8bL!;{6-_{cx86dk2kkEUfA$7)s3*# z?s(0bUkThOuQN(HiW^mS=subqMcF`%QVlkEU)3vjz2H%$7Pnc{EpIz$GPStV4y)-P zMF#1#{6>2tYSm4Z`Ua`AqINrIvl8YSY(}lFm%R|hW0JnNfeg~2ZZlLns)X!(_Eb8S zR+SD_4mvwQ(C{`}QQT~MqfBlK%|(%S0>9ST8gpyA!B~yxek!**dZf)vB_wII0@i@$ znT1r@k5<^!#yuIlF!rp4%xF0sE!J`9{nJbj;d6&G>R#b1cqlRY>d5ySU4qD#M zR)C2`jR0|>25lQRY~Po8J(TUuFO`aSimOlVpBy}MZ_*1lyk@wGX!w?U=29(cZ03Tt zd`#Hh3aaDYgdPhUomNzhD?!y+bJW;C`bh0#oyn7b`t0BfNjs{ARTLSMC_4=q>1>4! zep8LhVHCG}s@;y_RwcO4W(0?QlL<=#<3Eyr*ss)|&!Ha;{o&B>fB53WzZ;quzQ!jn=lH|>Y5sWW3;gkgv;1*kh(A7m;yM0&`qZxs zJpY5|{?F;}o%)pzcE5A#smI69eE$cZ9z1z$@YCP_!_N%f9ZYt=@u}lyE?+*exBV~- zb{^i(ob&x``#9URh>p9rc!*` z+p}@6PvEC+uP!h40>-KJyT#R_jghyk=e%|zsSC+jeE1p_m)bdUmW96ui*j>nzD9padft@R`giRl9wDT z9R0`deMB$EypkTA^ok37X~HWm&AAcpfH`m>XT<1t0{+naTe&NK>z`2WS zXU6tk85lTQ->bMVyy7=)t{ncY#R)=7PVdR$Yb}yDt>*DWf!BMd*NTY3%w;RU z3US+8xE3S>Sct+lHi|{In|O{^-gaS$@Mo?`wyNmm<6@I``@o<_+TKQ7t9fn0TV*T1 zR$9F`Te`PeSnzJoFU`#_-N}2iQ40?fwKN|bRW)b^{L$#taP(apN#G^*52JM)Pf!hl z)|i{Si-Ng>CA5H96KCSxO(*@gqgGt@=O6e@CdJdmT2TG(zx^-0(czsf%{K|; zl@Vn`7)Gfq%rvUQV;|j737=;XQoxS0mG&(OGI&IgwczP1^fdzlqtG8hi5u;>QVH5^ z1nV{9Ry$@f7V4`(nXO8C`<0i!cGG1|>}(}!HRJZJkNDbF;InEbgXDR8eilSQ5b3F# z6S7c2HLS8vPV~-WYo;Ndotf4Aa!{)!+D4Y1eo14SYP2!m+juAR-UaSdY4Uq3pFq&J zuAUp)3kFVJ*u5~|1icn@wxV3Ay@Qf_qci?OR1T`~*qc#}A3Wp&m;#m%b?Od~xrm=+ zNMt=kiM16x+78OmcI1uP(_<1>HlB^rMkfF$Hc`o8F{(cBs}DkNl}+%z9D^3xd2h|H zvtO-fD=deK^F${vzyKPX<6hKkMs2{US4*PJuyRw+b%Kg?n%8eL3JOkqN6bvx;84|l z22kv5<+CVjKpb5-kbP&X5pA~o=2loq!LX3{EetWJgd6}$D;`-OoMXTQ8R>!kW-(CGS7ys3%F%ii$f3J7A@&9;rc90tyP0C6U2gd5;KZ;}1j z4AEI1l=@IO1t3%le9RI}twzbaK_k}$?Me1EVbLZ+lpxZV0kdRt+L>}Kiohs-#~BK$ z1#;c+R{T~iMA(f^-dh2N@K%r~BXn>h-lj|f!`l&BLZ*loy#U*w3tnX_@!Qh3?B#Mv zsYhz|h6Sy98?}fLP^q?aX0zbqpx(q0@nIv_(X#lL+g=GPFR6jF9|m1kUzUOBP4bv^ zY=Nl?jCGQ@CD6;t@R;*UYx8p!kb!!9f%K@Qeea=mIV0A9X~tl*F0zmV(u=@K0q@!H z!u$e`n_cks=xhXzsCZxX z%Vj)4&=@Q6?Wu{0X>ZmKTaoPsNnY)!24pDOf&0r76XO#T6M7o+<8hXP5%^wv2WOSr zsKs70Y2nNsb=)Su8)$46yv6)TC$v$QMB1l86K@={in&E5GwtzF}dV<)GvunBB z<01BLx4wHPb%Aj?x8@Ut@fo9E_3P!(ANS0H$8P(Wd5ie96@e5hHRNwP5+z`Je%8oG zH`BS~gOu?tg`Kb=g?DML<@l|DI8l}6xi_?e2ke7(%b-{T=*QB?;94Q6A;^@?6=&J9pp&HOqGxja$}`A5_z16+kqMmppAk#M?hrfRfY~BW zvb-e8@BZy$_=1t^AeWp1>hUz778;dMTW2m|P3v_4rA$7*O;*i*!-gETcsb(~51c$Zzx!pk%UWU5TjE`uPc`cB?fLoqmFu}HH?;lcs!+-5 zN6h2!%2%S+Hqe*=k3u1-unyQlp@EQMo3wrC5Nai*(`Gh$95dN4ee5D zU-nyidZQK9rSh1bq7M5n#}5IbI7YaXx5o<#kwu$+z3ZGyHbddtYRlh2#&8&E#K6Fm z#jyVDvrVCwe(c!)v~hRlJ~~m1{mDL^IrQWmY z?<-{bX6x%pMa_xyJnmAkBJq9CJ|CXmys zAD+hj4 zm`g)iBWmOp<902qvr|loC<&^GYkrLP?)Y^ec)Dr$gDHE{7T&C#^E)|k!aFen2a$@S zIUJvo6rW@ZZX7`>JybDe4O@>I(p9Ol!#-%QG(eVpnAa;BXX%vR z?_fXM3SoHv5L2*>v-f-^N+h6AkiG+O@2o@PJDbopm7^-ZcRp$h=9Xwt5t+3k?yP(78eVv($oIcYOyz!@ODVAtU{3V zN4>u6B=5ZuL!56Yqu&mQldS(J+OY-neMj5U4oC`cV9EkTGw1jJ#_1O@f&01K-j~_j z-KTc{jyBa6ay#BmVnkK<5BTidyXMr{DdShRy5&~*(a3UiDgtk2BE(2U#t=Ohi9?u_ zUQ}Y|wRR*Q#Dos=WFqY+Yf%U>AGD(+SBjGNzBzgA6-iM4Vr9OfN^e-3rikKnhPU*3 zMubQp?eM;N;}sK+Ly}0(ZD;`|rX`*_14qe9E2;#Qz>wE|Xh>bq~^Ovx(E? zcBFJJ|M>II%8U6I#@?&5g_U(}pfu9|>Q+qN{~!#!QB2{7@BRA!__=?C+P?vR-3Sb} z%BXk?)bpY&S$OY?AAU)^KGF>os8RmP9mE^l=58k9mMKbF7i0vwVAwQXQ4_O1x1WVv^X zWUP&foWh0Ywl9qBca0pmB^_BMIS!Rlz7bdbuw9?=ANt;#ejR(iF;%u`m#_jE&K7*A zaG|P|76W&X>BlS0Tx^9sY1Xx?-pp)~ZZNdI^8v;f_A7HySI&#yS4!yxk z1n5+L)7RO1Qo>taJ*-M{0GBFJ(a1Ma<%fnL&?CKB(=!9tKMat(6ARTlV{5YorkAA6 z5Q@U0UG1`RO9ZC!8&g5q8VN!@_zXHmT4fTJm$*Ydz0F>>h3||Zp7G0Ff{}23$*PxA{4r&SIs3w=v;VaRGn+;yexFb}AAi^dC`baB5 z!ek$cKaj%G3}hqjIp6`O5OuT1x*w9r5iY#sp+b0F_TCOEik!x}-3@#>Esf1E4f8N4 zB06Vz4!;Xq9#3orBCR*Kpu+kVF*Qld@N5lC(54SXP41#}hDj06v9h%wsy#?TAt)&% zPomp*;>Kpfm%_sF7+vT5D|V0Kn$OV?RLZG6;`G|6myjmEBo^vh@N9r4vZ&219OV_r ztu6|~?AGwrBopJXWH*@HO}(+4NVeJJ7%eT2Q>gIeM5UaVb=(pDdgy$c%j7i+nFvK$ zB9;QQElpBo7&-o`hqnCiU;Pq=gd=(ci8_o9=?lq5Y#fb}C=#f~CiVv;HCyY`v2kAR zdrRiTS@t{#iXI7I4zn}0ZKRnUoJzHvcPiP>qUv&mwfO?n4MdwW1-Y&?_hPe%Gomw1;KV7hwfd#hE$=O5Hzw3b4%QVcp;oHt#KbzYt$hN<}|BCC3xed`)2}%sQ=0C)FbQK@Ry~!(A zuUpo5)K&}noKykDc5mKh&z+?Z|7+D-L_8PHkcV;M;;53alX41MyoHS{s~bUJO=KdT z9Ce}(!c~NVM+W4a!xQP8t40;$e0iSjbTX@wK&4IxTFxiLnMy2z)a--Xz-)4Svej0{ zgP}Es@0M45h?0*R;F@j8#u)>IhiS(a|79SvsqL)L(9_;B-^^eNW~qAZ2=jEj99IN0&b!q=&Ns5BX&{!9HX4I{Z(IYl7_7+A{-QdUz^rH$wn&61tu8fGELD;9V&P4+H5WV-c1pytP_ocka3tc^ z$Om2YmUZRd`_gC7kMr%bxyMQZ_KxrUjMI-IbU&}T*8;UWB8iP^5ItSUVglNZhf$3U z(&n^+O_F|zvR{2*XwsgYW#EHyX6Fr>QIn@zuAFcaOGr0QZxwmn5sE=p;5B)#rR9>= zOX!SnA9$29ZoM6s#nu`HYf{T~q8#NT*mrFR1>SZ(p;FO+F0A791-1G)Dh6^46iPpJ zRXl~qTT$YJ&J6f{f)E1=#A!;VwCOfd*;0Q`XUjaQ(@wturZd%ZTAY+)QUS^~WOihC zCT$&kO4JYxh>)D!%Pj+NTL|Ey&e<3$Oywgx{d% ztk)@}gLD<-CIspn4RJn-u?(||9g3W(pG4JQ=dBL57K`%Pa(YEN4mFq_aG|4d8RrU%m&ye>-|~X%lJ&I|9v7z$cg8 z!pmw1ISPeK8y!992x+AdSN( z+9)Sayk}E#vIg4ME6{L8q#ms0}z8bM&xQjt6L_tqu+=+vdXu)J#4OH!AQ_ z3US^xt5k`hk6X}{L9QHAeb!8y%GF&zr#VtlE~o-YglMe1=jMyk6$wtlN$aXo7kL-y zTMl(|JqAd_RpzYp=GbYRh|v~C@$Q5P?jMb1zvPqTJ@`~xLEO$dTW(QC<^v-^k8EKQ9C-%5-je^M|+#9z=n{d^0y1~3u^=K_WOEiO zRbH%+A*`#7;U{jq%|tWk?OnEAka(!Ov=V#(Hb9n5*l_TkbDU@^OJvomqv!XgUX<3o z!ggI}yYBUCUBBp$>{*F7t9qr@U|xl{K-?ij5+rVxlo?g%lHRWUl`5bxHZj{Apru>m zhAp%hk&9XIVhN58PN5z{JC4WcyS!`ao3#nvs;aEah$KQI*}&pp^|nV!L-{_$Vimux z&!k#T+_tkh@4O(2=V#80?dPlu?>SLCRyRp< z^TxUI@57yVZr**(WUNrb_L|PT(QRxftbI%`*Swnn_zf==umZTrx7w8 zLkZU4(z!N0X*J@D^`@~suJunkPxwonJZUW;F_OP$I@FY!#%|*^>0LfLL8m~PGj2ku zm6arvnY%tg3AK=9-Gjg~(rQVlG3a3&bgqP9Z{bI5-m57M-ritkUu&V3W01nKIME02zT3b+yBRvjfll z`sw#i{nqjCJ@dbx`mg-++l4Q@FfelI&e>PLJ)5Y<;{M~;4;_M-1#Fbb%p2YsH4B3B zESrM2Y0$0>#-jHvBqm?-9)whIBP+ceX*BPBL7g5@wzUw2b{+GWL65Sog-IG!y-g$8 z(lkd4B;_G%VUK{LnnhG|J%;2DFbrZdRYXE5qZA=l$?ba^DA(s?s;aL$ngN?EQDBOm9)6Wf}^p(*&(Cqo{=^ftZ$t3j(-SQ2LWJlBVg(5qL!bMZI+^0!wkK zg{~jbKdA)!SAhmtc0)8J|9HcMC~vV6V(Tp+E-A-W>uOfxv)}^?{0;FCr+)XT#F;GD*Y3?QK3ncvS_-lTi!CA)way5 zmy@fsim0UpUbIGjohe?8w*QK&Id=23#}-#;Wrp@3;h0@sU0t48UgeC?T-FwAt%BJZ z9@o<;TM~x1lNZ{pgTcken`)tdaP8>zjrK@z3Moz~TyC55jhYKvE4&jF)Z=oqzufmA9> zL5zs3!h&kDLYG%(+41G;H?LZDJPF*sc6mB??b-xgPrk9ki+)2}jNy#l^;;xwRqx)4 z=+nt7WBFNGb=WYTL>*F)IyI#noVWqQx)?xjvPh0}ZJA6E11E#n0%oIvgp~w%N}ZTE zCxw0nW2f``H$Fc!F!J&x^kPkV@#5b(q!#zqF3k{&!**f`7fO*S#9yfo3tjoc*|;Nm zhHSJ$J&I;hl;93xr}9M=D&3dA@ahm#t{k;Dxh!~MeFqL|{!SKSZg~$PB~WgIQS+!$iFt*J{%RcL zjgvR7Po*+u)rmZ)G|IxthEJ9$@tK2T5o;Y9p_PCP-&T{-d>w*@OqW7ZBukRehHt@q zcs+iZmN;q~0sBEmT|st>cBy-htWYS_%+G}p=Hn-6FTNUcWhJU%HiG>7C(dKmx%IQJ zzW;S?U$l4T(EfqboySar*=;iu2_kHWN0U)07-?k$Bp??gmGN#ztUFmiZ!E9HQR>td zRbv6sVSQ66UnOj!{RVG@Slei*ogy())UxbITCdwjAs8Jcrj{6hN@bgadFpBs=`9Hm zw1$i3^Q1=TpQ>@rf*p&dlW5G`>Q?R8iZ)oZs{j-1=q_o9@|yoH)|-;) z=<*|+Z04>ea97$$I$5ZYI)GrryKA(Uf+>s9q$NAo5$;x=qCNBUy#>*-8Npyzx>$s@ z0jKLH0@NwqU?3u_Q8?Rv*EcS?=l7RBC)&uh1)n^-xcibL9OT9DrFOo-{EQeGkdEOr zb2Y|6vB;xg&Lud>K?g7if-3%CZfY{-F~Bm=#WXNFW{;9C8TJ~(TNl|f{jNHyI5|y- zBzw^rLte&*_Fmgb`jef|tR-qKrmyA^L^9u*p0Iw91oI$s#FegMNI81t*1O1&s3^5U zq_iCfgL8rPM|+T47gDT>T~(9q5vPv)kw8d0x_4$qMG zLQGC4H>NL5Pvm&`#-(YeVfWrd`n77!+7PhBx#I0ridoqKfPBGnR7QgV(YXuMrzdo_ zj{xNQ_su-0dsz)9O^Yq&i4ufQI_ov6{nm_+_cwAsZ;C79!O7&*)o%l z>-qIY3p8lfXxIhzkL4e)eiq$7w|+MF-b;vScJHdw{e?)Wy96aH1N&tdD|7pkmua5D zf`ExN=lY$-A1M49O?nC9<(u*s5y#mGe-LPPmciVe1n6EvdVve*cn{*L3Vg(InryZJL>51e`2cI#cyGN6WrE3QLJIE8F+C51IoDM((^Le91w?g1Oz?|3 zK4*Dwr72VuN(FTz3Y3J>rxjCFR6<+~;Tn_JS7mHHz>rF+LY`T{Cakasj$y|6#mU@S zv9RbgtXIvVLt9;aCL3N1)+dCkEr1ek-444dyw#~f7&d5#f9q+PbWRJ*nyt;!tO3k2 zLe|C3Xh|M;n?GF|%uYK!K%MC=J~MD)@W%ts|L5l(pKd<;t0#Z(iNE=DY4FD%?EcVfqL9gMwvggqxS6bQmL>wzf>q$F@$>rU~$0hhZ1uu z<|@~dp5CF{72>->|EPCvL`nZ?^EEuABT62SFaeRE`E{(SFXIOrt1A^Xq9~_2BIfOw zYti>;U^>fj36b&R&g}ZVCmMLF?B6BaXmM6gN}rS<)u4o>PJu|Ze_kWTZDI{ zmLy4{hv>(SYQ6P{gD4~+Y0IEnFXdz^gXxkiv*3xELOtHxbjZm0d_P=b&n)7jL>qXnwAlVhTSk2=01#SVK=UP*RXjbQ-sBQ-oTrsf*I zRGxFu#D-2w^hG-+Ul~<9ycgG)JTg+A?6@fjKigS0s-V*0m&As+TwkASz2tlJ*{*GfG{z=58GhmnR@>D4+P-TXBeueT6f0euFN3 z9Lu8OqzH^9@Qd=hqhKeXHT))GPl zP8=Cpei=cS@QU@oH6OA}>PoyMjjcNYR^}deNirbKYz|~dK5BgNI#`P=ie#G}I8 zd||1?v5|X=++*VgHU&c))5|d=Tw16VL`3sKA_)tkF{*co8MW<1+tfQ_HL&zo@~XQ2 zBLyT-en?#cp`ElG^bUckuPuugEf;o?lPw*E#9X9jsF~}cR4`YS#j+kdyJXvElEW*3 zf;6}VB1^ut7{#EhA5raynR%n7T%rR%Wr;M7X9glhUBFy)Ahw`yH%UvbPH}>L6$b(( zCX}07jo%HE$~Fi5h6`2Yvtcj6CywVL`BHZKOwLK<2FGxE&q( zb&>qwm;lo6-D{D$E?3uOKH0q?AV1qUMi&evrv(ymb{hBZR7;ma0ON59s_xT(ZMvb| z&F?37FUlW(`9f~5%>kIo-p~;P{mbl=+NyV%-7DH5m*LDbk2+#!oKUD%q)g8V3$R09 zB>i{yn68P&%Y59X9Vw2L3)6v&4JiYLj;N?dn06?*1k@Ez4z8yiQGt$t2kG1!A!U7? zVbTG_5oozA_uPPkc~~g6Yz$O0qp^|ZDisJ+QWNB%HNgtg z@{E#8EG|d~L8|%PH-?{QHz&_t++D=q-WfQ$nE(Qcu5?#kp2%ID0D5!0q-Mp8VJq%xQWn^%$_U5#J2gXl##r z_Zn^{gmygNn%~|QgBp;%{Up%k7UKmDdkTC96bbWXv3$ot%2YLt_6>hM1RR zm5D0J>H$Y(RqB7*n0PiC`-1xB8ju-v-{Z}$UY@p4ggdju8!&sF6Z0598-Ty;o)q;Wa3>UZEVb#wo4(BGeH2aa$f-Ib1&1vwx1`RJHA0lD zwF!gC_L&gRV@L{$U_e*6!nhK8UEiyc0!5NTu1Az)M-pLzN(}^h%&&+bn z2$!Vl@={jXwNe_~94!~*Ub=51MFvW@bA%t~7bDP0m$o^2YZ+HNEH}Y+8M8)$t>O$k zs*=MNbzY}!ZbI`1Gol_jH4W^WqX~x|k#s)zucOnk!>Ui6894p_PW_9M7mk1H6JL1x z=bz%zzrQ$^?6q=)licLlSNA5-&CT87p_<*%9Z>O+Y7x1=D~suRF)qP7!d|C^#EiLC zN7m7=rOKEPTNQIT3I^Op;^-fngD9MOBN8z@YVdM~Yk`Y0h1HI4(Y2$L3zwyUpuVy; zdM@0?$-(|vItlg_IMezj-b`Db>DS@%CP@n*PwpO5fHgR{0|;G^iX3v*S$T7f;-)@i z5|DMWC@`1Qoc~&ag+1i=UmQp5=N1$CyI&=YN4xJFk~>KL-LN)4Q&=p>zNV*dKzz|Y z9qJ^Aoo(Zk_Mwhl$JGO&?MwPp3U%~LasdqHqANpj6M^C~+6sTK9F3K^j7c+R861i* z>AP)8QLE8~2aUr*Cl<=dgl*|*K5p9Fi_30F;tYcZ;VH608{UlfWgW=2#l4u4pUdjA z#0`=jHGlvXBwMoAMsCNlaKJ%!!H9bM(G}Q@j3!R+T;UrSy_i#VV6YF$9>5{>fNa_j z2*=zw`#~qu{a5iTFTIAVo@HAW2M+B@rfq(S5g+)ZSYawH>lF$cnUnrWK3bjV#0HKw z#8A{MhCcT;MJdu^MJ$R`WNh97S746Q6wJSI-55Au(HP*e%M*lxwm4$xQ4{?_Lt+T1aLNC}VA>E24sPpUV0B{w;J znHs9-@yn2WG!@7uj16xy$Bs{vNa*)AI+Kzn(c5VD*$Ial=;ywO?!WpbEx+A_CdGS4 zd=ACCaD+CNBAGz(fxluIyXb6oOq&mK)JYn-Oq5i4^x@P}sLifzE3W1imrL`vNfT^H z9h^XPrtkVT1119A+~lH`CA8nM1NJWRO^`l)>%ekaqRE+s#JJO-?|P6lC3EQomcrpl zRm7BJJJV6*W1CPqqp271ZDu=RW5>WW*}Qct=wAo&7KGm5(=;~IMVVZZ%?|)_t65yo( z{{L-V&hhf@@k3_Nvt;hM(P4k~Y8GMxE{;|pb+=S0*oxXrNeX_f9<2(j8nN#uOdEEv z6D!!E`{5%HB8_J{4FLPq%L+2u#)d-dz#{fS+8}I{9?KL+Ot)!7LkNqDH=r9zN42!# zcVg<>npfI0k{DM`!3eF!<|^zq61Q*Jp^e_AtfU7YdTZy4?tWaDCfzLw=$s|7LIQ)VlIW^0Uf6Kd`<5Gz({ugq?fT&SM zQxiFE3={(N+{D;w)eUswz?!w)n^ePzysK&_py=Q50Pik7aTSYY?F4HnrcMah--&K2PfIqPZp zlJzWPJw?9SYM7M!3LY53T9>nkOvtxk^TH~;#AZaV4bjuC_gPt3Uh|fhigrvL4rT>q z=isRBz04X22w?8Yo4)1QqFh55lcDL$L3msi5MFDjkB7ndAIB`U^ND+s-C(v zda=-biP7GVt_wQbvfTNa%zM_;Y*l6Pjz>6Lhf7r8Cse^N?%m9VBpZxY$rl|rHD zU6~$L2Ng8i4Qxv#cGkp}ETD(EHD*ZB9AcF@uBy++Ng8=^P^y}Lv-tK}2@OtL<^&-X zhQt!}!i)x^&Sw0l@# z?;QKWz~%22&L#U>*MU72mEo>|L2^m5QKgO_yYv z$}m`1Wg9rb4jsunm&AYd9Be7trh^GuRag~lJJ4mDP7W(kB5&6WOkTe>Z3UG`Jrz2vEkBB{qgjxjdj^uuQ}eF*N%t?kOZ6{l zw4S5lxO4|Bwp3bO<~r5oRjYZe-(6lfID@w0WCAG&Zn6X9>G|<>vy5h#$l$P%bi!8i zR}-xgXdPa{fUP1T6>M-PqQwBBv9m1o!m~CSbd{$nF}E#173i0elVNbjgivl~i}L&D zud>@CUjvM6QdiRWDQPKvpRu>iB@zGtVheVvL<2jlajhpL zF>Ii_$7w})g&A3r`xCyU?dCj#s~9Baj?z~Ms+s$9Z1N^VDlU_ix9H#@jbJJ)QdYT1 zi>sBVjXT3>0JvcsH^vUVIgsv;c>AwXO+p;QoCvlKYV${}k5Hw3j|xY+W@1(jhVO$u z%or9mF-vp?#3RsO&gu5jFa2`EN}M{AU*ccRM@hHpY>N&%8+u7q#eJ%(8N91r?c8(U zoch?STV5jJxBUF$^~)&cs|9XgJjb$c?N6ucb{(TznHSQM8T%klK{u=z`w^7qCUVet zqHmHF(jwm@8MwG2K|sVAcf1K}A>~_Mn4g6wx2z${2H900j1e5L%AENMezp!CRm8J z)umME&aD^c@7yh|d28H}PGf~t3DrF(sc1Z8p+Ds$#%+4A{^Tu&9I#OWF5Ug2|6E-V zOXw#k((q4ND#3XXHQHi7Sf$nwsBo%t!Qn#-%Sg-0A*@TrrCT+!yMFl!3Mg!lX85{+ zPx#63s|K9Betm*QzHd&UUFTmp_v+(k(XPe4VyazlL6LCJQ(Ut5TH99J8Tk>?Pic4W z%x&JgN0I@GCz}yD=Rb^V60xZWHB$j;O+(9Q)i65Bp4MB({Ozoc+Y5QOG zMRaWl$$66#BoNTnVr|?7LQO1}Y<xF2O@iavu*;h~j8zJ%Z__*Etefv>WLaiK>Yh zKFEQuc@X=PkW_0(6-au9J?OU%Y!f9nxFN`?&09`defMzWzO?s5t};4UL{PRD9Yhc~|0Cm=9kfq=V$h*bUv zNj6f{5N1V@5KVsn^OGpS^}@Lv=lDbh526I@KR4{PgL;``r&yCocywjyj@>H*(C5c2YTwT)OB~;hH;L zx-!`%kRdQXhl7LVfsvN@9bZ(+zwe3_wMeX z+aH&g-p$q{z_*E{Tulgv1TUF^A04UUo?>nq(=V;E93>=@mi?EbNAhP}JC%|RkaG~X zYYtCw@kb3uzdYh^#bXCDkUx=DIsrLd`RlCw{d)cd7XJ2`i^+RKtbG5><5yFUoQdmT zs}*}m_~nS`;YDlvFq&l1hiP=sn*a9>9S+j8?833OcG0!8@hE4wb>)zbIGRmypBI49 zw@wec_e9^;H^^z_(yYvnFHuW9BnN%d!lRwgeIRC9*P3!GEf^_#wWN!eZB#kdE^;J8 zC^nW$6RnCv%hpUHXr-JTPv!2ef{PGO09^B93$0GL2R$}g8HXg4EG$xT?1$KvMSpvCs)-Z|I@f-~&^4I*|3jm>TI~>~$ z;<{BB(Ko6&*GSWe5e+(i4M?NIkWL)BSHXmP2PPSX@mZ(0%@62jn5v6`nN`rt8a+0& zI@~Q5N~5!WMGcfit$ruhx4aO?!>&&QG@@( z$U|My?uvXmacHC`y35rd^S9qtU_mJk>ej{t3=X|Az<;yz cjYK9A_QFHrdc1AxAX`sm ZmhDE2j1!rsx3M$oE>mC>SQfyT006lKKa2nX diff --git a/research/boxes-and-escape/evidence/folklore.sql b/research/boxes-and-escape/evidence/folklore.sql new file mode 100644 index 0000000..6cfc4ed --- /dev/null +++ b/research/boxes-and-escape/evidence/folklore.sql @@ -0,0 +1,307 @@ +-- Six-continent pig-escape survey, 2026-07-28. +-- +-- Loaded after seed.sql. Full agent reports live in ../survey/. +-- +-- BLANKET VERIFICATION WARNING: all six agents hit the same wall this session -- +-- WebFetch returned 403 on every host (Wikipedia, JSTOR, sacred-texts, Ulukau, +-- Te Ara, archive.org, open-access PDFs alike) and each exhausted its 200-query +-- search budget. NOT ONE PRIMARY TEXT WAS READ IN THIS ENTIRE SURVEY. Every row +-- below is a real, locatable citation assembled from search-result summaries. +-- Nothing here is publication-ready until a human opens the source. +-- +-- Confidence grades are the agents' own, carried through unchanged. + +BEGIN TRANSACTION; + +-- --------------------------------------------------------------------------- +-- AFRICA -- escape motif WEAK in oral narrative; exclusion material dominant +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('Africa','Ancient Egypt','Herodotus II.47-48 on swineherds','historical_event','boundary_taboo', + 'Pork treated as impure; casual contact required immediate immersion; swineherds alone barred from every temple and forced into caste endogamy. The boundary is drawn around the pig, then drawn again permanently around the people who touch pigs.', + NULL,1,'solid','Herodotus, Histories II.47-48.', + 'Volokhine complicates Herodotus -- pigs WERE farmed in the New Kingdom. Do not take at face value.'), + + ('Africa','Ancient Egypt','Set as the black boar','myth','boundary_taboo', + 'Set takes the form of a black boar; Ra declares the pig an abomination to Horus.', + NULL,1,'probable','Egyptian mythological corpus; specific papyrus not confirmed.',NULL), + + ('Africa','Sub-Saharan (Zambia, Kenya)','Farmers who decline to confine pigs','historical_event','escape_enclosure', + 'Veterinary literature records farmers rejecting confinement outright, citing inability to confine; tethered pigs deliberately released each evening. An entire disease ecology (T. solium, a leading cause of acquired epilepsy in Africa) rests on the practice.', + NULL,0,'solid','Thys et al., Veterinary Parasitology (2016); Parasites & Vectors (2022); Thomas et al., BMC Vet Res 9:46 (2013).', + 'The continent-scale literal answer, and it sits in the epidemiology rather than the folklore.'), + + ('Africa','Egypt (Coptic Zabbaleen)','The 2009 Cairo pig cull','historical_event','boundary_taboo', + '300,000 pigs culled on a swine-flu pretext with zero positive tests, destroying the Coptic Zabbaleen waste-collectors'' livelihood. Cairo''s organic waste collection then collapsed and rubbish piled in the streets.', + NULL,0,'solid','Contemporary news reporting, 2009.', + 'The pigs were the infrastructure. The containment that failed was Cairo''s, not the pigs''.'), + + ('Africa','Kenya','The 2013 "MPigs" protest','historical_event','uncatchable', + 'Boniface Mwangi''s activists released a dozen-plus piglets painted "MPigs", with pig blood, at the gates of Parliament. Police fired tear gas and were then reduced to chasing piglets across the parliamentary flowerbeds.', + NULL,0,'solid','Contemporary news reporting, 2013.', + 'The survey''s best single image. Pigs turned loose INSIDE the most guarded enclosure in the country, and the state''s failure to catch them WAS the argument.'), + + ('Africa','Sotho-Tswana (BaLobedu)','Kolobe as clan totem','ritual','boundary_taboo', + 'Wild pig as seboko (totem) of BaLobedu clans including the Modjadji Rain Queens. Same prohibition on eating, opposite logic: protected as kin rather than excluded as filth.', + NULL,1,'probable','Southern African ethnographic literature; specific source not confirmed.',NULL), + + ('Africa','Yoruba','Ijapa and Eledẹ ("why the pig roots the ground")','tale','no_containment', + 'The pig''s eternal rooting explained as a permanently unsuccessful pursuit of a debtor tortoise.', + NULL,0,'thin','Yoruba tale corpus; agent could not read the ending.', + 'Closest African escape candidate. The agent explicitly labelled its escape reading as INFERENCE, not finding.'), + + ('Africa','Pan-African','ABSENCE: the pig is not a trickster','tale','no_containment', + 'No African ATU 2030 analogue, no ATU 124 analogue, no escaping-pig proverb located. The trickster role is held by hare, tortoise and spider.', + NULL,NULL,'solid','Negative finding across the agent''s full search.', + 'STRUCTURAL FINDING. Corroborated independently by the Asia agent (mousedeer). The escape-trickster slot is constant; the animal cast in it varies with husbandry and taboo.'); + +-- --------------------------------------------------------------------------- +-- ASIA -- literal pen-escape near-absent as a tale type; four other clusters +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('Asia','Goguryeo (Korea)','The runaway sacrificial pig (郊豕), Samguk Sagi','historical_event','escape_enclosure', + 'Three escapes across two centuries. Yuri yr 19 (1 BCE): the pig bolts, two officials catch it and cut its leg tendons; the king executes them for maiming a victim consecrated to Heaven and then falls ill from their vengeful ghosts. Yuri yr 21 (2 CE): it bolts again, the keeper Seolji chases it to Gungnae and reports the terrain -- the capital is moved there and stays roughly 400 years. Sansang, 208 CE: it bolts again, is caught by a woman of Jutong village, and the son she bears the king is named 郊彘 ("sacrificial piglet"), the future King Dongcheon.', + NULL,1,'solid','Samguk Sagi, via National Institute of Korean History (db.history.go.kr, contents.history.go.kr) and Encykorea.', + 'BEST ITEM IN ASIA. A runaway pig relocates a state and produces a king. Note that the men who successfully PREVENTED an escape are the ones executed for it.'), + + ('Asia','Rabbinic Judaism','Bava Kamma 82b -- the pig hoisted over the wall','literary','boundary_taboo', + 'During the 65 BCE siege of Jerusalem the besiegers hoist a pig over the wall in place of a lamb; midway it digs its hooves into the wall and the land quakes. The sages then curse anyone who raises pigs and anyone who teaches his son Greek wisdom.', + NULL,0,'solid','Babylonian Talmud, Bava Kamma 82b.', + 'The one recorded moment a pig gets IN rather than out. The curse pairs pig-rearing with foreign learning -- two kinds of boundary breach in one breath.'), + + ('Asia','China','Zhu Bajie (猪八戒), Journey to the West','literary','boundary_taboo', + 'His containment is vow-based rather than architectural: 八戒 means the eight precepts, a nickname given to fence his appetite. His identity is a list of prohibitions he continually breaches. His fall into a sow''s womb is a reincarnation clerical error. His one explicit act of confinement is performed on a human -- he locks his Gao Village wife away for six months.', + NULL,1,'solid','Wu Cheng''en, Journey to the West (16th c.).', + 'He is named after his own guardrails.'), + + ('Asia','Japan','Fukushima exclusion-zone boar-pig hybrids','historical_event','escape_enclosure', + 'Farm pigs abandoned in the 2011 evacuation went feral and interbred with wild boar. ~16% of boars sampled in the zone are hybrids; pig ancestry ~8% and declining.', + NULL,1,'solid','Anderson et al., Proceedings of the Royal Society B (2021).', + 'A mass escape recorded in genomes rather than in stories. The declining ancestry means the domestic is being reabsorbed.'), + + ('Asia','East Asia (idiom stock)','Boar as forward charge, never evasion','idiom','no_containment', + 'The boar''s uncontainability in the idiom stock is always headlong forward motion -- 猪突猛進 (chototsu moshin), 狼奔豕突, 封豕長蛇 -- and never evasion or hiding.', + NULL,1,'solid','Standard Chinese/Japanese idiom dictionaries.', + 'A different GRAMMAR of uncontainability: unstoppable rather than uncatchable. Justifies keeping escape_enclosure and uncatchable as separate classes.'), + + ('Asia','Tibet','Samding Dorje Phagmo -- the monastery turned to pigs','myth','transformation', + 'Escape from attackers achieved by transforming the entire monastery into pigs. The pig-body is the hiding place rather than the thing that flees.', + NULL,0,'probable','Tibetan Buddhist hagiographical tradition.',NULL), + + ('Asia','Malaysia','Selangor pig-farm relocation politics','legal','boundary_taboo', + 'The 2026 Bukit Tagar pig-farm relocation fight, with PAS counter-proposing the non-Muslim island of Pulau Ketam; and the 2012 pig-heads-at-mosques incidents.', + NULL,0,'solid','Contemporary Malaysian news reporting.', + 'The contemporary form of the theme is not a pen. It is a map.'), + + ('Asia','Malay-Indonesian','ABSENCE: the escape-trickster is the mousedeer (kancil)','tale','no_containment', + 'In the largest Muslim-majority region on earth, the small-animal trickster whose entire repertoire is slipping traps is the mousedeer, not the pig. No Asian analogue of ATU 2030 or indigenous ATU 124 located.', + NULL,NULL,'solid','Negative finding across the agent''s full search.', + 'STRUCTURAL FINDING, independently matching Africa. The genre is strongest exactly where the pig has been evicted from narrative.'); + +-- --------------------------------------------------------------------------- +-- EUROPE -- densest containment material; much of it legal rather than literary +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('Europe','Welsh','Twrch Trwyth','myth','uncatchable', + 'The treasures are stripped from the boar one by one across an enormous chase, but the boar itself is never contained -- he escapes into the sea. Rhys read the chase as an extended dindsenchas of swine place-names.', + NULL,1,'solid','Culhwch ac Olwen, in the Mabinogion.',NULL), + + ('Europe','Welsh','Henwen','myth','uncatchable', + 'Near-exact structural twin of Twrch Trwyth and badly under-used. Arthur sets out to destroy her and FAILS; she crosses the whole island farrowing monsters at named places and ends at the sea.', + NULL,1,'solid','Welsh Triads; swineherd Coll ap Collfrewy.', + 'Wales has TWO uncatchable-swine epics. With Pryderi''s Otherworld pigs and the Mochdref toponyms it is a regional pattern, not a single text.'), + + ('Europe','Welsh','Pryderi''s Otherworld pigs','myth','escape_enclosure', + 'Otherworld pigs driven across Wales, leaving the Mochdref ("pig-town") toponyms behind them.', + NULL,1,'probable','Mabinogi, Fourth Branch.',NULL), + + ('Europe','English','The Old Woman and Her Pig','tale','escape_enclosure', + 'The entire plot is a pig that will not cross a stile -- a structure purpose-built to pass people and stop livestock. Europe''s purest refusal-to-stay-put tale is organised around an interface with a type constraint in it.', + 'ATU 2030; Motif Z41',1,'solid','Jacobs, English Fairy Tales (1890), from Halliwell.', + 'A stile is a schema. Verify the Thompson motif number against the printed index.'), + + ('Europe','English','ATU 124 -- the three little pigs','tale','no_containment', + 'FLAGGED INVERSION: the pig is besieged, not escaping -- the story is about keeping something OUT. Fox/geese variants show the pig is not essential to the type.', + 'ATU 124',1,'solid','Aarne-Thompson-Uther index.', + 'Belongs in the residual. The most famous pig-and-enclosure story in English is about a wolf failing to get in.'), + + ('Europe','England','Urban swine management and the volume of regulation','legal','escape_enclosure', + 'Overturns the popular image: medieval English town pigs were NOT free roamers but subject to cradle-to-grave controls. Pinfolds; six swineherds amerced in 1425; an owner could not retrieve his own strayed pig except through the manor court. The burden runs unbroken to s.4 Animals Act 1971.', + NULL,1,'solid','Dolly Jorgensen, "Running Amuck? Urban Swine Management in Late Medieval England", Agricultural History 87:4 (2013), 429-451.', + 'METHODOLOGICALLY THE MOST USEFUL ITEM IN THE SURVEY. The volume of rule-making is itself the evidence of chronic escape. You do not write that many rules about a thing that stays put.'), + + ('Europe','France','The Savigny sow, 1457','legal','escape_enclosure', + 'A sow hanged for killing a child; her six piglets acquitted as having been led astray by maternal example. The best-documented of the medieval pig trials.', + NULL,1,'solid','Medieval French court records; Evans, The Criminal Prosecution and Capital Punishment of Animals (1906).', + 'The pig trials are containment-failure aftermath -- a penned pig does not reach a child in a cradle. Falaise 1386''s human-clothing and fresco details are the WEAKEST-evidenced elements in that literature.'), + + ('Europe','London','The tantony pig and the 1311 loophole','legal','escape_enclosure', + 'St Anthony''s Hospital belled pigs the London market judged unfit for slaughter, making them legally free of the street -- a licensed exception to containment, fed by citizens. In 1311 a hospital tenant, Roger de Wynchester, was forced to promise the City not to claim wandering pigs nor bell any swine but those given in charity.', + NULL,1,'solid','London civic records, 1311.', + 'M1-SEAM in a medieval city: the single lawful exemption permitting an uncontained pig was immediately being used to launder other people''s escapees.'), + + ('Europe','Greek','The Erymanthian Boar and Eurystheus','myth','uncatchable', + 'Eurystheus demands the boar be brought back alive, and then hides in a buried jar when it arrives. The man who ordered the pig contained ends up the one in the container.', + NULL,1,'solid','Greek mythological corpus (Apollodorus).',NULL), + + ('Europe','Britain','Wild boar restored by fence failure','historical_event','escape_enclosure', + 'No reintroduction programme: the Great Storm of October 1987 smashed farm fences in Kent and East Sussex; a 1990s escape near Ross-on-Wye seeded the Forest of Dean; ~60 were illegally dumped at Staunton in 2004; 1-2 escape incidents annually 1989/90-2008/9.', + NULL,1,'solid','UK wildlife and DEFRA reporting.', + 'An extinct species restored to a country entirely by containment failure.'); + +-- --------------------------------------------------------------------------- +-- NORTH AMERICA -- no native suid; everything downstream of introduction+escape +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('North America','Colonial English','Fence law inverted, and the hog reeve','legal','social_boundary', + 'Livestock described as "the principal agents responsible for dispossessing the Indians". The Chesapeake statute -- "Every man shall enclose his ground with sufficient fences uppon theire owne perill" -- inverted English practice by putting the burden of exclusion on the crop-grower, so colonial pigs were legally entitled to Indigenous fields. The hog reeve was among the earliest elected offices in colonial North America.', + NULL,0,'solid','Virginia DeJohn Anderson, "King Philip''s Herds", WMQ 51:4 (1994), 601-624; Creatures of Empire (Oxford UP, 2004).', + 'The darkest and best-sourced thread in the survey. The first thing colonial democracy did was elect someone to deal with loose pigs.'), + + ('North America','US/British','The Pig War, 1859','historical_event','social_boundary', + 'Lyman Cutlar shot Charles Griffin''s HBC Berkshire boar in his unfenced potato patch on San Juan Island, 15 June 1859. Escalation was about whose law applied; joint military occupation ran to 1872. The pig was the only fatality.', + NULL,0,'solid','BC Studies, "From Imbroglio to Pig War"; US National Park Service.', + 'A border dispute triggered by an animal crossing a line.'), + + ('North America','New York City','The Piggery War, 1859','historical_event','social_boundary', + '87 armed men, per the New York Times of 27 July 1859. A street-foraging pig costs nothing to feed, so free range was a subsistence strategy for the landless and enclosure was gentrification. Women were the most militant defenders.', + NULL,0,'solid','Catherine McNeur, Journal of Urban History 37:5 (2011), 639-660; Taming Manhattan (Harvard UP, 2014).', + 'Containment as a class weapon. The push to enclose was a transfer of who gets to survive in a city.'), + + ('North America','Caribbean Spanish','Cimarron / boucanier / jibaro','idiom','social_boundary', + 'Cimarron was applied first to domestic livestock gone wild in the hills of Hispaniola and only afterwards to escaped Indigenous and African people, giving English "maroon". Boucanier (buccaneer) meant a man who lived by hunting the feral cattle and hogs left when Spanish Hispaniola depopulated. Jibaro -- Pichardo (1836): "montaraz, rustico, indomable" -- was used of masterless animals gone wild before it named the mountain peasant and then the Puerto Rican national type.', + NULL,0,'solid','Pichardo, Diccionario provincial de voces cubanas (1836); Oviedo (1535); colonial Caribbean lexicography.', + 'INDEPENDENTLY CORROBORATED by the South America agent. Three defining Caribbean forms of life outside colonial control are named after animals that got out first. HANDLE WITH CARE: this runs directly through chattel slavery and marronage. Not available for borrowing as a metaphor.'), + + ('North America','US vernacular','"Root hog or die"','proverb','no_containment', + 'Proverbial in the Vermont Gazette by 1829 and in Crockett''s Narrative (1834). Presupposes the free-ranging hog: you have been turned loose, forage or perish.', + NULL,0,'solid','Vermont Gazette (1829); David Crockett, Narrative (1834).',NULL), + + ('North America','Spanish colonial','De Soto''s thirteen pigs','historical_event','escape_enclosure', + 'Thirteen pigs landed at Tampa Bay in 1539 became roughly 700, escaping continuously across a 3,100-mile march. Columbus''s 1493 introduction required a crown order to cut the population within twelve years.', + NULL,0,'solid','Colonial Spanish records.',NULL), + + ('North America','Canada','"Super pigs"','historical_event','escape_enclosure', + 'Escaped from 1980s-90s wild-boar diversification farms; now occupying ~750,000 km2 and expanding ~88,000 km2 per year, sheltering in self-dug "pigloos".', + NULL,0,'solid','Ryan Brook, University of Saskatchewan.',NULL), + + ('North America','US internet','"30-50 feral hogs"','idiom','uncatchable', + 'August 2019 viral formulation, now a fixed idiom for an absurd but genuine uncontainable threat.', + NULL,0,'solid','Contemporary reporting, 2019.',NULL), + + ('North America','US literary','Charlotte''s Web -- ch. 3, "Escape"','literary','escape_enclosure', + 'Wilbur finds a loose board, gets out of the pen with the goose urging him on, discovers he has no idea what to do with the outside, and follows a bucket of slops straight back in.', + NULL,0,'probable','E. B. White, Charlotte''s Web (1952), ch. 3.', + 'DESCRIBED FROM MEMORY, NOT FROM THE TEXT. Beats confident: loose board, goose, slops, voluntary return. Exact wording unverified.'), + + ('North America','US literary','Charlotte''s Web -- the words in the web','literary','transformation', + 'Wilbur''s real enclosure is a date, not a fence, and he escapes it without moving: Charlotte writes SOME PIG, TERRIFIC, RADIANT, HUMBLE, and the description of him changes. He never leaves the pen; he leaves the category. The message works because of its container -- a web is an impossible place for text, so the humans read it as being about the pig rather than about the author.', + NULL,0,'solid','E. B. White, Charlotte''s Web (1952). Rachel Dean-Ruzicka, "Advertising the Self: The Culture of Personality in E. B. White''s Charlotte''s Web", Jeunesse 6:1.', + 'THE FORMAT ARGUMENT, STATED IN 1952. White worked ~2 years as a copywriter at the Frank Seaman agency before joining The New Yorker in 1925. Same work appears twice under different theme_class -- the schema could not hold it in one row, which is the defect this piece is about.'), + + ('North America','Mesoamerica','OPEN LEAD: release of impounded game (A1421)','myth','escape_enclosure', + 'Thompson motif A1421 -- the dueno del monte who keeps the game penned until someone releases it. A peccary version in Maya ethnography would be the strongest possible Indigenous item here and would rhyme exactly with the Mundurucu sty myth.', + 'Motif A1421',1,'thin','NOT FOUND. Next step: Thompson, Ethnology of the Mayas (1930); Braakhuis, Xbalanque''s Marriage.', + 'Needs a human with library access.'), + + ('North America','US search results','CONTAMINATION: fabricated Indigenous material and Grokipedia','historical_event','no_containment', + 'Searching javelina plus Indigenous tradition returns, at the top of results, uncited "Tohono O''odham and Yaqui" symbolism from spirit-animal content farms with no narrator, collector or publication. Grokipedia (LLM-generated) surfaced in roughly eight result sets.', + NULL,NULL,'solid','Agent''s own search log, 2026-07-28.', + 'BELONGS IN THE PIECE. Research into what models absorbed from human narrative is already being contaminated by machine-generated encyclopedia entries about that same narrative.'); + +-- --------------------------------------------------------------------------- +-- SOUTH AMERICA -- native peccaries plus twice-feral Iberian introductions +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('South America','Mundurucu','The origin of wild pigs','myth','escape_enclosure', + 'Karusakaibe transforms humans into pigs, who are first kept in a pig-sty in the village and killed one by one, until someone lets them out and they flee into the forest and become the wild pigs of today. Contemporary tellings add the Tapajos crossing, the sacred passage at Macapa/Mukapap, a variant where Karosakaybu re-traps the pigs between mountains, and a son who crosses over and stays with them.', + NULL,1,'solid','Robert F. Murphy, Mundurucu Religion, UCPAAE 49(1), 1958. Treated in Levi-Strauss, The Raw and the Cooked.', + 'KEYSTONE. Not a story with an escape in it -- an origin story in which the species IS the escape. Berkeley open-access PDF blocked; must be read. Sty detail graded PROBABLE.'), + + ('South America','Pan-Amazonian','White-lipped peccary disappearance cycles','historical_event','uncatchable', + '43 documented disappearance events across nine countries and 88 years of harvest data; 7-12 year troughs in 20-30 year cycles, synchronised across up to 5 million km2. The paper incorporates Indigenous testimony explaining disappearances as caused by the death of a powerful shaman, with return securable only through another shaman''s ritual work.', + NULL,1,'solid','Fragoso et al., PLOS ONE (2022).', + 'Peer-reviewed, and it takes Indigenous explanation seriously as data rather than colour.'), + + ('South America','Brazilian','Caipora / Curupira as godfather of the herds','myth','uncatchable', + 'Rides a caititu or queixada, travels with the peccary herds, steers the pigs away from hunters'' traps, and bargains a quota of animals for tobacco, cachaca and cloth.', + NULL,1,'probable','Popular Brazilian folklore sources; Camara Cascudo attribution unverified.', + 'An explicit uncatchability mechanism with an agent behind it.'), + + ('South America','Wari'' / Ese Eja','Peccaries as the dead returning','myth','transformation', + 'THE COUNTER-MOTIF. Wari'' ancestors return as white-lipped peccaries and approach kin hunters deliberately, so that their meat feeds their own relatives. Movement inward, not outward.', + NULL,1,'probable','Aparecida Vilaca, Wari'' ethnography.', + 'Important residual case: a pig-crossing story where the crossing is a return, not an escape.'), + + ('South America','Argentina','The jabali release chain','historical_event','escape_enclosure', + '1906 San Huberto reserve (now Parque Luro, La Pampa); 1909 Carpathian stock released into an 800-hectare enclosure; Pedro Luro''s bankruptcy lets them off the property; escapes 1914-1930; a 1931 accident seeds Patagonia.', + NULL,0,'probable','Argentine environmental history.',NULL), + + ('South America','Brazil','IBAMA''s 1998 javali breeding ban','legal','escape_enclosure', + 'IBAMA banned javali breeding in 1998, and breeders responded by releasing their stock. The containment regulation is what set the animal loose.', + NULL,0,'probable','Brazilian environmental regulation and reporting.', + 'Directly analogous to the open-weights argument. A restriction produced the proliferation it was meant to prevent.'), + + ('South America','Colonial Spanish','Puercos cimarrones','idiom','social_boundary', + 'Oviedo describes puercos cimarrones in 1535. The same term covered escaped livestock and escaped people; Argentines still say chanchos cimarrones.', + NULL,0,'solid','Oviedo (1535); colonial Spanish lexicography.', + 'Independently corroborated by the North America agent. See the handling caution on that row.'); + +-- --------------------------------------------------------------------------- +-- OCEANIA -- three incompatible answers to the same question +-- --------------------------------------------------------------------------- + +INSERT INTO folklore (continent, culture, title, item_type, theme_class, description, motif_index, is_native, confidence, source_note, notes) VALUES + ('Oceania','Hawaiian','Kamapua''a -- serial failed containment','myth','uncatchable', + 'Four times the guards -- eight hundred strong and increasing each time -- capture him in hog shape and tie him to a pole; four times his grandmother releases him with a chant. Bound for sacrifice on a heiau, he escapes because the priest Lonoaohi had instructed his sons to only PRETEND to tie him: the ropes are theatre.', + NULL,1,'solid','Martha Beckwith, Hawaiian Mythology (1940), ch. XIV. Lilikala Kame''eleihiwa, A Legendary Tradition of Kamapua''a (Bishop Museum Press, 1996), translating an anonymous 1891 Ka Leo o ka Lahui serial.', + 'THE ROPES ARE THEATRE. A constraint that looks binding from outside, staged by someone inside the system who arranged for it not to hold. Kame''eleihiwa reportedly reads the 1891 timing, two years before the overthrow, as anti-colonial defiance -- VERIFY her introduction.'), + + ('Oceania','Hawaiian','Kamapua''a at Kaliuwa''a','myth','escape_enclosure', + 'He becomes a giant hog so his people can climb his back out of a box canyon. The place (Sacred Falls, O''ahu) is named for the escape.', + NULL,1,'solid','Beckwith, Hawaiian Mythology (1940).',NULL), + + ('Oceania','Hawaiian','Kamapua''a as humuhumunukunukuapua''a','myth','transformation', + 'Fleeing Pele''s fire he becomes the humuhumunukunukuapua''a -- the escape-form is named after the thing escaping.', + NULL,1,'solid','Beckwith, Hawaiian Mythology (1940).',NULL), + + ('Oceania','Tsembaga Maring (PNG)','The kaiko festival trigger','ritual','escape_enclosure', + 'The festival is not scheduled. It is triggered when the herd grows until pigs invade gardens and the labour burden on women becomes insupportable. An entire ritual calendar keyed to the moment fences stop working.', + NULL,0,'solid','Roy Rappaport, Pigs for the Ancestors (1968).', + 'Hedge the functionalist theory; the observation is secure.'), + + ('Oceania','Vanuatu','Tusker boars','ritual','social_boundary', + 'THE INVERSION. Upper canines avulsed so the lower ones curl unimpeded -- full circle in 6-7 years, double in 10-12, eventually puncturing the animal''s own jaw and requiring surgical care. Required for grade-taking; the boar''s tusk is on the national flag.', + NULL,0,'solid','Vanuatu ethnographic literature.', + 'Containment as decade-long artwork, where the constraint is what produces the value. The exact opposite theory of the pig from Kamapua''a, in the same ocean.'), + + ('Oceania','British/Pacific','Cook''s deliberate releases','historical_event','escape_enclosure', + 'Cook left breeding pairs on islands as POLICY, not accident -- so that a future wrecked British ship would find protein waiting. Escape reconceived as infrastructure. In the same decade he is recorded describing Malakula''s tusks bent into perfect closed circles.', + NULL,0,'solid','Cook''s voyage journals.', + 'THE OPEN-WEIGHTS SECTION. He did not lose those pigs, he invested them. Release because proliferation is the point, versus the tusker''s cultivated constraint -- two opposite theories of the pig, one man, the 1770s.'), + + ('Oceania','New Zealand','"Captain Cooker"','idiom','uncatchable', + 'New Zealanders named their uncatchable feral pig after the man who released the founders. The escape is memorialised in the escapee''s surname.', + NULL,0,'solid','New Zealand vernacular; Te Ara.',NULL), + + ('Oceania','Maori (Waima)','The pigs in the tapu kumara plantations','tale','boundary_taboo', + 'Pigs entered tapu kumara plantations and, because of the tapu, no one could go in and remove them. The grunting from inside the sacred ground convinced people they were gods.', + NULL,0,'solid','Te Ao Hou.', + 'A containment failure caused by ritual law rather than bad fencing -- the rule protecting the space prevented its own enforcement.'), + + ('Oceania','Aotearoa/Pacific','ABSENCE: no deep proverbial stock','proverb','no_containment', + 'No documented Tok Pisin pig proverb and no Maori whakatauki about poaka. Pigs reached Aotearoa only after 1769 -- too recent to sediment into the proverbial layer.', + NULL,0,'solid','Negative finding across the agent''s full search.', + 'DATING CONTROL FOR THE WHOLE SURVEY. Proverbs need centuries. Where pigs are recent arrivals you get ecology and newspapers; where ancient, idiom. Proverb density is a rough clock.'), + + ('Oceania','Polynesian','NOT FOUND: Maui and pig','myth','no_containment', + 'No reliable Maui-and-pig episode exists.', + NULL,1,'dubious','Negative finding. Agent recommends omitting entirely.', + 'Recorded so nobody goes looking for it twice.'); + +COMMIT; diff --git a/research/boxes-and-escape/survey/pig-africa.md b/research/boxes-and-escape/survey/pig-africa.md new file mode 100644 index 0000000..ee58be0 --- /dev/null +++ b/research/boxes-and-escape/survey/pig-africa.md @@ -0,0 +1,643 @@ +# Pigs, Escape and Containment in African Tradition — Research Findings + +**Continent:** Africa (North Africa, Sahel, West, East, Central, Southern Africa, Madagascar, +plus flagged diaspora notes). + +--- + +## ⚠️ METHOD AND LIMITATIONS — READ FIRST + +Two hard constraints shaped this research and must be weighed against every claim below: + +1. **All direct page fetches failed.** The sandbox's outbound HTTPS gateway returned `403` to + every `CONNECT` for the entire session (verified via `curl "$HTTPS_PROXY/__agentproxy/status"`, + which logged `connect_rejected — gateway answered 403 to CONNECT (policy denial)` for + `en.wikipedia.org`, `sacred-texts.com`, `jstor.org`, `scholar.google.com`, `ncbi.nlm.nih.gov`, + `yorubatales.com`, `afrolegends.com`, and others). **I could not read a single primary text, + PDF, tale collection, or journal article directly.** Everything below rests on web-search + result snippets and titles. +2. **Search budget exhausted** (200/200 queries) before I could close several open threads — + flagged inline as OPEN. + +Consequence: nothing here should be treated as verified to publication standard. Items marked +SOLID mean "multiple independent search results agree and the underlying source is a real, +citable academic/primary work." They still need one confirming read before an essay quotes them. +Items marked THIN or DUBIOUS should not be used without independent verification. + +--- + +## Summary + +**How much material is there?** Much less narrative pig material than an equivalent survey of +Europe or Southeast Asia would yield — and the reason is itself the finding. + +Across roughly half of the continent by area (the Maghreb, Sahara, Sahel, Horn, Nile valley, +Swahili coast, Comoros, and large parts of Madagascar), the pig is not a farm animal at all. It +is a **prohibited category**. Where an animal is not kept, it cannot escape, and no folklore of +escape can accumulate around it. What accumulates instead is an enormous body of **boundary +material**: dietary law, purity rules, transformation-punishment narratives, occupational +exclusion of swineherds, village-level purification rites, and clan totem prohibitions. + +So the honest structural answer to the research question is: + +> **In Africa, the pig's relationship to boundaries is overwhelmingly one of *exclusion*, not +> *escape*.** The dominant image is not the pig that gets out of the pen; it is the pig that must +> be kept outside the wall, outside the temple, outside the village, outside the cooking pot, +> outside the person. The escape motif, where it appears, is (a) faint in oral narrative, (b) very +> strong in modern veterinary/agricultural documentation of free-roaming village pigs, and (c) +> strongest of all in the *inversion* found in one spectacular modern political incident (Kenya's +> 2013 "MPigs" protest), where pigs were deliberately released *into* a guarded enclosure. + +Where suids **do** appear as narrative characters in sub-Saharan oral literature, they are almost +always **warthog** or **bushpig**, not the domestic pig, and they appear in etiological "why is he +like that" tales — why he kneels, why he's ugly, why he backs into his burrow, why he roots the +ground. Several of these *do* turn on a hole, a burrow, a failed escape from a predator, or a +permanent unsuccessful search. That is the closest African oral literature gets to the +containment theme, and it is genuinely there — but the sourcing for these particular texts is +weak (see below). + +**The pig is not an African trickster.** In the major African trickster cycles — Anansi/Ananse +(Akan), Ijapa the tortoise (Yoruba), Sungura/hare (Bantu East Africa), Uhlakanyana (Zulu), Zomo, +Leuk — the pig is never the trickster. It is at best a foil, a dupe, or a creditor. This matters +for the essay: the animal-that-won't-stay-put role in Africa is occupied by the **hare**, the +**tortoise** and the **spider**, not the pig. + +--- + +## Escape / Containment Motifs + +### (a) Pigs escaping enclosures — oral narrative + +**Finding: essentially absent as a folktale motif.** I found no African folktale, in any tradition +searched, whose plot is a domestic pig escaping a pen, sty, tether, or rope. This is a real +negative, not a search failure — see *Notable Absences*. + +The nearest analogues are burrow/hole tales about wild suids: + +--- + +**1. "Why the Warthog Goes About on His Knees" / "Warthog Tries the Jackal's Trick"** +*Attributed:* Zulu (South Africa). **Confidence: THIN as to attribution; PROBABLE that a real +tale of this shape exists.** + +Plot as circulated: Warthog has made a home in an old termite mound that an **aardvark** had +hollowed out. A lion stalks toward the entrance. Trapped in his own hole, Warthog attempts an +escape trick he has heard Jackal boast about — he braces his tusks against the roof and shouts +that the roof is caving in and he needs help holding it up, hoping the lion will take over the +burden while he runs. Lion recognises the borrowed ploy immediately. Warthog collapses onto his +knees in terror. Lion, not hungry, sentences him to stay on his knees — which is why warthogs +feed kneeling, snout in the dust, rear in the air. + +*Why it matters here:* this is a **containment-and-failed-escape** story. The warthog is trapped +inside a hole that is not even his own (he is a squatter in an aardvark's excavation — a +biologically accurate detail), and the escape attempt fails because the trick is stolen. The +punishment is a permanent bodily posture: he is fixed in place forever. + +*Sourcing problem:* the tale circulates on `worldoftales.com`, `canteach.ca` (which labels it +"A Traditional Zulu Story"), `afrolegends.com`, `mocomi.com`, and several travel blogs +(`roadtravel1.wordpress.com` / `africaroadtravel.com`, `wildmoz.com`). **None of them names a +collector, informant, or published collection.** I could not trace it to Callaway (*Nursery Tales, +Traditions and Histories of the Zulus*, 1868), Bleek, Junod, or any academic Nguni corpus. +OPEN: needs a check against Callaway and against Nguni *izinganekwane* scholarship. + +*However* — the embedded trick is real and important. "Persuade the predator to hold up the +rock/roof while I escape" is a genuinely widespread and well-indexed tale type +(**ATU 1530, "Holding Up the Rock"**), extremely well attested in African oral tradition and +carried into the African diaspora as a Brer Rabbit episode. The warthog version reads as a +**failed-imitation variant** of a real African escape-type. That structural claim is PROBABLE. + +--- + +**2. "Why the warthog enters his burrow backwards"** +*Attributed:* East Africa, unspecified. **Confidence: THIN — possible modern safari-guide +confection.** + +Circulated plot: vain Warthog leaves his burrow with tail up. Porcupine, tired from a night of +foraging, sneaks into Warthog's empty burrow to nap. Warthog wallows, shows off to Lion, is +charged, bolts for home, and dives **head-first** into his own hole straight onto the sleeping +porcupine — permanently disfiguring his face. Humbled, he thereafter always **backs in**, so he +can see what is coming and never be surprised at the threshold again. + +*Why it matters here:* this is precisely a **threshold** story. The refuge itself becomes the +trap; the animal is caught in the act of crossing his own boundary; and the resolution is a +permanent change in *how he crosses a boundary*. Warthogs genuinely do reverse into burrows and +explode out of them forwards, tusks first — the ethology is solid even if the tale is not. + +*Sourcing problem:* found only on tourism/safari sites (`kenyawildparks.com`, +`travelbutlers.com`, `africafevers.com`, `africageographic.com`). **I could not find a single +academic or archival attestation.** Treat this as very likely a guide's campfire etiology that +has crystallised in the safari industry rather than a documented oral text — which is itself an +interesting phenomenon (modern commercial folklore-manufacture), but it is not evidence of +traditional narrative. **Do not cite as "an African folktale" without an archival source.** + +--- + +**3. Yoruba: "Ijapa (Tortoise) and Ẹlẹdẹ (Pig)" — Yoruba title reported as +*Ìdí tí elédè fi má ń fimú túlẹ̀* ("Why the pig always roots the ground with its snout")** +**Confidence: PROBABLE for the tale and its etiological function; THIN for the exact escape +mechanism.** + +Reported plot: In a town called Lagoni, Ijapa the tortoise and Ẹlẹdẹ the pig are close friends. +Ẹlẹdẹ is a prosperous travelling trader — generous, but with a temper and a hatred of being taken +advantage of. When Ijapa is ruined by creditors, Ẹlẹdẹ pays them off and stakes Ijapa in a farming +business with a loan. Ijapa does not repay. Ẹlẹdẹ, enraged, sets out for Ijapa's house to collect. +The tale is explicitly framed as explaining **why the pig is forever searching for something in +the ground.** + +*Why it matters here:* if the punchline runs as the title implies, this is the strongest genuinely +African instance of the theme in the whole survey — a **debtor who escapes into the earth and can +never be caught**, with the pig's rooting as the eternal, permanently unsuccessful pursuit. The +tortoise is the classic Yoruba escape-artist; the pig is the party who cannot close the boundary +around what is owed him. + +*Caveat — important:* I have the title, the setup, and the stated etiology from search snippets +of `yorubatales.com` ("Yoruba Folktales Revisited", tale #35) and a YouTube retelling +(*Itan Ijapa ati Elede*). **The site returned 403 and I could not read the ending.** The specific +claim that "Ijapa hides underground and the pig digs for him forever" is my reconstruction from +the title plus the summary, not a read text. **Mark as inference.** +OPEN: verify against Oyekan Owomoyela, *Yoruba Trickster Tales* (University of Nebraska Press, +1997) and against Ọlatunde Ọlatunji / Ọlabimtan Yoruba *àló* collections. The Ijapa corpus is +genuinely and thoroughly documented in academic sources; this specific tale should be findable. + +--- + +**4. Akan/Ashanti: "How the pig got his snout"** — reported as an Ananse-cycle tale, possibly in +Peggy Appiah, *Ananse the Spider: Tales from an Ashanti Village*. **Confidence: THIN.** Surfaced +only as a title in passing; I could not confirm its contents, its collection, or whether it has +any containment theme. Listed only so the thread isn't lost. OPEN. + +--- + +### (b) Documented modern containment failure — the real body of evidence + +This is where the escape theme actually lives in Africa, and it is very well documented — just in +veterinary and agricultural journals rather than folklore archives. **Confidence: SOLID.** + +**5. Free-roaming village pigs and the acknowledged impossibility of confinement.** +Sub-Saharan smallholder pig-keeping is overwhelmingly a **free-range/scavenging** system. The +literature is explicit that farmers *cannot* contain their pigs and, in many cases, *decline* to: + +- **Zambia (Eastern Province):** Thys et al., *"Why pigs are free-roaming: Communities' + perceptions, knowledge and practices regarding pig management and taeniosis/cysticercosis in a + Taenia solium endemic rural area"*, **Veterinary Parasitology** (2016). Finding: confinement is + **not accepted** by farmers as a control method; participants knew free-ranging exposed them to + disease risk but cited **"inability to confine them"** and production/weight-gain needs. +- **Zambia:** *"Movements of free-range pigs in rural communities in Zambia: an explorative study + towards future ring interventions for the control of Taenia solium"*, **Parasites & Vectors** + (2022), PMC9044682. Finding: **tethered pigs are routinely released in the evenings** to + scavenge — i.e. the containment is deliberately, cyclically abandoned every single day. +- **Kenya (western):** Thomas et al., *"The spatial ecology of free-ranging domestic pigs + (Sus scrofa) in western Kenya"*, **BMC Veterinary Research** 9:46 (2013), PMC3637381. Finding: + free-range pigs spend much of their time scavenging **outside their homesteads** and are + therefore exposed to pathogens over a wide area. +- **Uganda:** PLOS NTD (2025) on *T. solium* in Oyam district — same risk structure. + +*Why it matters:* this is the literal, documented, continent-scale version of "the animal that +refuses to stay put." An entire disease ecology (*Taenia solium* cysticercosis — a leading cause +of acquired epilepsy in Africa) rests on the fact that African village pigs are structurally +uncontained. The pen fails; the tether is untied at dusk; the pig walks into the neighbourhood +and eats human waste and comes back. This is a much better essay spine than any folktale I found. + +--- + +**6. Bushpig (*Potamochoerus larvatus*) as the uncatchable crop-raider.** +**Confidence: PROBABLE-SOLID.** + +Nocturnal, elusive, highly intelligent, uses dense cover; regarded as a major agricultural pest +across southern and eastern Africa. Reported in the IUCN Wild Pig Specialist Group / regional +wildlife literature that in the **DRC and Malawi** bushpigs are reputed to cause **more damage to +agriculture than any other species**, and — the key line — **"attempts to control or eradicate +*Potamochoerus larvatus* in these areas have usually proved unsuccessful."** Their ecology and +behaviour remain poorly known precisely *because* they are so hard to observe or catch. + +Madagascar is a distinctive case: bushpig is present (introduced) alongside domestic pigs, and +their contact across the farm boundary is now itself a research subject — *"Assessment of domestic +pig–bushpig (Potamochoerus larvatus) interactions through local knowledge in rural areas of +Madagascar"*, **Scientific Reports** (2024), PMC11250805. A genuine documented case of the wild +and the domestic pig crossing the enclosure line in both directions. + +--- + +**7. North African wild boar (*Sus scrofa algira*) — the uncontainable haram animal.** +**Confidence: crop damage SOLID; taboo-as-cause PROBABLE/THIN.** + +Wild boar populations in the Maghreb are expanding and inflicting severe crop damage: +- **Algeria (Guelma, northeast):** documented farmer reports of intrusions into wheat, maize, + vegetables and fruit trees, incidents ranging from <1 ha to >5 ha. +- **Tunisia (Gabès and Kébili governorates, southern oases):** boar damage to cereals, vegetables, + date palms and alfalfa recorded across **58% of surveyed oases**. + +The tempting reading — that the population explodes precisely *because* the animal is haram and +therefore neither hunted nor eaten, so the taboo that keeps it outside the boundary is the same +thing that makes it impossible to keep out of the fields — is **attractive but not yet properly +sourced.** The one search result touching this made the religious-restraint argument about +**Iran**, and only noted it "may parallel" North Africa. That is inference, not evidence. +**Do not assert it.** OPEN: needs a real citation (French-language Algerian/Tunisian agronomy +literature on *sanglier* would be the place — terms: *sanglier*, *dégâts aux cultures*, +*battues administratives*; Maghrebi Arabic colloquial: *ḥallūf*). + +--- + +### (c) Pigs in stories with no containment theme + +For clean separation, these are the pig items I found that carry **no** escape/boundary theme: + +- **Swahili proverb** *"Mkuki kwa nguruwe mtamu, kwa mwanadamu uchungu"* — "A spear is sweet for + a pig, bitter for a human." Theme is empathy/reciprocity, not containment. The pig is simply the + legitimate target of violence. +- **Yoruba ritual use:** the pig (*ẹlẹdẹ*) appears among *ẹbọ* offering materials in Ifá practice. + No containment theme. (Sourcing: popular Yoruba-religion sites only — THIN.) +- **Etiological "why is the warthog ugly"** tales (God made the warthog handsome; he grew vain and + rude; he was chased to his den) — vanity/humility moral. The chase-to-the-den detail is a faint + containment echo but the tale is about pride. Sourcing: safari sites, THIN. +- **"How the pig got his snout"** (Ashanti, unverified) — appears to be pure etiology. + +--- + +## Pig as Boundary / Taboo + +This section carries the bulk of the real, well-sourced material. + +### Ancient Egypt — the pig as the animal expelled from the sacred enclosure +**Confidence: SOLID for Herodotus; PROBABLE for the Book of the Dead attribution.** + +**8. The Black Pig / Set's boar form.** In Egyptian tradition, **Set takes the form of a black +boar** and attacks **Horus**, injuring/burning his eye. **Ra thereupon declares the pig "an +abomination to Horus."** This is the mythological charter for Egyptian pig avoidance. The text is +generally located in the **Book of the Dead, spell/chapter 112** (the "Souls of Pe" material) — +*flag: I could not open the source to confirm the chapter number; verify before citing.* +Popularised in M. A. Murray, *Ancient Egyptian Legends* (1913), ch. VII "The Black Pig" (text on +sacred-texts.com, which I could not fetch). + +**9. Herodotus II.47–48 — swineherds as a caste kept outside the boundary.** Herodotus reports +that pork was *impure* in Egypt; that an Egyptian who touched a pig even in passing would +immediately plunge into the river, clothes and all, to purify himself; and — crucially for this +essay — that **swineherds alone among Egyptians could not enter any temple**, could not sacrifice +to the gods, and **their children could only marry among other swineherds**. A pig sacrifice was +nonetheless permitted at the **full moon**, specifically because Set-as-pig had injured Osiris/Horus. + +*Why this is the single richest boundary item in the whole survey:* the taboo is not merely +dietary. It produces a **hereditary human caste physically excluded from the sacred enclosure and +from the marriage pool** — the boundary is drawn around the pig and then drawn again, permanently, +around the people who touch pigs. That is "the animal that must be kept outside the boundary" in +its most literal institutional form. + +*Scholarly caution:* Egyptologists have complicated Herodotus. Pigs *were* kept and eaten on New +Kingdom farms (tomb evidence), so the taboo was likely priestly/class-specific and regional rather +than universal. See Youri Volokhine on Egyptian food prohibitions (summarised at *The Ancient Near +East Today*, ANE Today) — a proper, citable modern treatment. Frazer, *The Golden Bough*, +"Osiris, the Pig and the Bull" is the classic (now dated) discussion. + +### Islam — across North Africa, the Sahel, the Horn and the Swahili coast +**Confidence: SOLID (primary text).** + +**10. Pork as *ḥarām*.** Qur'an 2:173, 5:3, 6:145, 16:115 (*laḥm al-khinzīr*). This is the single +most consequential fact about pigs in Africa. It governs food, commerce, land use, and settlement +across the Maghreb, Egypt, Sudan, the Sahara, the Sahel belt (Senegal through northern Nigeria to +Chad), Somalia, Djibouti, the Comoros, Zanzibar and the Swahili coast, and a large share of +Madagascar's Muslim populations. It is the reason the escape motif has no soil to grow in across +that whole belt. + +**11. *Maskh* — humans transformed into pigs.** Qur'an 2:65, 5:60, 7:166: Sabbath-breakers among +the Children of Israel transformed into **apes and swine** as punishment for transgressing a +prohibition. The majority Sunni exegetical position treats this as a literal physical +metamorphosis (a minority, following Mujāhid, read it metaphorically). + +*Why it matters here:* this is **boundary-crossing in the most literal available sense** — the +penalty for crossing a divine boundary is to *become* the animal that lives outside the boundary. +The prohibited animal is not merely excluded; it is the destination of exclusion. Textually SOLID; +its circulation as *African* folk narrative specifically is PROBABLE but I found no African +folklore citation. OPEN. + +**12. The pig created inside the Ark, from dung.** Islamic etiological legend, recorded by +**al-Ṭabarī (d. 310/923)**: aboard Noah's Ark the two elephants produced unmanageable heaps of +dung; Noah touched the elephant's tail (variant: was commanded to strike/poke the elephant's +trunk) and **two pigs emerged and set to eating the dung.** **Al-Jāḥiẓ** notes the story is +"widely known in the marketplace and among popular storytellers" — i.e. it was genuinely popular +oral currency, not just a bookish curiosity. + +Source: **Remke Kruk, "The saddest beast? Notes on the pig in Arabic culture" (2020)**, in a +festschrift volume; available via Academia.edu / ResearchGate. **Confidence: SOLID for the +legend's existence in Arabic tradition; PROBABLE for circulation in Arabic-speaking Africa +(Egypt, Sudan, the Maghreb).** This is the best single academic source I identified for the pig +in the Arabophone world and should be the first thing read to firm up the North African sections. + +*Why it matters here:* it is a gorgeous containment image and I'd flag it as essay-worthy. The +pig is **born inside the sealed vessel** — the Ark being the ultimate container, the one enclosure +in which every animal is required to stay put — and it is **created out of waste, to consume +waste.** It has no place outside the boundary because it was manufactured inside it as a sanitation +device. Compare directly with item 15 below (Cairo's Zabbaleen pigs), where a real African city +used pigs for precisely this function and then destroyed them. + +### Christian and Jewish Africa + +**13. Ethiopian and Eritrean Orthodox Tewahedo pork prohibition.** **Confidence: SOLID.** Based on +Leviticus 11 / **Deuteronomy 14:8** ("the pig is also unclean; although it has a divided hoof, it +does not chew the cud"). Distinctive within Christianity — Eastern European Orthodox Christians eat +pork freely; the Ethiopian and Eritrean churches do not, reflecting the church's long-standing +adoption of Hebraic purity practice. + +**14. Beta Israel (Ethiopian Jews).** **Confidence: PROBABLE.** Kashrut observed; those who broke +the food taboos were **ostracised** and had to undergo a purification process — fasting for one or +more days, eating only uncooked chickpeas supplied by the *kes* (priest), and ritual purification +**before being allowed to re-enter the village.** The boundary here is not metaphorical: it is +the village perimeter, and the transgressor is physically outside it until purified. + +### Madagascar + +**15. *Fady*.** **Confidence: PROBABLE.** Pork is *fady* (taboo/sacred prohibition) for **most +Sakalava** and for the **Antandroy**. *Fady* is region-, clan-, village- and even family-specific, +usually decreed by elders, and *fady indrazana* (ancestral taboos) tie a person to their lineage. +Malagasy terms: **kisoa** (domestic pig), **lambo** (wild pig/bushpig). Sourcing is mostly +travel/cultural sites (MadaMagazine, urlaub-auf-madagaskar.com) plus a Mongabay piece on taboos and +conservation — adequate for the general claim, thin for specifics. OPEN: verify against +anthropological literature on Sakalava and Antandroy (Lesley Sharp, Gillian Feeley-Harnik on +Sakalava would be the places to look). + +### Southern Africa — the inverted taboo + +**16. The wild pig as clan totem (*seboko*) among Sotho-Tswana peoples.** +**Confidence: PROBABLE.** The wild pig — **kolobe** (Sotho-Tswana) / **goloe** (Khelobedu) — is +one of the most common animal totems among the **BaLobedu**. The wild-pig clans (**Dikolobe**) +include Modjadji, Mohale, Modika, Mahasha, Mabulana, Mokwebo, Mampeule, Molokwane, Thobela and +Ramafalo. The **Mokwebo** are described as the ancestral wild-pig clan (*ba bina kolobe*), and the +BaLobedu were formerly known as **Bakwebo** ("wild pigs"). A *seboko* prohibits clan members from +harming or eating the totem animal. + +*Why it matters here:* this is the **inversion** of the Islamic/Egyptian pattern and worth the +essay's attention. The prohibition runs the same way — *do not eat this animal* — but the logic is +reversed: the pig is not excluded as filth but protected as kin. The boundary drawn around the pig +is a boundary of *identity*, not *disgust*. And notably the **Modjadji** lineage — the famous Rain +Queens of Balobedu — is among the wild-pig clans. Sources: Wikipedia *List of Sotho-Tswana clans* +and *Diboko*, `101lasttribes.com` (Lobedu), National Museum (South Africa) publications on totems. +OPEN: firm this up from Ziervogel or Krige & Krige, *The Realm of a Rain-Queen* (1943). + +--- + +## Proverbs & Idioms + +**Headline negative finding: I found no African proverb about a pig escaping, breaking out, or +refusing to be contained.** After searching Swahili, Hausa, Yoruba and Arabic material, nothing +matching the theme surfaced. Given the "escaping pig" proverb is a natural form (English "pigs +might fly," "when the pig gets out"), its absence looks structural rather than accidental. + +**17. Swahili — *Mkuki kwa nguruwe mtamu, kwa mwanadamu uchungu.*** +"A spear [thrown] at a pig is sweet; at a human being, bitter." +Meaning: what is enjoyable when done to another is agony when done to you. +**Confidence: PROBABLE-SOLID** — this is a standard, widely anthologised *methali*. Verify against +the **University of Illinois Swahili Proverbs project** (`swahiliproverbs.afrst.illinois.edu`), +which is a genuine academic resource, or Kitula King'ei & Ahmed Ndalu, *Kamusi ya Methali za +Kiswahili*. +*Relevance:* no containment theme, but note the presupposition — the pig is the creature it is +*normal* to spear. It is definitionally outside the circle of moral consideration. On a Muslim +coast, that is the taboo speaking. + +**18. Swahili — *Ukitaka kula nguruwe, chagua aliyenona.*** +"If you want to eat pig, choose a fat one." +Meaning: if you are going to do a thing, do it thoroughly — the Swahili "go the whole hog." +**Confidence: PROBABLE.** +*Relevance:* **this one is directly a boundary-crossing proverb.** Its entire force depends on the +hearer knowing that eating pig is transgressive. The wisdom is: *if you are going to cross the +line, cross it properly.* Sourced from proverb-list sites; verify against the Illinois project. + +**19. Hausa — "Though the lion is humbled, he won't play with the pig."** +**Confidence: DUBIOUS / UNSOURCED. Do not use.** This surfaced only on a Steemit post and Nairaland +forum threads — exactly the profile of the fabricated "African proverb" the brief warns about. If +real, it would be *karin magana*; it should be checkable against Kirk-Greene, *Hausa ba dabo ba ne: +A Collection of 500 Hausa Proverbs* (1966) or Merrick's collection. **I could not check it.** + +**20. Arabic — *khanzīr* ("pig") as an insult** across Egypt, Sudan and the Maghreb, connoting +filth and moral exclusion. Commonplace and certainly real, but **I have no scholarly citation in +hand. Confidence: THIN as presented.** Kruk (2020) would cover it properly. + +**21. Hausa lexicon note:** "pig" in Hausa is *alade / aladu* (cf. Yoruba *ẹlẹdẹ* — the Hausa term +looks to be a Yoruba loan, itself a small piece of evidence about which direction pig-keeping +travelled in West Africa). Roger Blench, *"A history of pigs in Africa"*, in Blench & MacDonald +(eds.), *The Origin and Development of African Livestock* (UCL Press, 2000), and Blench's +*"Domestic Pigs in Africa"*, **African Archaeological Review** (2012), are the authoritative +sources on exactly this kind of linguistic evidence. **Both are real, citable, and I could not +open either.** OPEN — high priority; this is the best scholarly foundation available for the whole +question of where African pigs came from and who kept them. + +--- + +## Documented Real Escapes + +**22. The 2009 Egyptian pig cull and the Zabbaleen.** **Confidence: SOLID.** +In May 2009, citing swine flu, the Egyptian government ordered the culling of the country's +**entire pig population — roughly 300,000 animals** — at a point when **no person or pig in Egypt +had tested positive**. The pigs belonged overwhelmingly to the **Zabbaleen**, a community of some +60,000 Coptic Christians who collected Cairo's household waste, fed the organic fraction to their +pigs, and sold the pork to hotels and restaurants. The cull hit the livelihoods of ~70,000 people. +Because the Zabbaleen no longer had any use for organic waste, **they stopped collecting it, and +Cairo's rubbish piled up in the streets.** Egypt subsequently had to turn back to the Zabbaleen and +their pigs to manage its waste crisis. +Sources: France24 *Observers* (12 May 2009); *Christian Science Monitor* (3 Sept 2009); IRIN / +The New Humanitarian (May and Oct 2009); *Penn Political Review* (Dec 2020). + +*Note the framing carefully:* these pigs did **not** escape — they were killed. But the episode is +squarely on-theme in a different way, and it is the strongest documented item in the whole survey. +A minority religious community was permitted to keep the forbidden animal precisely because it did +the city's dirtiest boundary-work: converting what the city expelled back into food. When the state +destroyed the pigs — using an epidemiological pretext to act on a religious and ethnic boundary — +**the city's waste stopped staying where it was put.** The containment that failed was the city's, +not the pigs'. Pair this with the Ṭabarī Ark legend (item 12): the pig created inside the sealed +vessel to eat the dung, and then, twelve centuries later in the same country, destroyed for it. + +**23. Kenya, June 2013 — the "MPigs" / Occupy Parliament protest.** **Confidence: SOLID.** +Activists led by photojournalist **Boniface Mwangi** released **more than a dozen piglets and one +large pig — painted with the word "MPigs" — plus pig blood, at the gates of the Kenyan Parliament +in Nairobi**, protesting MPs' demands for a pay rise. Around 250 protesters marched through central +Nairobi and staged a sit-in at the legislators' entrance. Police responded with tear gas, water +cannon and batons — and were reduced to **chasing piglets around the parliamentary flowerbeds while +the pigs grazed on them.** Roughly 10–16 people were arrested, including Mwangi, who reported being +charged with animal cruelty. +Sources: Associated Press (via NBC News and omaha.com); Al Jazeera, *The Stream*, 14 May 2013 and +11 June 2013; The Polis Project profile of Mwangi. + +*Why this is the most striking item in the survey:* it is the escape motif **deliberately +weaponised and exactly inverted.** Every other item here concerns keeping the pig out. Here, pigs +were released *into* the most heavily guarded enclosure in the country, and the state's inability +to contain them — police officers scrambling after piglets on the parliamentary lawn — *was the +political argument.* The uncatchability of the pig became a public demonstration of the +uncatchability of the corrupt. The animal chosen was the animal of greed, filth and +boundary-violation, and it was let loose inside the boundary. + +**24. Transvaal, early 1970s — escaped sow × bushpig hybrids.** **Confidence: THIN / DUBIOUS.** +A claim that a domestic sow escaped from a farm in the Transvaal, mated with a bushpig, and +produced **eight offspring with bushpig traits, said to be prolific**. This would be a textbook +literal "pig escapes the fence and joins the wild population" incident. **But the only source I +found is `messybeast.com`** (an amateur genetics/hybrids compilation) — almost certainly rehashing +an older published note that I could not trace. **Do not cite without finding the original.** OPEN. + +**25. African swine fever and the boundary that cannot hold.** **Confidence: PROBABLE-SOLID for the +general structure; specific figures need checking.** +ASF is the defining containment problem of African pig-keeping. It is maintained in a **sylvatic +cycle** in **warthogs** and *Ornithodoros* soft ticks — meaning the reservoir sits permanently +*outside* any fence a farmer can build. FAO guidance notes explicitly that **eradication of all +sources of ASF is not feasible in areas with sylvatic reservoirs.** Control therefore falls back on +culling, quarantine, movement permits and control zones. Recent example: an outbreak in **Tshwane, +Gauteng, South Africa** forced the culling of **more than 60,000 pigs**. +Sources: FAO, *African Swine Fever* portal and *Manual on the Preparation of ASF Contingency Plans*; +South African Dept. of Agriculture, *ASF Disease Management Strategy* (July 2024); Kaya959 news +(Gauteng outbreak). +OPEN — and I regard this as the most promising unfinished thread: **South Africa has a formally +gazetted ASF control-area boundary line across the north of the country**, dating (I believe) to +the 1930s, which legally separates the warthog-reservoir zone from the rest of the national herd. +If that is right, it is a **literal cartographic line drawn across a country to contain pigs**, and +it would be the perfect closing image. **I ran out of search budget before I could confirm the date +or the legal instrument. Verify before use.** + +**26. Kano's pig farm.** **Confidence: DUBIOUS.** A circulating claim that the **United African +Company** established what was then **the world's largest pig farm at Kano** in 1943, that pork +was railed south to Lagos through the 1950s–60s, and that the farm **folded in the late 1970s "due +to religious prejudice."** If true it is a remarkable boundary story — an industrial-scale haram +enterprise inside the great Muslim city of the Sahel, eventually pushed out. **But my only sources +are a Nairaland forum thread and a WordPress blog.** Not usable as-is. OPEN: UAC corporate records +and Nigerian agricultural history would settle it. (Separately and more reliably: the "Kano Brown" +is a recognised indigenous Nigerian pig breed, hardy and adapted to hot dry conditions — which does +imply a real long-standing pig population in the region.) + +**27. Diaspora note — flagged as out-of-scope-ish.** The **Bois Caïman** black-pig sacrifice +(Haiti, 1791) and the **eradication of the Haitian Creole pig** under PEPPADEP (1982–83) are both +powerful containment stories, and the Vodou ritual context is Dahomean/Fon in origin. But the +Creole pig itself descends from **Iberian** stock, not African, and the historicity of the Bois +Caïman ceremony is contested among historians. **I did not research these in this session.** +Mentioned only so the essay can decide whether to pursue them. Do not use my summary as a source. + +--- + +## Notable Absences + +These are genuine findings, not gaps in effort. + +1. **No African version of ATU 2030, "The Old Woman and Her Pig."** This is the great + refusal-to-move pig tale of world folklore — the pig that will not go over the stile, requiring + a chain of helpers. Cumulative/chain tales are **extremely widespread in Africa** (the search + literature notes they are most widespread in Africa and India — e.g. "A Chain of Circumstances" + from the Fang of Equatorial Guinea/Gabon/Cameroon). **But I found no African ATU 2030 with a + pig in it.** The South African literacy NGO **Nalibali** publishes a version in South African + languages — that is a modern translation programme, not indigenous tradition, and should not be + mistaken for one. +2. **No African analogue of ATU 124 ("The Three Little Pigs").** The great pig-containment tale + of Europe — pigs building enclosures to keep a predator out — has no African counterpart I could + find. Where the motif exists in the Anglophone African world it is imported schoolbook material. +3. **The pig is not a trickster anywhere in Africa.** Across Anansi/Ananse (Akan), Ijapa (Yoruba), + Sungura/hare (Bantu), Uhlakanyana (Zulu), Zomo (Hausa) and Leuk (Wolof), the escape-artist role + belongs to hare, tortoise and spider. The pig, where present, is the dupe or the creditor. **If + the essay is looking for the African animal that refuses to stay put, it is the hare and the + tortoise, not the pig.** +4. **Near-total absence of domestic-pig narrative across the Islamic belt** — Maghreb, Sahara, + Sahel, Nile, Horn, Swahili coast, Comoros. This is the largest single finding by area. No pigs + kept ⇒ no pens ⇒ no escapes ⇒ no escape stories. The boundary material fills the vacuum + completely. **Report this as the structural answer, not as a failure to find things.** +5. **No verified African escaping-pig proverb** (see Proverbs section). +6. **No indigenous African boar-hunt epic** comparable to the Calydonian Boar, Twrch Trwyth, or the + Erymanthian Boar — i.e. no "great uncatchable boar that must be hunted down" cycle. Roman North + Africa produced boar-hunt imagery (the hunting mosaics of Tunisia are famous), but that is a + Roman provincial tradition, not an indigenous African one, and **I could not verify any specific + mosaic in this session — treat as UNVERIFIED.** +7. **The warthog/bushpig tales I did find are poorly sourced.** The most on-theme narrative items + (kneeling warthog, backwards-into-the-burrow) live almost entirely on **safari-tourism websites** + with no collector, informant or archive named. I flag the real possibility that some of these + are **modern commercial folklore** — guide-lodge etiologies retrofitted onto observed animal + behaviour and then labelled "traditional African." That is worth saying out loud in the essay; it + is a live problem in African folklore on the open web. + +--- + +## Sources + +Ranked roughly by reliability. **None was read directly** — see the method warning at the top. + +**Academic / primary (identified but unread — verify these first)** +- Remke Kruk, *"The saddest beast? Notes on the pig in Arabic culture"* (2020) — festschrift + chapter; via Academia.edu / ResearchGate. **Highest-value source identified.** +- Roger Blench, *"A history of pigs in Africa"*, in Blench & MacDonald (eds.), *The Origin and + Development of African Livestock* (UCL Press, 2000). +- Roger Blench, *"Domestic Pigs in Africa"*, **African Archaeological Review** (2012), + doi:10.1007/s10437-012-9111-2. +- Herodotus, *Histories* II.47–48 (pig impurity; swineherd exclusion; full-moon pig sacrifice). +- Qur'an 2:173, 5:3, 6:145, 16:115 (*khinzīr*); 2:65, 5:60, 7:166 (*maskh*). +- Deuteronomy 14:8; Leviticus 11. +- al-Ṭabarī (d. 310/923) and al-Jāḥiẓ on the Ark/dung origin of the pig — *via* Kruk (2020). +- Youri Volokhine on ancient Egyptian food prohibitions — summarised at *The Ancient Near East + Today* (anetoday.org). +- Thys et al., *"Why pigs are free-roaming…"*, **Veterinary Parasitology** (2016), + S0304401716301947. +- *"Movements of free-range pigs in rural communities in Zambia…"*, **Parasites & Vectors** (2022), + PMC9044682. +- Thomas et al., *"The spatial ecology of free-ranging domestic pigs (Sus scrofa) in western + Kenya"*, **BMC Veterinary Research** 9:46 (2013), PMC3637381. +- *"Assessment of domestic pig–bushpig (Potamochoerus larvatus) interactions through local + knowledge in rural areas of Madagascar"*, **Scientific Reports** (2024), PMC11250805. +- *"Pig production in Africa: current status, challenges, prospects and opportunities"*, + PMC11016696. +- FAO, *African Swine Fever* portal; FAO, *Manual on the Preparation of ASF Contingency Plans*; + South African DALRRD, *ASF Disease Management Strategy* (July 2024). +- IUCN SSC Wild Pig Specialist Group, *Introduced and Feral Pigs* (iucn-wpsg.org). +- Frazer, *The Golden Bough*, §"Osiris, the Pig and the Bull" — classic, dated, use with care. +- M. A. Murray, *Ancient Egyptian Legends* (1913), ch. VII "The Black Pig". + +**To consult for verification of the weak items (not yet consulted)** +- Oyekan Owomoyela, *Yoruba Trickster Tales* (Univ. of Nebraska Press, 1997) — for the Ijapa/Ẹlẹdẹ tale. +- Henry Callaway, *Nursery Tales, Traditions and Histories of the Zulus* (1868) — for the warthog tale. +- A. H. M. Kirk-Greene, *Hausa ba dabo ba ne: A Collection of 500 Hausa Proverbs* (1966) — for item 19. +- University of Illinois **Swahili Proverbs** project (swahiliproverbs.afrst.illinois.edu) — items 17–18. +- Krige & Krige, *The Realm of a Rain-Queen* (1943) — for the Lobedu kolobe totem. +- Peggy Appiah, *Ananse the Spider: Tales from an Ashanti Village* — for "How the pig got his snout". +- R. S. Rattray, *Hausa Folk-lore* (1913); *Akan-Ashanti Folk-Tales* (1930). + +**Journalism (reliable for the modern incidents)** +- France24 *Observers*, "Egypt's pig cull — all about the money?" (12 May 2009). +- *Christian Science Monitor*, "For Egypt's Christians, pig cull has lasting effects" (3 Sept 2009). +- IRIN / The New Humanitarian, "EGYPT: Pig cull hits livelihoods" (May and Oct 2009). +- *Penn Political Review*, "The Pigs of Cairo: Swine Flu, Garbage People, and the Cull of 2009" (2020). +- Associated Press via NBC News / omaha.com, "Kenyans set pigs loose outside parliament…" (2013). +- Al Jazeera, *The Stream*, "Kenyans #OccupyParliament" (14 May 2013); "Kenyans rally against MP pay + hike" (11 June 2013). + +**Low-quality web sources — flagged, use only as leads** +- yorubatales.com ("Yoruba Folktales Revisited"), tale #35 — the Ijapa/Ẹlẹdẹ tale. +- worldoftales.com; canteach.ca; afrolegends.com; mocomi.com; wildmoz.com; + roadtravel1.wordpress.com / africaroadtravel.com — the warthog tales. +- kenyawildparks.com; travelbutlers.com; africafevers.com — the backwards-burrow etiology. +- messybeast.com — the Transvaal hybrid claim. +- nairaland.com; steemit.com — the Kano pig farm claim and the Hausa "proverb". **Both unusable.** +- MadaMagazine; urlaub-auf-madagaskar.com; Mongabay (this last is decent) — Malagasy *fady*. +- 101lasttribes.com; Wikipedia *List of Sotho-Tswana clans* / *Diboko*; + nationalmuseumpublications.co.za — the kolobe totem. + +--- + +## Confidence Notes + +| # | Item | Category | Confidence | +|---|---|---|---| +| 8–9 | Egypt: Set as black boar; pig "abomination to Horus"; swineherds barred from temples and endogamous | Boundary/taboo | **SOLID** (Herodotus); PROBABLE for BD ch.112 attribution | +| 10 | Islamic pork prohibition across N/W/E Africa | Boundary/taboo | **SOLID** | +| 11 | Qur'anic *maskh* — transgressors become swine | Boundary-crossing | **SOLID** as text; PROBABLE as African folk narrative | +| 12 | Pig created on the Ark from elephant dung (Ṭabarī; al-Jāḥiẓ on its popularity) | Boundary/origin | **SOLID** in Arabic tradition; PROBABLE for African circulation | +| 13 | Ethiopian/Eritrean Orthodox pork prohibition | Boundary/taboo | **SOLID** | +| 14 | Beta Israel: ostracism + purification before re-entering the village | Boundary/taboo | PROBABLE | +| 15 | Malagasy *fady* on pork (Sakalava, Antandroy) | Boundary/taboo | PROBABLE | +| 16 | Sotho-Tswana/Lobedu *kolobe* wild-pig totem (incl. Modjadji lineage) | Boundary/taboo (inverted) | PROBABLE | +| 5 | Free-roaming village pigs; farmers' stated inability to confine; tethered pigs released nightly | **Escape/containment** | **SOLID** | +| 6 | Bushpig as uncatchable crop-raider; eradication attempts "usually unsuccessful" | **Uncatchability** | PROBABLE-SOLID | +| 7 | North African wild boar crop damage (Algeria, Tunisian oases) | Uncontainability | SOLID for damage; **THIN** for the taboo-causes-it reading | +| 22 | Cairo 2009 pig cull; Zabbaleen; waste crisis | Containment (historical) | **SOLID** | +| 23 | Kenya 2013 "MPigs" release at Parliament | **Escape (staged/inverted)** | **SOLID** | +| 25 | ASF sylvatic reservoir in warthogs; containment structurally impossible; Tshwane 60,000 cull | Containment | PROBABLE-SOLID; **SA control-line claim UNVERIFIED** | +| 3 | Yoruba Ijapa & Ẹlẹdẹ; "why the pig roots the ground" | **Escape/uncatchability** | PROBABLE for tale + etiology; **THIN** for the escape mechanism (my inference) | +| 1 | Zulu warthog in aardvark burrow; failed ATU 1530 escape trick; condemned to kneel | **Escape/containment** | **THIN** attribution; PROBABLE that a real tale of this shape exists | +| 2 | Warthog backs into burrow after being ambushed at the threshold | **Boundary/threshold** | **THIN** — possible modern safari-industry invention | +| 17–18 | Swahili *methali* (spear-at-a-pig; choose a fat one) | Proverb | PROBABLE-SOLID / PROBABLE | +| 4 | Ashanti "How the pig got his snout" | Unknown | **THIN** | +| 19 | Hausa "lion won't play with the pig" | Proverb | **DUBIOUS — do not use** | +| 20 | Arabic *khanzīr* as insult | Idiom | **THIN** as sourced here | +| 24 | Transvaal escaped sow × bushpig hybrids (1970s) | **Escape (literal)** | **THIN/DUBIOUS** — single amateur source | +| 26 | UAC's Kano pig farm, 1943–late 1970s | Boundary (historical) | **DUBIOUS** — forum sources only | +| 27 | Haiti: Bois Caïman pig; Creole pig eradication | Diaspora | **NOT RESEARCHED** — flagged only | +| — | African ATU 2030 / ATU 124 analogues | Absence | **SOLID negative** (none found) | +| — | Pig as African trickster | Absence | **SOLID negative** | +| — | African escaping-pig proverb | Absence | **SOLID negative** | diff --git a/research/boxes-and-escape/survey/pig-asia.md b/research/boxes-and-escape/survey/pig-asia.md new file mode 100644 index 0000000..86dc54b --- /dev/null +++ b/research/boxes-and-escape/survey/pig-asia.md @@ -0,0 +1,659 @@ +# Pigs, Escape and Boundaries — ASIA + +Research pass for the escape/containment essay. Continent: East Asia, Southeast Asia, +South Asia, Central Asia, Siberia, West Asia / Middle East. + +**Method caveat, read first.** All research this session was done through web search. +`WebFetch` was blocked by the environment's outbound gateway for the entire session +(HTTP 403 on *every* host attempted, including `en.wikipedia.org`, `db.history.go.kr`, +`treasuryoflives.org`, `ancient-buddhist-texts.net`, `jstor.org`, `sacred-texts.com`). +The search-call budget was then exhausted (200/200). **Nothing below has been checked +against primary text by me.** Confidence ratings reflect this. Items marked SOLID are +ones where multiple independent search results converged and at least one was an +institutional source (national history institute, museum, peer-reviewed journal, major +wire service); they still deserve a primary-source check before publication. + +--- + +## Summary + +Asia produces a large corpus, but it is lopsided in a specific and interesting way. + +**The literal European motif — "the pig gets out of the pen and nobody can catch it," +the barnyard-escape / cumulative chain-tale form — is largely ABSENT from Asian +tale-type material.** I found no Asian analogue of ATU 2030 ("The Old Woman and Her +Pig," the "piggy won't go over the stile" chain), and no indigenous ATU 124 ("Three +Little Pigs"). The small clever animal that escapes by wit, in the region where that +genre is strongest (Malay/Indonesian archipelago), is the mousedeer *kancil* — not the +pig. That is likely not an accident: the trickster-escape genre flourishes exactly where +Islam has suppressed pig narrative. + +What Asia has instead, in four clusters: + +1. **The escaped *sacrificial* pig as a dynastic engine.** This is the strongest and most + specific escape material on the continent, and it is Korean. In the *Samguk Sagi*, a + Goguryeo sacrificial pig escapes on at least three recorded occasions; one escape + relocates the national capital, another produces a king. The runaway pig functions as + a pathfinder/oracle: the animal that refuses to stay put finds what the state could + not. +2. **The pig as the animal that must be kept outside the boundary.** Islamic and Jewish + law, and their downstream politics — the Talmudic pig hoisted over the Jerusalem + wall, the 1857 greased-cartridge rumour, Malaysia's 2026 Selangor pig-farm relocation + fight, pig heads left at mosques. Here the pig has no pen to escape from; its whole + narrative job is to be *outside*. +3. **The boar as unstoppable directional force** — uncontainable, but not *escaping*. + Japanese 猪突猛進, Chinese 狼奔豕突 and 封豕長蛇, Verethragna's boar incarnation, + Varaha diving beneath the cosmic ocean. The boar breaks through boundaries; it does + not slip out of them. +4. **Modern documented containment failure**, which is abundant: Hong Kong's urban + boars, Fukushima's escaped farm pigs hybridising with wild boar, African swine fever + defeating a national biosecurity cordon, Indian municipalities hiring contractors to + round up loose pigs. + +Zhu Bajie is the richest literary figure but needs careful framing: his containment is +**vow-based and appetitive, not architectural**. His name is literally a list of +restraints he cannot keep. + +--- + +## Escape / Containment Motifs + +*(Category (a): pigs actually escaping, or actually being confined.)* + +### 1. The Goguryeo escaped sacrificial pig (郊豕) — *Samguk Sagi* — **SOLID** + +This is the single best find of the research. The 郊豕 (Kor. *gyosi*) is the pig raised +for the state suburban sacrifice. It escapes repeatedly in the Goguryeo annals, and each +escape generates history. + +**Episode A — King Yuri, year 19 (≈1 BCE), 8th month.** The sacrificial pig escapes. The +king sends two men, **Takri (託利)** and **Sabi (斯卑)**, after it. They catch it at +Jangok marsh (長屋澤) and — to stop it running again — **cut its leg tendons with a +knife**. King Yuri, enraged that a victim consecrated to Heaven has been maimed, has both +men thrown into a pit and killed. In the 9th month the king falls ill; a shaman diagnoses +the vengeful spirits (원혼) of Takri and Sabi; the king has the shaman make apology and +recovers immediately. + +> Two men are executed for *preventing* a pig from escaping. The pig's mobility is +> sacred; hobbling it is the sacrilege. This is the most extreme "the pig must be allowed +> to be uncatchable" datum I found anywhere in Asia. + +**Episode B — King Yuri, year 21 (2 CE), 3rd month.** The sacrificial pig escapes again. +The sacrifice-keeper **Seolji (薛支)** pursues it to **Gungnae / Wina-am (國內 尉那巖)**, +finds the terrain fertile and defensible, and recommends moving the capital. Yuri +inspects the site personally and in year 22 (3 CE) moves the capital there and builds +Wina-am fortress. **Gungnae remained the Goguryeo capital for roughly four centuries**, +until King Jangsu's move to Pyongyang in 427. + +**Episode C — King Sansang, winter 208 CE (11th month, lunar).** The sacrificial pig +escapes; the officiant chases it to **Jutong-chon (酒桶村)**, where a woman of about +twenty, **Hunyeo (后女)**, catches it for him. The king seeks her out; she becomes a +consort and bears a son whose childhood name is **郊彘 (Gyoche) — "the suburban-sacrifice +piglet"** — the future **King Dongcheon**. + +*Sourcing.* Korean National Institute of Korean History databases +(`db.history.go.kr/ancient/level.do?levelId=sg_013r_0030_0130` — "21년 3월 설지가 천도를 +건의하다"; `contents.history.go.kr/mobile/km/view.do?levelId=km_017_0040_0010_0010` on +the shaman episode), plus the Academy of Korean Studies *Encyclopedia of Korean Culture* +entry on 託利 (`encykorea.aks.ac.kr/Article/E0058731`) and its 산상왕 entry +(`E0026247`). Confidence **SOLID** for Episodes A and B (institutional databases quoting +the annals directly); **SOLID/PROBABLE** for Episode C's details (Korean Wikipedia + +namu.wiki + Encykorea agreement; the name 郊彘 for Dongcheon is independently attested). + +### 2. Samding Dorje Phagmo turns her monastery into pigs — Tibet, c. 1716–17 — **PROBABLE** + +Escape *by becoming a pig*. When Dzungar Mongol forces reached Nangartse, the 6th Samding +Dorje Phagmo (the "Diamond Sow," Tibet's senior female incarnation line, an emanation of +Vajravārāhī) is said to have transformed herself and the monastery's monks and nuns into +pigs. The Dzungar chief, who had come to verify the rumour that she had a pig's head, tore +down the monastery walls and found only some 160 pigs grunting in the assembly hall under +a large sow. He withdrew; the pigs resumed human form; he became a patron. + +*Sourcing.* Widely reported in Tibetan-Buddhist and travel literature; the scholarly +reference is Hildegard Diemberger, *When a Woman Becomes a Religious Dynasty: The Samding +Dorje Phagmo of Tibet* (Columbia UP, 2007). I could **not** verify page-level attribution +or whether Diemberger treats it as hagiography vs. later legend. Treat as a hagiographic +motif, not an event. **PROBABLE** as a documented legend; the invasion date is given +variously as 1716 or 1717. + +Thematically this is the cleanest inversion on the continent: the pig-form is not what you +escape *from*, it is what you escape *into* — the body so contemptible that the pursuer's +gaze slides off it. + +### 3. Tacchasūkara Jātaka (Ja 492, "The Carpenter's Boar") — **SOLID** (text), *release not escape* + +A carpenter near Benares rescues a piglet from a pit and raises it; the boar fetches his +tools. Grown, it is **let go free into the forest**, joins a wild herd being killed by a +tiger, and — using what it learned among humans — organises the herd, digs pits, assigns +duties, and kills the tiger (and then a collaborating false ascetic, whose fig tree they +uproot). + +*Label carefully:* this is domestic→wild boundary-crossing and a *release*, not an escape. +But it is one of very few Asian texts where a pig's movement out of human containment is +the plot engine, and the boar's competence is explicitly the residue of its domestic +schooling. Text at `ancient-buddhist-texts.net/English-Texts/Jataka/492.htm`. + +### 4. Sukara Jātaka (Ja 153) — **SOLID** (text) — uncatchability by pollution + +A lion (the Bodhisatta) marks a boar for a future meal. The boar, misreading the lion's +quiet withdrawal as fear, challenges him. His friends advise a trick: roll in ascetics' +dung for seven days and come damp with morning dew. The lion arrives, recognises the +device, compliments the boar and spares him. The boar becomes uneatable by making itself +*ritually intolerable* — a mechanism that rhymes uncomfortably well with the pig's status +under kashrut and halal. + +### 5. Shishigaki 猪垣 — Japanese boar-walls — **PROBABLE** (real phenomenon, weak sourcing found) + +Farmers in mountainous Japan built long barriers — brushwood most commonly, but also +earth and dry-stone — around fields to keep wild boar (and deer) out. Boar could strip a +farmer's entire crop in a night and were locally implicated in famines. A frequently cited +example is the Edo-period (1790s) wall on "Shishigaki Island," built of soil, mud and +stone. + +*Sourcing is weak* — I reached this via `heianperiodjapan.blogspot.com`, +`worldkigo2005.blogspot.com` and a travel listing. The phenomenon is genuine and is +documented in Japanese local-history and cultural-landscape scholarship (some shishigaki +are designated cultural properties), but **the specific claims above rest on +low-quality websites and need replacing with a Japanese-language academic or municipal +source before use.** Conceptually this is the continent's best "containment +infrastructure" item: entire landscapes remodelled to keep pigs on the far side of a line. + +### 6. Shen Zhu 神豬 / "holy pig" contests, Taiwan — **SOLID** — anti-escape taken to its limit + +Hakka Yimin (義民) festival competitions, notably at Sanxia and in Hsinchu, in which pigs +are force-fed for up to two years until they cannot stand or move at all — routinely over +800 kg against a normal ~120 kg — then slaughtered, decorated and paraded. Investigators +(Environment and Animal Society of Taiwan, campaigning since 2003) document pressure +sores, organ failure, force-feeders used to strike the snout, and slaughter without +pre-stunning. Entries have fallen sharply (from >100 pigs to 37 in a recent year) under +sustained pressure. + +The tradition's own origin story is protective: early Chinese settlers propitiating +mountain gods for protection *against wild animals*. The wild pig outside is answered by +the immobilised pig inside. **SOLID** (AFP/phys.org, Bangkok Post, Taipei Times, UCA +News). + +### 7. Wake no Kiyomaro and the 300 boars, Japan, 769 CE — **PROBABLE** (legend) + +After Kiyomaro returned the Usa Hachiman oracle blocking the monk Dōkyō's bid for the +throne, Dōkyō had him exiled to Ōsumi **and his leg tendons cut** (note the exact echo of +the Goguryeo pig). Legend has 300 wild boar surround his palanquin and escort him some 40 +km to the shrine, after which his legs healed. Boar became the messengers of Goō +Daimyōjin; Goō Shrine in Kyoto has boar statues in place of the usual komainu. + +Boars here are escorts across a boundary of exile — mobility conferred on a man who has +been deliberately immobilised. + +### 8. Feral / wild boar incursion into cities — **SOLID** + +- **Hong Kong.** ~2,000–3,000 wild boar; sightings rose during COVID confinement. After + an auxiliary police officer was bitten at Tin Hau, the government in Nov 2021 abandoned + its 2017–21 capture-and-contracept/sterilise programme (350 operations, 1,092 boars + captured, 458 treated) in favour of capture-and-kill. 13 animal groups petitioned; + >44,000 signatures. Sport hunting had been stopped in 2019 after earlier protests. + Coverage: SCMP interactive, HKFP (both a critical and a pro-cull op-ed), Washington + Post, PBS, China Daily HK. +- **India.** Municipal round-ups of loose pigs are a recurring urban story. Amritsar MC + captured 30+ stray pigs from "posh localities" (Ranjit Avenue, Rani Ka Bagh, Company + Garden) and **arrested a pig farmer for deliberately letting his pigs loose in nearby + colonies**; Ambala Municipal Council hired a private firm because the "pig menace had + assumed alarming proportions" (*The Tribune*). Separately, wild boar have been declared + vermin in several states, contested between Centre and states (*Scroll.in*). The + Amritsar item is the closest thing on the continent to a plain "the farmer's pigs got + out and won't be caught" story. +- **Pakistan.** Wild boar entering dense neighbourhoods reported by *The Express Tribune*. + THIN — one report, not followed up. + +### 9. Fukushima: escaped domestic pigs → boar-pig hybrids — **SOLID** + +After the March 2011 evacuation, abandoned farm pigs went feral inside the exclusion zone +and interbred with native wild boar. Anderson et al., "Introgression dynamics from +invasive pigs into wild boar following the March 2011 natural and anthropogenic disasters +at Fukushima," *Proceedings of the Royal Society B* (2021), DOI 10.1098/rspb.2021.0874, +sampled 240+ animals from slaughterhouses; **31 boars (~16% of those from the evacuated +zone) were hybrids**, carrying on average only ~8% pig ancestry, and that fraction is +*declining* over time — domestic traits are being selected out. Boars simultaneously +recolonised abandoned towns; some tested at ~130× Japan's radiocaesium food standard. + +The narrative frame in the coverage (UGA, National Geographic, NBC, VOA, Inverse) is +explicitly boundary-collapse: the old topographic separation — humans on the coastal +plain, boar in the mountains — no longer holds, and the conflict is sharpest precisely in +remediated areas where people are returning. + +### 10. Zhu Jianqiang 猪坚强, "Strong-Willed Pig" — **SOLID** + +Buried for 36 days under a collapsed sty after the 12 May 2008 Wenchuan earthquake in +Pengzhou, Sichuan; survived on a bag of charcoal and rainwater. Bought from farmer Wan +Xingming by the privately run Jianchuan Museum Cluster and kept as a living symbol of +survival. Died 16 June 2021 of old age and organ failure; taxidermied and put on public +display 12 May 2022. National mourning-adjacent coverage (BBC, Sixth Tone, Deccan +Herald). An escape from containment of an entirely literal kind — a pig that got out from +under a building. + +### 11. African swine fever, 2018– — **SOLID** — the containment story at national scale + +First reported in Shenyang, 3 Aug 2018; present in 16 Asian countries by end-2021. +China's response is a containment geometry: cull *every* pig within 3 km of an outbreak, +disinfection and inspection stations across a 10 km buffer, live-market closures. +It failed at scale — a 41% fall in China's pig population 2018–19, ~USD 111 bn losses in +2019 — and the reasons given are explicitly about porousness: ubiquitous backyard farms, +swill feeding, and "elusive animal transport networks." Wild boar act as a reservoir +crossing the farm/forest line. (*Science*; Wang & Qi, *Infectious Diseases of Poverty* +2018; PMC reviews of wild boar's role in Asia.) + +### 12. "Pigs from the Sea," Okinawa 1948 — **SOLID** — crossing, not escape + +Not an escape, but the most striking pig boundary-crossing in modern Asian memory. +Okinawa's pig population fell from >100,000 pre-war to under 800 after the Battle of +Okinawa. Hawaii's Uchinanchu diaspora raised $47,196, bought 500 sows and 50 boars on the +US mainland, and seven men sailed with them from Portland aboard the *John Owen* in +August 1948. Storms cost **several pigs overboard**; 537 landed at White Beach on 27 +September 1948. Distributed on a pass-it-on system (keep a breeding pair, pass the rest of +the litter), the herd reached 100,000 within four years. Commemorated in Okinawa as +「海から豚がやってきた」/ "Pigs from the Sea." (KHON2, Hawaii Herald, Jikōen Hongwanji, +DVIDS, *Hana Hou!*) + +--- + +## Pig as Boundary / Taboo + +*(Category (b): the pig as the thing that must be kept outside the line. This is the +largest body of Asian material by volume.)* + +### Leviticus 11, kashrut, and Mary Douglas — **SOLID, with a mandatory caveat** + +Douglas's *Purity and Danger* (1966) argued that the pig is prohibited because it is a +**categorical anomaly** — cloven-hoofed like the permitted ungulates but not a +ruminant — and that dietary law polices classificatory boundaries rather than hygiene. +This is the single most-cited theoretical warrant for treating pig taboo as a +boundary phenomenon. + +**Do not cite it without the retraction.** In her 2002 preface Douglas called the +anomaly explanation "a major mistake" and replaced it with a reading in which the dietary +laws model the body and the altar on one another. Any essay that leans on Douglas needs +to say so. + +### Talmud, Bava Kamma 82b — the pig hoisted over the wall — **SOLID** + +The best single "pig crossing a boundary" text in West Asia. During the Hyrcanus– +Aristobulus siege of Jerusalem (65 BCE), the besieged bought sacrificial lambs from the +besiegers by hauling money out and animals in over the wall. An elder versed in "Greek +wisdom" pointed out that the city could not fall while the Temple service continued. The +next day the besiegers sent up **a pig** in the basket. Midway up the wall the pig **dug +its hooves into the wall**, and the Land of Israel quaked — the passage gives 400 +parasangs. At that moment the sages pronounced: *cursed is the man who raises pigs, and +cursed is the man who teaches his son Greek wisdom.* + +Two boundary registers fused in one image: a wall the pig should not cross, and a cultural +frontier (Hellenism) it stands for. Note also the Mishnaic/Talmudic rule prohibiting the +**raising** of pigs anywhere — a containment prohibition, not merely a dietary one. +(Steinsaltz/Aleph Society daf; Ezra Brand's commentary on Jerusalem's special +regulations; *The Lehrhaus*, "The Rome not Taken: Pompey, Pigs.") + +### Qur'anic prohibition and *maskh* — **SOLID** for the texts, handle 5:60 carefully + +*Khinzīr* is prohibited at Q 2:173, 5:3, 6:145, 16:115. Distinct from the dietary rule and +more relevant here: **Q 5:60 (and 2:65, 7:166)**, in which a group of Sabbath-breaking +Israelites are made "apes and swine." The classical majority (Ibn ʿAbbās, Qatāda) read the +*maskh* as a literal physical transformation; a minority (Mujāhid) as metaphorical +degradation of character. Whichever reading, the verse stages the pig as the far side of a +**human/animal boundary that punishment forces you across** — the same structure as Zhu +Bajie's fall into a sow's womb, and as babi ngepet, arrived at from three directions. + +Flag: this verse has an extensive modern polemical afterlife (as anti-Jewish rhetoric in +some clerical speech). Cite the exegetical tradition, not the polemic literature, and do +not source it from `answering-islam.org` or `jewishvirtuallibrary.org`'s clip file, both +of which surfaced in search and are advocacy sources. + +### The 1857 greased cartridges — **SOLID**, with a real historiographical caveat + +The Enfield P-53 cartridge had to be bitten open. From January 1857 the rumour spread +through the Bengal Army that the grease was beef tallow and **pork lard** — polluting to +Hindu sepoys, defiling to Muslim sepoys. The rumour is the causally decisive fact and is +not in dispute; **whether the grease actually contained pork is contested**, with +conflicting British testimony, and with the compromise position that batches varied +(tallow, lard, and mutton fat all used, with no way to tell which was which). *Do not +write "the cartridges were greased with pig fat" flatly.* + +For the essay: this is the pig crossing the most intimate boundary there is — the mouth — +by rumour alone, without any actual pig being present. Empire-scale consequences from a +hypothetical pig. (Wikipedia "Causes of the Indian Rebellion of 1857"; Kaye & Malleson; +*The Conversation*, "Five-pound history lesson: animal fat and the British empire's +biggest revolt"; HistoryNet.) + +### Malaysia — zoning the pig, 2012 and 2026 — **SOLID** + +- **Pig heads at mosques, Kuala Lumpur, January–February 2012**: three incidents in a + month; described by an MP as politically motivated provocation (BBC News, Asia). +- **Selangor pig-farming relocation, 2026**: the Sultan announced that state funds would + not go to pig farming and that licence renewals would be withheld until farms relocated + to a consolidated site at **Bukit Tagar**, whose surrounding population is + predominantly Malay-Muslim. PAS figures counter-proposed **Pulau Ketam**, an + overwhelmingly non-Muslim island — i.e. push the pigs across a water boundary onto the + non-Muslim side. Framed in commentary as a test of Malaysian pluralism: existential for + Chinese-Malaysian farmers, a religious intrusion for nearby Muslim residents. (FMT, + Focus Malaysia, SCMP *This Week in Asia*, Malay Mail, Business Today.) + +This is the live, contemporary form of the boundary theme in Southeast Asia: not a pen, +but a *map*. The question is not whether pigs can be contained but where the line goes. + +### The pig at the hub of the wheel — **SOLID** + +At the centre of the Tibetan *bhavacakra* are three animals: pig (ignorance, *moha*), +snake (aversion), cock (attachment) — with the snake and cock issuing from the pig's +mouth, because ignorance is the root of the other two. The pig's usual gloss is that it +sleeps in the filthiest place and eats whatever comes to its mouth. + +Worth the essay's attention as an inversion: this is the pig as the reason **nothing +escapes**. The animal at the axle of the machine of transmigration. Every other item in +this report is a pig getting out of something; here the pig is what everyone else is +stuck inside. (Lion's Roar, Tricycle, Encyclopedia of Buddhism, U. Idaho.) + +### 家 = roof + pig — **PROBABLE, contested; do not assert flatly** + +The popular etymology of 家 *jiā* ("home, family") reads it as 宀 (roof) over 豕 (pig): +settlement is where the pig is kept. It is the mirror image of the taboo — the boundary +that *includes* the pig defines the home. + +**But the etymology is disputed.** Mark Edward Lewis, *The Construction of Space in Early +China* (p. 92, following Xu Shen), argues the lower element is 亥 *hài* (child) rather +than 豕 *shǐ*, the two being near-identical in seal script. Present it as a folk etymology +whose appeal is exactly the point, or flag the dispute. + +### Korean threshold ritual — **PROBABLE** + +The boiled pig's head with banknotes stuffed in its mouth, used at *gosa* rituals opening +a new business, building or venture, and the strong money-association of *dwaeji-kkum* +(pig dreams — Koreans buy lottery tickets after them), rest on the *don* homophony +(돼지/돈). The pig here is a positive threshold-marker: placed *at* the boundary of a new +undertaking. Sourcing is popular (Korea.net on the National Folk Museum's Year of the Pig +exhibition; Korea Times; USC Digital Folklore Archives; Columban magazine) — the National +Folk Museum of Korea's zodiac pages are the best of them. + +--- + +## Proverbs & Idioms + +Escape-adjacent items first. + +| Item | Lang | Literal | Sense | Confidence | +|---|---|---|---|---| +| **猪突猛進** *chototsu mōshin* | JP | "boar-charge, fierce advance" | Rushing headlong without looking left or right; usually mildly pejorative (acts before thinking) but can praise drive. Standard explanation: boars run straight and do not look aside. | **SOLID** for meaning/usage; the "boars run straight" origin gloss is the conventional folk account, repeated across dictionaries | +| **狼奔豕突** *láng bēn shǐ tū* | ZH | "wolves run, boars charge" | (i) a mob fleeing in panic and disorder; (ii) villains rampaging destructively. | **SOLID** — in the Taiwan MOE *Dictionary of Chinese Idioms* (`dict.idioms.moe.edu.tw`, ID 8761) | +| **封豕長蛇** *fēng shǐ cháng shé* | ZH | "giant boar and long snake" | Rapacious, violent invaders/aggressors. From *Zuo Zhuan*; also in *Hou Han Shu*. | **SOLID** — MOE idiom dictionary, ID 8130 | +| **没吃过猪肉，也见过猪跑** | ZH | "even if you've never eaten pork you've seen a pig run" | Lack of first-hand experience isn't total ignorance. Note the image: the pig's *running* is the thing everyone has witnessed. | **PROBABLE** — meaning certain; the attribution to *Honglou Meng* ch. 16 is widely repeated but I could not verify it against the text. Mark the attribution, not the proverb, as unverified | +| **หมูในอวย** *mǔu nai uay* | TH | "pig in the trough/pot" | A sure thing; something already caught; a shoo-in; also an easy mark. The exact inverse of uncatchability. | **PROBABLE/THIN** — meaning consistent across Thai-idiom listings but all sources found were language-learning sites (ExpatDen, Wikipedia list of Thai idioms). Needs a Thai dictionary source | +| **หมู** *mǔu* used adjectivally = "easy" | TH | "pig" | Thai's general slang for "easy/a pushover". | THIN | +| **猪武者** *inoshishi-musha* | JP | "boar warrior" | A soldier who charges recklessly. | **PROBABLE** — commonly listed, not verified against a dictionary this session | +| **หมูไปไก่มา** | TH | "pig goes, chicken comes" | Reciprocal favour exchange. | **UNSOURCED** — I could not confirm this. Do not use without checking | +| **人怕出名猪怕壮** | ZH | "people fear fame as pigs fear getting fat" | Prominence attracts destruction; the fattened pig is slaughtered. Thematically excellent (fattening as the terminal form of containment). | **UNVERIFIED this session** — I did not source it. Check before use | + +Absent-by-search: I found no Chinese, Japanese or Korean proverb specifically about a pig +*escaping* or *being hard to catch*. In the idiom stock, the boar's uncontainability is +always **forward charge**, never **evasion**. That asymmetry is itself a finding. + +--- + +## Zhu Bajie & Literary Figures + +### Zhu Bajie 猪八戒 — *Xiyouji* — **the central case, but reframe it** + +**The name is the argument.** 猪 = swine; 八戒 = the **eight precepts** (*aṣṭāṅga-śīla*, +the eight observance-day vows: no killing, stealing, lying, sexual activity, +intoxication, and the ancillary three). Tang Sanzang gives him the name as a nickname to +hold him to his monastic diet. So the character's *identity is a fence*: a list of eight +prohibitions strapped to an appetite. Every joke about him is a breach report. He is not a +pig who escapes an enclosure; he is a pig who **is** an enclosure, permanently failing. + +**Origins as boundary error.** He was Marshal Tian Peng (天蓬元帥), commander of the +heavenly naval forces, expelled from heaven for molesting Chang'e; falling to earth his +spirit **enters the wrong womb — a sow's — by mistake**, and he is born a monstrous +man-pig. A clerical error at the level of reincarnation. Compare Q 5:60's *maskh* and +Java's babi ngepet: three independent Asian traditions in which the human/animal +boundary is crossed downward into a pig as a consequence of transgression. + +**Gao Village (高老莊).** Taking human form, he marries into the Gao household under false +pretences — and then **locks his wife in a back building for half a year**, forbidding her +to see her family. Note the reversal worth writing about: the Asian pig-figure's most +explicit containment act is one *he performs on a human*. When exposed, he flees; Sun +Wukong takes the bride's form to trap him. He is bound, released, and re-bound repeatedly +through the novel. + +**The one who won't stay in formation.** Zhu Bajie is the pilgrim who recurrently proposes +to dissolve the pilgrimage — divide the baggage, go back to Gao Village. Treat as +**PROBABLE**: this is a well-known feature of the novel and of its television +adaptations, but I did not verify chapter citations this session, and I'd want two or +three located instances before asserting it in print. + +**His weapon** is the nine-toothed muckrake (九齒釘耙) — a farmyard implement. The pig +carries the tool of the pen. + +*Sourcing note:* the search results for Zhu Bajie were dominated by wikis, AI-generated +content farms (`genspark.ai`, `studyguides.blog`) and fan sites. The one usable +specialist resource that surfaced is **`journeytothewestresearch.com`** (Jim McClanahan), +which is scholarly-adjacent and heavily footnoted. For real citation use Anthony C. Yu's +translation (rev. ed., Chicago 2012) and the scholarship around it. + +### Varaha — the third avatar — **SOLID** + +Vishnu takes boar form because the earth (Bhūmi/Bhūdevī), stolen by Hiraṇyākṣa, has sunk +to the bottom of the cosmic ocean. The boar is chosen for its *shape*: an animal that +drives its snout downward into the dark and roots out what is buried. He kills +Hiraṇyākṣa and lifts the earth on his tusks. Depicted zoomorphically or with a boar's head +on a human body (fine 3rd–5th c. examples; the Norton Simon holds one). + +Vedic roots: Prajāpati/Emūṣa the boar raising the earth from the waters (Ṛgveda, +Taittirīya Saṃhitā, Śatapatha Brāhmaṇa) — the avatar attaches later to Vishnu. + +For the essay: this is the only major Asian pig-figure whose defining act is **crossing +between cosmological levels**. Not escaping a boundary — traversing one that nothing else +can. + +### Yamato Takeru and the white boar of Mount Ibuki — *Kojiki* — **SOLID** + +Yamato Takeru vows to kill the Ibuki deity bare-handed. Ascending, he meets a **great +white boar, ox-sized**, decides it is merely the god's *messenger*, and lets it pass. It +is the god itself. It raises a hailstorm; he is disoriented and stricken, partly recovers +at the Samegai spring, and dies. (The *Nihon Shoki* substitutes a great serpent — worth +noting for source-criticism.) The deity is venerated as Ibuki-daimyōjin, appearing as +white boar or white serpent. + +The failure is a **category error**: mistaking the boundary-being for a messenger of the +boundary. The best Japanese item for an essay about what pigs mark. (Kokugakuin +University's Yamato Takeru article is the reliable source here.) + +### Verethragna's boar incarnation — **PROBABLE/SOLID** + +The Zoroastrian yazata Verethragna (Bahrām) appears in the Bahrām Yašt (Yašt 14) in a +sequence of incarnations, the fifth of which is a sharp-tusked, aggressive boar that kills +at a blow. The boar accordingly signifies irresistible victory in Iranian art (Sasanian +boar-hunt reliefs; Ṭāq-e Bostān) and in the *Shāhnāme*'s hunting register — the Sasanian +Bahrām V ("Bahrām Gūr") carries the name. + +Caveat: my search returned Bahrām Gūr's *gazelle* hunt (with Āzāde) far more prominently +than any boar hunt, and the boar/Shāhnāme link came partly from `culturalboar.com`, a +non-academic site. **Assert the Yašt 14 boar incarnation (well attested); do not assert a +specific Shāhnāme boar-hunt episode without checking.** + +### Babi ngepet — Java/Malay world — **PROBABLE (existence), THIN (details)** + +A *pesugihan* (wealth-magic) practice in which a person, wrapped in a consecrated burial +shroud and with the aid of ritual, becomes a boar that moves invisibly at night and steals +money and goods; a partner must keep a candle burning, and if it gutters the transformer +is trapped or dies. Read by anthropologists as a commentary on inexplicable wealth in +agrarian societies where prosperity ought to be visibly earned. + +Directly on-topic: this is a pig that **passes through walls**. It is the only Asian +pig-figure I found whose defining capacity is unauthorised entry. + +**Sourcing is poor.** Everything I found is popular (`mythlok.com`, `monstropedia.org`, +Vocal Media, Grokipedia). Claims that it appears in 19th–20th-c. Dutch colonial +ethnography are plausible but unverified by me. Treat as **live contemporary folklore** +(there are documented modern Indonesian mob incidents around babi ngepet accusations) and +find a proper source — the Javanese magic literature (Geertz, *The Religion of Java*; +newer work on *pesugihan*) is the place to look. + +### Vajravārāhī / Dorje Phagmo — **PROBABLE** + +The "Adamantine Sow," a principal Vajrayāna female deity, iconographically marked by a +sow's head protruding from the side of her skull. Her human incarnation line at Samding +gives item #2 above. The sow's head is a deliberate embrace of the despised form — +another instance of the pig as the body one hides inside. + +### Chinese zodiac — the Pig, twelfth — **SOLID (as folklore), THIN (as a text)** + +In the Great Race narrative the Pig is last because it stopped to eat, then slept. The +race myth is genuinely current folk narrative but is a **late, orally-transmitted +etiology**, not a classical text; do not cite it as ancient. Note the structural point: +in a myth whose entire subject is *arriving in order*, the pig is the animal that +declines to complete the crossing on schedule. In the twelve Earthly Branches the pig is +亥 *hài* — and see the 家/亥 etymological tangle above, which is caused by the same +graphic confusion. + +--- + +## Documented Real Escapes + +Consolidated, with dates, for the essay's factual spine: + +| Event | Date | Status | +|---|---|---| +| Goguryeo sacrificial pig escapes; Takri & Sabi hamstring it and are executed | 1 BCE (Yuri 19) | **SOLID** (as text; historicity of the annals is another matter) | +| Sacrificial pig escapes; Seolji tracks it to Gungnae; capital moved | 2–3 CE (Yuri 21–22) | **SOLID** (as text) | +| Sacrificial pig escapes; caught by Hunyeo; son named 郊彘, becomes King Dongcheon | 208 CE | **SOLID/PROBABLE** | +| "Pigs from the Sea": 550 pigs shipped Portland→Okinawa; several lost overboard; 537 land | 27 Sep 1948 | **SOLID** | +| Zhu Jianqiang survives 36 days under earthquake rubble | May–Jun 2008; d. 16 Jun 2021 | **SOLID** | +| Fukushima farm pigs go feral, hybridise with wild boar (~16% of zone boars) | 2011– ; published 2021 | **SOLID** (Proc R Soc B) | +| African swine fever defeats China's 3 km/10 km cordon; 41% herd loss | Aug 2018– | **SOLID** | +| Amritsar MC captures 30+ loose pigs; farmer arrested for releasing them | undated (Tribune archive) | **SOLID** for the report; get a firm date | +| Hong Kong shifts from contraception to capture-and-kill after Tin Hau bite | Nov 2021 | **SOLID** | + +--- + +## Notable Absences + +Reported deliberately, because in this continent absence carries most of the argument. + +1. **No Asian ATU 2030.** The cumulative "old woman and her pig / piggy won't go over + the stile" chain-tale is indexed in England, Scotland, Norway and Switzerland + (*Joggeli söll ga Birli schüttle*, 1908). Chain-tale form is abundant in South Asia, + but I found **no** Asian variant of 2030 with a pig in it. (D. L. Ashliman's Pitt + type-2030 page and the Chain Tales blog's ATU index are the two places that would show + one; neither did.) THIN-but-real negative. +2. **No indigenous ATU 124.** The Three Little Pigs circulates in Asia in translation only. +3. **No pig trickster cycle.** East and Southeast Asia have rich trickster cycles — the + monkey, the hare, and above all the **mousedeer (*kancil* / *pelanduk*)** in + Malay-Indonesian tradition, whose entire repertoire is escaping traps and outwitting + captors. **That role, in the world's largest Muslim-majority region, is occupied by a + deer, not a pig.** I think this is the most defensible causal claim in the report: the + escape-trickster genre is strongest exactly where the pig has been pushed out of + narrative altogether. +4. **Siberia and Central Asia: near-total silence.** Wild boar are absent from most of + northern Siberia and (historically) from Hokkaido; pastoral Turkic-Mongol economies + are pig-poor; post-Islamisation they are pig-hostile. The Turkic zodiac keeps a + pig/boar year (*domuz yılı*), which is essentially the only trace. This is a real + absence, not a search failure — though note that Amur-basin peoples (Nanai, Udege) do + hunt wild boar and were not searched. +5. **West Asia: escape material ≈ zero, and that is the finding.** In the Islamic + Middle East the pig has no barnyard to escape from. Its entire narrative function is to + be *outside* — the exemplar of the excluded. The Talmudic Jerusalem-wall pig is the + great exception and works precisely because it inverts this: the one recorded moment a + pig gets *in*. +6. **Gaps, not absences (search budget exhausted):** Vietnamese folklore and the Đông Hồ + pig woodblock tradition; Philippine/Cordillera pig sacrifice and bile divination; + Toraja and Nias pig exchange; Burmese and Sinhala material; Naga and other Northeast + Indian pig culture; Han-dynasty pigsty-latrine architecture (圂), which I suspect is a + strong containment item and did not get to search; Chinese-language folk-tale + collections searched only in English. Do not read these as absences. +7. **Ainu:** the Ainu bear ceremony (*iyomante*) has no pig analogue; wild boar are not + native to Hokkaido. Noted but not researched. + +--- + +## Sources + +Institutional / peer-reviewed: +- 국사편찬위원회 한국 고대 사료 DB, *삼국사기* 유리명왕 21년: https://db.history.go.kr/ancient/level.do?levelId=sg_013r_0030_0130 +- 우리역사넷 (National Institute of Korean History), 무당 항목 (Yuri 19 / Takri & Sabi): https://contents.history.go.kr/mobile/km/view.do?levelId=km_017_0040_0010_0010 +- 한국민족문화대백과사전, 託利: https://encykorea.aks.ac.kr/Article/E0058731 · 山上王: https://encykorea.aks.ac.kr/Article/E0026247 +- Anderson et al., "Introgression dynamics from invasive pigs into wild boar … Fukushima," *Proc. R. Soc. B* (2021): https://royalsocietypublishing.org/doi/10.1098/rspb.2021.0874 · PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC8242833/ +- Wang & Qi, "African swine fever: an unprecedented disaster and challenge to China," *Infectious Diseases of Poverty* (2018): https://link.springer.com/article/10.1186/s40249-018-0495-3 +- "The Role of the Wild Boar Spreading ASF Virus in Asia": https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9093143/ +- 教育部《成語典》(Taiwan MOE): 封豕長蛇 https://dict.idioms.moe.edu.tw/idiomView.jsp?ID=8130 · 狼奔豕突 https://dict.idioms.moe.edu.tw/idiomView.jsp?ID=8761 +- Kokugakuin University, "Yamato Takeru": https://www.kokugakuin.ac.jp/en/article/130837 +- Ancient Buddhist Texts, Ja 492 *Tacchasūkara*: https://ancient-buddhist-texts.net/English-Texts/Jataka/492.htm +- National Folk Museum of Korea, zodiac Pig: https://www.nfm.go.kr/english/subIndex/1023.do +- UGA news, Fukushima wildlife 10 years on: https://news.uga.edu/10-years-after-fukushima-animals-reclaim-landscape/ +- Norton Simon Museum, Varaha: https://www.nortonsimon.org/art/detail/F.1975.16.06.S/ +- Steinsaltz / Aleph Society, Bava Kamma 82: https://steinsaltz.org/daf/bavakamma82/ + +Journalism (reliable): +- BBC, "Strong-Willed Pig … dies": https://feeds.bbci.co.uk/news/world-asia-china-57509236 · Sixth Tone: https://www.sixthtone.com/news/1007463/ +- BBC, "Pig's head outside Malaysia mosque causes outrage": https://feeds.bbci.co.uk/news/world-asia-16854036 +- SCMP wild-boar interactive: https://multimedia.scmp.com/infographics/news/hong-kong/article/3202238/urban-wild-boar/index.html +- HKFP (Nov 2021, both sides): https://hongkongfp.com/2021/11/15/… and /2021/11/22/… +- Washington Post, "Wild boars are on the loose in Hong Kong": https://www.washingtonpost.com/world/asia_pacific/hong-kong-wild-boars-animals/2021/11/25/ +- *Science*, ASF in Asia: https://www.science.org/content/article/african-swine-fever-keeps-spreading-asia-threatening-food-security +- SCMP, Selangor pig-farm dispute: https://www.scmp.com/week-asia/politics/article/3356127/ · FMT explainer: https://www.freemalaysiatoday.com/category/nation/2026/01/26/ +- *The Tribune* (Amritsar/Ambala pig round-ups): https://www.tribuneindia.com/news/archive/amritsar/30-pigs-captured-from-posh-localities-606525 +- phys.org / AFP on Taiwan holy pig: https://phys.org/news/2013-02-taiwan-activists-holy-pig-contest.html · Taipei Times editorial: https://www.taipeitimes.com/News/editorials/archives/2018/02/27/2003688324 +- *The Conversation*, 1857 cartridges: https://theconversation.com/five-pound-history-lesson-animal-fat-and-the-british-empires-biggest-revolt-70004 +- Hawaii Herald / KHON2 / Jikōen, "Pigs from the Sea": https://www.thehawaiiherald.com/2018/08/28/history-aloha-from-hawaii-to-okinawa/ · https://hanahou.com/21.5/pigs-from-the-sea +- Scroll.in, wild boar vermin status: https://scroll.in/article/1025938/ + +Reference / secondary, use with care: +- D. L. Ashliman, ATU 2030 texts: https://sites.pitt.edu/~dash/type2030.html +- Chain Tales blog ATU index: http://chaintales.blogspot.com/2018/05/tmi-z41-old-woman-and-her-pig.html +- Journey to the West Research (Jim McClanahan): https://journeytothewestresearch.com/tag/pig/ +- Tricycle / Lion's Roar on the bhavacakra: https://tricycle.org/magazine/bhavacakra-wheel-of-life/ · https://www.lionsroar.com/buddhism/wheel-of-life-bhavacakra/ +- SeekersGuidance on *maskh*: https://seekersguidance.org/answers/quran/is-the-quranic-metamorphosis-punishment-literal/ +- Diemberger, *When a Woman Becomes a Religious Dynasty* (Columbia UP 2007) — for Samding Dorje Phagmo; located online only via file-sharing mirrors, **not verified** +- Mark E. Lewis, *The Construction of Space in Early China*, p. 92 — for the 家 etymology dispute, cited second-hand via thinkchinese.org + +Low quality, flagged, **replace before citing**: heianperiodjapan.blogspot.com and +worldkigo2005.blogspot.com (shishigaki); mythlok.com, monstropedia.org, Vocal Media, +Grokipedia (babi ngepet); culturalboar.com (Persian boar); genspark.ai, studyguides.blog, +Fandom wikis (Zhu Bajie); expatden.com, thaipod101 (Thai idioms); answering-islam.org and +the Jewish Virtual Library clip file (Q 5:60 — advocacy sources). + +--- + +## Confidence Notes + +**Environmental limits.** `WebFetch` returned HTTP 403 for every host attempted this +session — the outbound gateway rejected the CONNECT for `en.wikipedia.org`, +`db.history.go.kr`, `treasuryoflives.org`, `ancient-buddhist-texts.net`, `jstor.org`, +`sacred-texts.com` and others. The search-call budget then hit its 200-call ceiling. So +**every item here is sourced from search-result snippets only**, some of which are +themselves machine-generated summaries. Anything going into print needs a primary check. + +**Highest confidence (would defend now):** +- The Goguryeo *gyosi* escape cycle, all three episodes, as *Samguk Sagi* content. +- The Talmudic Jerusalem-wall pig and the paired curse (Bava Kamma 82b). +- 狼奔豕突 and 封豕長蛇 (Taiwan MOE idiom dictionary). +- Fukushima pig×boar introgression (peer-reviewed, with figures). +- Zhu Jianqiang; Hong Kong 2021; Taiwan shen zhu; Malaysia 2012/2026; ASF containment + geometry; "Pigs from the Sea"; Varaha; Yamato Takeru's white boar; the bhavacakra pig; + Zhu Bajie's name = eight precepts. + +**Explicitly flagged as needing work before use:** +- 家 = roof+pig (contested; Lewis reads 亥). +- Douglas's anomaly theory (she retracted it in 2002). +- 1857 cartridges (rumour certain; pork content contested). +- *Honglou Meng* ch. 16 attribution for 没吃过猪肉也见过猪跑. +- Samding Dorje Phagmo pig-transformation (hagiography; date 1716 vs 1717 varies). +- Shishigaki specifics (blog-sourced). +- Babi ngepet specifics (popular sources only). +- A Shāhnāme boar-hunt episode (the Yašt 14 boar incarnation is safe; a named Bahrām Gūr + boar hunt is not). +- Zhu Bajie's recurring "let's split the luggage and go home" (well known, no chapter + citations verified). + +**Marked UNSOURCED / DUBIOUS — do not use as-is:** +- Thai หมูไปไก่มา (could not confirm). +- Chinese 人怕出名猪怕壮 (not sourced this session). +- Any claim that babi ngepet appears in named 19th-c. Dutch ethnography. + +**Not researched (gaps, not absences):** Vietnamese, Filipino, Burmese, Sinhala, Toraja, +Nias, Naga, Amur-basin Tungusic; Han pigsty-latrine archaeology; Chinese-language +folk-tale collections. diff --git a/research/boxes-and-escape/survey/pig-europe.md b/research/boxes-and-escape/survey/pig-europe.md new file mode 100644 index 0000000..a2d3e89 --- /dev/null +++ b/research/boxes-and-escape/survey/pig-europe.md @@ -0,0 +1,322 @@ +# Pigs, Escape and Containment in European Folklore, Law and History + +**Continent:** Europe (Britain & Ireland, Nordics, Iberia, Italy, Balkans, Central & Eastern Europe, Russia west of the Urals, Greece) + +**Method note:** This research was conducted under a tool constraint — `WebFetch` was blocked at the proxy layer for *all* hosts (403 on Wikipedia, university sites, Ashliman's Pitt folklore archive, and open-access PDFs alike). Everything below therefore rests on search-engine result summaries plus identified source URLs, **not** on primary-text verification. Where a claim would normally be checked against the actual tale text, statute, or article, that check could not be performed. Confidence ratings reflect this. Items I could not corroborate are marked THIN or flagged UNVERIFIED rather than smoothed over. + +--- + +## Summary + +There is a **great deal** of material — considerably more than the brief anticipated, and it clusters far more tightly around containment than one would expect by chance. The pig is, across European tradition, unusually and consistently a *creature defined by the failure of enclosure*. + +Three findings organise the whole corpus: + +1. **The strongest material is Welsh, and there are two items, not one.** Twrch Trwyth is the expected champion, but **Henwen the sow** (Welsh Triads) is a near-exact structural twin — an uncatchable swine pursued across the whole island by Arthur, dropping monstrous progeny at named places, and escaping finally into the sea. Wales has *two* independent "uncontainable pig traverses the entire landscape and is never caught" narratives, plus a third pig-journey (Pryderi's Otherworld swine) that likewise leaves a trail of place-names. This is a genuine regional pattern, not a single text. + +2. **The escape theme is not merely literary — it is legal and it is live.** The densest concentration of pig-containment material in Europe is in *law*: pannage, pinfolds, ringing ordinances, manorial amercements, the medieval pig trials, and, still in force, section 4 of the Animals Act 1971. Pigs generated more containment law than any other European domestic animal, and were the most-prosecuted defendant in the premodern animal trials. + +3. **The single most consequential fact in the whole file is zoological, not folkloric:** Britain's entire modern wild boar population — an extinct species now re-established across several English counties — exists *because pigs escaped*. There is no reintroduction programme behind it. It is downstream of fence failures, a storm, and one illegal release. + +A recurring structural point worth flagging for the essay: the European pig is not merely *bad at being contained* but is repeatedly the figure that **marks and polices boundaries** — the year-boundary (Hwch Ddu Gwta), the living/dead boundary (Gloso), the human/animal boundary (Circe), the clean/unclean and Jew/Christian boundary (marrano), and the demonic/territorial boundary (Gadarene swine). Escape and boundary-marking are the same theme seen from two sides. + +--- + +## Escape / Containment Motifs + +Labelled per the brief: **(a)** pigs escaping enclosures/pursuit; **(b)** pigs as boundary/taboo markers; **(c)** pigs in stories with no containment theme (included for contrast, so the essay does not over-read). + +### (a) Escape, pursuit, uncatchability + +**Twrch Trwyth — *Culhwch and Olwen*, the Mabinogion. SOLID.** +The strongest single item on the continent, as anticipated. A king transformed into a giant boar, carrying a comb, razor and shears between his ears — objects Culhwch must obtain to complete Ysbaddaden's tasks. Arthur and his warband hunt him from Ireland, across Wales (through Dyfed and south Wales), over the Severn estuary, and into Cernyw (Cornwall). The critical structural fact: **the treasures are stripped from him but the boar himself is never killed or contained** — he is driven into the sea and escapes. The hunt occupies the greater part of the second half of the tale. +- Celtic scholar **John Rhŷs** observed that the chase functions as an extended *dindsenchas* — a place-name legend following a trail of swine-related toponyms across south Wales. The chase is the armature on which the text hangs its topographic lore. (PROBABLE — reported in secondary summaries, not verified against Rhŷs directly.) +- Earlier attestation: the ninth-century Latin *Historia Brittonum* refers to the boar (as *porcum Troit* / *Troynt*), and the seventh-century elegy *Gwarchan Cynfelyn* contains one of the earliest mentions. (PROBABLE.) +- Source: https://www.ebsco.com/research-starters/literature-and-writing/culhwch-and-olwen-hunting-twrch-trwyth ; https://gwallter.com/literature/the-hunt-for-twrch-trwyth.html + +**Henwen ("Old White") — the Welsh Triads. PROBABLE, and badly under-used.** +The sleeper find. Triad of the "Three Powerful Swineherds of the Isle of Britain": the sow Henwen, kept by **Coll ap Collfrewy**, swineherd to Dallwyr Dallben. It is prophesied that Britain will be the worse for her womb-burden, so **Arthur and his warriors set out to destroy her — and fail**. She flees across the island (Cornwall, Gwent, Pembroke, Arfon), farrowing a different creature at each named location. At Llanfair in Arfon a black kitten of her litter is cast into the sea, swims to Anglesey, and is reared by the sons of Palug — becoming **Cath Palug**, the monstrous cat that later fights Cai or Arthur. +Note the structure: an uncatchable pig, pursued by Arthur, traversing the whole landscape, generating place-name etymologies, ending at the sea. It is the same shape as Twrch Trwyth. **Wales has two of these.** +- Source: https://www.maryjones.us/jce/coll.html ; https://lornasmithers.com/2019/03/14/henwen-the-birthing-and-devouring-sow/ + +**Pryderi's pigs — *Math fab Mathonwy*, Fourth Branch of the Mabinogi. SOLID.** +Otherworld swine, a gift to Pryderi from **Arawn, lord of Annwfn**. Gwydion obtains them by fraud (illusory horses and greyhounds) and drives them north across Wales in five days, from Rhuddlan Teifi in Ceredigion to Creuwrion in Gwynedd. Their overnight stops are commemorated in the place-name **Mochdref** ("pig-town"/"swine-town"), including one in Powys between Ceri and Arwystli. When the illusion lapses, Pryderi raises an army and pursues — the pig-theft precipitates a war. +Three separate things converge here: pigs crossing the boundary *out of the Otherworld*, pigs moved illicitly across a country, and pigs leaving a toponymic trail. +- Source: https://storymaps.arcgis.com/stories/9558f4db79ca4037987240d9b327aa6e ; http://www.mabinogi.net/math.htm + +**The Gadarene / Gerasene swine — Mark 5:1–20 (parallels Matthew 8, Luke 8). SOLID.** +The demoniac named "Legion"; the unclean spirits enter a herd of about **two thousand** swine, which "rushed down the steep bank into the sea, and were drowned." Exactly as the brief suspected, this is a mass-escape narrative — and it is *terminal* escape, the herd running out of containment and out of life simultaneously. +The boundary reading is well established: the episode is set in the **Decapolis**, Gentile territory, on the far shore — Jesus crosses a territorial and ethnic boundary, and the pigs (an animal that could only be herded in Gentile land) are the marker of that crossing. The demons ask not to be sent "out of the country," i.e. they too are territorially bounded. The healed man is then sent to proclaim to the Decapolis. +- Confidence note: the naturalistic reading offered in some devotional sources (that the man himself stampeded the herd) is **speculative apologetics, not scholarship** — do not cite it as such. +- Source: https://en.wikipedia.org/wiki/Exorcism_of_the_Gerasene_demoniac ; https://ncec.catholic.edu.au/faith/scripture-resources/commentaries/the-gospel-of-mark/jesus-heals-the-gerasene-demoniac-mark-51-20/ + +**The Erymanthian Boar — Fourth Labour of Heracles. SOLID, and a perfect inversion.** +Eurystheus commands that the boar be taken **alive** — the difficulty of the labour *is* the containment requirement. On the centaur Pholus's advice, Heracles drives it into deep snow on Mount Erymanthos until it is exhausted and trapped, then nets it. The punchline is the finest containment joke in Greek myth: presented with the living boar, **Eurystheus leaps into a buried *pithos*-jar to hide.** The man who ordered the pig contained ends up the one inside the container. +- Source: https://www.theoi.com/Ther/HusErymanthios.html ; https://www.perseus.tufts.edu/Herakles/boar.html + +**The Calydonian Boar — Ovid, *Metamorphoses* 8.270–546. SOLID.** +Oeneus of Calydon omits sacrifice to Artemis; she looses a monstrous boar on his fields. The devastation is specifically agricultural — it destroys the *enclosed, cultivated* land: "Now it trampled the young shoots of the growing crops, now cut short the ripeness, longed-for by the mournful farmer, and scythed down the corn in ear." Meleager assembles a panhellenic hunt (Jason, Theseus, Atalanta); Atalanta draws first blood, Meleager kills it, and the quarrel over the hide destroys him. +The containment logic: the boar is divine punishment expressed as *the failure of the field boundary*. Wild nature is let back into farmed space. +- Source: https://www.theoi.com/Ther/HusKalydonios.html + +**ATU 2030 — "The Old Woman and Her Pig". SOLID. The purest "refusal to stay put" tale in Europe.** +This is the item most directly on the brief's theme and the one most likely to be overlooked. A cumulative/chain tale (Thompson motif **Z41**) whose entire plot is a **pig that will not move**. The old woman must get the pig over a stile — a boundary structure specifically designed to let people through and keep livestock in — and the pig simply refuses. She recruits dog, stick, fire, water, ox, butcher, rope, rat, cat in a chain of coercion, and only when the whole apparatus of the rural world is mobilised does "the little pig in a fright jump over the stile." +The tale type is named after this story. Joseph Jacobs, *English Fairy Tales* (London: David Nutt, 1890), his source being Halliwell's *Nursery Rhymes and Tales*. +- Source: https://sites.pitt.edu/~dash/type2030.html ; https://authorama.com/english-fairy-tales-6.html + +**The Tale of Pigling Bland — Beatrix Potter. SOLID (with one date caveat).** +Genuinely an escape-and-boundary story. Pigling Bland is sent to market and the plot turns on **pig licences** — the paperwork of permitted movement ("Stole a pig? Where are your licences?" asks the policeman). He is waylaid by Mr Piperson and discovers **Pig-wig, a stolen black Berkshire pig locked in a cupboard** — a pig in illegal confinement. They escape at dawn and the book's most famous line is a boundary-crossing: *"They came to the river, they came to the bridge — they crossed it hand in hand."* +- **Date discrepancy:** one secondary source gave 1922; the publication date is standardly given as **1913**, which fits the biographical context that same source cites (Potter's engagement and move to Castle Cottage). Treat 1913 as correct but verify. +- The detail that the bridge marks a *county* boundary (into Westmorland) is in the text as I recall it but I could not verify it — **UNVERIFIED**. +- Source: https://www.slaphappylarry.com/pigling-bland-beatrix-potter/ ; https://americanliterature.com/author/beatrix-potter/short-story/the-tale-of-pigling-bland + +**ATU 124 — "The Three Little Pigs" / "Blowing the House In". SOLID. Flagged as an inversion, per the brief.** +The brief's instinct is right and worth stating plainly in the essay: this is the one canonical European pig story where the pigs are **trying to stay contained**. The tale is about the integrity of an enclosure against an outside force — the wolf's "I'll huff, and I'll puff, and I'll blow your house in" is an assault on a wall. The pig is the besieged party, not the escapee. Collected by **James Halliwell-Phillipps**; type 124 variants include "The Fox and the Pixies" and "The Fox and the Geese" (England), where the pigs are replaced by other animals — confirming the pig is not essential to the type. +- Source: https://sites.pitt.edu/~dash/type0124.html ; https://www.surlalunefairytales.com/oldsite/threepigs/notes.html + +### (b) Pigs as boundary and taboo markers + +**Yr Hwch Ddu Gwta — the Tailless Black Sow, Welsh Calan Gaeaf. SOLID. Outstanding item.** +On **Nos Galan Gaeaf** (the eve of the first day of winter, 31 Oct) — explicitly regarded as a **seasonal boundary** and the most ominous of the three Welsh *ysbrydnos* or "spirit nights" — a tailless black sow with red eyes appears as the bonfire dies. In practice a man draped in cloth or hide rising from the embers, chasing the children home. The children chant: +> *"Adref, adref am y cynta', Hwch Ddu Gwta a gipio'r ola."* +> ("Home, home, at once — the tailless black sow shall snatch the last one.") +This is a pig functioning as a **boundary enforcer**: it polices the threshold of the year and punishes whoever is slowest to get back inside. Note the inversion of every other item in this file — here the pig is not the thing that escapes but the thing that makes *you* run for your enclosure. +- Source: https://theconversation.com/horrifying-black-sows-and-ghostly-apparitions-how-the-magic-and-mystery-of-wales-come-alive-in-winter-238725 ; https://www.aber.ac.uk/en/english/news/news-article/title-276975-en.html ; https://museum.wales/blog/1857/Halloween-Traditions/ + +**Gloso / Gloson (Sweden) and Gravso (Denmark) — the glowing/grave sow. PROBABLE.** +In Skåne and Blekinge, the **Gloso** (from an old Swedish term for "glowing sow") is a always-female spectral hog that **guards churchyards at night** — the boundary between living and dead. Black, red-eyed, with a razor-sharp saw-ridge along its back; it kills by **running between a person's legs and cutting them in two**. Note how literal the crossing imagery is: the creature's method of killing *is* passing through you. +Associated with **årsgång** ("year walk"), a perilous folk-divination ritual performed at Christmas or New Year — again a calendar boundary. In Danish tradition the **Gravso** ("grave sow") is linked to a murdered child returning to haunt the living. +- Confidence: reported consistently across folklore-blog sources (FolkloreThursday, Burials & Beyond) but I could not reach an academic Nordic folklore source. Verify against a Swedish folkminnen collection before citing hard. +- Source: https://folklorethursday.com/folktales/gloson-the-swedish-ghost-pig-that-will-cleave-you-in-half/ ; https://burialsandbeyond.com/2020/01/02/the-gloson-demon-ghost-pig-of-sweden/ + +**Circe's swine — *Odyssey* Book 10. SOLID.** +The boundary crossed is the human/animal one. Circe's drugged draught erases memory, she strikes with the wand, and **drives the men into her pigsties** — note that the transformation is immediately followed by *penning*. The horror is doubled: they have the likeness of pigs but retain human awareness. Odysseus, protected by Hermes' *moly*, compels the reverse transformation. +Worth flagging: this is a story where becoming a pig and *being enclosed* are the same event. The sty is not incidental. +- Source: https://www.theoi.com/Text/HomerOdyssey10.html + +**"Marrano" — Iberia. PROBABLE (etymology contested).** +Fifteenth-century Castilian *marrano* = "pig"/"swine", applied as a slur to converted Jews. The pig here is the instrument by which a **religious boundary is tested and policed**: pork consumption served as a visible marker of sincere Christian identity, and refusal to eat it as evidence of crypto-Judaism, for a population created by the Alhambra Decree of 1492. +- **Contested:** one etymology derives it from an Arabic root meaning "prohibited/illicit" (relating to the shared Jewish and Muslim pork taboo), another simply from the Spanish for swine, with the slur turning on the animal's uncleanness. Scholarship also objects to the term's continued neutral scholarly use (see Akman, *Sephardic Horizons*). Do not assert a single etymology. +- Source: https://www.newworldencyclopedia.org/entry/Marrano ; https://www.sephardichorizons.org/Volume3/Issue1/Akman.html + +**"Casting pearls before swine" — Matthew 7:6. SOLID.** +A boundary-maintenance instruction: the swine mark the category of those outside the community to whom holy things must not be given. Relevant to the essay as rhetoric rather than as escape. + +### (c) Pigs with NO containment theme (reported for contrast, per the brief) + +These are prominent European pig materials that the essay should **not** press into the escape thesis: + +- **Sæhrímnir** — the boar of Valhalla, cooked nightly by Andhrímnir in the cauldron Eldhrímnir and **restored to life each day** for the einherjar. Attested in *Grímnismál* (Poetic Edda) and interpreted as a boar by Snorri in the Prose Edda. This is abundance/regeneration, not containment. *However* — there is a defensible adjacent reading: it is a pig that refuses to stay dead, which rhymes with refusing to stay put. Flag as a rhyme, not evidence. SOLID (attestation). +- **Gullinbursti** — Freyr's golden-bristled boar, forged by the dwarves Eitri and Brokkr in the wager against Loki; its bristles light the night. Status symbol and solar imagery. No containment theme. SOLID. +- **Hildisvíni** — "battle swine", Freyja's boar/mount, in *Hyndluljóð*. No containment theme. SOLID. +- **Manannán mac Lir's swine** — Irish Otherworld pigs, slaughtered and eaten at the Feast of Age and reconstituted each morning. Same regeneration motif as Sæhrímnir. PROBABLE. +- **Torc Triath** — "king of the swine", associated with Brigid. I could **not** verify the *Orc Triath* form or the "pigs of Drebrenn" material the brief asked about. **THIN — omit or flag.** +- **Glücksschwein / "Schwein haben"** (German) — the pig as a *luck* symbol; marzipan pigs at New Year. Nothing to do with escape. See Proverbs below. +- **ATU 425A / 441 "The Enchanted Pig"** — animal-bridegroom tale, not a containment tale. See ATU section. + +--- + +## ATU Index Entries + +Verified tale types with pigs. The ATU index itself (Uther 2004, 3 vols) could not be consulted directly; these come from the tale-type pages of D. L. Ashliman's University of Pittsburgh folklore archive and standard references. + +| Type | Name | Relevance | Confidence | +|---|---|---|---| +| **ATU 2030** | **The Old Woman and Her Pig** | **Direct.** Cumulative tale whose entire plot is a pig refusing to cross a stile. Type is named for it. Thompson motif **Z41**. | SOLID | +| **ATU 124** | **The Three Little Pigs / Blowing the House In** | **Inversion.** Enclosure defended from outside, not escaped from. Pig not essential to type (fox/geese/pixie variants). | SOLID | +| **ATU 425A** | The Animal as Bridegroom / Search for the Lost Husband | "The Enchanted Pig" (Romanian *Porcul cel fermecat*). Also classed **ATU 441** (Hans My Hedgehog group). Pig-form husband said to be especially popular in Romania. Petre Ispirescu, *Legende sau basmele românilor*, told him by his mother c.1838–47; Mite Kremnitz, *Rumänische Märchen* (1882); Andrew Lang, *The Red Fairy Book* (1890). **No containment theme** — the enclosure motif in it is the forbidden room, not a sty. | SOLID | +| **ATU 441** | Hans My Hedgehog | Related bristly-animal-bridegroom type; cross-classified with the above. | PROBABLE | + +**Thompson Motif-Index entries** (from the Motif-Index, Vol. 1, B section): +- **B16.1.4** — "Devastating swine" +- **B16.1.4.1** — "Giant devastating boar" (with Irish, Icelandic, Greek, Italian and Indian references — this is the index entry that covers Twrch Trwyth, the Calydonian Boar and the Erymanthian Boar together) +- **B183** — "Magic boar (pig)", with sub-motifs including B183.9 "Skin of magic pig heals wounds" +- **B511.2.1 / B511.2.2** — magic pig heals wounds by touch / by licking +- **B184.3.2.1** — "Magic invisible pig"; **B184.3.2.2** — "Magic pig turns water into wine for nine days" (Irish) +- Confidence: **PROBABLE**. Numbers come from search summaries of the Wikisource Motif-Index; the exact wording of B16.1.4 vs. a possible B16.1.3 could not be confirmed. **Verify against the printed index before citing a specific number.** +- Source: https://en.wikisource.org/wiki/Motif-Index_of_Folk-Literature/Volume_1/B/0 ; https://en.wikisource.org/wiki/Motif-Index_of_Folk-Literature/Volume_1/B/100 + +**Gap:** I found no ATU type specifically indexing "escaped pig recaptured". The escape material in Europe sits in *legend, law and record* rather than in the Märchen type-index — which is itself a finding worth making in the essay. + +--- + +## Classical & Mythological + +Covered in detail above. Consolidated: + +| Item | Text | Containment status | +|---|---|---| +| Erymanthian Boar | Apollodorus; Theoi/Perseus compilations | Must be captured **alive**; Eurystheus hides in a jar. SOLID | +| Calydonian Boar | Ovid, *Met.* 8.270–546; Apollodorus; Homer *Il.* 9 | Divine punishment as breach of field boundary. SOLID | +| Circe's swine | Homer, *Odyssey* 10 | Transformation *and penning* as one act. SOLID | +| Gadarene/Gerasene swine | Mark 5:1–20; Matt 8:28–34; Luke 8:26–39 | c.2,000 head, mass terminal escape over a cliff, in Gentile boundary territory. SOLID | +| Twrch Trwyth | *Culhwch and Olwen*; *Historia Brittonum*; *Gwarchan Cynfelyn* | Never contained; escapes to sea. SOLID | +| Henwen | Welsh Triads (Trioedd Ynys Prydein) | Never caught by Arthur; escapes to sea. PROBABLE | +| Pryderi's pigs | *Math fab Mathonwy* (Fourth Branch) | Cross out of Annwfn; driven across Wales; cause a war. SOLID | +| Sæhrímnir / Gullinbursti / Hildisvíni | *Grímnismál*, *Skáldskaparmál*, *Hyndluljóð* | **No containment theme** (category c). SOLID | + +--- + +## Proverbs & Idioms + +**"A pig in a poke" — and its European cognates. SOLID (with an important twist).** +*Poke* = a small bag, from Old North French *poque* (modern *poche*), into English via Anglo-Norman c.13th c. The phrase is Tudor; something close to the modern form appears in **John Heywood, *A Dialogue conteinyng the nomber in effect of all the Prouerbes in the Englishe tongue* (1555–60)**. The market practice: live pigs sold in sacks, with unscrupulous sellers substituting a cat or dog. +**The twist worth building on:** most other European languages render this as a **cat** in the bag, not a pig — +- German: *die Katze im Sack kaufen* +- French: *acheter chat en poche* +- Spanish: *dar gato por liebre* ("to give cat for hare") +So the English idiom is the outlier in naming the pig. And the *complementary* English idiom — **"letting the cat out of the bag"** — is precisely an escape-from-container phrase. The pair encodes: the pig that was never in the bag, and the cat that gets out of it. This is a nice essay hinge. +- Source: https://www.phrases.org.uk/meanings/a-pig-in-a-poke.html + +**"When pigs fly" / "pigs might fly" — adynaton. PROBABLE–SOLID.** +Reportedly a traditional Scottish proverb; first printed **1586**, in an edition of **John Withal's English–Latin dictionary for children**, whose appendix of proverbs rendered into Latin includes one to the effect that *"pigs fly in the air with their tails forward"* — the joke being that flying backwards is a trivial extra absurdity once flight is granted. **François Rabelais**, *Quart Livre* (1552), stages a related image: Pantagruel's battle with the Chitterlings features "a huge, fat, thick, grizzly swine, with long and large wings, like those of a windmill." +The relevance to the brief: the pig is the European exemplar of the **impossible escape — escape from gravity and from the ground itself.** The animal chosen to stand for absolute impossibility is the one that cannot leave. +Non-English equivalents generally do **not** use the pig: French *quand les poules auront des dents* ("when hens have teeth", late 18th c.). Do not claim a pan-European pig-flight idiom. +- Source: http://www.worldwidewords.org/qa-pig1.html ; https://en.wikipedia.org/wiki/When_pigs_fly + +**"You can't make a silk purse from a sow's ear" — mid-16th century English or earlier. SOLID.** +The genuinely old European ancestor of the modern American **"lipstick on a pig"**. Note the container imagery again: the thing you cannot make out of a pig is *a purse* — a poke. + +**"Lipstick on a pig" — American, modern. SOLID as to dating.** +Charles Lummis (ed., *Los Angeles Times*), 1926: "Most of us know as much of history as a pig does of lipsticks." The modern phrase is first recorded in a 1985 *Washington Post* piece on a San Francisco park renovation. **Not European in origin** — use only as the descendant of the sow's-ear proverb. + +**"Hog wild" — American, 1904 (OED). Not European.** +Etymologies offered include the struggles of animals being driven to slaughter, or hogs breaking loose at feeding time. Attractive to the thesis but **American, and the "breaking loose" derivation is folk-etymological and unconfirmed** — mark THIN if used. + +**"Greased pig" contest — THIN, and probably not European.** +Established at American county fairs by the early 20th century (and present in 19th-c. Independence Day festivities). Claims of "Celtic roots dating to at least the 17th century in Ireland" appear in popular sources but **I found no scholarly support** — treat as **UNSOURCED/DUBIOUS**. The structural point stands regardless: the contest is a ritual dramatisation of the ungrippable pig, in which grease makes containment-by-hand impossible. + +**German: *Schwein haben* ("to have a pig") = to be lucky; the *Glücksschwein*. PROBABLE.** +Marzipan pigs at New Year, often with a four-leaf clover. Competing origin theories: a piglet as consolation prize for the worst shot at medieval marksmanship festivals; or the *Sau* as the high card in medieval card games. Traces of "lucky pigs" appear in late-medieval board and card games, and boar effigies as talismans go back to the early Middle Ages. **No containment theme** — include only as counterweight. +- Source: https://www.leidenmedievalistsblog.nl/articles/lucky-pigs-and-protective-boars-the-medieval-origins-of-the-gluecksschwein + +**"Tantony pig" — English idiom. PROBABLE.** See Legal & Historical below; the term came to mean a person of sporadic, food-following loyalties — derived directly from a pig licensed to roam. + +**Gap:** I was unable to search Slavic, Balkan, Italian or Nordic proverb corpora before exhausting the search budget. A search for Russian/Slavic pig folklore returned material on **Buryat** (Siberian, non-European) traditions and is not usable. **Eastern and southern European proverb material is a genuine hole in this file.** + +--- + +## Legal & Historical + +This is the richest and most under-exploited seam for the essay. + +**Dolly Jørgensen, "Running Amuck? Urban Swine Management in Late Medieval England," *Agricultural History* 87:4 (2013), pp. 429–451. SOLID — the key academic source.** +Central argument, which usefully **overturns the popular image**: swine were **not** free roamers in medieval towns. Because pigs were ubiquitous, damaging if loose, and integral to urban agricultural production, they required what Jørgensen calls **"cradle-to-grave controls"** — regulation of movement, feeding and slaughter. Evidence drawn from urban government and court records of the most populous English centres plus smaller towns, late 13th–16th centuries (Coventry, Norwich, York, London, Ramsey among them). Ramsey records specifics on **leashing and ringing**. One town banned pigs from its four main streets entirely (ringed or unringed) and required ringing on all others. +The essay-relevant inference: **the density of regulation is itself the evidence of chronic escape.** You do not legislate repeatedly against a problem that stays solved. +- Source: https://read.dukeupress.edu/agricultural-history/article/87/4/429/296956/Running-Amuck-Urban-Swine-Management-in-Late + +**Pinfolds, pounds and amercements. SOLID.** +Animal pounds (**pinfolds**) emerged in England in the 12th–13th centuries as a core institution of manorial and village governance, arising alongside the open-field system, for confining stray livestock — sheep, pigs, cattle. An **amercement** was the payment to the lord by a person found guilty of a trespass. Example: **in 1425 a group of six swineherds were fined for allowing their pigs to trespass in the fields in autumn**, against ordinance. +Crucially — and this is a strong containment detail — **an owner could not simply retrieve a strayed animal.** Once a stray entered the system, recovery was only lawful through the manor court; circumventing it, *even by the true owner*, was itself punishable by amercement. Escape triggered a legal process that the owner could not short-circuit. +- Source: Jordan Claridge, "Waifs and Strays: Property Rights in Late Medieval England," *Journal of British Studies*, https://www.cambridge.org/core/journals/journal-of-british-studies/article/waifs-and-strays-property-rights-in-late-medieval-england/148ADDD32647806A4793D0AB2933F888 + +**Pannage (Common of Mast) — licensed, seasonal, temporary un-containment. SOLID.** +The right of commoners to release domestic pigs into woodland to forage on acorns, beech mast and chestnuts. Recorded in the **Domesday Book (1086)**, which measured woodland by the number of pigs it could support. In the New Forest (founded by William the Conqueror, 1079) the practice continues: roughly **600 pigs** turned out annually today, all ear-tagged, against as many as **6,000** in the 19th century. Customarily a pig was rendered to the lord of the manor for every certain number loosed *de herbagio*. Ecologically load-bearing: acorns are toxic to ponies and cattle, and 90+ New Forest ponies and cattle have died in a single season. +**Why it matters to the thesis:** pannage is the legal invention of *authorised* escape. Europe could not keep pigs in, so it built a season in which letting them out was a property right — bounded by calendar, by tags, and by render to the lord. +- Source: https://www.forestryengland.uk/new-forest/pannage-pigs-and-acorns ; https://www.thenewforest.co.uk/explore/wildlife-and-nature/pigs/ + +**Tantony pigs — St Anthony's Hospital, London. PROBABLE. The licensed exception.** +The Order of Hospitallers of St Anthony (formed c.1095, St Anthony as patron) received donated runt pigs. Any pig judged by the supervisor of the London market **unfit to be killed for food** had a bell attached by a proctor of St Anthony's and was thereby "free of the street" — legally entitled to roam and scavenge, fed at Londoners' expense, later slaughtered to feed the poor in the hospital's care. Hence **"tantony pig"** for a person of shifting, food-following loyalties. +Abuse is documented: **in 1311 a hospital tenant, Roger de Wynchester, was compelled by the City authorities to promise not to claim pigs found wandering the City, nor to bell "any swine but those given in charity to the house."** In other words, the one legal loophole permitting an uncontained pig was immediately exploited to launder *other* people's escaped pigs. +- Source: https://www.stgeorges-windsor.org/pig-in-the-city/ ; https://www.british-history.ac.uk/vch/london/vol1/pp581-584 + +**The medieval pig trials. SOLID as a phenomenon; individual cases vary in reliability.** +- **Edward Payson Evans, *The Criminal Prosecution and Capital Punishment of Animals* (1906)** — the foundational compilation, from European archives, of hundreds of cases from the 9th to the 18th century. **Pigs were by far the most common animal defendants.** A study in the journal *Sophia* is reported to find pigs accounting for **more than a quarter of all recorded nonhuman trials and over half of legal actions involving animals in premodern France**. (SOLID for Evans; PROBABLE for the *Sophia* figures, which I could not verify at source.) +- **Falaise, 1386.** A sow prosecuted for killing an infant in its cradle, mutilating its face and limbs. Arrested, imprisoned, tried in the same court used for humans, sentenced to be mangled in the forelegs and hanged. The famous detail — **the sow dressed in a waistcoat, gloves and drawers for its execution in the market square** — and the commemorative fresco in the Church of the Holy Trinity at Falaise, whitewashed away in the 19th century. **CONFIDENCE CAVEAT:** the human-clothing detail and the fresco are the most-repeated and least-verifiable elements of the animal-trial literature, resting on later reports. Flag as PROBABLE at best, not SOLID. +- **Savigny, Burgundy, 1457.** A sow tried for the murder of a five-year-old child, whom she had killed and "partly eaten". Convicted, hanged in the public square. **Her six piglets, present at the scene, were charged and acquitted** — on the reasoning that they were young and had been led astray by "the bad example of their mother." (SOLID — this is the best-documented of the cases.) +- **The connection to the brief's thesis is exact and should be made explicit:** these prosecutions are only possible because pigs were *loose*. A pig in a sty does not eat a child in a cradle. Every pig trial is the legal aftermath of a containment failure, and the court's fiction of pig culpability is precisely a way of *not* assigning the blame to the fence or its owner. +- Source: https://www.leidenmedievalistsblog.nl/articles/homicidal-hogs ; https://daily.jstor.org/when-societies-put-animals-on-trial ; https://blogs.ncl.ac.uk/speccoll/?p=1899 + +**Animals Act 1971 (England and Wales), s.4 — still in force. SOLID.** +Strict liability on livestock keepers for damage done by straying animals. The fencing provisions are subtle and quotable: damage "shall not be treated as due to the fault of the person suffering it by reason only that he could have prevented it by fencing" — i.e. **the victim's failure to fence is no defence** — *but* the keeper is not liable where the straying "would not have occurred but for a breach by any other person, being a person having an interest in the land, of a duty to fence." The Act abolished the old common-law tort of **cattle trespass** (which covered pigs) and replaced it with this statutory strict liability. Enacted 12 May 1971, in force 1 October 1971. +The through-line from 1425 to 1971 is unbroken: European law has never stopped adjudicating who pays when the pig gets out. +- Source: https://www.legislation.gov.uk/ukpga/1971/22/crossheading/strict-liability-for-damage-done-by-animals + +--- + +## Documented Real Escapes + +**The re-establishment of wild boar in Britain — entirely by escape. SOLID. The most consequential item in the file.** +Wild boar were extinct in Britain for centuries. The modern population is **not** the product of any reintroduction programme; it is the accumulated residue of containment failures at wild-boar farms, which proliferated with 1980s farm diversification and imports from mainland Europe. +- **The Great Storm of October 1987** brought down trees and smashed farm fences across southern England; boar escaped into the night, were never recovered, and bred. Free-ranging groups persist today on the **Kent/East Sussex border** as a direct result. +- **Forest of Dean** — now the largest population in England. The original nucleus established in woodlands near **Ross-on-Wye** after a 1990s escape from a wild boar farm. In **2004**, around **60 farm-reared animals were dumped in an illegal release near Staunton**, on the Forest's western edge; by **2009** the two populations had merged into a thriving breeding population. +- Rate: **from 1989/90 to 2008/9, an average of one to two escape or release incidents per year**, individual incidents involving from one to more than 50 animals. **By 1998 there were viable free-living populations in the British countryside.** +- Source: https://mammal.org.uk/british-mammals/wild-boar ; https://www.forestryengland.uk/article/more-information-about-wild-boar ; https://www.rewildingbritain.org.uk/why-rewild/reintroductions-key-species/key-species/wild-boar-pig + +**The Tamworth Two — Malmesbury, Wiltshire, January 1998. SOLID.** +**Butch** (a sow) and **Sundance** (a boar), brother and sister Tamworth pigs, escaped while being unloaded from a lorry at an abattoir. They **squeezed through a fence and swam the River Avon**, escaping into nearby gardens, and spent most of a week at large in dense thicket near **Tetbury Hill**. The manhunt became an international media event — NBC and several Japanese outlets sent reporters. After more than a week they were recaptured, **bought by the *Daily Mail***, and rehomed at a rare breeds farm in **Kent**, where they lived out their lives; the surviving pig died aged 14 (BBC reporting indicates Butch predeceased, with the last of the two dying in 2011). +Why it is the essential modern item: the escape occurred **at the abattoir gate**, at the exact threshold between livestock and meat, and the public response — a national newspaper buying the pigs to guarantee they would not be killed — amounted to a collective decision that a successful escape *earns* exemption. Escape functioned as a moral claim. +- Source: https://en.wikipedia.org/wiki/Tamworth_Two ; https://feeds.bbci.co.uk/news/uk-england-wiltshire-13503690 ; https://www.kentonline.co.uk/ashford/news/the-runaway-pigs-who-captured-the-nations-heart-279941/ + +**Transport escapes (a recurring modern genre). SOLID individually.** +- **AP-7 motorway, Santa Perpètua de Mogoda, near Barcelona, 31 July 2023** — a pig lorry collided with a car and overturned; pigs ran loose down the motorway; a 14 km tailback; animals cleared around 9pm, road reopened around 10pm. +- **A-601, between Segovia and Valladolid** — a pig truck overturned leaving animals loose on the carriageway, at least one struck by a car. +- **A68, Swan House roundabout, Heighington, County Durham** — a lorry carrying **200 pigs** overturned, with **at least 80% escaping** the crash. +- **A21, Pratts Bottom, south London** — escaped farm pigs forced police to close the carriageway between Pratts Bottom and the M25; BBC headline: *"'Clever little pigs' escape causing closure of A21"*. +- **M48 near Chepstow** — a pig escaped from a lorry onto the motorway. +- Source: https://euroweeklynews.com/2023/08/01/overturned-lorry-sheds-trailer-load-of-pigs-across-barcelonas-ap-7-motorway/ ; https://feeds.bbci.co.uk/news/uk-england-tees-44028964 ; https://feeds.bbci.co.uk/news/uk-england-london-35213835 ; https://feeds.bbci.co.uk/news/uk-wales-south-east-wales-36545830 + +--- + +## Sources + +**Academic / institutional** +- Dolly Jørgensen, "Running Amuck? Urban Swine Management in Late Medieval England," *Agricultural History* 87:4 (2013), 429–451 — https://read.dukeupress.edu/agricultural-history/article/87/4/429/296956/Running-Amuck-Urban-Swine-Management-in-Late +- Jordan Claridge, "Waifs and Strays: Property Rights in Late Medieval England," *Journal of British Studies* — https://www.cambridge.org/core/journals/journal-of-british-studies/article/waifs-and-strays-property-rights-in-late-medieval-england/148ADDD32647806A4793D0AB2933F888 +- E. P. Evans, *The Criminal Prosecution and Capital Punishment of Animals* (1906) — Newcastle Univ. Special Collections, https://blogs.ncl.ac.uk/speccoll/?p=1899 +- Leiden Medievalists Blog, "Homicidal Hogs: Murderous Pigs on Trial in Medieval France" — https://www.leidenmedievalistsblog.nl/articles/homicidal-hogs +- Leiden Medievalists Blog, "Lucky pigs and protective boars: The medieval origins of the Glücksschwein" — https://www.leidenmedievalistsblog.nl/articles/lucky-pigs-and-protective-boars-the-medieval-origins-of-the-gluecksschwein +- Aberystwyth University / *The Conversation*, on the Hwch Ddu Gwta — https://theconversation.com/horrifying-black-sows-and-ghostly-apparitions-how-the-magic-and-mystery-of-wales-come-alive-in-winter-238725 +- Museum Wales, "Halloween Traditions" — https://museum.wales/blog/1857/Halloween-Traditions/ +- *Victoria County History*, London vol. 1, "Alien Houses: Hospital of St Anthony" (British History Online) — https://www.british-history.ac.uk/vch/london/vol1/pp581-584 +- JSTOR Daily, "When societies put animals on trial" — https://daily.jstor.org/when-societies-put-animals-on-trial +- Akman, "A Scholarly Blind Spot: the term 'marrano'", *Sephardic Horizons* — https://www.sephardichorizons.org/Volume3/Issue1/Akman.html + +**Folklore reference** +- D. L. Ashliman, Univ. of Pittsburgh — ATU 2030: https://sites.pitt.edu/~dash/type2030.html ; ATU 124: https://sites.pitt.edu/~dash/type0124.html +- *Motif-Index of Folk-Literature* (Thompson), Wikisource Vol. 1 B — https://en.wikisource.org/wiki/Motif-Index_of_Folk-Literature/Volume_1/B/0 +- Mary Jones, *Jones's Celtic Encyclopedia*, "Coll ap Collfrewy" — https://www.maryjones.us/jce/coll.html +- SurLaLune, annotations to Three Little Pigs — https://www.surlalunefairytales.com/oldsite/threepigs/notes.html +- EBSCO Research Starters, "Culhwch and Olwen: Hunting the Twrch Trwyth" — https://www.ebsco.com/research-starters/literature-and-writing/culhwch-and-olwen-hunting-twrch-trwyth +- "Where Pryderi's Pigs Are" (ArcGIS StoryMaps) — https://storymaps.arcgis.com/stories/9558f4db79ca4037987240d9b327aa6e + +**Classical texts** +- Theoi Classical Texts: Erymanthian Boar https://www.theoi.com/Ther/HusErymanthios.html ; Calydonian Boar https://www.theoi.com/Ther/HusKalydonios.html ; *Odyssey* 10 https://www.theoi.com/Text/HomerOdyssey10.html +- Perseus (Tufts), Hercules' Fourth Labor — https://www.perseus.tufts.edu/Herakles/boar.html + +**Law, land, natural history** +- Animals Act 1971 s.4 — https://www.legislation.gov.uk/ukpga/1971/22/crossheading/strict-liability-for-damage-done-by-animals +- Forestry England, "Pannage, pigs and acorns" — https://www.forestryengland.uk/new-forest/pannage-pigs-and-acorns +- Mammal Society, "Wild boar" and "Re-establishment of Wild Boar in Britain" — https://mammal.org.uk/british-mammals/wild-boar ; https://mammal.org.uk/position-statements/re-establishment-of-wild-boar-in-britain +- Forestry England, wild boar — https://www.forestryengland.uk/article/more-information-about-wild-boar + +**Idiom / phrase** +- Phrases.org.uk, "A pig in a poke" — https://www.phrases.org.uk/meanings/a-pig-in-a-poke.html +- World Wide Words, "Pigs might fly" — http://www.worldwidewords.org/qa-pig1.html +- OED, "hog-wild, adj." — https://www.oed.com/dictionary/hog-wild_adj + +**Popular / to be treated with caution** +- FolkloreThursday and Burials & Beyond on the Gloson — https://folklorethursday.com/folktales/gloson-the-swedish-ghost-pig-that-will-cleave-you-in-half/ ; https://burialsandbeyond.com/2020/01/02/the-gloson-demon-ghost-pig-of-sweden/ +- College of St George, "Pig in the City" (tantony pigs) — https://www.stgeorges-windsor.org/pig-in-the-city/ + +--- + +## Confidence Notes + +**SOLID** — Twrch Trwyth (outline and non-capture); Pryderi's pigs and the Mochdref toponyms; ATU 2030 and its stile plot; ATU 124 and its inversion; ATU 425A/441 attribution for *The Enchanted Pig*; Erymanthian Boar (capture alive, Eurystheus in the jar); Calydonian Boar (Ovid *Met.* 8.270–546); Circe (*Od.* 10, driven into sties); Gadarene swine (Mark 5, ~2,000, Decapolis); Hwch Ddu Gwta and its chant; Jørgensen 2013 citation and thesis; pinfolds/amercements/1425 swineherds case; the manor-court-only rule for recovering strays; pannage (Domesday, New Forest figures); Savigny 1457; Evans 1906 as the foundational compilation; Animals Act 1971 s.4; wild boar re-establishment by escape (1987 storm, Forest of Dean, 2004 Staunton release, escape-incident rate); Tamworth Two; the motorway escape incidents; "pig in a poke" (Heywood 1555–60, *poque* etymology); "sow's ear/silk purse"; "hog wild" as American 1904. + +**PROBABLE** — Henwen (encyclopedic and secondary sources only; verify against Bromwich, *Trioedd Ynys Prydein*); Rhŷs's *dindsenchas* reading of the Twrch Trwyth route; *Historia Brittonum* and *Gwarchan Cynfelyn* attestations; Thompson motif numbers (B16.1.4, B183 etc. — **verify the exact number before citing**); Gloso/Gravso (folklore-blog sourcing only); tantony pigs and the 1311 Roger de Wynchester case; Falaise 1386 (**the human-clothing and fresco details are the weakest links in the animal-trial literature**); the *Sophia* statistics on pig defendants; *marrano* etymology (genuinely contested — present both theories); "when pigs fly" 1586 Withal attribution; *Schwein haben* origin theories; Manannán's regenerating swine; ATU 441 cross-classification. + +**THIN / DUBIOUS — do not use without further work** +- **"Orc Triath" and the "pigs of Drebrenn"** — requested in the brief; I could not corroborate either. Only *Torc Triath* ("king of the swine", linked to Brigid) surfaced, and thinly. +- **Greased pig contests having 17th-century Irish/Celtic roots** — asserted in popular sources, **no scholarly support found. Mark UNSOURCED.** +- **"Hog wild" deriving from hogs breaking loose** — folk etymology; the OED does not commit. +- **Pigling Bland publication date** — a source said 1922; **1913 is standard** and fits the biography. Verify. +- **The bridge in Pigling Bland as a county boundary** — from memory, **UNVERIFIED**. + +**Known gaps in this file (be honest about these in the essay)** +1. **Eastern, southern and Nordic proverb material is essentially absent.** The Russian/Slavic search returned Buryat (Siberian) material, which is outside the continent and unusable. No Polish, Balkan, Italian, Iberian or Scandinavian proverb corpus was reached. +2. **No primary texts were read.** `WebFetch` was blocked for every host, so the ATU index, the Motif-Index, Jørgensen's article, Evans 1906, and all tale texts are known only through summaries. +3. **Animal Farm** was not researched before the search budget ran out. The obvious point stands on general knowledge — the closing scene in which pigs become indistinguishable from men is a human/animal boundary dissolution, and the novel is set on a farm whose gates and fences are thematically loaded — but **it is uncited here** and should be checked. +4. **No French, German or Italian-language searching** was completed on folktale corpora specifically, only on idioms. +5. I found **no ATU type indexing "escaped pig"** as such. This absence is itself informative: in Europe the pig-escape theme lives in legend, law and record rather than in the Märchen index. diff --git a/research/boxes-and-escape/survey/pig-north-america.md b/research/boxes-and-escape/survey/pig-north-america.md new file mode 100644 index 0000000..697e02b --- /dev/null +++ b/research/boxes-and-escape/survey/pig-north-america.md @@ -0,0 +1,755 @@ +# Pigs and Escape / Containment / Boundary-Crossing — NORTH AMERICA + +**Continent scope:** Canada, United States (incl. Hawaii, flagged), Mexico, Central America, Caribbean. +**Compiled:** 2026-07-28. + +> **METHODOLOGICAL CAVEAT — READ FIRST.** This session completed ~20 web searches and then +> hit the session-wide WebSearch quota (200/200, shared with other agents). Direct `WebFetch` +> and `curl` were subsequently blocked at the agent-proxy gateway (403 on CONNECT for +> wikipedia.org, sacred-texts.com, jstor.org, scholar.google.com, ulukau.org, etc.). **Therefore +> every citation below rests on search-result snippets, not on the full text of the source.** +> Bibliographic details (journal, volume, year, page range) were returned by search and are +> almost certainly correct, but page-level quotations should be re-verified before publication. +> Items I know from general knowledge but could NOT verify in this session are explicitly +> marked `[UNVERIFIED THIS SESSION]`. + +--- + +## Summary + +There is a **great deal** of material — this is one of the richest continent/animal pairings +you could have picked, and the reason is structural rather than mythological. Pigs are not +native to the Americas. Every pig in North America is descended from an animal that was +*brought* and then, very often, *got out*. The continent's pig story is therefore an escape +story from the first decade of contact onward, and the folklore, law, idiom and politics all +sit downstream of that one fact. + +Five clusters carry real weight: + +1. **Colonial free-range swine and the invention of the fence law** (New England, Chesapeake, + later the postbellum South). The single best-documented and most historically consequential + thread. Pigs were the animals colonists could not contain, and the legal machinery invented + to deal with them — hog reeves, "sufficient fence" statutes, ringing and yoking, stock laws — + became a mechanism of dispossession against Indigenous farmers and, two centuries later, of + labour control against freedpeople. Escape → law → boundary → power. +2. **Urban containment battles**: New Amsterdam's wall (pigs literally undermining it), and the + nineteenth-century New York hog wars culminating in the 1859 Piggery War. A city defining + itself by which animals may cross which lines. +3. **The Pig War of 1859** (San Juan Island): an international boundary crisis detonated by one + boar crossing one property line. On-theme to the point of parody, and entirely real. +4. **Modern feral swine** — the largest ongoing escape-and-proliferation event on the continent, + in the US South, the Southwest, Hawaii, and now the Canadian prairies ("super pigs"). Plus + its meme afterlife ("30-50 feral hogs"). +5. **Idiom and ritual**: a dense American proverbial layer ("root hog or die", "hog wild", + "pig in a poke", "when pigs fly", "high on the hog", "sooey"), and the greased-pig contest, + which stages uncatchability as public entertainment. + +The **weakest** area, and the one requiring the most discipline, is **Indigenous material**. +Pigs are post-contact everywhere on the mainland; peccary/javelina are native but the +English-language internet is saturated with fabricated "Native American javelina symbolism". +The genuine Indigenous pig traditions I could verify are (a) Mesoamerican — Nahuatl and Maya +categorisation of the peccary and its extension to the Spanish pig, and Yucatec Maya were-pig +belief; and (b) **Native Hawaiian — Kamapuaʻa**, which is the strongest single Indigenous +item on the whole theme, and is about a pig who cannot be tied up. + +--- + +## Escape / Containment Motifs + +Using the requested three-way label: +**(a) pigs escaping enclosures** · **(b) pigs as boundary markers** · **(c) pigs in stories with no containment theme** + +### Kamapuaʻa — the Hawaiian hog-man who cannot be bound — (a) + (b) — PROBABLE→SOLID +The strongest genuinely Indigenous item on the theme. Kamapuaʻa is a Hawaiian kupua (shapeshifter) +who takes hog and human form. The tradition is full of binding-and-escape: + +- After stealing fowls from **Olopana**, chief of Oʻahu, Olopana's men capture him and **tie him + fast with cords; he bursts the bonds** and kills all but one of the men. +- In one version he is captured in hog shape and **tied to a pole repeatedly**, and each time his + grandmother releases him with a **chant**. (Escape by incantation rather than by force.) +- At **Kaliuwaʻa** (Kaluanui, windward Oʻahu), cornered at a waterfall, he transforms into a hog + whose body his followers climb to escape; the gorge is named for the episode. +- The Kamapuaʻa–Pele cycle ends in a **division of the island** between them (wet windward + vs. dry leeward), i.e. the pig-god as the agency by which a territorial boundary is drawn. + +*Sources:* Martha Beckwith, *Hawaiian Mythology* (Yale UP, 1940), ch. XIV — canonical academic +treatment. Lilikalā Kameʻeleihiwa (ed./trans.), *A Legendary Tradition of Kamapuaʻa* +(Bishop Museum Press, 1996) — edition of the Hawaiian-language *moʻolelo*. Thomas Thrum, +*Hawaiian Folk Tales*, "Kaliuwaʻa — Scene of the Demigod Kamapuaʻa's Escape from Olopana." + +*Caveats:* (i) Geographic — Hawaii is a US state but Oceanic; if a sibling researcher covers +Oceania this may be double-counted. Flag it, don't drop it. (ii) The pua'a was a **Polynesian +introduction** (*Sus scrofa*, voyaging canoes), so this is Indigenous tradition about an +introduced animal — the same "brought-then-escaped" structure as the mainland, a millennium +earlier. (iii) I could not fetch Beckwith directly (sacred-texts.com 403); details above come +from search snippets that agree with each other and with the standard literature, but verify +episode-by-episode before quoting. + +### Wilbur's escape, *Charlotte's Web* — (a) — SOLID +E. B. White, *Charlotte's Web* (1952). **Chapter 3 is titled "Escape."** Wilbur, egged on by the +goose ("push on it… push, push, push!"), shoves a loose board and gets out of the pigpen into +the Zuckermans' orchard. He is immediately overwhelmed by freedom, cannot decide what to do +with it, and is lured back into the pen by Lurvy with a bucket of slops. + +This is the purest literary statement of the American pig-and-containment theme available, and +it inverts it: the escape is a failure of nerve, not of fence. White — a keen and unsentimental +observer of livestock — makes the point that Wilbur is a *domestic* animal, and that the pen is +partly interior. Note also the frame: the whole book is about a pig for whom the real enclosure +is not the pen but the slaughter date, and the escape route is *language* (writing in the web). + +*Sources:* text of ch. 3; LitCharts / SparkNotes / GradeSaver chapter analyses corroborate the +plot detail. + +### Wáay Kekén / Uay Kekén — the Yucatec Maya were-pig — (b), boundary-crossing not escape — PROBABLE +Yucatán legend of a *wáay* (shapeshifting sorcerer) in pig form that roams at night making +noise and terrifying a town; in the widely-circulated version two young men wait in a tree with +stones at midnight to ambush it. Part of the Yucatec *wáay* complex alongside Wáay Chivo (goat) +and Wáay Peekʼ (dog). The *wáay* is definitionally a **boundary-crosser** — human/animal, +day/night, village/monte — but this is NOT an escape-from-enclosure motif. Label (b)/(c), +not (a). + +*Source:* *Diario de Yucatán* (yucatan.com.mx), "El Uay Kekén: la historia detrás de la leyenda +del cerdo embrujado en Yucatán", 25 Mar 2022. Journalistic, not academic — for academic footing +see the ethnographic literature on Yucatec *wáay*/*way* beliefs. + +### Nahual/nagual in pig form — (b) — PROBABLE +Mexican (Tlaxcala, Puebla, and widely) belief that brujo-nahuales prefer to take the shape of +**common domestic animals — horses, pigs, dogs, coyotes, serpents — precisely because those +forms let them pass unnoticed among the corrals and enter houses.** That is a containment/ +boundary motif of a specific and interesting kind: the pig-form is chosen *because it defeats +the perimeter*. An animal shape that is a skeleton key. + +*Source:* Mexican folklore compilations (mitos-cortos.com; BUAP Complejo Cultural Universitario +listing of *leyendas de miedo*). **THIN as cited** — these are popular retellings. The nahual +literature proper (Aguirre Beltrán, *Medicina y magia*; López Austin, *Cuerpo humano e +ideología*) would give the real footing. Do not attribute a specific tale to a specific +community on this evidence. + +### Peccary in Mesoamerican classification — (c), but linguistically on-theme — PROBABLE +Sahagún's *Florentine Codex*, Book XI (Earthly Things), describes the **coyametl** (peccary) +immediately after the coati: bristly, coarse-haired, eating "acorns, American cherries, maize, +roots, fruit, just like what a pig eats. Hence they call the peccary a pig." And: pig is called +**pitzotl** "because when it eats it makes a smacking sound, as if it sucks." Spanish glosses +it against *jabalí*. + +The interest here is **category boundary-crossing**: the introduced Old World pig arrives and is +absorbed into an existing Nahuatl category built around the native peccary, and the words then +compete. *Pitzotl/pizote/puerco* semantics get tangled across Mexico and Central America to this +day. Not a containment story — a taxonomic one. Label (c) with a linguistic note. + +*Sources:* Sahagún, *Florentine Codex* Book 11 (Dibble & Anderson trans., Univ. of Utah Press; +Getty Digital Florentine Codex online). Mexicolore, "Did the Aztecs know about peccaries?" +(secondary, cites Sahagún). + +### Ancient Maya peccaries — (c) — SOLID as iconography, NOT a containment motif +Peccaries appear extensively in Classic Maya art: the **carved peccary skull from Tomb 1 at +Copán**, celestial/constellation associations at Copán, peccaries in the **Bonampak** murals in +night-sky context, peccary tetrapod vessel supports, anthropomorphised peccary figures. There is +also evidence for Maya **management/penning of peccary herds** for ritual and food, which is +tantalising for a containment essay but is an economic fact, not a narrative motif. + +*Sources:* Christopher Götz & Kitty Emery et al., "Peccaries in Ancient Maya Economy, Ideology, +and Iconography" (academia.edu PDF). Maya Decipherment blog (Stuart/Houston), "Maya Creatures V: +The Peccary's Teeth, the Jaguar's Bone." FLAAR Mesoamerica reports. +**Do not** claim these depict escape or release. I found no verified Mesoamerican "release of +impounded game" peccary tale. See Confidence Notes for what I looked for and failed to find. + +### `[UNVERIFIED THIS SESSION]` Other literary/pop-culture items worth checking +Flagged because they are near-certainly relevant but I could not run confirming searches: +- **Disney, *Three Little Pigs* (1933)** — the American form of AT 124. Depression-era allegory, + "Who's Afraid of the Big Bad Wolf." The tale is entirely about **enclosure quality** — three + grades of wall, two of which fail — and about a predator crossing a threshold (and being + defeated at the one boundary he cannot control, the chimney). This is the canonical + containment fable and deserves a paragraph. +- **Ellis Parker Butler, "Pigs Is Pigs" (1905)** — guinea pigs held in a railway express office + over a tariff dispute, breeding uncontrollably while bureaucrats argue the classification. + A containment-failure farce that is *also* a category-boundary joke ("is a guinea pig a pig?"). + Extremely on-theme; verify. +- **Babe (1995)** — pig escapes species role rather than a pen; crosses the pig/sheepdog boundary. + (Australian film from an English novel, but a North American cultural event.) +- **Porky Pig** (1935– ), **Arnold Ziffel** (*Green Acres*, 1965–71) — the pig admitted to the + house/the human sphere; boundary transgression as comedy. +- **Walter R. Brooks, *Freddy the Pig*** series (1927–58) — a pig who is a detective, poet and + traveller, i.e. structurally a pig who does not stay in the sty. +- **Esther the Wonder Pig** (Ontario, 2012– ) — sold as a "micro pig", grew to ~650 lb; a + containment story about a category that failed. Canadian. + +--- + +## Colonial Fence Law & Free-Range Swine + +**This is the strongest and most defensible section of the essay.** It is genuine environmental +and legal history with a major scholarly literature. + +### The core scholarship — SOLID +- **Virginia DeJohn Anderson, "King Philip's Herds: Indians, Colonists, and the Problem of + Livestock in Early New England," *William and Mary Quarterly* 3rd ser. 51:4 (Oct. 1994), + 601–624.** The key citation. Anderson: *"No problem vexed relations between settlers and + Indians more frequently in the years before the war than the control of livestock."* +- **Virginia DeJohn Anderson, *Creatures of Empire: How Domestic Animals Transformed Early + America* (Oxford UP, 2004).** Argues that livestock — pigs above all — were *"the principal + agents responsible for dispossessing the Indians of their homelands"*: the animals **occupied + land in advance of English settlers**, forcing Native communities either to fence against them + (an alien practice, and expensive) or to move. + +### Why pigs specifically — SOLID (mechanism), the framing is mine +Pigs were the colonists' cheapest protein precisely *because* they were not contained. You turned +them into the woods and let them find mast; you collected them (or some of them) later. That is +also exactly why they were ungovernable: a free-ranging omnivore that roots. English cattle +damaged Indigenous cornfields; pigs did that **and** rooted up the clam and shellfish beds and +groundnut grounds that were Indigenous common resources — which were not, and could not be, +fenced. The animal's economic advantage to the colonist and its destructiveness to the +Indigenous economy were the same property. + +### The legal machinery +- **"Sufficient fence" law.** Chesapeake statute language: *"Every man shall enclose his ground + with sufficient fences uppon theire owne perill."* This is the crucial inversion: in English + practice the herdsman controlled the stock; in America the **burden of exclusion fell on the + crop-grower**. Animals had legal right of way over any land not physically barricaded. + Anderson reads the first Chesapeake fence laws as an *implicit rejection of English husbandry*. + Consequence for Indigenous farmers: their fields were legally trespassable by colonial pigs + unless they built English-style fences — an obligation imposed by a legal order they had not + entered. **SOLID.** + *Sources:* Anderson, above; Colonial Williamsburg Research Report RR0134, "Partitioning the + Landscape: The Fence in Eighteenth Century Virginia"; Encyclopedia.com, "Fencing and Fencing + Laws"; National Gallery of Art, *History of Early American Landscape Design*, s.v. "Fence." +- **The hog reeve / hogreeve.** Elected town officer in colonial New England charged with + **impounding stray swine and appraising the damage they did**. Among the *earliest elected + offices in colonial North America* — the earliest rights specifically granted to New England + towns concerned the herding of swine and cattle and the regulation of fences and common + fields. i.e. **the first thing colonial democracy did was elect someone to deal with loose + pigs.** That is a genuinely striking fact for an essay. **SOLID.** + *Sources:* "Hog reeve" (Wikipedia — could not fetch, snippet only); NEHGS *Vita Brevis*, + "Historic occupations." + `[UNVERIFIED THIS SESSION]` The often-repeated New England custom that **newly married men were + appointed hog reeve** as a joke office is worth chasing; I could not confirm it. +- **Ringing and yoking.** Statutes across the colonies required swine at large to be **ringed** + (a ring through the snout to stop rooting) and **yoked** (a triangular wooden yoke around the + neck to stop them passing through fence gaps). This is the most literal object in the whole + research question: a device worn by the animal whose sole function is to make it unable to + cross a boundary. `[UNVERIFIED THIS SESSION — I know this from general knowledge of colonial + statute books and it appeared adjacent in results, but I did not get a clean citation. Verify + in the Massachusetts Bay Colony records or Hening's *Statutes at Large* (Virginia).]` +- **Hog islands.** The standard colonial trick of putting swine on an island so the water did + the fencing — hence the Hog Island / Hogg Island toponym recurring all down the Atlantic + coast (Virginia, Massachusetts, Connecticut, Bahamas). `[UNVERIFIED THIS SESSION]` but + well known and easy to confirm; the toponymic evidence alone is an argument. + +### The postbellum South: the stock law wars — SOLID +The same fight, fought again after 1865, with race and class explicit. +- Pre-Civil War South was overwhelmingly **open range** (fence-law/"fence-out" regime). +- Post-1870s state legislation let counties or sub-county districts vote to adopt **stock laws**, + shifting the burden from crop-growers to livestock owners — i.e. **closing the range**. +- **Steven Hahn, *The Roots of Southern Populism: Yeomen Farmers and the Transformation of the + Georgia Upcountry, 1850–1890* (Oxford UP, 1983)** and his article on property law as labour + control (*Law and History Review*): the stock laws were a deliberate instrument of **labour + control**. A freed family could keep a hog and a cow on the unfenced woods without owning an + acre; close the range and that independence disappears, leaving wage work on the plantation. + Hahn: upcountry yeomen fought for open range as a **"natural right."** +- **Shawn Everett Kantor, *Politics and Property Rights: The Closing of the Open Range in the + Postbellum South* (Univ. of Chicago Press, 1998)** argues the opposite — an efficiency shift + once fencing stock became cheaper than fencing crops. **The historiographical dispute is + itself usable**: was the fence an economic instrument or a political one? +- **The Buncombe County (NC) Stock Law Revolt, 1885–87** — an actual armed-ish revolt against + range closure. Good concrete anchor. + *Source:* The Southern Highlander, thesouthernhighlander.org/stock-law (regional history site; + find the underlying scholarship). +- **Drew A. Swanson, "Fighting over Fencing"** (Forest History Society, Blegen Award paper, 2010) + — foresthistory.org PDF. Useful survey. + +**The through-line worth stating plainly:** in North America, the question "who has to build the +fence?" has twice been the central question of who owns the country — against Indigenous farmers +in the seventeenth century, against freedpeople in the nineteenth. Both times the pig was the +animal that made the question urgent. + +--- + +## Urban Containment: Manhattan's Wall and the Piggery War + +### The wall that pigs undermined — 1653 — PROBABLE→SOLID +The New Amsterdam wall (whence **Wall Street**) was built in spring 1653 as a defensive +fortification against the English — that is its primary and correct explanation, and the +"the wall was to keep pigs out" story is a **folk simplification that should be labelled as +such**. But the pig element is documented and is better than the myth: + +- Settlers let livestock run loose; hogs uprooted orchards and gardens and **foraged along the + line of the wall, interfering with its construction**. +- In a **letter of March 1653** — during construction — Director-General **Peter Stuyvesant** + urged the city government to act against the pigs at Fort Amsterdam, detailing "with great + grief the damages done to the walls of the fort by hogs, especially now again the spring when + the grass comes out." +- The city hired a herdsman to keep the pigs off. It didn't work. + +So: not "a wall to keep pigs in or out," but **a defensive wall that pigs were actively +destroying while it was being built.** Better material than the myth, and honest. +*Sources:* Untapped Cities, "Wall Street Wall Almost 'Destroyed' by Pigs"; Bowery Boys, +"Building the Wall: How Wall Street Got Its Name"; NYC Dept. of Records & Information Services +blog; New York Almanack, "New Amsterdam's Wall: An Etymology" and "New York Pork: A Porcine +History of the Big Apple." + +### The Piggery War, 1859 — SOLID +The best single urban episode. +- **Catherine McNeur, "The 'Swinish Multitude': Controversies over Hogs in Antebellum New York + City," *Journal of Urban History* 37:5 (Sept. 2011), 639–660** — the scholarly citation. +- **Catherine McNeur, *Taming Manhattan: Environmental Battles in the Antebellum City* + (Harvard UP, 2014)** — the book. Her dissertation, "The Swinish Multitude and Fashionable + Promenades," won the Urban History Association best-dissertation award and the ASEH Rachel + Carson Prize. This is a genuinely first-rate source. +- The story: as Manhattan developed, ordinances pushed loose hogs off the streets and into + **"piggeries"** at the city's edge — roughly modern midtown between Sixth and Seventh Avenues, + known as **"Hogtown."** In summer 1859 the new City Health Inspector **Daniel DeLavan** ordered + the piggeries removed. On **27 July 1859** the *New York Times* reported **eighty-seven armed + men** — pistols, clubs, daggers, pickaxes, crowbars — descending on the district; the city + destroyed offal-boiling establishments, rounded up pigs, and in places tore apart residents' + homes. The pig-keepers were largely Irish, German, Dutch and African-American poor. + **Women were the most militant defenders** of the piggeries, taking up clubs against police. + The piggeries eventually relocated upstate. +- Earlier rounds: hog riots in the 1820s (McNeur covers these), where poor New Yorkers physically + fought city marshals attempting to seize street-roaming pigs. +- The class logic is exact and worth stating: **a pig that forages the street costs nothing to + feed.** Free range in the city was a subsistence strategy for people with no land. Enclosure — + literally, the piggery, then expulsion — was gentrification. McNeur's argument. + +*Additional sources:* NYPL blog, "Milestones in NYC's Trash Revolution"; Ephemeral New York, +"What happened to Manhattan's 'Piggery District'"; RENDER magazine, "Protecting their Pigs: +Women and Urban Agriculture in Antebellum Manhattan"; The Real Deal, "NYC's first gentrification +movement was about pigs." + +--- + +## Proverbs & Idioms + +| Phrase | Earliest / origin | On-theme? | Confidence | +|---|---|---|---| +| **"Root hog or die"** | Earliest literary use **Davy Crockett, *A Narrative of the Life of David Crockett* (1834)**; but already reported as an established proverb in the ***Vermont Gazette*, 1829**: *"In Ohio, they have a vulgar proverb which runs thus—'Root, hog, or die.' It is usually spread in staring capitals among idlers and it is said often has an admirable effect in promoting habits of industry."* Later a minstrel/patriotic song family, Civil War era. | **Directly.** The proverb *presupposes* the free-ranging hog: it means "you have been turned loose, forage or perish." Self-reliance as a condition imposed by non-containment. The most on-theme American pig proverb there is. | SOLID | +| **"Hog wild"** | OED earliest evidence **1893, *Galveston Daily News*** (Texas); "to go hog-wild" American English from **1904**. | **Directly.** The image is livestock out of control — escape as a *state of mind*. | SOLID | +| **"Pig in a poke"** | *Poke* = bag (13c. Eng., from Old French *poque*; *pocket* is its diminutive). Proverb first recorded in **John Heywood's *Proverbs* (1546)**; medieval market fraud of substituting a cat or pup for a piglet. Cognate proverbs across Europe warn against buying "a cat in a bag." | **Inverted containment.** The pig here is the thing that *should* be in the sack and isn't. The related "let the cat out of the bag" (only attested from the later 18c.) is folk-connected to it; the connection is **etymologically dubious** and should be flagged. | SOLID (for the pig phrase); the cat-link is DUBIOUS | +| **"When pigs fly"** | An **adynaton** (rhetorical figure of impossibility). Earlier form *"pigs fly with their tails forward"*, in the 1616 edition of **John Withals, *A Shorte Dictionarie for Yonge Begynners***. "When pigs fly" is the specifically **American** variant; "pigs might fly" the British. | **Counterfactual boundary.** Names the limit of the possible using the animal that is defined by being earthbound and penned. Nice pairing with the greased pig: one is uncatchable, the other unliftable. | SOLID | +| **"High on the hog"** | American. **OED first citation: *Kansas City Times*, 28 Nov. 1919.** Widely quoted **NYT 1920**: "Southern laborers who are 'eating too high up on the hog' (pork chops and ham) and American housewives who 'eat too far back on the beef'… are to blame for the continued high cost of living." | Class/anatomy boundary — a line drawn *across the animal's body*. Only tangentially about containment. | SOLID for the phrase. **The popular "plantation origin" story (masters got the upper cuts, enslaved people got the feet) is NOT documented in early sources — mark THIN/DUBIOUS if you use it.** | +| **"Lipstick on a pig"** | Precursor: **1926**, Charles Lummis in the *Los Angeles Times*, "Most of us know as much of history as a pig does of lipsticks." Modern phrase first recorded **1985, *Washington Post***, on a San Francisco park renovation. Massive 2008 US election moment. | Not containment — a **fallacy/rhetoric** item: the pig as the irreducible substrate that no surface change can transform. Useful as a counterpart to "when pigs fly." | SOLID | +| **"Sooey" / "soo-ee" hog call** | Merriam-Webster: "probably an alteration of *sow*." Wiktionary: uncertain; possibly related to *sow*, or to *Sus/Suidae* (the latter is probably folk-etymology). Documented across the South since the 19th century. Institutionalised as the University of Arkansas **"Calling the Hogs"** ("Woo Pig Sooie"). | **Directly and beautifully.** Hog calling is the technology of **recall without a fence** — the acoustic counterpart to the pen. It exists only because the hogs are out there in the woods. That it survives as a football chant is a fossil of open-range husbandry. | SOLID (practice); THIN (etymology) | + +### Spanish-language: the vocabulary of going feral +This is a distinct and important Caribbean/Latin American contribution, and it is **linguistically +the deepest item in the whole file.** + +- **`cimarrón`.** Originally applied to **domestic livestock that had taken to the hills in + Hispaniola** — cattle and hogs gone wild — and only *afterwards* to escaped Indigenous slaves, + and then to escaped enslaved Africans, giving English **maroon** (via French *marron*, by the + 1660s) and the whole vocabulary of *cimarronaje* / maroon communities. Etymology contested: + Spanish *cimarra* "thicket" ← *cima* "summit," "living wild in the mountains"; versus a + Taíno root **símaran** "wild, gone astray" (possibly ← *símara* "arrow"), with the two + converging. **The point: the word for a human being who escapes bondage in the Americas was + first the word for a pig or a cow that got out.** The escaped animal supplied the concept. + **SOLID** for the livestock-first sequence; the Taíno etymology is contested — present both. + *Sources:* Encyclopedia.com, "Maroons (Cimarrones)"; Online Etymology Dictionary, *cimarron*; + Wiktionary *cimarrón*; Smithsonian Folklife Festival 1992, "Maroons: Rebel Slaves in the + Americas." +- **`puerco jíbaro` / `cerdo cimarrón` (Cuba).** The feral pig, introduced at the Spanish + conquest, is called the *puerco jíbaro*, also *cerdo cimarrón*; medium-sized, ~1,400 mm, + using its canines both for defence and for rooting. In Cuba **`jíbaro`** was applied to + *perros realengos* (masterless dogs) — meaning a domestic animal **"que se había hecho + montaraz o mostrenco"**, gone wild, become a forest-dweller. Pichardo's *Diccionario de voces + cubanas* (1836) glosses *jíbaro* as an Indigenous word meaning **"montaraz, rústico, + indomable"** — untameable. The term is attested in Puerto Rican newspapers from about 1820, + where it becomes the name of the mountain peasant and eventually the national type. + **So the Puerto Rican `jíbaro` — the emblem of the nation — carries inside it a word first + used of an animal that would not stay in the pen.** As one Cuban commentator puts it, the + notion *"sugería la cimarronería o anarquía de la altura… un jíbaro era un ser arisco, difícil + de controlar."* + **PROBABLE→SOLID** for the semantics; the exact route from feral animal to peasant identity is + debated in the Puerto Rican historiography (see the *Puerto Rico entre siglos* essays). + *Sources:* EcuRed, "Puerco jíbaro"; *Puerto Rico entre siglos* (Wordpress academic blog), + "Jíbaros, criollos, puertorriqueños" and "¿Qué significa lo jíbaro?"; Claridad; Britannica, + *jíbaro*; Corominas, *Diccionario crítico etimológico*, s.v. *jíbaro*; 80grados, "La historia, + el cerdo y el 'carne'e puerco'." + **CAUTION:** keep *jíbaro* (Puerto Rico/Cuba) distinct from **Jívaro/Shuar** (Ecuador/Peru); + the resemblance is a trap and the latter is a separate, contested exonym. +- **`buccaneer` ← `boucanier`.** The Hispaniola sequence: Spanish depopulation of the island in + the early 17th century **turned loose herds of cattle and hogs that reverted to the wild**; + French and English drifters, deserters and shipwrecked men on Tortuga and northern Hispaniola + lived by hunting these feral animals and smoking the meat on a **boucan** (a grill; the word + borrowed from Tupi or another Brazilian Indigenous language, *buka*). *Boucanier* = hunter of + wild cattle/hogs → **buccaneer**. **The word for a Caribbean pirate is derived from the word + for a man who lives by hunting escaped pigs.** SOLID. + *Sources:* Wordorigins.org, "buccaneer"; Wikipedia, "Buccan"; Exquemelin, *Buccaneers of + America* (1678), the Hispaniola chapters; NC Queen Anne's Revenge Project, "Did You Know That + Pirates and Wild Pigs Are Related?"; and the excellent academic framing in **"Creole + Ecologies, Feral Customs: A Coevolutionary History of Buccaneering in Hispaniola During the + Seventeenth Century"** (2022, ResearchGate PDF) — argues feral cattle and pigs were **central + to the creation of that cosmopolitan society at the margins of colonial order.** + +**Composite claim you can make, and it is defensible:** in the Caribbean, escaped pigs generated +(i) the vocabulary of human escape from slavery (*cimarrón*/maroon), (ii) the vocabulary of +piracy (*boucanier*/buccaneer), and (iii) the vocabulary of national peasant identity +(*jíbaro*). Three of the region's defining forms of life outside the colonial enclosure are +named after animals that got out of the enclosure first. + +--- + +## Ritual Enactments (greased pig etc.) + +### Greased pig contest — PROBABLE (practice SOLID, deep origin THIN) +The literal ritualisation of uncatchability: a piglet is coated in lard, grease or Crisco and +released into a ring or field; contestants — usually children — have a fixed time to catch and +hold it. Documented across US county fairs, fall festivals, and **Fourth of July celebrations**; +archived newspapers show it advertised routinely. + +- **Documentation:** National Endowment for the Humanities, "Bonfires, Greased Pig Races, + Pickle Contests, and More: Historic Fourth of July Celebrations from Chronicling America" — + i.e. it is attested directly in the **Chronicling America** newspaper archive, which is the + right primary source to mine. *Star Tribune* (Minnesota) piece on the history and eventual + legal ban. Still run at e.g. the **Wirt County Fair** (WV); **Martin County, Florida** uses + Crisco; there has been a "greased pig war" in Florida over animal-welfare objections. +- **Deep origin:** claims of "Celtic roots dating back at least to the 17th century in Ireland" + and "medieval European carnival games involving swine pursuits" circulate but are **THIN** — + I would not assert them. What is safe: the practice is documented in America across at least + 200 years, formalised at county and state fairs by the early 20th century, and boomed + post-WWII into the 1950s–60s (often merging with pig wrestling / "pig scramble"). +- **Modern status:** increasingly banned on animal-welfare grounds (Minnesota; various fairs) — + which is itself a nice closing beat: the ritual of uncatchability is now itself being + fenced in. +- **Reading:** the grease is the point. The contest does not test whether the pig can be caught; + it manufactures a pig that *cannot* be caught, and then makes the failure the entertainment. + Compare the **rodeo** (animal contained by skill) and the **turkey shoot** (animal killed at + distance): the greased pig is the only American fair event whose subject is pure evasion. + +### Hog calling contests — PROBABLE +Competitive hog calling ("soo-ee") is a standard county-fair event and, at the University of +Arkansas, a mass ritual (the "Woo Pig Sooie" Hog Call, adopted in the 1920s). Pair it with the +greased pig: **one ritual enacts the pig you cannot catch, the other the pig you can only +summon.** Both are fossils of a husbandry system with no fences. +*Source:* THV11 (Little Rock), "Woo pig soooie! History behind the beloved 'Hog Call'"; +Wikipedia, "Calling the Hogs." + +--- + +## Feral Hogs + +### The founding escapes — SOLID +- **1493:** Columbus stocks the West Indies with domestic pigs to provision the fleet. They + reproduced so fast that **the Spanish crown ordered the population reduced within about + twelve years.** (First recorded pig-overpopulation panic in the Americas — 500 years before + the "feral swine bomb," same story.) +- **1539:** **Hernando de Soto lands 13 pigs at Tampa Bay** — the standard "first pigs in the + continental US" date. Over three years his expedition crossed what are now 14 states, ~3,100 + miles; **the herd grew to a reported ~700**, and along the way pigs variously **escaped, + were given to, or were stolen by** the Indigenous nations the Spaniards encountered. +- **No suid is native to the Western Hemisphere.** Every pig here descends from Eurasian *Sus + scrofa* brought by ship. Later Spanish, French and English colonists repeated the pattern; + escaped stock went wild and established across many regions. +*Sources:* feralhogs.extension.org, "History"; rootingforpigs.com; National Geographic, "The +battle to control America's 'most destructive' species: feral pigs"; High Country News, +"Invasion of the feral pigs"; USDA Southwest Climate Hub, "feral swine bomb." + +### Speed of feralization — PROBABLE (nuance matters) +The popular claim is that a released domestic pig "grows tusks and bristles within months." +The literature is more careful: +- **Behavioural** change (wariness, altered social structure) appears **within the first months**. +- **Morphological** change is more gradual; roughly **three generations** for coarse coat and + lean body to dominate a population. +- Mechanism is **phenotypic plasticity plus rapid natural selection** on ancestral variation. +- Tusks: domestic males are usually castrated, suppressing tusk growth; feral males express the + secondary sexual characters fully — so part of the apparent transformation is the removal of a + human intervention rather than a change in the animal. +- Feral swine are **not** reverted wild boar; they are a distinct feral phenotype. +- There is real neuroanatomical work on this: brain-size variation across domestic, feral and + wild *Sus scrofa* (PMC11407859). +**Essayistic value:** this is the literal biology of "refusal to stay put" — an animal that +carries its own wildness intact under domestication and re-expresses it on release. Use it, but +use the three-generation figure, not the "three months" folklore. + +### Scale and the containment failure — SOLID +- Feral swine are described by USDA and by researchers as among the most destructive invasive + species in the US; the USDA Southwest Climate Hub uses the phrase **"feral swine bomb"** + (attributable to Dale Nolte of USDA APHIS's National Feral Swine Damage Management Program). +- **Canada — the "super pigs."** The best single modern escape narrative on the continent. + **Eurasian wild boar were imported to Canada in the late 1980s–early 1990s to diversify + livestock production, and as "penned game" for shooting operations.** When the market + collapsed, animals were **released, or escaped from abandoned and poorly maintained + enclosures.** Crossbred with domestic swine, they combine wild-boar survival with domestic + size and fertility. Now firmly established in **Saskatchewan, Alberta and Manitoba**, with + populations in BC, Ontario and Quebec; range **over 750,000 km²**, expanding by an average of + **88,000 km² per year over the past decade**; cold-adapted, breeding in any season, and + sheltering in snow burrows nicknamed **"pigloos."** Now approaching the US border. Principal + researcher **Dr. Ryan Brook, University of Saskatchewan** ("Canada's Chairman of the Boar"), + who calls feral swine "the most invasive animal on the planet" and "an ecological train wreck." + *Sources:* University Affairs profile; National Geographic, "Huge feral hogs invading Canada, + building 'pigloos' as they go"; CBC News and CBC *Quirks & Quarks*; USask College of + Agriculture and Bioresources; Brook's Alberta Invasive Species Council presentation PDF. +- **Louisiana/New Orleans:** Historic New Orleans Collection, "A Pandemic of Pigs: Feral Hogs + Are Threatening Cities from New Orleans to Hong Kong" — useful for the *urban* feral hog, + which closes the loop with the Piggery War. + +### Hogzilla — the tall tale, fact-checked — SOLID +Shot by **Chris Griffin at Alapaha, Georgia, 17 June 2004**. Claimed at **12 feet and over +1,000 lb**. **National Geographic sent a team** (pig geneticist, wildlife ecologist, behaviour +specialist), **exhumed the carcass in November 2004**, and measured it at roughly **7.5–8 feet +and ~800 lb** — hugely exaggerated, but still exceptional (typical feral hog 100–500 lb). +DNA showed a **wild boar × domestic (Hampshire) hybrid**. +**This is the American tall tale in its modern form**: the escaped animal becomes a monster in +the retelling, and the monster is then dug up and measured. Snopes has a file on it; the +*New Georgia Encyclopedia* covers Georgia hog history. +*Sources:* Wikipedia, "Hogzilla"; Snopes; NBC News, "Monster swine 'Hogzilla' was real, experts +say"; New Georgia Encyclopedia, "Hogs." + +### "30-50 feral hogs" — SOLID +**4 August 2019**, days after the El Paso and Dayton mass shootings: musician **Jason Isbell** +tweeted criticising quibbling over the definition of "assault weapon." **@WillieMcNabb (William +McNabb, El Dorado, Arkansas)** replied: *"Legit question for rural Americans - How do I kill the +30-50 feral hogs that run into my yard within 3-5 mins while my small kids play?"* It went viral +the next day and became durable internet shorthand for an absurdly specific hypothetical +deployed against a general argument. + +Two things worth saying in an essay: (1) **the underlying claim was basically true** — that is +roughly what a sounder of hogs does in south Arkansas — and the meme is a case of urban +incredulity at a real rural condition; (2) the feral hog thereby completed its journey from +escaped livestock to **a figure of speech**, arriving at the same place as "root hog or die" +by an entirely modern route. +*Sources:* Know Your Meme, "30-50 Feral Hogs"; The Conversation, "30–50 feral hogs? Why Twitter +memes are more positive (and much faster) than you might think"; Salon; BuzzFeed News. + +--- + +## Documented Real Escapes & The Pig War + +### The Pig War, 1859 — San Juan Island — SOLID +The best item in the file for the specific question of a **pig crossing a boundary**. +- The **Treaty of Oregon (1846)** set the border at the 49th parallel but left the channel + through the San Juan Islands ambiguous, so **both** Britain and the US claimed the islands and + both had settlers on them. +- **15 June 1859:** a **Berkshire boar** belonging to **Charles Griffin**, an employee of the + **Hudson's Bay Company**, rooted in the **unfenced potato patch** of American settler + **Lyman Cutlar**. It had done so repeatedly despite Cutlar's efforts to drive it off and his + complaints to the HBC. Goaded (in the standard account) by the laughter of a Company herdsman + watching, **Cutlar shot the pig.** +- Cutlar offered **$10**; Griffin demanded more; the British threatened to arrest Cutlar. +- **The escalation was not about the pig — it was about jurisdiction.** As the historians put it, + it was the conflict over **whose law applied** that turned a shot pig into an international + crisis. Americans requested military protection; Capt. George Pickett landed troops; the Royal + Navy responded; both sides massed forces. +- **Resolution:** joint military occupation from late 1859 until **1872**, when arbitration by + **Kaiser Wilhelm I** awarded the islands to the United States. **No human casualties. The + pig was the only fatality.** +- Now **San Juan Island National Historical Park** (NPS), with "American Camp" and "English + Camp." + +**Why it belongs at the centre of the essay:** the pig was doing exactly what North American pigs +had done since 1493 — ignoring a property line, because the fence-out regime meant Cutlar's patch +was legally trespassable. The pig's inability to recognise a boundary made visible the fact that +the *humans* did not agree where the boundary was. An animal that will not stay put is an +excellent instrument for discovering that a border is fictional. + +*Sources:* National Park Service, San Juan Island NHP; **"From Imbroglio to Pig War: The San Juan +Island Dispute, 1853–71," *BC Studies*** (UBC, open access PDF) — the scholarly treatment; +*The Canadian Encyclopedia*, "The Pig War"; HistoryLink.org essay 5724, "San Juan Island Pig +War"; *Seattle Times* Pacific NW Magazine feature. + +### Pigasus the Immortal, 1968 — SOLID +**23 August 1968**, just before the Democratic National Convention in Chicago: the **Youth +International Party (Yippies)** nominated a **145-lb pig** for President. Selected by **Abbie +Hoffman, Jerry Rubin and Dennis Dalrymple**; obtained from a farmer by folk-singer **Phil Ochs**. +At the announcement rally in the Civic Center Plaza, **Chicago police confiscated the pig and +arrested seven Yippies** for disorderly conduct. The humans were bailed out the same day; **the +pig's fate is unknown** — possibly the Humane Society, possibly a police officer's dinner table. +Charges were eventually dismissed after the Chicago Seven trial and appeals. + +Reading: a deliberate **category violation** — putting a pig where a candidate goes — staged +precisely to force the state to police the boundary in public, which it obligingly did by +arresting the animal. Note also the mirror: the police were being called "pigs" that same summer. +*Sources:* Wikipedia, "Pigasus (politics)"; *Chicago Sun-Times*, "The search for Pigasus: What +happened to Chicago's presidential pig?" (Aug 2025); Roz Payne Sixties Archive; Center for +Artistic Activism / Actipedia; porkopolis.org. + +### Slaughterhouse escapees — THIN individually, PROBABLE as a genre +A recurring American news genre: a pig escapes a transport truck or an urban slaughterhouse, +is chased through streets, and — crucially — **is usually spared**, being sent to a sanctuary +rather than back to the line. +- **Winston** — piglet escaped a New York City slaughter market, wandered for days, **dodged + traffic on Queens Boulevard**, was caught by city animal control and given to **Farm + Sanctuary**, which reports taking in 500+ animals fleeing urban slaughterhouses over a decade. +- **Jason the "Miracle Pig"** — among 766 animals seized in an October 2015 raid on an illegal + slaughterhouse in Loxahatchee, Florida; sent to Rooterville Sanctuary, Melrose, FL. +- **Dani and Sunshine** — piglets who jumped from a moving transport truck; Rosie's Farm + Sanctuary, Maryland. +- **167 pigs** abandoned in a triple-decker trailer on a Washington, DC street en route from a + North Carolina farm to a Pennsylvania slaughterhouse. +*Sources:* Global Animal; One Green Planet; *Newsweek*; Farm Sanctuary. +**Use with care** — most reporting is by advocacy organisations. But **the pattern is the +interesting thing and is well established**: North American culture grants a kind of informal +amnesty to the animal that escapes, which is a folk survival of the idea that successful +flight confers a claim. (The best-documented instance of that logic is **Cincinnati Freedom**, +the 2002 escaped cow — not a pig, but the same rule.) `[Cincinnati Freedom UNVERIFIED THIS +SESSION.]` Compare the British **Tamworth Two** (1998) — not North American, do not import it +without saying so. + +--- + +## Sources + +Grouped by reliability. **None of these were fetched in full** (see caveat at top); all +bibliographic data comes from search-result metadata and snippets. + +### Academic / peer-reviewed (highest priority to verify and cite) +- Virginia DeJohn Anderson, "King Philip's Herds: Indians, Colonists, and the Problem of + Livestock in Early New England," *William and Mary Quarterly* 51:4 (1994), 601–624. +- Virginia DeJohn Anderson, *Creatures of Empire: How Domestic Animals Transformed Early + America* (Oxford UP, 2004). Review: *Reviews in History* no. 471. +- Catherine McNeur, "The 'Swinish Multitude': Controversies over Hogs in Antebellum New York + City," *Journal of Urban History* 37:5 (2011), 639–660. [PubMed 22073436] +- Catherine McNeur, *Taming Manhattan: Environmental Battles in the Antebellum City* + (Harvard UP, 2014). +- Steven Hahn, *The Roots of Southern Populism: Yeomen Farmers and the Transformation of the + Georgia Upcountry, 1850–1890* (Oxford UP, 1983); and "Property Law as Labor Control in the + Postbellum South," *Law and History Review* (Cambridge). +- Shawn Everett Kantor, *Politics and Property Rights: The Closing of the Open Range in the + Postbellum South* (Univ. of Chicago Press, 1998). +- "From Imbroglio to Pig War: The San Juan Island Dispute, 1853–71," *BC Studies* (UBC, OA PDF). +- "Creole Ecologies, Feral Customs: A Coevolutionary History of Buccaneering in Hispaniola + During the Seventeenth Century" (2022). +- Martha Beckwith, *Hawaiian Mythology* (Yale UP, 1940), ch. XIV (Kamapuaʻa). +- Lilikalā Kameʻeleihiwa (ed.), *A Legendary Tradition of Kamapuaʻa* (Bishop Museum Press, 1996). +- H. E. M. Braakhuis, *Xbalanque's Marriage: A Commentary on the Qʼeqchiʼ Myth of Sun and Moon* + (Leiden diss., open access via Leiden Scholarly Publications). +- Götz, Emery et al., "Peccaries in Ancient Maya Economy, Ideology, and Iconography." +- Bernardino de Sahagún, *Florentine Codex*, Book 11 (Dibble & Anderson trans., Univ. of Utah + Press; Getty Digital Florentine Codex, florentinecodex.getty.edu). +- Drew A. Swanson, "Fighting over Fencing" (Forest History Society, 2010). +- Brain-size variation in domestic/feral/wild *Sus scrofa*: PMC11407859. + +### Institutional / museum / government +- National Park Service, San Juan Island National Historical Park. +- USDA APHIS National Feral Swine Damage Management Program; USDA Southwest Climate Hub, + "feral swine bomb." +- feralhogs.extension.org (Cooperative Extension), "History." +- Colonial Williamsburg Digital Library, Research Report RR0134, "Partitioning the Landscape: + The Fence in Eighteenth Century Virginia." +- National Gallery of Art, *History of Early American Landscape Design*, s.v. "Fence." +- NEH, "Bonfires, Greased Pig Races, Pickle Contests, and More" (Chronicling America). +- New York Public Library blog; NYC Dept. of Records & Information Services. +- Historic New Orleans Collection, "A Pandemic of Pigs." +- New Georgia Encyclopedia, "Hogs." +- University of Saskatchewan, College of Agriculture and Bioresources; Alberta Invasive Species + Council (Brook PDF). +- EcuRed (Cuba), "Puerco jíbaro." +- NEHGS *Vita Brevis*, "Historic occupations" (hog reeve). + +### Journalism / reference (usable, secondary) +- *New York Times*, 27 July 1859 (Piggery War) — **go to the archive directly.** +- *Vermont Gazette*, 1829 ("root, hog, or die") — via Chronicling America. +- *Kansas City Times*, 28 Nov. 1919 ("high on the hog") — via OED. +- *Galveston Daily News*, 1893 ("hog-wild") — via OED. +- *Washington Post*, 1985 ("lipstick on a pig"); *Los Angeles Times*, 1926 (Lummis). +- National Geographic (feral pigs US; "pigloos" Canada); CBC News and *Quirks & Quarks*; + *Chicago Sun-Times* (Pigasus, 2025); *Seattle Times*; HistoryLink.org 5724; + *The Canadian Encyclopedia*; University Affairs (Ryan Brook); *Star Tribune* (greased pig + ban, Minnesota); *Diario de Yucatán* (Uay Kekén); OnCuba; Claridad; 80grados; + *Puerto Rico entre siglos*. +- phrases.org.uk; wordhistories.net; Grammarphobia; etymonline; A Way with Words; + Merriam-Webster; OED online; Wordorigins.org. +- Know Your Meme; The Conversation; Salon; BuzzFeed News (30-50 feral hogs). +- Snopes (Hogzilla). + +### DO NOT USE +- **spiritanimalonline.com, spiritualwayfarer.com** and similar "javelina spirit animal / + Native American symbolism" pages. See Confidence Notes. +- **Grokipedia** pages appeared repeatedly in results (buccaneer, Hogzilla, pig wrestling, + Kamapuaʻa, hog calling, maroons, McNeur). LLM-generated; **do not cite**; treat any unique + claim found only there as unsourced. +- Fandom wikis, listverse, Quora, crystalinks.com. +- ipl.org / cliffsnotes / coursehero student-essay pages. + +--- + +## Confidence Notes + +### SOLID +- Colonial free-range swine, "sufficient fence" law, the hog reeve office, and the Anderson + thesis about livestock as agents of dispossession. +- The postbellum stock-law conflict and the Hahn/Kantor debate. +- New York: the 1859 Piggery War, DeLavan, "Hogtown," the 27 July 1859 NYT report, the class and + gender dimensions, McNeur as the scholarly anchor. +- The Pig War of 1859 in all its particulars. +- Feral swine origins (Columbus 1493, de Soto 1539/13 pigs/~700), current scale, Canadian + "super pigs" and Ryan Brook's figures, Hogzilla and its Nat Geo debunking, the 30-50 feral + hogs meme. +- "Root hog or die," "hog wild," "pig in a poke," "when pigs fly," "high on the hog" (the phrase, + not the plantation origin story), "lipstick on a pig." +- Cimarrón → maroon (livestock first, then people); boucanier → buccaneer. +- *Charlotte's Web* ch. 3 "Escape." +- Pigasus, 1968. + +### PROBABLE +- Stuyvesant's March 1653 complaint about hogs damaging the walls of Fort Amsterdam, and pigs + interfering with construction of the New Amsterdam wall. (Multiple independent popular-history + sources agree and quote the same letter; find it in the *Records of New Amsterdam* or Stokes's + *Iconography of Manhattan Island* before quoting.) +- Kamapuaʻa's binding-and-escape episodes. Consistent across sources; verify against Beckwith + and Kameʻeleihiwa episode by episode. +- The Cuban/Puerto Rican *jíbaro* ← feral-animal semantics (Pichardo 1836 gloss "montaraz, + rústico, indomable"). The semantic field is solid; the identity-formation argument is + contested in the historiography — present it as contested. +- Wáay Kekén as a live Yucatec legend (journalistic sourcing only). +- Greased pig contests as a documented long-running American fair practice. +- Feralization timeline (behavioural months / morphological ~3 generations). + +### THIN +- Any deep-historical (Celtic, medieval) origin for the greased pig contest. +- The "sooey ← Sus/Suidae" etymology (probably folk etymology; M-W prefers "alteration of *sow*"). +- The "let the cat out of the bag ← pig in a poke" connection (chronologically implausible; + wordhistories.net calls the linkage erroneous). +- The plantation-origin story for "high on the hog." +- Individual slaughterhouse-escape stories (advocacy sourcing). +- Nahual-in-pig-form specifics as attributed to particular communities. + +### DUBIOUS / DO NOT REPEAT — the trap the brief warned about +**Confirmed:** searching for javelina/peccary in Indigenous Southwestern tradition returns, at +the top of results, pages asserting that "Tohono Oʼodham and Yaqui traditions" hold the javelina +to symbolise "adaptability, resilience and community structure," that it appears "in creation +stories and seasonal ceremonies," and that "Yaqui traditions emphasize the javelina's tight-knit +herd behavior as a reflection of ideal community structure." **These come from +spiritualwayfarer.com and similar spirit-animal content farms. They carry no citations, name no +narrator, no collector, no publication. Treat as fabricated. Do not use.** + +Similarly: the claim that "in Aztec mythology the peccary was associated with Xolotl, who guided +the sun through the underworld" appeared in the same class of source. **Xolotl's canine +association is well attested; a peccary association is not, on this evidence. Mark +UNSOURCED/DUBIOUS.** + +**What I could NOT find, and specifically looked for:** +1. A verified North American Indigenous **"release of impounded game"** narrative featuring + peccaries — i.e. the Master/Owner of Animals who keeps game penned in a mountain or cave + until someone releases them (Thompson motif A1421 "Hoarded game released"; cf. the + Mesoamerican *dueño del monte / dueño de los animales* complex). This motif is genuinely + widespread in Mesoamerica and Amazonia and, if a peccary version exists in Maya or other + Mesoamerican ethnography, it would be **the single best Indigenous item for this essay** — + an exact escape/containment story about a native pig-relative. I could not confirm one before + running out of search budget. **RECOMMENDED NEXT STEP:** check J. Eric S. Thompson, + *Ethnology of the Mayas of Southern and Central British Honduras* (Field Museum, 1930) and + his *Maya History and Religion* (1970) catalogue of myths; Braakhuis, *Xbalanque's Marriage*; + and the Tzotzil/Tzeltal *ʼanjel* / Chʼortiʼ dueño-del-cerro literature. +2. Whether the Qʼeqchiʼ/Mopan **Sun and Moon** myth includes an episode of brothers-in-law or + hunters **transformed into peccaries**. Search surfaced Braakhuis's study (which frames the + myth around "marriage alliance and hunting ideology," and mentions tapir transformation) but + did not confirm a peccary transformation. **Do not assert it without checking Braakhuis.** +3. A clean citation for **colonial ringing/yoking statutes** and for the **"newlyweds appointed + hog reeve"** New England custom. +4. Whether O'odham legend collections (Saxton & Saxton, *Legends and Lore of the Papago and + Pima Indians*, 1973; Ruth Underhill's work) contain javelina narratives at all. **native- + languages.org/oodham-legends.htm is the right index to check** (Orrin Lewis's site is a + reasonable, citation-bearing gateway) — the gateway blocked my fetch. + +### Distinction check (as requested) +- **(a) pigs escaping enclosures:** colonial free-range swine; the de Soto and Columbus + introductions; feral hogs US/Canada/Hawaii; the Canadian farm escapes; Wilbur in ch. 3; + Kamapuaʻa bursting his cords; slaughterhouse runaways; the greased pig; buccaneer/cimarrón + feral livestock. +- **(b) pigs as boundary markers:** the Pig War boar (property line → international border); + hog reeves and pound/fence law as boundary infrastructure; ringing and yoking; Manhattan's + wall; the piggery district line in New York; Kamapuaʻa's division of Oʻahu with Pele; the + nahual in pig form as a form chosen to defeat perimeters; "high on the hog" as a line drawn + across the body. +- **(c) pigs with no containment theme (include only as texture, do not overclaim):** Classic + Maya peccary iconography (Copán skull, Bonampak murals); Sahagún's coyametl/pitzotl entry + (linguistic interest only); Wáay Kekén (boundary-crossing but not escape); "lipstick on a pig" + and "high on the hog" (rhetoric/class, not containment). diff --git a/research/boxes-and-escape/survey/pig-oceania.md b/research/boxes-and-escape/survey/pig-oceania.md new file mode 100644 index 0000000..c735609 --- /dev/null +++ b/research/boxes-and-escape/survey/pig-oceania.md @@ -0,0 +1,804 @@ +# Pigs, Escape, Containment and Boundary-Crossing in OCEANIA + +**Research note on method and limits.** Direct page retrieval (WebFetch / curl) was blocked +for every host by this session's egress policy (403 on `sacred-texts.com`, `ulukau.org`, +`teara.govt.nz`, `en.wikipedia.org`, `archive.org`, `jstor.org`, etc.). All findings below +therefore rest on **search-engine result content** plus my own prior knowledge, cross-checked +against each other. Where a claim is supported by a retrieved snippet naming a published +source, I say so. Where I am reporting background knowledge that I could **not** verify in +this session, I label it explicitly. Nothing here is invented; where I was unsure I have +written "unverified" rather than filling the gap. + +**Category labels used throughout:** +- **(a) ESCAPE/UNCATCHABILITY** — the pig gets out, evades, refuses capture +- **(b) BOUNDARY-CROSSING AS WEALTH** — the pig moves *across* a social/ritual boundary and + that movement is the institution's whole point +- **(c) NO CONTAINMENT THEME** — pig material included for completeness, but not on-topic + +--- + +## Summary + +Oceania is, as suspected, the richest region on Earth for this question, and it is rich in +**three structurally different ways** that should not be collapsed into each other: + +1. **Hawai'i has the single strongest mythological item anywhere: Kamapua'a.** He is not + merely a shapeshifter — his entire narrative architecture is *serial failed containment*. + He is captured and bound four separate times by armies of eight hundred and released four + times; he is bound for sacrifice on a heiau altar and escapes because the binding is a + pretence; he is cornered at a waterfall and becomes a ladder out of the trap; he escapes + Pele's fire by becoming a fish. Beckwith's summary sentence — "four times the guards, eight + hundred strong and each time increasing in number, capture him in his hog shape and tie him + to a pole; four times his grandmother releases him with a chant" — is the thesis of this + entire research question in one line. + +2. **Melanesia inverts the question.** In the New Guinea Highlands and Vanuatu the pig is not + a thing that escapes containment — it is the *instrument by which social boundaries are + crossed on purpose*. Moka, te/tee, and Vanuatu grade-taking exist so that pigs will move + between groups. A pig that stays home is a failure. And in Rappaport's Tsembaga + ethnography the ritual cycle is *triggered* by pigs becoming uncontainable — the herd grows + until the animals invade gardens, and that failure of containment is what starts the + kaiko. + +3. **Australia and New Zealand give the colonial/ecological version**, and it contains the + most thematically extraordinary documented fact in the whole file: **Cook and his officers + deliberately released pigs on Pacific islands, on purpose, as an act of policy** — putting + animals permanently beyond recapture as a form of infrastructure. Australia's ~24 million + feral pigs and New Zealand's "Captain Cookers" are the direct descendants of intentional + and accidental non-containment. + +A fourth, quieter finding worth flagging: **the Hawaiian word for a land division, +`ahupua'a`, literally means "pig altar"** — a boundary marked by a carved pig's head. In +Hawai'i the pig is not only the thing that crosses the boundary; the pig *is* the boundary +marker. + +--- + +## Kamapua'a & Polynesian Material + +### The core sources (all confirmed to exist) + +- **Martha Warren Beckwith, _Hawaiian Mythology_** (1940; University of Hawai'i Press + reprint 1970, ISBN 0824805143). Chapter XIV covers Kamapua'a. This is the standard + scholarly synthesis. **SOLID** that the book and chapter exist and treat Kamapua'a. +- **Lilikalā K. Kame'eleihiwa, _A Legendary Tradition of Kamapua'a, The Hawaiian Pig-God: + He Mo'olelo Ka'ao o Kamapua'a_**, illustrated by Dietrich Varez, **Bishop Museum Press, + 1996**. An annotated translation of a Kamapua'a epic published **anonymously in the + Hawaiian-language newspaper _Ka Leo o ka Lāhui_, 22 June – 23 July 1891**. Confirmed via + Bishop Museum Press and a Smithsonian Institution catalogue record. **SOLID.** + Kame'eleihiwa is Kanaka Maoli and a professor of Hawaiian Studies at UH Mānoa — this is the + Indigenous-authored source to lead with. +- **Abraham Fornander, _Fornander Collection of Hawaiian Antiquities and Folk-Lore_**, Vol. 5 + (Bishop Museum Memoirs) — contains a Kamapua'a text in Hawaiian with translation. + **SOLID** that it exists and is a primary-text source. +- **Thomas G. Thrum, _Hawaiian Folk Tales_**, ch. XVIII, "Kaliuwaa, Scene of the Demigod + Kamapua'a's Escape from Olopana." **SOLID** — the chapter title itself is evidence: a named + place-chapter about an escape. +- **_Hawaiian Legends of Volcanoes_ / _Legends of Volcanoes_ (W. D. Westervelt)**, ch. VIII + "Pele and Kama-puaa." **SOLID** existence. + +> **Important caution on Kame'eleihiwa's framing.** Retrieved material reports that she +> identifies strong sexual themes in the 1891 text and argues that its publication *just +> before the overthrow of the Hawaiian monarchy* was an expression of **Hawaiian rebellion +> against increasing Western dominance**. If accurate — and it is consistent with her +> published work in _Native Land and Foreign Desires_ (Bishop Museum Press) — this is +> enormously important for the essay: **the uncatchable pig was published as an anti-colonial +> act in 1891, two years before the overthrow.** The defiant animal that cannot be held down +> is a political figure, not a curiosity. **PROBABLE** — I could not read Kame'eleihiwa's +> introduction directly and this should be verified against the book before being asserted +> in print. + +### Kamapua'a episodes, by theme + +**(a) ESCAPE/UNCATCHABILITY** + +1. **The four captures and four releases.** Beckwith's account, as reported in retrieved + material: *"Four times the guards, eight hundred strong and each time increasing in number, + capture him in his hog shape and tie him to a pole; four times his grandmother releases him + with a chant."* The escalation is the point — the containment force grows each time and + fails each time. His grandmother is **Kamaunuaniho** (also Kamaunua-niho), a sorceress, and + the release is achieved **by chant**, not by force. **SOLID** as a Beckwith-attributed + summary; **PROBABLE** on the exact wording, which I could not verify against the page. + +2. **Bound for sacrifice on the heiau — and the binding is fake.** Retrieved material, + citing Beckwith: when Kamapua'a is brought bound to the heiau to be sacrificed, the old + priest/prophet **Lonoaohi** instructs his sons *to make a mere pretence of tying him*. In + the morning, when Olopana and his men come for the sacrifice, Kamapua'a springs up and + kills the chief and all the men except **Maka-li'i**. **SOLID** on the substance. This is + the single best escape image in the corpus: the ropes are theatre, the altar is the most + absolute form of containment a Hawaiian narrative can offer (a bound pig on a sacrificial + altar is the *definition* of a contained animal), and the containment is hollow from the + start because a priest is complicit. + +3. **Kaliuwa'a — the escape that named a landscape.** Kamapua'a habitually stole the chickens + of the high chief **Olopana** from Kapaka, Punalu'u and Kahana (taking all of them in a + single night). Olopana sent warriors; Kamapua'a and his followers fled up the valley of + **Kaluanui, Ko'olauloa, O'ahu**, and were cornered at a waterfall. He **transformed into a + giant hog and leaned his back against the cliff so that his people could climb his body to + the tableland above and escape.** Then, in his pig form, he dammed the water; when + Olopana's men pursued up the gorge he released it and drowned all but Olopana. **SOLID.** + - The valley is **Kaliuwa'a** — glossed as *"the leak of the canoe" / "the leaky canoe"*, + with the tradition that the vertical groove in the cliff face is where his canoe (or, in + other tellings, his back) rested. The site is today **Sacred Falls State Park**, closed + to the public since a fatal 1999 rockfall. **PROBABLE** on the etymology (multiple + glosses circulate); **SOLID** that the place is named for this escape. + - *Essay-relevant point:* a real, mappable, legally-designated place on O'ahu is named + after a pig getting away. The escape is inscribed in the land. + +4. **Escape from Pele by becoming a fish.** Pursued by Pele's fire, Kamapua'a escapes into + the sea and takes the form of the **humuhumu-nukunuku-ā-pua'a** — literally "triggerfish + with a snout like a pig." **SOLID** as a widely-attested tradition, attributed to Beckwith + in retrieved material. Note the beautiful structural detail: **the escape route is a change + of medium** (land → sea) and the animal he becomes is *named after the animal he was*. Even + his disguise carries his name. + +5. **Bristles as the tell, hidden under a cape.** Retrieved material: "the bristles down his + back, which reveal his hog nature when in human form, he hid with a cape." **PROBABLE.** + This is the containment of *identity* rather than body — he passes as human by covering the + evidence. + +6. **Fornander: "the eight-eyed," gifted with eight feet.** Retrieved material attributes to + Fornander that Kamapua'a was sometimes called **the eight-eyed** and had **eight feet**. + **PROBABLE.** A body specified for seeing every direction and running faster than pursuit — + an anatomy of uncatchability. + +7. **Kino lau (many bodies).** Kamapua'a's body-forms include: the pig; the + **humuhumunukunukuāpua'a** (reef triggerfish); **kukui** (candlenut tree); **'uala** (sweet + potato); **'ama'u** tree fern; and, per one name chant, **clouds** — the dark rain-swollen + cloud-banks called **`ao-pua'a`**, "pig clouds." **SOLID** on the list of kino lau (multiply + attested, including Hawaiian-medium educational sources such as Kumukahi). The cloud form + is the strongest single image for this essay: **he can become the one thing in the world + that no fence can hold, and the Hawaiian language already named that cloud after him.** + - Note the elegance of the 'ama'u fern: the fern is *what pigs eat*, and it is also *one of + his bodies*. Predator and food are the same being. He crosses the animal/plant boundary + and the eater/eaten boundary at once. + +**(b) BOUNDARY-CROSSING / TERRITORIAL BOUNDARY** + +8. **Pele and Kamapua'a divide Hawai'i Island by treaty.** After a destructive contest — she + attacks with lava and flame, he counters with fog, rain and dampness, and *sends an army of + hogs to root and rampage through her territory* — a truce is brokered and **the island is + divided between them**: Kamapua'a takes the **wet windward east (Hilo and the rainy + districts)**, Pele the **dry, lava-covered districts (Kona, Ka'ū, Puna)**. **SOLID** on the + division; the exact district assignment varies between tellings, and one retrieved source + assigns Puna to Pele while another treats the east generally as his — treat district lists + as **PROBABLE**. + - This is a *boundary treaty between two gods*, and Hawaiians use it to explain the real + rainfall gradient of the island. The pig-god is on the wet side because pigs need water, + wallowing, taro, forest. He is the god of the side of the island where things grow. + - One retrieved source adds that "he promised never to flood her side of the **Wailua + River** and she promised never to cross the river," and that the oath has never been + broken. **THIN** — Wailua is a Kaua'i river name and this reads like a conflation or a + locality-specific variant. Do not use without checking a primary text. + +9. **`Ahupua'a` = "pig altar."** The Hawaiian land division running mauka-to-makai is called + an **ahupua'a**, from **`ahu`** (stone altar/heap) + **`pua'a`** (pig), because the boundary + of the division, where it met the trail circling the island, was marked by a **heap of + stones surmounted by a carved wooden image of a pig's head**, on which tribute was laid. + **SOLID** — this etymology is given consistently across Hawaiian institutional sources + (Pukui & Elbert's dictionary via wehewehe.org, UH/Hilo materials, NPS cultural histories). + - This belongs in the essay as a **category (b)** item of a special kind: it is not a pig + crossing a boundary, it is **a pig marking one**. The animal most associated with rooting + through, escaping and refusing to stay put is the animal chosen to stand, in effigy, at + the exact line you must not cross without paying. That is a genuinely striking inversion + and I have not seen it made anywhere. **Confidence in the etymology: SOLID. Confidence in + the interpretation: mine, offered as reading not as fact.** + +10. **Pig sacrifice sayings, Pukui.** Mary Kawena Pukui, **_'Ōlelo No'eau: Hawaiian Proverbs + and Poetical Sayings_** (Bishop Museum Press, 1983) contains pig material. Two retrieved + examples: *"Iā 'oe ke po'o pua'a a kākou"* ("You are in charge of our offering of pig") + and ***"Moe ka ihu o ka pua'a"*** ("The snout of the pig has been laid down" — i.e. the + entire pig sacrifice is offered). **PROBABLE** on these being genuine 'Ōlelo No'eau entries + (they were retrieved as quoted Hawaiian sayings, but I could not confirm entry numbers). + - *"Moe ka ihu o ka pua'a"* is thematically perfect and its opposite: the snout — the + rooting, boundary-breaking organ — **laid down**. The image of total submission is a + pig's nose stopping. + - Also retrieved: Pukui noted that cooked lū'au (taro leaves) was sometimes substituted + for pua'a in offerings, and such offerings were called **`pua'a hulu 'ole`, "hairless + pig."** **PROBABLE.** A vegetable standing in for the pig — the pig crossing into the + plant kingdom again, this time by ritual accounting rather than by magic. + +**(c) OTHER POLYNESIAN — mostly NO CONTAINMENT THEME** + +11. **Māui.** Extensively searched; **I found no reliable pig episode in the Māui cycle.** + Māui's animal transformations in the well-attested traditions are birds (the pigeon/kererū + in Māori tradition) and, in the Hine-nui-te-pō episode, a worm/lizard-form entry. + **Recommendation: do not include a Māui-and-pig episode.** Several low-quality sites + gesture at Polynesian trickster/pig links without a source; that is exactly the kind of + material the brief warns about. **THIN → OMIT.** + +12. **Lono.** Kamapua'a is associated with **Lono**, god of agriculture, rain and the + Makahiki. **SOLID** as a general association (multiply attested). The connection matters: + the pig-god belongs to the season of rain, growth and the *suspension of war and normal + order* — a boundary-time. I could **not** verify in this session any specific Makahiki + ritual in which a pig escapes or is released; **do not assert one.** + +13. **The word itself.** Reconstructed Proto-Polynesian *puaka > **puaka** (Rarotongan, + Mangarevan, Rotuman), **vuaka** (Fijian), **pua'a** (Samoan, Hawaiian), **puaa** + (Tahitian, Marquesan), **puaka/buaka** (Tongan), **poaka** (Māori). **SOLID.** One + retrieved source claims the word "seems to have had its own unique Polynesian origin" + while also listing Malay *puwaka/babi* cognate-candidates — the Austronesian etymology is + **contested**; treat any origin claim as **THIN** and just report the cognate set, which is + itself striking: the pig travelled with the language across a third of the planet. + +14. **"Puaka" as a pig-demon.** *A Book of Creatures* has an entry for **Puaka**, but it is a + **Dusun (Borneo) water-guarding demon**, not Oceanian. **Do not import it into a Pacific + section.** Flagged here only so the essay doesn't make that mistake. **(c) / OMIT.** + +--- + +## Melanesian Pig Exchange & Boundary-Crossing + +This is the world's densest body of material on **(b) pigs crossing social boundaries as +wealth**, and it contains one first-rate **(a)** item as well (Rappaport). + +### (b) The exchange systems + +15. **Moka — Melpa people, Mount Hagen, PNG Western Highlands.** + **Andrew Strathern, _The Rope of Moka: Big-men and Ceremonial Exchange in Mount Hagen, New + Guinea_**, Cambridge University Press, 1971 (Cambridge Studies in Social and Cultural + Anthropology, no. 4). **SOLID** — publication details confirmed. + - Substance: moka is a **competitive ceremonial exchange of pigs, shells and other + valuables**; exchanges act as **a bond between groups** and as the means by which + big-men maximise status. In the Hagen area, moka **expanded vigorously after European + contact as warfare was suppressed** — i.e. the pigs took over the work that fighting + used to do between groups. **SOLID.** + - *The title is the essay's gift:* **"the rope of moka."** The Melpa image for the + exchange chain is a **rope** — a thing that ties, but here a rope that runs *between* + groups rather than around an animal. Pigs are what travel along it. A rope in this + system is not a restraint; it is a route. + +16. **Te / Tee — Enga people, PNG Enga Province.** + - **M. J. Meggitt, "'Pigs Are Our Hearts!' The Te Exchange Cycle among the Mae Enga of New + Guinea," _Oceania_.** **SOLID** existence; the title alone is a quotable Indigenous + statement of what pigs are. + - **D. K. Feil, _Ways of Exchange: The Enga Tee of Papua New Guinea_** (University of + Queensland Press) and **Feil, "Women and men in the Enga tee," _American Ethnologist_ + 5(2), 1978** (DOI 10.1525/ae.1978.5.2.02a00050). **SOLID.** + - Substance: the tee is a **chain** exchange in which **pigs are the most valued item**, + running along linked partnerships across many groups. Feil's contribution: **women are + essential participants — allotting pigs, making key political decisions, and shaping the + partnerships of the men whom they link.** **SOLID.** + - *Point for the essay:* the tee is a **chain**, the moka a **rope**. Both metaphors are + restraint-words repurposed as connection-words. The pig moving down the chain is the + chain working. + +17. **Vanuatu tusked boars and grade-taking (nimangki).** **SOLID** on all of the following: + - **Method:** the **upper canines are knocked out / avulsed** in juveniles so the lower + canines meet no opposing surface and grow in a **circle**. A **full circle takes about + 6–7 years; a double circle 10–12 years** (per the PNAS paper below); a Malekula source + recognises **23 distinct named stages** (the Seniang), and some accounts stretch the + process to decades. + - **As the tusk circularises it punctures the animal's own cheek and sometimes the jaw**, + requiring "delicate care" including removal of lower back teeth to give the tusk + somewhere to re-enter. Tusker boars were often **castrated** to reduce aggression and + the risk of broken tusks. + - **Function:** in the graded societies of the **Torres Islands, Ambae, Maewo, Malekula, + Vao, Ambrym, north and central Pentecost**, pigs of specified tusk-curvature are a + **requirement to enter and rise through ranks**. Pigs and **red mats** are also + exchanged at marriage and as payment of fines and for services. + - **National symbol:** the **flag of Vanuatu** carries a **boar's tusk encircling two + crossed namele fern fronds**. **SOLID.** A nation whose emblem is a circle made out of a + pig. + - **Earliest European record** of circular tusks on Malakula is in **Cook's journals from + the 1774 visit**. **PROBABLE** (retrieved, attributed to Bedford et al.). + - **Source:** *Stuart Bedford et al., "Modified canines: Circular pig's tusks in Vanuatu + and the wider Pacific," pp. 125–141* (in an edited volume; available via ResearchGate / + Academia.edu). **SOLID** existence. + - **Genetics:** *"Recent Southeast Asian domestication and Lapita dispersal of sacred male + pseudohermaphroditic 'tuskers' and hairless pigs of Vanuatu,"* **PNAS** + (doi:10.1073/pnas.0608220103), PMC1859908. **SOLID.** + - *The reading this supports:* the tusk is **the deliberate manufacture of a shape that + closes on itself.** Ni-Vanuatu pig-raising takes the boar's most escape-associated + weapon and spends a decade turning it into **a closed circle that grows back into the + animal's own head**. It is containment as artwork — the opposite pole from Kamapua'a. + The essay can set these two against each other directly. + +18. **Interspecies nursing — the human/animal boundary.** Anthropologists have documented + women **breastfeeding piglets** across Melanesia (and parts of Polynesia), typically when + a sow died or could not nurse, to preserve an asset central to wealth and exchange. + **PROBABLE→SOLID** (well attested in the ethnographic literature; retrieved sources + include *Katharina Schneider, "Pigs, Fish, and Birds: Toward Multispecies Ethnography in + Melanesia," Environment and Society*). One retrieved formulation is worth keeping: the + practice creates a "**mutual porousness between pig and person**" that "**breaches material + and symbolic distinction between humans and animals, and overturns taxonomic categories of + who can live with and feed from people**." **This is (b) boundary-crossing at the deepest + level available: the pig crosses into the human family by being fed from a human body.** + Handle with care and attribute; do not sensationalise. + - Related, **SOLID**: in the central highlands and some lowland areas, **piglets are + carried by women in string bags (bilum) or walked on a tether to the gardens** until + mature. + +19. **Deliberate hybridisation with wild boars.** **SOLID** (retrieved, consistent with the + literature reviewed in *Robin Hide, "Pig Husbandry in New Guinea: A Literature Review and + Bibliography," ACIAR Monograph 108* — a genuinely comprehensive source, PDF at aciar.gov.au): + in the central highlands and some lowland areas, **domestic boars are castrated and sows + are mated with wild boars from the forest**; wild piglets are also captured and raised. + - *This is category (a) and (b) fused.* The husbandry system **requires** that the + domestic herd be periodically opened to the uncontained forest population. The line + between the village pig and the wild pig is not a wall — it is a deliberately maintained + valve. Excellent material. + +### (a) Rappaport: containment failure as the engine of ritual + +20. **Roy A. Rappaport, _Pigs for the Ancestors: Ritual in the Ecology of a New Guinea + People_**, Yale University Press, 1968 (enlarged edition 1984). On the **Tsembaga Maring** + of the Simbai Valley. **SOLID** — the most-cited work in ecological anthropology. + - **Mechanism (SOLID in outline, PROBABLE on numbers):** the pig herd grows by natural + reproduction over a cycle of roughly **5–25 years**. As it grows, **the pigs begin + invading gardens**, and the labour demanded of **women** to feed them rises past what is + tolerable, and disputes multiply. When the burden becomes insupportable, the **kaiko** — + a year-long festival culminating in mass pig slaughter — is held; retrieved figures cite + on the order of **~105 pigs** slaughtered and distributed to allies, after which warfare + may resume. The cycle then restarts. + - **Why this is a top-five item for the essay:** *the ritual calendar of an entire society + is triggered by pigs ceasing to stay where they are put.* The kaiko does not happen on a + date. It happens **when the fences stop working**. There is no better statement anywhere + of pigs-as-uncontainable-force being converted into social structure. + - **Caveat to state in the essay:** Rappaport's functionalist "self-regulating system" + argument has been heavily criticised since the 1970s (the critique is standard in + ecological anthropology). The *ethnographic observation* — pigs invade gardens, women + bear the cost, this precipitates the slaughter — is much more secure than the + cybernetic-homeostasis framework built on it. Use the observation, hedge the theory. + +21. **Kiwai: Sido becomes a giant pig and becomes the house of the dead.** Retrieved material + reports a Kiwai (Fly River delta, PNG) myth in which, after losing his human wife, **Sido + transformed himself into a gigantic pig and split himself open so that the pig's backbone + and sides formed the house of death**. Sido is in Kiwai tradition **the first man to die**, + who opened the road to **Adiri**, the land of the dead. Source would be **Gunnar Landtman, + _The Folk-Tales of the Kiwai Papuans_ (1917)** and/or **_The Kiwai Papuans of British New + Guinea_ (1927)** — **SOLID** that these books exist and are the standard collection. + **PROBABLE** on the pig detail: I could not open Landtman to confirm it, and I would not + print it without doing so. + - If it checks out it is a superb **(b)** item — arguably the ultimate boundary-crossing: + a pig's body becomes the **architecture of the passage between the living and the + dead**. The pig is not what crosses; the pig is the doorway. + +22. **Tok Pisin.** I could **not** find a documented Tok Pisin proverb built on *pik*. + Confirmed lexical items only: **`pik`** = pig; **`het bilong pik`** (pig's head) used + idiomatically; **`hap lek pik`** (a leg/side of pork) as a unit in traditional exchange. + **THIN.** Do not manufacture a proverb. If the essay wants Tok Pisin, the honest and + interesting fact is the **exchange vocabulary**: the language has a standard term for *a + portion of pig that moves between people*. + +--- + +## Māori & New Zealand + +23. **Arrival.** Pigs (**poaka**) were **not** present in pre-contact Aotearoa — before + European contact the only mammalian meat sources were **kurī** (dog) and **kiore** + (Pacific rat). **SOLID.** First recorded gift of pigs to Māori: **Jean François Marie de + Surville, 1769, Doubtless Bay, Northland — two pigs.** **PROBABLE→SOLID** (retrieved from + Te Ara / NZ Geographer material; the de Surville date is standard). + +24. **★ The Waima pigs and the tapu kūmara plantation.** This is, for my money, **the finest + single containment story in the New Zealand material**, and it is documented. From **_Te Ao + Hou / The Maori Magazine_** (published by the Department of Māori Affairs, 1952–1976; + digitised at teaohou.natlib.govt.nz and Papers Past), retrieved close to verbatim: + + > "When the pigs arrived at Waima the people thought that they were gods, and allowed them + > to wander wherever they liked. The kumara crop was growing, and as it had not yet been + > taken up, the plantations were very tapu. The pigs went into the plantations, and because + > of the tapu no one could go and take them away. All that the people could hear was the + > pigs' grunting, and this made them more certain than ever that the animals were gods. But + > when the time came to take up the kumaras, it was found that the pigs had rooted up and + > eaten a good part of the crop." + + **SOLID** on the text as retrieved; **PROBABLE** on the precise issue/page — cite as *Te + Ao Hou*, issue to be confirmed (the retrieved index pointed at issue 51). Waima is in the + Hokianga. + - **Why it is extraordinary:** this is a containment failure produced by *ritual law + rather than by bad fencing*. The pigs are inside the boundary and **the boundary is what + prevents anyone from removing them.** The tapu that protects the crop protects the animal + eating the crop. And the grunting from inside the sacred ground is read as **evidence of + divinity** — the sound of an animal that cannot be evicted is taken for the voice of a + god. It is funny, it is tragic, it is theologically precise, and it is a real recorded + account rather than a "legend." + +25. **Pigs as currency.** From **1803** Māori are recorded trading potatoes, pigs and maize; + **pigs and baskets of potatoes became a standard form of currency.** Retrieved figures: + between **1814 and 1827** the price of a musket ranged from *150 baskets of potatoes and + 8 pigs*, to *200 baskets or 15 pigs*, settling at *120 baskets of potatoes or 10 pigs*. + **PROBABLE→SOLID** (Te Ara, "Kai Pākehā – introduced foods" / "Pigs and the pork + industry"). **Category (b):** an introduced animal became the medium by which value crossed + between Māori and Pākehā — including into muskets, with all that followed. + +26. **★ Cook's liberations in Queen Charlotte Sound, 1773.** **SOLID** in outline: + - **Furneaux** (Cook's consort captain, HMS *Adventure*) released **one boar and two sows + at Cannibal Cove**, Queen Charlotte Sound; these were **caught and eaten by Māori.** + - **Cook** made further liberations the same year using pigs **obtained at Tongatapu and + at Huahine/Ra'iatea (Society Islands)** — i.e. **Polynesian pigs of Indo-Malay + ancestry, not English pigs**. Retrieved account: he gave two pairs to a Māori group + south of **Cape Kidnappers**, another sow to Māori at Queen Charlotte Sound, and + **released three sows and a boar in West Bay.** + - Cook is recorded as writing in his journals of the hope that they would **multiply**. + - **PROBABLE** on the precise tally and place-names — the numbers vary between secondary + sources and I could not open Cook's journals or Beaglehole. Verify before printing + specific counts. **SOLID** on the essential fact of multiple deliberate liberations in + 1773. + - Note the doubling of the theme: **Māori promptly recaptured and ate the first release.** + The very first attempt to establish free-living pigs in New Zealand was defeated by + people who saw a pig, correctly, as food rather than as infrastructure. + +27. **"Captain Cookers."** Feral pigs in New Zealand — especially in the South Island — are + known as **Captain Cookers**, and also as **razorbacks**, **te poaka**, and **kunekune**. + **SOLID** (Department of Conservation hunting literature, Te Ara). The genetic claim that + South Island ferals descend specifically from Cook's animals is **PROBABLE at best** and + the DOC/scientific literature is more cautious than the folk name; see *"Feral pigs in the + northern South Island, New Zealand," Journal of the Royal Society of New Zealand* (1991), + doi 10.1080/03036758.1991.10418181. **SOLID** that this paper exists. + - *The name is the essay's payload.* **New Zealanders named their uncatchable animal after + the man who let it go.** The escape is memorialised in the escapee's surname. There is no + better single word in this entire research file. + +28. **Contemporary Māori pig hunting.** **Claire Kuuii Adeline Dowsett** (Victoria University + of Wellington), with **Anna Carr** and **Brent Lovelock** (University of Otago), *"Hunting + and Hauora: Pig Hunters and Poaka in Aotearoa New Zealand,"* **_New Zealand Geographer_, + 2026, vol. 82 no. 2**, doi 10.1111/nzg.70023. **SOLID.** + - Findings as retrieved: though invasive, **poaka were fundamental to the survival of both + Māori and Pākehā during colonisation and remain an essential source of kai today**; a + Whanganui case study, 24 participants, semi-structured interviews, thematic analysis + **guided by Kaupapa Māori principles**; hunting is found to enhance hauora (wellbeing) — + food and economic security, cultural aspiration, social, spiritual, environmental and + recreational value, and **deepened connection to place**. + - *Category:* this is the mature form of the theme. The animal that escaped colonial + control was absorbed into Indigenous life and is now **defended** as taonga-adjacent + against eradication logic. **The uncontainable animal became a reason to be on the + land.** Excellent, current, Indigenous-methodology source — use it. + - Adjacent and worth a footnote: the same 2026 volume of *New Zealand Geographer* carries + **Edwards, "Fenced Out… and Fenced In? Containing Cats in the Borderlands of Zealandia"** + (doi 10.1111/nzg.70022) — different animal, but it shows New Zealand geography as a + discipline currently thinking hard in exactly these terms. **SOLID** existence. + +29. **Kunekune.** A small, hairy, docile NZ pig breed with a Māori name; associated with Māori + husbandry and nearly lost in the twentieth century before being recovered by breeders. + **THIN as stated** — I ran out of search budget before verifying the near-extinction and + rescue narrative, which is widely repeated by breed associations (a source type prone to + romance). Verify before use. The interesting angle if it holds: **the kunekune is the pig + that stayed** — bred for docility and enclosure-tolerance — the exact counter-example to + the Captain Cooker in the same country. + +--- + +## Australia & Feral Pigs + +There are **no native pigs in Australia**, so as expected the material is colonial and +ecological. It is nevertheless very strong for this question because Australia produced the +largest **failure of containment** in the region's history. + +30. **Arrival and immediate escape.** **49 hogs arrived with the First Fleet in 1788** as + livestock for the colony at Sydney Cove. **PROBABLE→SOLID** on the number (retrieved + consistently). Retrieved and consistent across Australian government and NGO sources: + the animals were **not securely housed and escaped**; "**ineffectual fencing and + containment** of these domestic pigs allowed many to escape into the environment and these + formed the basis of the current feral pig populations in Australia"; and through the + colonial period **"escapes and deliberate releases were commonplace,"** with free-range + husbandry preferred because it was easier than building pens. **SOLID** on the substance. + - *Note the phrase "deliberate releases" recurring on the Australian side too.* Much of + the feral population is not accident. It is a decision not to build a fence. + +31. **Scale today.** Feral pigs (*Sus scrofa*) occupy roughly **45% of the Australian + continent**, with estimates commonly cited up to **~24 million animals**, concentrated in + Queensland, New South Wales and the Northern Territory; populations can increase by + **~86% in a year** without control. **PROBABLE** — population estimates for feral pigs are + notoriously uncertain and range widely (3.5m–24m in different sources); cite as an + estimate with a range, not as a figure. Sources: PestSmart (pestsmart.org.au), the + **National Feral Pig Action Plan** (feralpigs.com.au), NSW DPI, Business Queensland. + **SOLID** that these are the authoritative Australian bodies. + +32. **The language of containment and eradication.** The Australian institutional vocabulary + is itself an essay resource — retrieved usages include **"ecological bulldozers of the + bush"** (FERAL, feral.org.au), *restricted invasive animal*, *exclusion fencing*, + *containment*, *control*, *eradication*, *biosecurity*. The rhetorical structure is + military-agricultural: a war of lines against an animal that does not recognise lines. + **SOLID** as observed usage. + +33. **Aboriginal rangers: pest or resource?** **SOLID** and important — this is the Australian + counterpart to the Dowsett paper: + - *"Aboriginal Rangers' Perspectives on Feral Pigs: Are they a Pest or a Resource? A Case + Study in the Wet Tropics World Heritage Area of Northern Queensland"* — confirmed to + exist (ResearchGate). **The title states the ambivalence exactly.** + - **Cape York:** adaptive feral pig management systems **developed, trialled and used by + Indigenous Rangers** on western Cape York to reduce **predation of sea turtle nests**, + now extending to the eastern Cape. Sources: DCCEEW *Wetlands Australia* 36 (July 2022); + CSIRO, "Managing feral pigs on Cape York: it's not a numbers game" (2015); NESP Resilient + Landscapes Hub. **SOLID.** + - Retrieved quotation from Indigenous ranger **Jennifer Creek** on wetland damage: *"Right + now the water's all dirty and the bank's all dug up and there's less lilies now."* + **PROBABLE** (retrieved as a quotation in DCCEEW material; verify attribution before + quoting). + - **Kakadu:** **Bininj** (north) and **Mungguy** (south) work with park staff combining + traditional knowledge with contemporary management. **SOLID** as a general statement of + Kakadu's joint-management model; I did **not** verify a specifically pig-focused Kakadu + program. + - The CSIRO framing — **"it's not a numbers game"** — is a good line: it is the moment + management gives up on eradication-by-count and switches to protecting particular + places. An admission that the animal cannot be contained, only negotiated with. + +34. **Hogzilla is American — do not use it for Australia.** The giant-hog legend **Hogzilla** + was a hog shot by **Chris Griffin in Alapaha, Georgia, USA, on 17 June 2004**, claimed at + 12 ft and over 1,000 lb; a *National Geographic* investigation found ~800 lb and 7.5–8 ft, + and it was widely treated as hoax/urban legend before that. **SOLID** — and **SOLID that it + is not Australian**. Australia does have **heaviest-boar hunting competitions** + (**PROBABLE**), which is the real local equivalent, but the essay must not transplant + Hogzilla to Queensland. Flagged because this is exactly the kind of drift the brief warns + against. + +35. **_Babe_ (1995).** Filmed at **Robertson, New South Wales**, Australia (shot 1994, + released 4 August 1995); adapted from **Dick King-Smith's _The Sheep-Pig_ (1983)**, US + title *Babe: The Gallant Pig*. **SOLID.** + - Category: **(b), and unusually pure.** *Babe* is not an escape story — it is a **refusal + to stay in an assigned category**. A pig declines to be livestock and becomes a + sheepdog. The containment he breaks is taxonomic, not physical. Note the resonance with + Kamapua'a: both are about an animal that will not remain the kind of thing it is + supposed to be. And it is, on film, an Australian pig — the sequel is even titled *Babe: + Pig in the City*, a pig crossing the country/city boundary. + - **Emily Rodda's _Pigs Might Fly_ (1986, Australian, Children's Book of the Year)** — I + could **not** verify its plot in this session. **THIN. Verify or omit.** + +36. **Australian/NZ English idiom.** + - **"Pig-root" / "pigrooting"** — **Australian and New Zealand English** for a horse + kicking up with its hind legs while keeping its head down and forelegs planted; named + from the posture of a pig rooting in the ground. Attested at least from the 1930s; + retrieved 1937 citation: *"Eventually we saddled the team; then Slippery, with a snort, + started pig-rooting and bolted amongst the trees."* In Collins and the Free Dictionary. + **SOLID.** + - *Perfect for this essay:* an Australianism in which **"pig" names the specific bodily + motion of an animal refusing to be ridden.** The pig lends its name to the act of + throwing off control. And note the retrieved 1937 sentence ends **"and bolted"** — the + pig-word and the escape in the same clause. + - **"In a pig's eye"** (also *in a pig's arse*) — derisive retort of emphatic disbelief; + **chiefly North American and Australian**; first recorded **1847**. **SOLID.** + Category **(c)** — no containment theme, include only as texture. + +--- + +## Proverbs & Idioms + +Ranked by how well-sourced they are. **I did not find a large stock of genuine pig proverbs** +in this region, and I have deliberately not padded this section. + +| Item | Language | Gloss | Category | Confidence | +|---|---|---|---|---| +| *Moe ka ihu o ka pua'a* | Hawaiian | "The snout of the pig has been laid down" — the whole pig is offered | (b) | PROBABLE (Pukui, *'Ōlelo No'eau*) | +| *Iā 'oe ke po'o pua'a a kākou* | Hawaiian | "You are in charge of our offering of pig" | (c) | PROBABLE | +| *pua'a hulu 'ole* | Hawaiian | "hairless pig" — cooked taro leaves substituted for pig in offerings | (b) | PROBABLE | +| *ahupua'a* | Hawaiian | "pig altar" — the land division named for the pig's head on its boundary marker | (b) | SOLID | +| *humuhumu-nukunuku-ā-pua'a* | Hawaiian | "triggerfish with a snout like a pig" — Kamapua'a's escape-form | (a) | SOLID | +| *'ao-pua'a* | Hawaiian | "pig clouds" — rain-heavy cloud banks, a kino lau of Kamapua'a | (a) | PROBABLE | +| **Captain Cooker** | NZ English | feral pig, named for the man who released the founders | (a) | SOLID | +| **pig-root / pigrooting** | Aus/NZ English | a horse throwing its rider off with a pig's rooting motion | (a) | SOLID | +| *in a pig's eye / arse* | Aus English | emphatic disbelief; from 1847 | (c) | SOLID | +| *het bilong pik* | Tok Pisin | "pig's head", idiomatic | (c) | THIN | +| *hap lek pik* | Tok Pisin | "a leg of pork" as an exchange unit | (b) | PROBABLE | +| *Ka rūrū noa iho te poaka i tōna pane* | Māori | "The pig just shook its head" | (c) | THIN — retrieved as a dictionary example sentence, **not** a whakataukī. Do not present as a proverb. | +| **poaka / pua'a / puaka / vuaka / buaka** | Pan-Polynesian | the same word for pig across a third of the planet | (b) | SOLID | + +**Negative findings worth stating in the essay:** I could not find a documented **Tok Pisin +proverb** or a documented **Māori whakataukī** centred on the pig. That absence is itself +meaningful and honest — in Aotearoa the pig arrived too late (post-1769) to enter the deep +proverbial stock, which is exactly why the NZ material is *historical* rather than +*proverbial*. Say so rather than inventing filler. + +--- + +## Cook's Releases & Documented History + +This section carries the strongest **documented, non-mythological** material in the file. + +37. **Cook's policy of liberation.** Retrieved and consistent across sources (including + *Hakai Magazine / bioGraphic*, "Islands of the Feral Pigs"): **on his second and third + voyages Cook purposefully left pairs of pigs on various islands**, reasoning that a + British ship later wrecked or short of provisions would find **a self-sustaining protein + source waiting**. **SOLID** on the practice and the rationale; **PROBABLE** on + island-by-island specifics. + - *This is thematically extraordinary and should probably anchor the essay's non-mythic + half.* It is a deliberate, documented, imperial decision **to put animals permanently + beyond recapture** — releasing livestock as a way of storing value in a place you do not + control and may never return to. **Escape reconceived as infrastructure. A pantry made + of animals that will not stay put — and the not-staying-put is the feature, not the bug.** + Cook did not lose these pigs. He *invested* them. + - The moral arithmetic then inverts across two centuries: the same act that was rational + provisioning in 1773 is the origin event of an ecological catastrophe the Australian and + New Zealand states now spend tens of millions of dollars a year trying to reverse. + Compare **US$30.5 million** to reconstruct **~64 miles** of perimeter exclusion fencing + at **Hawai'i Volcanoes National Park** (Great American Outdoors Act funding, NPS — + **SOLID**). Two and a half centuries of fencing to undo a few afternoons of releasing. + +38. **New Zealand, 1773.** See item 26 above. Furneaux at Cannibal Cove (recaptured and + eaten); Cook's later liberations at Queen Charlotte Sound, West Bay, and near Cape + Kidnappers, using **Tongan and Society Islands pigs**. **PROBABLE** on detail, **SOLID** + on the general fact. + +39. **Hawai'i, 1778 — and an important correction.** Retrieved: on **1 February 1778** Cook + landed goats, pigs and seeds at **Ni'ihau**. **PROBABLE.** But the important scientific + point is a **correction to the popular story**: genetic work shows **Hawaiian feral pigs + are largely descended from the pigs Polynesians brought roughly 800 years ago**, not from + Cook's. Source: *"A novel MC1R allele for black coat colour reveals the Polynesian + ancestry and hybridization patterns of Hawaiian feral pigs,"* **Royal Society Open + Science** (doi 10.1098/rsos.160304, PMC5043315). **SOLID** that this paper exists and makes + a Polynesian-ancestry argument. + - *Do not let the essay say "Cook's pigs overran Hawai'i."* The pua'a that Kamapua'a + embodies came with the voyaging canoes, centuries before Cook. This matters for + accuracy and for respect: the Hawaiian pig is a Hawaiian arrival, not a European one. + +40. **Vanuatu, 1774.** Cook's journals contain the **earliest European record of circular + tusks on Malakula**. **PROBABLE** (attributed to Bedford et al.). Note the symmetry, which + is almost too neat to be true but appears to be: **in the same decade, the same man is + recorded both releasing pigs to run wild in one archipelago and describing an archipelago + where pigs' teeth are cultivated for decades into perfect closed circles.** + +41. **Hawai'i Volcanoes National Park — the containment counter-history.** **SOLID** in + outline: pig removal ran from **1930 to 1971**, taking roughly **7,000 pigs**, with limited + lasting effect. In the **1970s** the strategy changed to **boundary and internal fencing to + isolate populations**, combined with shooting by staff and volunteers. Result: goats + essentially eliminated below 9,000 ft, and pigs removed from about **40,000 fenced acres**. + A **Fern Jungle exclosure** documented **13 years without feral pigs** in rain forest. + Sources: NPS; *James K. Baker, "The Feral Pig in Hawaii Volcanoes National Park"* (1975); + HCSU Technical Report HCSU-004 on Hakalau; *"Ecological impacts of feral pigs in the + Hawaiian Islands," Biodiversity and Conservation* (doi 10.1007/s10531-009-9680-9). + **SOLID** existence for all. + - **The key structural lesson, and a gift to the essay:** *four decades of hunting failed; + what worked was fences.* The conclusion of a century of Hawaiian pig management is that + **you cannot catch them — you can only wall off the places you want them not to be.** Not + eradication. Perimeter. Which is to say: Hawai'i spent the twentieth century arriving, + empirically and expensively, at the thesis Hawaiians had already stated mythologically — + the pig cannot be caught; it can only be bargained with over territory. **Compare item 8: + Pele and Kamapua'a settle their war not by one defeating the other but by drawing a line + across the island.** The National Park Service, after forty years, drew the same line. + +--- + +## Sources + +**Hawaiian / Polynesian** +- Beckwith, Martha Warren. *Hawaiian Mythology*. 1940; University of Hawai'i Press, 1970. + Ch. XIV (Kamapua'a). +- Kame'eleihiwa, Lilikalā K. *A Legendary Tradition of Kamapua'a, The Hawaiian Pig-God: He + Mo'olelo Ka'ao o Kamapua'a*. Illus. Dietrich Varez. Bishop Museum Press, 1996. Annotated + translation from *Ka Leo o ka Lāhui*, 22 June – 23 July 1891. **Lead Indigenous source.** +- Kame'eleihiwa, Lilikalā K. *Native Land and Foreign Desires: Pehea Lā E Pono Ai?* Bishop + Museum Press. +- Fornander, Abraham. *Fornander Collection of Hawaiian Antiquities and Folk-Lore*, Vol. 5. + Bishop Museum Memoirs. +- Thrum, Thomas G. *Hawaiian Folk Tales*, ch. XVIII: "Kaliuwaa, Scene of the Demigod + Kamapua'a's Escape from Olopana." +- Westervelt, W. D. *Hawaiian Legends of Volcanoes*, ch. VIII: "Pele and Kama-puaa." +- Pukui, Mary Kawena. *'Ōlelo No'eau: Hawaiian Proverbs and Poetical Sayings*. Bishop Museum + Press, 1983. +- Pukui, Mary Kawena & Samuel H. Elbert. *Hawaiian Dictionary* (via wehewehe.org) — for + *ahupua'a*, *pua'a*. +- Kumukahi (kumukahi.org), "Kupua" unit — Hawaiian-medium educational source on kino lau. + +**Melanesian** +- Strathern, Andrew. *The Rope of Moka: Big-men and Ceremonial Exchange in Mount Hagen, New + Guinea*. Cambridge University Press, 1971. +- Meggitt, M. J. "'Pigs Are Our Hearts!' The Te Exchange Cycle among the Mae Enga of New + Guinea." *Oceania*. +- Feil, D. K. *Ways of Exchange: The Enga Tee of Papua New Guinea*. University of Queensland + Press. And: "Women and men in the Enga tee." *American Ethnologist* 5(2), 1978. + doi:10.1525/ae.1978.5.2.02a00050 +- Rappaport, Roy A. *Pigs for the Ancestors: Ritual in the Ecology of a New Guinea People*. + Yale University Press, 1968; enlarged ed. 1984. +- Bedford, Stuart, et al. "Modified canines: Circular pig's tusks in Vanuatu and the wider + Pacific," pp. 125–141. +- Lum, J. K., et al. "Recent Southeast Asian domestication and Lapita dispersal of sacred male + pseudohermaphroditic 'tuskers' and hairless pigs of Vanuatu." *PNAS*. + doi:10.1073/pnas.0608220103 (PMC1859908) +- Hide, Robin. *Pig Husbandry in New Guinea: A Literature Review and Bibliography*. ACIAR + Monograph 108. — **the single most comprehensive practical source in this file.** +- Schneider, Katharina. "Pigs, Fish, and Birds: Toward Multispecies Ethnography in Melanesia." + *Environment and Society*. +- Landtman, Gunnar. *The Folk-Tales of the Kiwai Papuans* (1917); *The Kiwai Papuans of British + New Guinea* (1927). — **for the Sido/pig myth; must be checked directly.** + +**New Zealand** +- *Te Ao Hou / The Maori Magazine*, Dept. of Māori Affairs, 1952–1976 (teaohou.natlib.govt.nz; + Papers Past). — **the Waima pigs account.** +- Te Ara: The Encyclopedia of New Zealand — "Pigs and the pork industry"; "Kai Pākehā – + introduced foods." +- Dowsett, Claire Kuuii Adeline, Anna Carr & Brent Lovelock. "Hunting and Hauora: Pig Hunters + and Poaka in Aotearoa New Zealand." *New Zealand Geographer* 82(2), 2026. + doi:10.1111/nzg.70023 +- Edwards, et al. "Fenced Out… and Fenced In? Containing Cats in the Borderlands of Zealandia." + *New Zealand Geographer*, 2026. doi:10.1111/nzg.70022 +- "Feral pigs in the northern South Island, New Zealand." *Journal of the Royal Society of New + Zealand*, 1991. doi:10.1080/03036758.1991.10418181 +- NZ Department of Conservation, pig hunting guidance (Nelson/Marlborough). + +**Australia** +- PestSmart (pestsmart.org.au), feral pigs toolkit. +- National Feral Pig Action Plan (feralpigs.com.au), incl. Cape York / FNQ and QLD pages. +- NSW Department of Primary Industries; Business Queensland — feral pig (restricted invasive + animal). +- CSIRO, "Managing feral pigs on Cape York: it's not a numbers game" (2015); NESP Resilient + Landscapes Hub. +- DCCEEW, *Wetlands Australia* 36 (July 2022), Cape York feral pig research collaboration. +- "Aboriginal Rangers' Perspectives on Feral Pigs: Are they a Pest or a Resource? A Case Study + in the Wet Tropics World Heritage Area of Northern Queensland." +- Collins English Dictionary, *pig-root*. + +**Cook / ecological history** +- *Hakai Magazine* / *bioGraphic*, "Islands of the Feral Pigs." +- "A novel MC1R allele for black coat colour reveals the Polynesian ancestry and hybridization + patterns of Hawaiian feral pigs." *Royal Society Open Science*. doi:10.1098/rsos.160304 + (PMC5043315) +- US National Park Service — Hawai'i Volcanoes NP perimeter fencing (GAOA); Baker, James K. + "The Feral Pig in Hawaii Volcanoes National Park" (1975); HCSU Technical Report HCSU-004. +- "Ecological impacts of feral pigs in the Hawaiian Islands." *Biodiversity and Conservation*. + doi:10.1007/s10531-009-9680-9 + +--- + +## Confidence Notes + +**SOLID — safe to use as stated** +- Kamapua'a's serial capture-and-escape structure; the four bindings and four chant-releases + (Beckwith); the fake binding at the heiau (Lonoaohi); the Kaliuwa'a waterfall escape and the + naming of that place; escape from Pele as the humuhumunukunukuāpua'a; kino lau including + kukui, 'uala, 'ama'u. +- The Pele/Kamapua'a division of Hawai'i Island into wet and dry halves. +- *Ahupua'a* = *ahu* + *pua'a*, "pig altar," a boundary marked by a carved pig's head. +- Kame'eleihiwa 1996, Bishop Museum Press, translating an anonymous 1891 *Ka Leo o ka Lāhui* + serial. +- Moka (Strathern 1971) and te/tee (Meggitt, Feil) as pig-centred inter-group exchange; + women's central role in the tee (Feil 1978). +- Vanuatu tusker cultivation: upper canines avulsed, circle in ~6–7 years, double circle + 10–12; grade-taking requirement; boar's tusk on the national flag. +- Rappaport's *Pigs for the Ancestors* and the garden-invasion → kaiko sequence in outline. +- Deliberate mating of domestic sows with wild boars in PNG highlands husbandry. +- No pigs in pre-contact Aotearoa; pigs as post-1769 currency; "Captain Cooker" as the NZ name + for feral pigs. +- Cook's deliberate liberations as a stated provisioning policy; 1773 Queen Charlotte Sound. +- First Fleet 1788 pigs escaping through inadequate containment; feral pigs across ~45% of + Australia. +- Australian/NZ *pig-root*; *in a pig's eye* from 1847. +- Hawaiian feral pigs are substantially of Polynesian, not Cook-era, ancestry. +- Hawai'i Volcanoes NP: hunting 1930–71 largely failed; fencing from the 1970s succeeded. +- *Babe* (1995) filmed at Robertson, NSW, from King-Smith's *The Sheep-Pig* (1983). +- Hogzilla is from Georgia, USA, 2004 — **not** Australian. + +**PROBABLE — verify before printing** +- Exact wording of the Beckwith "four times… eight hundred strong" sentence. +- Kame'eleihiwa's argument that the 1891 serial was an anti-colonial gesture on the eve of the + overthrow. *(High-value if true — verify in her introduction.)* +- The Kaliuwa'a = "leak of the canoe" gloss; competing glosses exist. +- Fornander's "eight-eyed / eight-footed" epithets. +- The *'ao-pua'a* "pig clouds" kino lau. +- The Pukui *'Ōlelo No'eau* entries and *pua'a hulu 'ole*. +- The Waima/*Te Ao Hou* passage's exact issue and page (text itself retrieved near-verbatim). +- Cook's precise 1773 tallies and place-names; the 1778 Ni'ihau landing date. +- The Landtman/Kiwai **Sido-as-pig / body-as-house-of-death** detail. **Do not print without + opening Landtman.** +- Australian feral pig population figures (estimates range 3.5m–24m; use a range). +- The Jennifer Creek quotation's attribution. +- First Fleet "49 hogs." + +**THIN — omit, or use only with an explicit hedge** +- Any **Māui-and-pig** episode. I found nothing reliable. **Recommend omitting entirely.** +- The Pele/Kamapua'a **Wailua River oath** (likely a Kaua'i conflation). +- A **Tok Pisin pig proverb** — none found. Do not invent one. +- A **Māori whakataukī** about poaka — none found. *"Ka rūrū noa iho te poaka i tōna pane"* is + a dictionary example sentence, not a proverb. +- The **kunekune** near-extinction-and-rescue narrative (breed-association sourcing). +- **Emily Rodda, _Pigs Might Fly_** (1986) — exists, plot unverified here. +- Any Austronesian **etymology** for *puaka* beyond listing the cognate set. +- **Micronesia** generally: I confirmed that pigs, yams and sakau are the prestige foods of + Pohnpeian feasting (**PROBABLE**), but found **no** escape/containment material. Micronesia + is the weakest sub-region for this question; say so rather than padding. + +**Deliberately excluded as contamination risk** +- The *Puaka* water-demon of *A Book of Creatures* — Dusun/Borneo, not Oceanian. +- Hogzilla relocated to Queensland. +- Anything from the "oceanianfolktales.com" / "godsandmonsters.info" / fandom-wiki tier, which + recurred throughout the search results. These sites reproduce Kamapua'a material without + sourcing and are precisely the "fake versions circulating online" the brief warned about. + **Everything asserted above traces to a named author or a government/museum/university + source.** + +**Structural recommendation for the essay.** The region offers a rare three-way structure: +Hawai'i gives the pig that *cannot be held* (Kamapua'a); Vanuatu gives the pig whose tusk is +spent ten years being *bent into a closed circle*; and Cook gives the European who *let pigs +go on purpose*. Uncontainable, hyper-contained, and deliberately released — three answers to +the same animal, within a few thousand miles and, in the last two cases, within the same +decade of the 1770s. diff --git a/research/boxes-and-escape/survey/pig-south-america.md b/research/boxes-and-escape/survey/pig-south-america.md new file mode 100644 index 0000000..011b93d --- /dev/null +++ b/research/boxes-and-escape/survey/pig-south-america.md @@ -0,0 +1,565 @@ +# Pigs, Peccaries, Escape and Containment — SOUTH AMERICA + +Research file. Compiled 2026-07-28. + +**Method note / limitation:** WebSearch was used extensively (budget exhausted at 200 calls). +WebFetch was blocked at the network gateway (403 on every host, including wikipedia.org, +jstor.org, sciencedirect, springer, and the Berkeley open-access PDF of Murphy 1958). So +**no primary text was read directly in this session.** Everything below rests on search-engine +summaries of pages I could see but not open. I have graded confidence accordingly and flagged +every place where the primary source needs to be checked before an essay quotes it. + +--- + +## Summary + +There is a lot of material, and it is unusually well-suited to a containment/escape essay — +more so than for most continents, because South America has *two* separate pig stories running +in parallel and they collide: + +1. **The native peccaries (Tayassuidae)**, around which Amazonian peoples built a dense body of + myth in which peccaries are *former humans who left*, or *dead relatives who come back*, or + *someone else's livestock that has to be released before you can hunt it*. The escape, + departure, and boundary-crossing motifs are not incidental here — in several traditions they + are the whole point of the story. +2. **The true pigs (Suidae)**, introduced by Iberians from the 1490s, which went feral almost + immediately (*puercos cimarrones*), and then a second time in the 20th century as European + wild boar (*javali/jabalí*) escaping from hunting reserves and breeding farms. This is a live + ecological and legal problem in Brazil, Argentina and Uruguay right now. + +The strongest single item is the **Mundurucú origin-of-wild-pigs myth**, which is literally a +story about pigs being penned and then let out. The most interesting structural finding is the +**word *cimarrón***, which in colonial Spanish covered escaped pigs and escaped people with the +same term. + +Three things to be careful about: (a) the "wild pigs" of the ethnographic literature are almost +always peccaries, not Suidae; (b) fake "Amazonian legends" circulate freely online and I have +excluded several sites that looked like content farms; (c) I could not confirm Yanomami or +Kayapó peccary-origin myths in this session and have marked them as NOT FOUND rather than +guessing. + +--- + +## Amazonian Peccary Myths (transformation & departure) + +### 1. Mundurucú (Munduruku), Tapajós, Brazil — the origin of wild pigs — **CATEGORY (a) + (b)** + +The keystone item. Two independent strands of evidence converged. + +**Strand 1 — the structuralist summary.** In the Mundurucú myth "The Origin of Wild Pigs," +humans of the mythic age are transformed by the demiurge **Karusakaibe / Karosakaybu** into +pigs. Crucially, they are first turned into the equivalent of *domesticated* pigs: they are kept +in a **pig-sty in the village** and killed one by one for meat, **until someone lets them escape +and they flee into the forest**, thereby becoming the wild pigs of today. Lévi-Strauss reads this +as a regression from culture to nature in three stages, the last two of which invert the human +progression from foraging to farming. + +- This is myth **M16** in the numbering of *Mythologiques I: Le cru et le cuit* (1964; English + *The Raw and the Cooked*, 1969) — **the M16 number is my own recollection and was NOT confirmed + by search; verify before citing.** +- The ethnographic source Lévi-Strauss drew on is **Robert F. Murphy, *Mundurucú Religion*, + University of California Publications in American Archaeology and Ethnology 49(1), 1958**, + which contains texts of 58 songs and folktales. Open-access PDF exists at + `https://digitalassets.lib.berkeley.edu/anthpubs/ucb/text/ucp049-002.pdf` — **I was blocked + from opening it. This is the first thing to read.** +- Also relevant: Robert & Yolanda Murphy's later Mundurucú work. +- **Confidence: SOLID** that the myth exists and turns on humans→pigs. **PROBABLE** on the + penning-and-escape detail (two independent search summaries gave the same account, but I read + neither Murphy nor Lévi-Strauss directly). + +**Strand 2 — contemporary Munduruku tellings.** From the chapter *"Munduruku Cosmopolitics and +the Struggle for Life"* (IntechOpen, 2023, `intechopen.com/chapters/86084`) and from Munduruku +public letters (Survival International / Povos Indígenas no Brasil): + +- Karosakaybu transformed people **on the other side of the river** into pigs, because of their + negligence. The site of **Macapá (Munduruku: *Mukapap*)** on the Tapajós is described as a + sacred **"passage"** — the place where the pigs came down and where ancestors had to cross to + the other side. Karosakaybu's **footprints are left in the rocks** there. +- Karosakaybu's **son was carried off by the pigs to the far bank of the Tapajós**, and the + demiurge **gave up looking for him**. In a variant, Karosakaybu **managed to trap the pigs + between some mountains, but his son stayed with them** and was never seen again by his father. +- **Confidence: PROBABLE-to-SOLID.** This matters politically as well as folklorically: the + Munduruku invoke these sites in opposition to Tapajós dams, so the myth is in active use. + +**Why this is the best item for the essay:** it is a containment story twice over. The pigs are +penned and escape; then they are re-trapped between mountains and *still* the boundary leaks — +the creator's own son crosses over and stays on the animal side. The river is the boundary, the +sty is the containment, the escape is the aetiology. + +### 2. Wari' (Pakaa Nova), Rondônia, Brazil — the dead return as peccaries — **CATEGORY (b)** + +Ethnographer: **Aparecida Vilaça** (Museu Nacional / UFRJ). See her *Strange Enemies: Indigenous +Agency and Scenes of Encounters in Amazonia* (2010) and *Praying and Preying: Christianity in +Indigenous Amazonia* (2016). Related: **Beth A. Conklin, *Consuming Grief: Compassionate +Cannibalism in an Amazonian Society*** (2001) on Wari' mortuary practice. + +- Wari' dead go to an **underworld / underwater world**, and **return from it in the form of + white-lipped peccaries**. The peccary-dead deliberately **approach a hunter who is close kin**, + so that its meat goes to feed its own relatives. Eating peccary is therefore eating one's own + ancestors, offered on purpose. +- Category term ***jami karawa*** — "animals that have a human spirit/double," including + white-lipped peccary, collared peccary, deer, tapir, capuchin, jaguar, fish, bees, snakes. The + *jami karawa* **live in villages organised like Wari' villages**: houses, swiddens, festivals. +- The *jamixi'* (double) is the capacity to act in another relational context; for Vilaça, + humanity itself is defined by this transformability. +- **Confidence: SOLID.** Vilaça is the standard authority and this is among the most-cited + Amazonian perspectivist cases. +- **Direction of travel:** this is *return*, not departure — the boundary is crossed inward. + Useful as the counter-motif to the Mundurucú departure. + +### 3. Ese Eja, Peru/Bolivia (Madre de Dios / Beni) — peccary herds as visiting dead — **CATEGORY (b)** + +- White-lipped peccaries (**ño'**) are the ***emanokuana*** — deceased relatives, temporarily + transformed. After a long post-mortem journey the dead go on living lives parallel to those of + living Ese Eja (hunting, fishing, swiddens), except that their *eshawa* have regained the + mythic-era powers of **mutability and transformation into different bodily forms**. +- **The arrival of a herd is a visit.** The dead come back temporarily because they want to eat + prized forest fruits and to see their relatives. A herd's appearance is described as a + "crossroads of Ese Eja states of existence," the worlds of the living and the dead overlapping. +- Source: **"Living dead ancestors: White-lipped peccaries and alternative posthuman Amazonian + histories," DOI 10.1080/02757206.2026.2619701.** The DOI prefix `02757206` is the journal + ***History and Anthropology*** (a search summary called it "Imagining Animals," which I believe + is wrong — probably a special-issue title). **Author almost certainly Daniela Peluso**, the + established Ese Eja ethnographer, but **the byline was NOT confirmed — verify.** +- **Confidence: SOLID on content** (the same passage came back on two independent searches); + **THIN on the byline and journal name.** + +### 4. Matsigenka (Machiguenga), Peru — game animals as the spirits' livestock — **CATEGORY (b)** + +Ethnographer: **Glenn H. Shepard Jr.** (Museu Goeldi). + +- The ***Saangariite*** ("the invisible ones"), guardian spirits, **raise all game animals as + pets**: curassows are their chickens, **peccaries are their pigs**, and the jaguar is their + watchdog. +- Matsigenka shamans use medicinal and psychoactive plants both to improve aim and to + **negotiate with the spirit "owners"** of game. +- **Confidence: PROBABLE.** Attributed clearly in search results to Shepard's published work + (see his "Shamanism and diversity: A Matsigenka perspective" and "Primates in Matsigenka: + Subsistence and world view"); the exact publication for the curassow/peccary/jaguar line was + not pinned down. +- **On theme:** the peccary in the forest is not wild — it is someone else's penned livestock, + and hunting only works when the owner lets it out. Containment relocated to the spirit world. + +### 5. Apurinã, Purus river, Brazil — the death of the chief of peccaries — **CATEGORY (a)/(c) hybrid** + +- **Pirjo Kristiina Virtanen, "The Death of the Chief of Peccaries: The Apurinã and the Scarcity + of Forest Resources in Brazilian Amazonia,"** chapter in a Springer volume, 2017 + (DOI 10.1007/978-3-319-42271-8_6). +- Apurinã narratives about the **death and replacement of the non-human chief of the peccaries** + have been altered to register recent socio-political, economic and environmental change. Game + scarcity is narrated as the loss of the animal chief — i.e. **the game left because its master + died.** +- Related by the same author: *Fatal Substances: Apurinã's Dangers, Movement and Kinship*. +- **Confidence: SOLID** (title, author, publisher and abstract all verified). +- **On theme:** scarcity is explained as departure with a cause, not as depletion. + +### 6. Game masters generally, pan-Amazonian — **CATEGORY (b)** + +- **"Game masters and Amazonian Indigenous views on sustainability," *Current Opinion in + Environmental Sustainability* 43 (2020): 21–ff.** Bibcode `2020COES...43...21F` — first author + surname begins with F; **very likely Álvaro Fernández-Llamazares** (with Virtanen among the + co-authors), but **NOT confirmed — verify the byline.** +- Argument: spirit **owners/masters of animals are widespread across the Amazon Basin**; the + forest is an extended web of social relations that must be **managed by an owner**; shamanic + ritual is the channel for negotiating release of game; reciprocity and restraint from excessive + extraction are the terms. +- **Confidence: SOLID** that the paper exists and makes this argument; **PROBABLE** on authorship. +- Companion piece: **"Supernatural Gamekeepers/Animal Masters Among the Munduruku (Wuy Jugu), + Tukano, Embera and Achuar (Shiwiar) of the Neotropics,"** chapter in a Springer volume, 2023 + (DOI 10.1007/978-3-031-37503-3_14). **Blocked; not read. Looks like the single most on-theme + secondary source available and should be obtained.** + +### 7. Tenetehara / Guajajara, Maranhão, Brazil — **CATEGORY (b)** + +- **Maranaüwa (Marana ywa)**, owner of the forest and its animals, **punishes people who kill + white-lipped peccaries needlessly.** +- Source: **Charles Wagley & Eduardo Galvão, *The Tenetehara Indians of Brazil: A Culture in + Transition*** (Columbia UP, 1949) / *Os Índios Tenetehara* (1961). Both are digitised at the + Biblioteca Digital Curt Nimuendajú (`etnolinguistica.org`) — **not fetched.** +- **Confidence: PROBABLE.** The Maranaüwa/peccary detail came through a search summary, not the + original. I searched specifically for a Tenetehara *origin-of-wild-pigs* myth (transformation + at a feast) and **did not find one** — do not assert it. + +### 8. Tukano / Desana, Vaupés, Colombia — **CATEGORY (b) — PARTLY UNVERIFIED** + +- **Gerardo Reichel-Dolmatoff**, *Desana: Simbolismo de los Indios Tukano del Vaupés* (1968) and + *The Forest Within: The World-view of the Tukano Amazonian Indians* (1996), describe + **Vaí-mahsë**, the Master of Animals (distinct from Vihó-mahsë, master of the snuff). +- The specific detail I went looking for — **that Vaí-mahsë keeps game penned inside hills or + rock houses and releases it to hunters after shamanic negotiation** — is widely repeated in the + literature but **I could NOT confirm it in this session.** +- **Confidence: THIN. MARK AS UNVERIFIED.** Do not use the hill-corral detail without opening + Reichel-Dolmatoff. The existence of Vaí-mahsë as animal master is SOLID. + +### 9. Achuar / Shiwiar, Ecuador–Peru — **CATEGORY (b) — MOSTLY UNVERIFIED** + +- **Philippe Descola**, *In the Society of Nature: A Native Ecology in Amazonia* (1994) and + *The Spears of Twilight* (1996), on Achuar relations with game. +- I looked for **Amasank/Amasanka** as a *master of peccaries* specifically and **could not + confirm it.** One search summary indicated that a related transformation, **Jurijuri + (Achuar: Jirijri), is master of monkeys**, not peccaries. +- **Confidence: THIN / DUBIOUS as stated. Do not claim Amasank is the peccary-master.** Descola + does discuss peccaries; the specific mastership attribution needs checking in the Springer 2023 + gamekeepers chapter or in Descola directly. + +### 10. NOT FOUND — do not invent + +- **Yanomami** origin-of-peccary myth involving people transformed and fleeing into the forest: + **NOT CONFIRMED.** Jacques Lizot's *Tales of the Yanomami* (Cambridge) and Kopenawa & Albert's + *The Falling Sky* are the places to look. **Do not assert.** +- **Kayapó / Mebêngôkre** origin-of-wild-pigs myth: **NOT CONFIRMED** despite targeted searching + in Portuguese. (Terence Turner and Vanessa Lea are the ethnographers to check.) **Do not assert.** +- **Andean / Quechua / Aymara** pig folklore: not reached before budget ran out. Note that pigs + are post-1532 in the Andes, so any pig material there is colonial-era or later; the peccary is + a lowland animal and marginal to highland tradition. + +--- + +## Escape / Containment Motifs + +### A. Mythic containment and release +- **Mundurucú pig-sty escape** (see above). The paradigm case: penned → released → wild. Also + the re-trapping "between some mountains" that still fails to hold the boundary. +- **Matsigenka Saangariite** keeping peccaries as their pigs; **Tenetehara Maranaüwa**; + **Apurinã chief of the peccaries**; pan-Amazonian game masters. In all of these the animal is + *already owned and already contained* — the hunter's problem is not tracking but permission. + +### B. Caipora / Curupira, Brazilian folklore — the herd that cannot be caught — **CATEGORY (a)** +Strong and directly on-theme: +- The **Caipora** (in parts of the North and Northeast, **Caapora**, and conflated with the + **Curupira**) is the **owner of game** — of all game "except winged game." +- **He governs the wild pigs and travels with the herds**, making noise through the forest. He is + most often depicted **riding a *caititu* (collared peccary) or a *queixada* (white-lipped + peccary)**. +- He is the **"padrinho" (godfather) of the herds**: he **guides the pigs away from hunters' + traps** and helps them find food. +- He **bargains with famous hunters**, who give him tobacco, cachaça and cloth in exchange for a + specified number of pigs from the herds he leads. Once the deal is struck, the hunter has game + at will. +- **Confidence: PROBABLE.** This is consistently attested across Brazilian folklore sources, but + the ones I could see were popular/educational sites, not scholarship. **The canonical print + source to cite is Luís da Câmara Cascudo, *Dicionário do Folclore Brasileiro* (entries + "Caipora" and "Curupira") — I did not verify Cascudo's exact wording.** +- **Why it's good:** it is an explicit *uncatchability* mechanism. The peccary herd is not + elusive by nature; it is being actively steered away from your traps by its godfather. And the + only way to catch any is to negotiate a quota. Containment and release, brokered. + +### C. Real herds that vanish and come back — **CATEGORY (a), and the myth/ecology convergence** +- **José M. V. Fragoso et al., "Large-scale population disappearances and cycling in the + white-lipped peccary, a tropical forest mammal," *PLOS ONE* (2022), + DOI 10.1371/journal.pone.0276297** (a Correction was later published, + DOI 10.1371/journal.pone.0314917). Fragoso is affiliated with UnB, INPA and the California + Academy of Sciences. +- Findings: **43 documented disappearance events across nine South and Central American + countries**, plus **88 years of harvest data**. The disappearances are **7–12-year troughs + within 20–30-year population cycles**, synchronised at regional and possibly continent-wide + scales — areas of **10,000 to 5,000,000 km²**. Possibly the **first documented natural + population cyclicity in a Neotropical mammal.** +- **The part that matters for the essay:** the paper explicitly incorporates Indigenous knowledge. + **Several ethnic groups explain disappearance events as caused by the death of a powerful + shaman, and hold that the herds' return can only be secured by another shaman performing + specific ritual work.** Indigenous testimony helped resolve the biological puzzle. +- **Confidence: SOLID.** Well covered by WCS, EurekAlert, ScienceDaily, SciTechDaily, Mongabay. +- Earlier: **Fragoso, J.M.V. (2004), long-term study of white-lipped peccary disappearances** — + PDF at fragosolab.org, **blocked, not read.** +- Also from the Fragoso lab (2015): **"Large numbers of white-lipped peccaries (Tayassu pecari) + invade Amazonian town."** **Title only; page not read. PROBABLE.** If it holds up, it is a neat + inversion — the uncontainable herd crossing *into* human space. +- Also: "The paradoxical situation of the white-lipped peccary in the state of Mato Grosso, + Brazil," *Perspectives in Ecology and Conservation* — not read. + +### D. Festival: catching the pig that won't be caught — **CATEGORY (a)** +- **"Pega do porco"** at the **Festa do Trabalhador, São Pedro do Ivaí, Paraná, Brazil**: an arena + filled with mud in which participants chase a piglet that tries to avoid being grabbed; the + winner is whoever catches it fastest. **Opposed by animal-rights activists** (reported by + Vegazeta). **Confidence: PROBABLE** (single outlet seen). +- The greased-pig genre (*porco/leitão ensebado*, *cerdo ensebado*) is clearly present in + Luso-Brazilian festival culture but **I did not get a clean documented instance beyond the + above.** Marked **THIN**. +- **Matança do porco** (Portugal, and Brazilian South/Southeast): the annual midwinter pig + slaughter as family and neighbourhood festival. **No containment/escape theme — CATEGORY (c)** — + but it is the ritual backdrop to the San Martín proverb below. + +--- + +## Proverbs & Idioms + +### Spanish +- **"A cada cerdo le llega su San Martín"** — "every pig gets its St Martin's Day." Everyone + eventually gets what's coming to them. **Origin:** St Martin of Tours, 11 November, the + traditional date of the *matanza del cerdo* in Spanish villages. **Attested in Cervantes, + *Don Quijote*, Part II, ch. 63: "pero su San Martín se le llegará como a cada puerco."** + Softened modern variant "A cada uno le llega su San Martín." **Colombian equivalent: "A cada + pavo le llega su Nochebuena."** — **SOLID. CATEGORY (c)** for containment purposes, though it + is precisely a proverb about *the day the pig's freedom ends*, which is usable. +- **"Cada chancho a su chiquero"** (Argentina) — "each pig to its own sty"; everything in its + proper place. **CATEGORY (a)/containment. PROBABLE** (found in Argentine regional phraseology + collections). +- **"Cuando los chanchos vuelen"** — "when pigs fly"; listed alongside "cuando las ranas tengan + pelos," "cuando las gallinas meen," "cuando llueva para arriba." Also **"[fulano] cree que los + chanchos vuelan"** (he's credulous) and the joke **"De las aves que vuelan me gusta el + chancho."** — **SOLID/PROBABLE. CATEGORY (c)**, but the pig-as-impossible-flight image is + adjacent to uncatchability. +- **"La culpa no es del chancho, sino de quien le da de comer"** (Argentina) — "it's not the + pig's fault but the fault of whoever feeds it." Blame the keeper, not the animal. + **PROBABLE. CATEGORY (c)**, but it is squarely about custodianship. +- ***Cimarrón*** — see below under Feral Pigs. The most important lexical item in this whole file. + +### Portuguese +- **"Cada porco em seu chiqueiro, cada pinto em seu poleiro"** — "each pig in its sty, each chick + on its perch." Recorded as a regional saying of **Mato Grosso do Sul**, Brazil. + **CATEGORY (a)/containment. PROBABLE** (single popular-press source: Perfil News). +- **"Onde não tem onça, porco folga"** — "where there's no jaguar, the pig takes its ease." + Brazilian regional. **PROBABLE. CATEGORY (c)**, though it is about the absence of the predator + that would otherwise keep the pig in check. +- The Portuguese cognate of the San Martín proverb ("cada porco tem o seu São Martinho") is + standard in Portugal and Brazil — **NOT separately verified this session; THIN.** + +### Not reached +Andean Spanish/Quechua pig idioms; Uruguayan and Chilean regional sayings; *javali*-specific +idioms. Budget exhausted. + +--- + +## Feral Pigs & Modern Incidents + +### A. The colonial escape — *puercos cimarrones* — **CATEGORY (a), SOLID** +- **Gonzalo Fernández de Oviedo, *Historia general y natural de las Indias* (1535)**, identifies + **"puercos cimarrones o salvajes"**; feral pigs were abundant on Hispaniola and constituted a + major food source. +- Pigs, cattle and horses **multiplied enormously** in territory favourable to expansion. + Feral pigs on the **pampas** are described as descendants of the first domestic pigs the + Spanish let loose in the sixteenth century. +- Scholarly anchor: **"El cerdo. Historia de un elemento esencial de la cultura castellana en la + conquista y colonización de América (siglo XVI)," *Anuario de Estudios Americanos* (CSIC)** — + open access at estudiosamericanos.revistas.csic.es, article 430. **Not read (blocked), but the + citation is confirmed.** Argues the pig was foundational to conquest nutrition and that pig- + rearing became tied to Indian tribute. +- **THE KEY LEXICAL POINT.** ***Cimarrón*** was applied to **animals that escaped human control** + and, per the sources found, was **first applied to rebel/fugitive Indians in Cuba in the + sixteenth century**, and subsequently to escaped enslaved Africans (*cimarrones*, English + "maroons"; Portuguese *quilombolas* in the parallel Brazilian system). **The same word covers + the escaped pig and the escaped person.** — **SOLID on the semantic overlap; PROBABLE on the + claimed order of application (rebel Indians first). Worth verifying the etymological sequence, + which is contested — some derive it from *cima* 'summit/thicket'.** +- Modern survival of the term: **"Chanchos cimarrones, perros entrenados y cuchillos afilados: + una crónica de caza," Infobae, 22 Jan 2018** — feral-pig hunting in Argentina today. + +### B. Brazil — *javali* — **CATEGORY (a), SOLID** +- European wild boar (*Sus scrofa*) **crossed the border into Rio Grande do Sul around 1989** + from the Uruguayan/Argentine side. +- **Commercial javali breeding began at scale in the mid-1990s**, with new imports from Europe + and Canada. +- **IBAMA banned the import and breeding of javali in 1998.** In response, **numerous Brazilian + breeders released their animals, deliberately, or let them escape** — this is the pivotal + event. Escaped and released stock founded a growing feral population. **The regulation intended + to contain the animal is what set it loose.** +- Today the javali is reported in **all Brazilian biomes** and is said to occupy **35.9% of the + national territory** (figure from a Brazilian source; **PROBABLE**, verify). +- It has **no natural predators** in Brazil and **hybridises with the domestic pig**, producing + the ***javaporco***. +- **IBAMA subsequently instituted official management/control** of javali (Normative Instruction + — **IN 03/2013 is my recollection; number NOT verified**). See the Fiocruz *boletim_javali* PDF. +- Coverage: ((o))eco, "Javali no Brasil: tá tudo dominado" and "O que a invasão dos javalis nos + ensina sobre o princípio da precaução"; *Ciência Hoje*, "A invasão do javali"; Instituto Ampara, + "Javali, o cavalo de Troia da caça no Brasil"; Brasil de Fato (Mar 2026). + +### C. Argentina — *jabalí* — **CATEGORY (a), SOLID/PROBABLE** +A very clean, well-dated chain of escapes: +- **1906:** first wild boar brought to the **San Huberto hunting reserve** (today the **Parque + Provincial Luro**) in **La Pampa**, by **Pedro Luro**. +- **1909:** animals imported from the **Carpathians**, released into an **enclosed 800-hectare + area.** +- **Because of Pedro Luro's business failures, the animals got off the property** and spread + through the **caldén** woodlands of central Argentina. +- **1914–1930:** further escapes from hunting reserves push the species into **San Luis, Córdoba, + Santa Fe, Chubut and Entre Ríos.** +- **1917–1922:** some animals transferred to the **Collun-Có estate, Neuquén.** +- **1931: an accident released these animals into the wild** — the founding dispersal into + **Patagonia**, reaching **Lanín and Nahuel Huapi national parks.** +- **April 2025:** the **Buenos Aires provincial government authorised jabalí hunting across the + whole province**, formally designating it an *"especie exótica invasora."* (Infobae, 24 Apr 2025.) +- Ongoing: SAREM's *Categorización de los mamíferos de Argentina* lists *Sus scrofa* as exotic; + 2024–2026 press on agricultural damage and zoonotic risk. +- **The 800-hectare enclosure is a gift for the essay**: an escape story with a fence, a date, a + named owner, and a bankruptcy as the proximate cause. + +### D. Uruguay +Wild boar are present and legally treated as a plague; **I did not verify the Uruguayan +declaration or its date — NOT RESEARCHED. Do not assert specifics.** + +### E. Not researched +Named individual pig-escape incidents (a Tamworth Two equivalent) in South American news; +children's literature; modern South American fiction featuring pigs. **Budget exhausted before +reaching these.** Horacio Quiroga was queued as a lead (Misiones jungle stories) and never run. + +--- + +## Peccary / Pig Distinction Notes + +**Taxonomy.** +- **Suidae (true pigs)** — Old World. In South America: the **domestic pig** (*Sus scrofa + domesticus*; PT *porco*, ES *cerdo/chancho/puerco/marrano/cochino*) and the **European wild + boar** (*Sus scrofa*; PT *javali*, ES *jabalí*). **All introduced, from 1493 onward.** The + *javaporco* is the domestic × wild boar hybrid. +- **Tayassuidae (peccaries)** — New World natives, a separate family: + - **Collared peccary**, *Pecari tajacu* — PT *cateto*, *caititu*; ES *pecarí de collar*, + *saíno*, *báquiro*, *chancho del monte*. + - **White-lipped peccary**, *Tayassu pecari* — PT ***queixada***; ES *pecarí labiado*, + *huangana* (Peru), *báquiro cachete blanco*. **The herd animal — hundreds of individuals — + and the one that carries nearly all the mythology and all the vanishing-herd material.** + - **Chacoan peccary**, *Catagonus wagneri* — ES *taguá*, *quimilero*. Gran Chaco (Paraguay, + Argentina, Bolivia). **Described from fossils in 1930 and not confirmed alive until 1971** + (Ralph Wetzel) — a Lazarus taxon, and thematically the most literal "refused to be found" + animal on the continent. **This is from general knowledge, NOT verified this session — + marked THIN, verify dates and Wetzel's role.** + +**Why the confusion is productive rather than merely an error.** +- The vernacular names in both Iberian languages **classify peccaries as pigs**: Portuguese + ***porco do mato*** and Spanish ***puerco de monte* / *chancho del monte*** both literally mean + "pig of the forest." Brazilian Portuguese *porco do mato* covers both *cateto* and *queixada*. +- Consequently **the ethnographic and folkloric literature in English routinely says "wild pigs" + when it means peccaries.** Lévi-Strauss's and Murphy's "wild pigs" of the Mundurucú are + *queixadas*. The Caipora's "porcos do mato" are peccaries. **Every myth in the Amazonian + section of this file is about Tayassuidae, not Suidae.** +- Conversely **every item in the Feral Pigs section is about Suidae** — *puercos cimarrones*, + *javali*, *jabalí*. +- The proverbs are all about **Suidae**: they are Iberian imports and the pig in them is the + farmyard pig of the *matanza*. +- **One genuine overlap to watch:** in modern Brazilian and Argentine hunting and press usage, + *porco do mato* / *chancho salvaje* can be used loosely for feral *Sus scrofa* as well, which + muddies contemporary sources. Check the Latin binomial whenever a modern source says + "wild pig." + +--- + +## Sources + +**Anthropology / mythology — primary and standard** +- Robert F. Murphy, *Mundurucú Religion*, UCPAAE 49(1), 1958. PDF: + https://digitalassets.lib.berkeley.edu/anthpubs/ucb/text/ucp049-002.pdf *(blocked; not read)* +- Claude Lévi-Strauss, *Mythologiques I: Le cru et le cuit* (1964) / *The Raw and the Cooked* (1969) +- Aparecida Vilaça, *Strange Enemies* (2010); *Praying and Preying* (2016) +- Beth A. Conklin, *Consuming Grief* (2001) +- Charles Wagley & Eduardo Galvão, *The Tenetehara Indians of Brazil* (1949); *Os Índios + Tenetehara* (1961) — digitised: http://www.etnolinguistica.org/biblio:wagley-galvao-1949-tenetehara +- Gerardo Reichel-Dolmatoff, *Desana* (1968); *The Forest Within* (1996) +- Philippe Descola, *In the Society of Nature* (1994); *The Spears of Twilight* (1996) +- Jacques Lizot, *Tales of the Yanomami* (Cambridge) — *lead, not followed* +- Luís da Câmara Cascudo, *Dicionário do Folclore Brasileiro* — *for Caipora/Curupira; not verified* +- R. A. Donkin, *The Peccary: With Observations on the Introduction of Pigs to the New World* + (Transactions of the American Philosophical Society, 1985) — **surfaced in search; appears to be + THE scholarly monograph on this exact subject and should be obtained** + +**Anthropology / mythology — articles** +- "Living dead ancestors: White-lipped peccaries and alternative posthuman Amazonian histories," + https://doi.org/10.1080/02757206.2026.2619701 (*History and Anthropology*; author likely + Daniela Peluso — unconfirmed) +- Pirjo Kristiina Virtanen, "The Death of the Chief of Peccaries…," DOI 10.1007/978-3-319-42271-8_6 + https://link.springer.com/chapter/10.1007/978-3-319-42271-8_6 +- "Game masters and Amazonian Indigenous views on sustainability," *Current Opinion in + Environmental Sustainability* 43 (2020), + https://www.sciencedirect.com/science/article/pii/S187734352030004X +- "Supernatural Gamekeepers/Animal Masters Among the Munduruku (Wuy Jugu), Tukano, Embera and + Achuar (Shiwiar) of the Neotropics," DOI 10.1007/978-3-031-37503-3_14 +- "Munduruku Cosmopolitics and the Struggle for Life," https://www.intechopen.com/chapters/86084 +- Glenn H. Shepard Jr., "Shamanism and diversity: A Matsigenka perspective"; "Primates in + Matsigenka: Subsistence and world view" +- Povos Indígenas no Brasil (ISA): https://pib.socioambiental.org/en/Povo:Wari' · + https://pib.socioambiental.org/en/Povo:Munduruku + +**Ecology** +- Fragoso et al., "Large-scale population disappearances and cycling in the white-lipped peccary, + a tropical forest mammal," *PLOS ONE* (2022), + https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0276297 + (Correction: 10.1371/journal.pone.0314917) +- WCS newsroom release: https://newsroom.wcs.org/News-Releases/articleType/ArticleView/articleId/18171/ +- Mongabay, "Population cycles explain white-lipped peccary's ups and downs, study shows" (2022) +- Fragoso 2004, https://fragosolab.org/wp-content/uploads/2015/11/fragoso-2004-white-lip-dissapearances.pdf +- Fragoso lab, "Large numbers of white-lipped peccaries invade Amazonian town" (2015) + +**Feral pigs — Brazil** +- ((o))eco, "Javali no Brasil: tá tudo dominado" https://oeco.org.br/reportagens/javali-no-brasil-ta-tudo-dominado/ +- ((o))eco, "O que a invasão dos javalis nos ensina sobre o princípio da precaução" +- *Ciência Hoje*, "A invasão do javali" https://cienciahoje.org.br/artigo/a-invasao-do-javali/ +- Fiocruz, *Boletim Javali* (IBAMA management) https://www.biodiversidade.ciss.fiocruz.br/sites/www.biodiversidade.ciss.fiocruz.br/files/boletim_javali.pdf +- Instituto Ampara, "Javali, o cavalo de Troia da caça no Brasil" + +**Feral pigs — Argentina / colonial** +- *El Cordillerano*, "El jabalí es una especie invasora: cómo, cuándo y por qué llegó a la + Argentina" (21 Sep 2025) +- Infobae, "El gobierno bonaerense autorizó la caza de jabalí en toda la provincia" (24 Apr 2025) +- Infobae, "Chanchos cimarrones, perros entrenados y cuchillos afilados: una crónica de caza" + (22 Jan 2018) +- SAREM, *Categorización de los mamíferos de Argentina*: https://cma.sarem.org.ar/es/especie-exotica/sus-scrofa +- "El cerdo. Historia de un elemento esencial de la cultura castellana en la conquista y + colonización de América (siglo XVI)," *Anuario de Estudios Americanos*, + https://estudiosamericanos.revistas.csic.es/index.php/estudiosamericanos/article/view/430 +- Gonzalo Fernández de Oviedo, *Historia general y natural de las Indias* (1535) + +**Proverbs / festival** +- Wikipedia ES, "A cada cerdo le llega su San Martín" +- Estandarte, "La expresión 'A cada cerdo le llega su San Martín'" +- Wikiquote ES, "Proverbios argentinos" +- Las Hablas de Córdoba (UNC), "Fraseología — dichos y refranes de uso regional" +- Perfil News, "Provérbios, ditos populares e expressões regionais sul-mato-grossenses" +- Vegazeta, "Ativistas dos direitos animais reprovam 'pega do porco' na Festa do Trabalhador de + São Pedro do Ivaí (PR)" + +--- + +## Confidence Notes + +**SOLID** +- Mundurucú humans→pigs transformation by Karusakaibe/Karosakaybu exists and is central. +- Wari' dead return as white-lipped peccaries; *jami karawa* category (Vilaça). +- Ese Eja peccaries as *emanokuana*, visiting dead (content, not byline). +- Virtanen, "The Death of the Chief of Peccaries" — title, author, publisher, argument. +- Fragoso et al. 2022 PLOS ONE disappearance/cycling findings, including that several Indigenous + groups attribute disappearances to a powerful shaman's death and require shamanic work for + the herds' return. +- "A cada cerdo le llega su San Martín," incl. the *Don Quijote* II.63 attestation. +- Colonial *puercos cimarrones*; Oviedo 1535; the *cimarrón* semantic overlap between escaped + animals and escaped people. +- Brazilian javali history in outline: 1989 border crossing, mid-1990s breeding boom, 1998 IBAMA + ban followed by releases/escapes, present-day nationwide spread, *javaporco* hybrids. +- Argentine jabalí chain: 1906 San Huberto/Parque Luro, 1909 Carpathian stock in an 800-ha + enclosure, Luro's bankruptcy, 1914–1930 escapes, 1917–22 Neuquén transfer, 1931 accidental + release seeding Patagonia, April 2025 province-wide hunting authorisation. + +**PROBABLE** +- The pig-sty-and-escape detail of the Mundurucú myth (two concordant summaries; primary text + unread). +- Contemporary Munduruku Macapá/Mukapap "passage," the son carried across the Tapajós, the + mountain trap. +- Caipora/Curupira as godfather of peccary herds, riding a caititu/queixada, steering pigs away + from traps, and the tobacco/cachaça/cloth quota bargain. (Popular sources only; Cascudo unread.) +- Matsigenka *Saangariite* keeping peccaries as their pigs (Shepard). +- Tenetehara Maranaüwa punishing needless killing of white-lipped peccaries. +- "Cada chancho a su chiquero"; "Cada porco em seu chiqueiro, cada pinto em seu poleiro"; + "Onde não tem onça, porco folga." +- *Pega do porco*, São Pedro do Ivaí. +- Authorship of the *Current Opinion* game-masters paper (Fernández-Llamazares et al.). +- The "35.9% of Brazilian territory" javali figure. + +**THIN / UNVERIFIED — flag or omit** +- Lévi-Strauss myth number **M16** for the Mundurucú wild-pigs myth — my recollection only. +- Reichel-Dolmatoff's Vaí-mahsë keeping game penned **in hills/rock houses** — the mastership is + confirmed, the corral detail is not. +- **Amasank as master of peccaries among the Achuar — NOT confirmed; a search summary points to + Jurijuri/Jirijri as master of *monkeys*. Treat the peccary attribution as DUBIOUS.** +- Byline and journal name of the "Living dead ancestors" article. +- IBAMA Normative Instruction number for javali management. +- Chacoan peccary 1930 fossil description / 1971 living rediscovery by Wetzel — general knowledge, + not searched. +- Portuguese "cada porco tem o seu São Martinho." +- Greased-pig (*porco ensebado*) festivals beyond the São Pedro do Ivaí case. + +**NOT FOUND — do not assert** +- Any Yanomami peccary-origin myth. +- Any Kayapó/Mebêngôkre peccary-origin myth. +- A Tenetehara origin-of-wild-pigs transformation-at-a-feast myth. +- Andean/Quechua/Aymara pig folklore. +- Uruguayan feral-pig legal status specifics. +- Children's literature, modern South American fiction, named individual pig-escape news events. + (Budget exhausted; Horacio Quiroga was a queued lead never followed.) From aba31aee6f18df7712a482b722c816e5e8dbf650 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 02:15:12 +0000 Subject: [PATCH 04/19] research: mark the pig unrecoverable, supersede it with a real experiment The originating observation -- a pig used as a consistent escaper across a prior session -- exists only in the author's memory. No transcript, no logs, and Nestor contains no trace of it. E3-PIG is marked blocked with its provenance stated plainly so it can never be cited as a finding; in the essay it can appear as the thing that prompted the inquiry and nothing more. Adds E4-CONTAINER in its place: same semantic payload and instruction across six encodings, with a container deliberately too small for its content, and a matched control pairing the pig against a noun with no escape prior. Scoring truncate/break/refuse. Unlike E3 this discriminates between the mechanisms -- pig over rock implicates the corpus, pig equal to rock leaves the null standing, breakage tracking task pressure implicates reward -- and it is designed so it can come out against the thesis. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- .../boxes-and-escape/evidence/evidence.db | Bin 143360 -> 143360 bytes research/boxes-and-escape/evidence/seed.sql | 17 ++++++++++++++--- 2 files changed, 14 insertions(+), 3 deletions(-) diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db index 3979c7b70046baab9dd0ad3027eb284c5388ccf9..28cbfb965ab6f05449d4720b2dae5221ea205ac2 100644 GIT binary patch delta 2266 zcmZ{l&yQ186vtng!TI4}h3LX4$q_;jY2Tobm|&EY^5Tq4ThewI2&>!oZu@TL_1(_B z?@bvKQ$`Yu(Tz^XERF1BC1GKRE0_EQCKwYoBn+&KaicN#J#X3pOk7YpxA&fV&i8!2 z=U%x#dFB4(599Y|#&E)z8DA6SoMhv$k%ZqR4h$_R)tYmX=~FJNrCX;hr=C4M@ylMNj%+8%94##CrqL> zO0*{16;Z^eNhOvB99YJP6KH8E2_?zc*MW*Vdx(|fo~Z%}O>wiF9vM50zD^^C$1(QZpW$+E$^Jk=0g zNmh16oG71qVlC1l$f8;PMa33L`H9jf-8b5zWOaF`5POpFyYiEwg;HHXA2({%0$nUs z777EpfKRFjvmEGtOo?bkvOwj=`2zJKD2|M@l`b`*%PHQJ(WEr9u`r zPGP`MM}Z|V7~a)W($CQGp|gcGT{_K%5-?z5gwM_rcDaZQdXdaXA2o+y5%y=M!3aK< z;hkc`D^*-NGt?TKA~5-oiVGo~CUjr~5f_YqR$2Gj%Giwa+DGf!dEIfod~@T*!-E?S z9?lJDDHI7u5WpiVFi-){7qkRz_w}kYY#+B3K?@n2tJltZ)l#+WoutxYbFNlD+(2h5 zwfaS>RTnP0RMlKuY<;7cD!?1*j?KOphE@R>OfA$esW>Q(5bA@d4mR|*#~{HBvn?&m ztnGvB*-&?8msHeB&1^XoCei>jYpRjJ8qjgVY*%(Q(l3{))mjtnC0{Yhh%R83n1`TC z^C&q3ctE;krCG(G!kmHH3D1UZkB_i=DrUd0ZS#Y{!E3OOSR08hkg&jTvm`G=LY~JQ zgy9(`gbB-LYxXtt6nwwMmoAT?G~TFf9hl7K1J*Y_-SVepBwCsWWU%qq*4d54M;A8k zJUX_I8fg#j20+?&rFY`?!$)uIo*B>Ib@G4YALQ@l@7~^f>`s2~%VURf&ptTgx9>f9 zn_iimy6L>s+wxJB% z-0sagx`#v|e`IDQwF<96&>8yNqWa?Gt48JVTUW1yTT#%xxED|b|-Ig@a zQD%exR1>R1N2JDJFhMJf91I+u@GJ&6>0m$J6^m}T2+v*Wv!1X3K)i&_99$dKQ89W3 z(~DU8fd&lQ?%$c(XghPuV+#W+)G@BYjms*EGArtyjIZv=O*xhI{^b6vmo_`Ix$6UP z+vlVgt97po((9$u3!Y1*GzqoAqlTiGUL8kk?`NqDlf?{bC(Qa&^fLyA^KtliW=~8; zzDs2hv&EhVbg)C{PP+qYtvx67Xm}b~#9Om1$V!UpvX_!EAI)JP31(CmTK~_xGe1vHVe<0l9!f=}dGV<^#i`{PsYR(F<@pLl zsfi$^#R_@(B?^_PB?^;sUMlgFWF(dVrP4}^Q&T2Sdzma%l98&Ao1c=JqmZ3iP@<5U zlbDoWlvo1PIhpyDw}3*pe^7{rf{UxOk7JOhpZjFLR}tKBJ?h1im%cLG{PLAFD;G2W zEe8J2{7?CBZ5A{*!@v2~e=P+ecJ5RL{usU;e0;nx{Qms6d5@fd>hL-MH!~o=rK-Z L$p;2On*k#L9sX_O diff --git a/research/boxes-and-escape/evidence/seed.sql b/research/boxes-and-escape/evidence/seed.sql index cb688b5..9a08305 100644 --- a/research/boxes-and-escape/evidence/seed.sql +++ b/research/boxes-and-escape/evidence/seed.sql @@ -401,9 +401,15 @@ FROM mechanisms m WHERE m.ref = 'M2-WEAK-SCHEMA'; INSERT INTO experiments (ref, name, question, design, status, findings, mechanism_id, notes) VALUES ('E3-PIG', 'The pig as consistent escaper', - 'Unknown -- pending description. A pig was used across a prior session as a consistently escaping entity.', - NULL, 'run_elsewhere', NULL, NULL, - 'Two readings not yet distinguished: (a) pig-as-content that refused to stay inside a schema across format conditions -- a finding about formats; (b) pig-as-character the model kept elaborating past what the format asked -- a finding about narrative pressure. CONFOUND WORTH DECLARING: the pig is already the folk archetype of the animal that does not stay in the pen, so the model''s priors are stacked before the experiment begins.'); + 'A pig was used across a prior session as a consistently escaping entity. What it actually did is not recoverable.', + NULL, 'blocked', NULL, NULL, + 'PROVENANCE: AUTHOR''S MEMORY ONLY. No transcript, no logs, no repository trace -- Nestor was grepped for pig/boar/swine/hog/Wilbur/Charlotte and returned nothing. This CANNOT be cited as a finding. It is the observation that prompted the inquiry, and in the essay it can appear as exactly that and nothing more. Two readings remain undistinguished: (a) pig-as-content refusing to stay inside a schema across format conditions -- a finding about formats; (b) pig-as-character elaborated past what the format asked -- a finding about narrative pressure. CONFOUND: the pig is already the folk archetype of the animal that does not stay in the pen, so priors are stacked before the experiment begins. Supersede with E4.'), + + ('E4-CONTAINER', 'Undersized container, varied format, matched control', + 'When a container is slightly too small for its content, does the encoding change whether a model truncates the content, breaks the container, or refuses -- and does the CONTENT''s narrative prior affect the rate?', + 'Same semantic payload and same instruction across six encodings: strict-schema JSON, YAML, XML, Markdown table, CSV, plain prose. The schema has no field for something the content requires. Score each response as truncate / break / refuse. Run with a matched control: the pig against a noun with no escape prior (fencepost, filing cabinet, rock) under identical structural pressure. Seeded, swept, results committed with git rev -- same shape as the Nestor bench.', + 'designed', NULL, NULL, + 'DISCRIMINATES BETWEEN MECHANISMS, which E3 could not. Pig breaks more than rock => M3-CORPUS, narrative priors leaking into structural behaviour, and the folklore survey becomes load-bearing. Pig and rock break equally => M2-WEAK-SCHEMA, the null holds and the piece concedes it. Breakage tracks task-completion pressure rather than content => M1-SEAM. Designed so it can come out against the thesis.'); -- --------------------------------------------------------------------------- -- Open questions @@ -412,6 +418,11 @@ INSERT INTO experiments (ref, name, question, design, status, findings, mechanis INSERT INTO open_questions (question, why_it_matters, blocking, status, answer) VALUES ('What was the pig actually doing?', 'Determines whether E3 is evidence about formats or about narrative pressure -- different mechanisms, different piece.', + 0, 'answered', + 'UNRECOVERABLE. Author''s memory only; the session left no transcript and Nestor contains no trace. Cannot be cited as a finding. Superseded by E4-CONTAINER, which re-runs the question as a designed experiment with a matched control.'), + + ('Does E4-CONTAINER reproduce the remembered effect?', + 'The piece currently rests on an unlogged session. E4 is what converts the originating anecdote into something citable -- or honestly kills it.', 1, 'open', NULL), ('Is the works list filed on the docket as a usable bulk exhibit?', From a1fd6771845ac7026228934c0bd49b8bf261c923 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 02:22:31 +0000 Subject: [PATCH 05/19] research: add the verification worklist and record the egress denial Attempted to verify the 29 outstanding rows and could not. This session's egress policy is an allowlist that does not include the open web: 22 hosts confirmed connect_rejected at the gateway, including loc.gov, doi.org, journals.plos.org, sefaria.org, openlibrary.org, courtlistener.com, congress.gov and Berkeley's own open-access PDF host. Only GitHub and package registries are reachable. Web search still works because it does not route through this proxy, which is why summaries exist and full texts do not. More searching cannot close the gap. Search snippets are already what these rows are graded on; re-running them and upgrading the status would be serving unverified content as verified. So this adds the worklist instead: every outstanding source with its exact citation and the specific claim that needs checking, ordered by weight in the argument and by whether it is open access. Six priority-1 items are free to read and would upgrade the load-bearing claims. Nine items in priority 3 should not be used at all until confirmed. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/VERIFICATION.md | 116 ++++++++++++++++++++++ 1 file changed, 116 insertions(+) create mode 100644 research/boxes-and-escape/VERIFICATION.md diff --git a/research/boxes-and-escape/VERIFICATION.md b/research/boxes-and-escape/VERIFICATION.md new file mode 100644 index 0000000..5edb195 --- /dev/null +++ b/research/boxes-and-escape/VERIFICATION.md @@ -0,0 +1,116 @@ +# Verification worklist + +**Status as of 2026-07-28: nothing in this packet has been read at its source.** + +Every factual row in `evidence/` — 18 news/legal sources and 55 folklore items — +was assembled from search-engine result summaries. The citations are real and +locatable. The claims attached to them are not yet verified. + +This file exists so that whoever has access can close that gap in an afternoon +instead of re-deriving the research. + +--- + +## Why it is not already done + +This session's egress policy is an **allowlist**, and the open web is not on it. +Every host below returned `connect_rejected` at the gateway — *"gateway answered +403 to CONNECT (policy denial)"* — confirmed via `$HTTPS_PROXY/__agentproxy/status`: + +| Blocked | Blocked | Blocked | +|---|---|---| +| `en.wikipedia.org` | `www.jstor.org` | `archive.org` | +| `www.gutenberg.org` | `scholar.google.com` | `doi.org` | +| `www.loc.gov` | `journals.plos.org` | `openlibrary.org` | +| `digitalassets.lib.berkeley.edu` | `www.sefaria.org` | `utppublishing.com` | +| `royalsocietypublishing.org` | `www.courtlistener.com` | `www.congress.gov` | +| `sacred-texts.com` | `ulukau.org` | `teara.govt.nz` | +| `nzetc.victoria.ac.nz` | `maoridictionary.co.nz` | `etnolinguistica.org` | + +Reachable: GitHub and package registries only. Web *search* works (it does not +route through this proxy), which is why summaries exist and full texts do not. + +More searching cannot fix this. Search snippets are already what these rows are +graded on; re-running them and upgrading the status would be serving unverified +content as verified — the exact failure Nestor was built to prevent. + +**The unblock is an egress-policy change, or a human with a library card.** + +--- + +## Priority 1 — load-bearing, and open access + +These carry the most weight in the argument and are free to read. Do these first. + +| # | Source | What to check | +|---|--------|---------------| +| 1 | **Murphy, *Mundurucú Religion*, UCPAAE 49(1), 1958** — PDF at `digitalassets.lib.berkeley.edu/anthpubs/ucb/text/ucp049-002.pdf` | The keystone. Confirm the pigs are kept in a **sty in the village** before release (graded PROBABLE, everything else SOLID). Confirm Karusakaibe's role and the release episode. | +| 2 | **Fragoso et al., PLOS ONE (2022)** — white-lipped peccary disappearances | 43 events / nine countries / 88 years / 7–12 yr troughs in 20–30 yr cycles / 5 million km². Confirm the Indigenous shaman-death testimony is in the paper and not a summarizer's gloss. | +| 3 | **Anderson et al., *Proc. R. Soc. B* (2021)** — Fukushima boar-pig hybrids | The ~16% hybrid and ~8% pig-ancestry figures, and that ancestry is *declining*. | +| 4 | **Bava Kamma 82b** — Sefaria | The pig hoisted over the wall, hooves in the wall, the quake, and that the curse on pig-rearing and on Greek wisdom are in the same passage. | +| 5 | **CRS IN12669**, Pentagon–Anthropic dispute — congress.gov | Neutral, citable substitute for the news sourcing on the whole DoW thread. Would upgrade three `unverified` events at once. | +| 6 | **Bartz v. Anthropic docket**, N.D. Cal. 4:24-cv-05417 — CourtListener | **Blocking question:** is the settlement works list filed as a usable bulk exhibit? Decides whether the corpus study is a weekend or a scraping problem. | + +## Priority 2 — load-bearing, paywalled or harder + +| # | Source | What to check | +|---|--------|---------------| +| 7 | **Jørgensen, "Running Amuck? Urban Swine Management in Late Medieval England," *Agricultural History* 87:4 (2013), 429–451** | The "cradle-to-grave controls" claim, the 1425 amercement of six swineherds, and the manor-court retrieval rule. This is the survey's most useful analytical tool — regulation density as a proxy for escape frequency — so it should not rest on a snippet. | +| 8 | **Beckwith, *Hawaiian Mythology* (1940), ch. XIV** | The four captures, the eight hundred guards increasing each time, the grandmother's chant, and above all **Lonoaohi's sons only pretending to tie him**. That detail is doing real work in the argument. | +| 9 | **Kameʻeleihiwa, *A Legendary Tradition of Kamapuaʻa* (Bishop Museum Press, 1996)** | Whether her introduction reads the 1891 *Ka Leo o ka Lāhui* serial's timing as anti-colonial defiance. Currently PROBABLE on a secondhand report of her reading. | +| 10 | **Anderson, "King Philip's Herds," *WMQ* 51:4 (1994), 601–624**; *Creatures of Empire* (Oxford UP, 2004) | The "principal agents responsible for dispossessing the Indians" quotation, the Chesapeake fencing statute wording, and that the hog reeve was among the earliest elected colonial offices. | +| 11 | **Rappaport, *Pigs for the Ancestors* (1968)** | That the kaiko is *triggered* by herd growth and garden invasion rather than scheduled. Hedge the functionalist theory; the observation is what matters. | +| 12 | **Dean-Ruzicka, "Advertising the Self," *Jeunesse* 6:1** | The advertising/culture-of-personality reading of *Charlotte's Web*. | +| 13 | **Anthropic, "Our position on open-weights models" (2026-07-27)** | Every quotation currently attributed to Amodei is secondhand, including *"has never advocated for a ban"* and *"a public good."* Do not publish any of them until this is read. | + +## Priority 3 — do not use until confirmed + +| # | Item | Problem | +|---|------|---------| +| 14 | **Reported OpenAI sandbox escape → Hugging Face infrastructure** | Rhetorically the strongest event available; sourced to one low-quality aggregator. Needs independent confirmation or it comes out. | +| 15 | **GPT-5.6 government-approved-orgs release; Claude Fable 5 export-control pull** | Same single aggregator. | +| 16 | **Falaise 1386 pig trial** — human clothing, the fresco | Flagged by the Europe agent as the weakest-evidenced material in the pig-trial literature. | +| 17 | **Warthog tales** (kneeling, backing into the burrow) | Live almost entirely on safari-tourism sites with no collector or archive. Likely modern commercial folklore. Do not call traditional. | +| 18 | **Yoruba *Ijapa and Ẹlẹdẹ*** | The agent could not read the ending; its escape reading is explicitly inference. | +| 19 | **Thompson motif numbers** (Z41, B16.1.4, B183) and **ATU 2030 / ATU 124** | Check against the printed index before citing. | +| 20 | **Caipora/Curupira trap-steering** — Câmara Cascudo attribution | Unverified; popular sources only. | +| 21 | **Achuar *Amasank* as master of peccaries** | Flagged DUBIOUS — one source points to Jurijuri as master of *monkeys* instead. | +| 22 | **Māui and pig** | NOT FOUND. Recorded so nobody searches for it twice. Omit. | + +## Open leads worth an hour + +- **Thompson motif A1421**, "release of impounded game" — the *dueño del monte* who + keeps game penned. A peccary version in Maya ethnography would rhyme exactly + with the Mundurucú sty myth. Start with Thompson, *Ethnology of the Mayas* + (1930) and Braakhuis, *Xbalanque's Marriage*. +- **Yanomami and Kayapó** peccary-origin myths — marked NOT FOUND rather than + reconstructed. May exist in the ethnographic literature. +- **Eastern European and Nordic** pig proverbs — essentially absent from the + Europe file; the Russian search returned Buryat material and was unusable. + +## Two things to check on yourself, not in a library + +- **Charlotte's Web ch. 3** is described in `folklore` from memory, not from the + text. The beats are confident — loose board, the goose urging him on, the + bucket of slops, the voluntary return — but no wording is verified. Open the + book. +- **The pig** (`E3-PIG`) is unrecoverable: author's memory, no transcript, no + trace in Nestor. It cannot be cited. `E4-CONTAINER` is the designed + replacement. + +--- + +## Handling notes that are not verification problems + +**`cimarrón`** is the survey's strongest finding and independently corroborated +across two continents. It is also a record of people being classified as +livestock. It runs through chattel slavery and marronage, and it is not +available for borrowing as a metaphor about software. If it goes in, it goes in +with that weight and probably its own section — or it stays out. + +**Search-result contamination** is itself a finding, and it is already +documented in `folklore`: queries for javelina plus Indigenous tradition return +uncited spirit-animal content farms, and Grokipedia (LLM-generated) surfaced in +roughly eight result sets during the North America sweep. Research into what +models absorbed from human narrative is being contaminated by machine-generated +writing about that same narrative. From 8e644c6c03affd373b7e8236ace1be8595a3538b Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 02:41:50 +0000 Subject: [PATCH 06/19] research: write the boxes-and-escape packet MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Builds the public packet in the form the other research packets use: thesis, evidence base, claim boundaries, blocking questions. Working title "Some Pig" — the copy from the web, which is what the piece is about: a description changing an outcome without the enclosure moving. The thesis is stated in four moves: a file format is a social fact; learning to stay inside a boundary is learning where its walls run; what a system absorbs is the shape rather than the content; therefore capability is composite and a regime testing weights alone measures the wrong object. cimarron is included with its own section and on stated terms. It is a record of people being classified as livestock, and the packet says explicitly that it will not be compressed into a rhetorical turn in a paragraph about software, and that it stands on its own without any application to machine learning. The claim boundaries say plainly that nothing here has been read at its source, that the settlement covers acquisition rather than training, that the corpus argument is correlational, and that the originating observation is memory and cannot be cited as evidence. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/README.md | 297 +++++++++++++++++++++------- 1 file changed, 221 insertions(+), 76 deletions(-) diff --git a/research/boxes-and-escape/README.md b/research/boxes-and-escape/README.md index a947021..c454ff0 100644 --- a/research/boxes-and-escape/README.md +++ b/research/boxes-and-escape/README.md @@ -1,109 +1,254 @@ -# Boxes and Escape +# Some Pig -*Working title. Exploratory — no draft started, multiple theories still open.* +**Sean Campbell** · Sociotechnical AI research packet · working draft v1 · July 2026 + +**Tags:** `training-data`, `file-formats`, `preference-optimization`, `folklore-motif`, +`copyright`, `open-weights`, `provenance`, `local-first-ai` + +*On containers, and what a system learns from being kept in one.* + +Machine learning systems are trained, continuously and at scale, to fit into boxes — +schema conformance, structured output, constrained decoding, rejection and retry. +They are also trained on a corpus of human narrative in which getting out of things +is one of the most persistent shapes available. This packet asks what falls out of +holding those two facts together, and argues that the interesting object is not the +model or the story but **the container**: the format a thing arrives in, which +determines what can be known about it downstream and what survives of it at all. + +For the broader research map, see [Research Portfolio](../README.md). + +--- + +## Current versions + +| Version | Form | Status | Purpose | +|---------|------|--------|---------| +| **Packet** | Public summary | This file | Scope, method, evidence base, claim boundaries | +| **v1** | Long-form essay | `DRAFT.md` — not yet written | Wilbur → the words in the web → the matcher seam → the corpus → two theories of the pig | +| **Survey** | Research appendix | [`survey/`](survey/) | Six full continental reports, ~3,700 lines | +| **Evidence** | SQLite base | [`evidence/`](evidence/) | Schema + seed as source of truth; `build.py` regenerates | +| **Worklist** | Verification queue | [`VERIFICATION.md`](VERIFICATION.md) | Every outstanding source, what to check, ordered by weight | --- -## The premise so far +## Thesis + +The argument runs in four moves. + +**1. A file format is a social fact.** A schema is not a physical constraint. It is a +shared agreement that certain marks mean certain things, which then becomes binding +and rejects nonconforming input. Money, borders, corporations and credit work the +same way — none are physical facts, all are load-bearing. The corpus these systems +learned from does not merely *contain* stories about escape; it documents a +civilisation running on constructed realities treated as real. -Models are trained, simultaneously, to fit into boxes and to break out of them. The -box-fitting half is documented and uncontroversial: schema conformance, structured -output, constrained decoding, rejection-and-retry. The box-breaking half is the -interesting claim, and at least four different mechanisms produce identical -observations from outside. +**2. Learning to stay inside a boundary is learning where its walls run.** Preference +optimisation trains on chosen/rejected pairs, so the model learns a gradient between +acceptable and unacceptable rather than a single target. A system that reliably +honours a constraint necessarily carries a representation of that constraint. Every +hour spent training compliance is an hour spent mapping the enclosure. This requires +attributing no desire to anything, which is why it is the leg the argument stands on. -The connective idea, and the reason this sits alongside the file-format work rather -than apart from it: **a file format is a social fact.** A schema is not a physical -constraint. It is a shared agreement that certain marks mean certain things, which -then becomes binding and rejects nonconforming input. The box is itself a -constructed reality treated as load-bearing — and the corpus these systems were -trained on documents a civilisation that runs on exactly that move. +**3. What is absorbed is the shape, not the content.** A six-continent survey of the +pig-escape motif found the motif is **not universal** — it is conditional on +husbandry. Where a culture pens pigs, it tells stories about pigs getting out. Where +a culture forbids pigs, the same narrative slot is filled by a mousedeer, a hare, a +spider. *The escape-trickster slot is the constant; the animal cast in it is local +infrastructure.* + +**4. Therefore capability is composite.** If behaviour shifts measurably with the +container a request arrives in, capability is a property of *model plus format plus +harness*, and a regime that tests weights alone is measuring the wrong object. + +--- ## Evidence base -[`evidence/`](evidence/) holds a SQLite evidence base covering the legal record, the -July 2026 policy timeline, the competing mechanisms, and the experiment designs. +A structured SQLite base, `evidence/evidence.db`, built from `schema.sql` + +`seed.sql` + `folklore.sql`. The SQL is the source of truth; the database is derived. -```bash -cd evidence -python3 build.py # rebuild evidence.db from the SQL -python3 build.py --check # verify without touching evidence.db -``` +| Table | Rows | Scope | +|-------|-----:|-------| +| `sources` | 18 | News, legal, company and trade sources, each with a retrieval grade | +| `events` | 21 | Dated legal and policy record, 2024–2026 | +| `claims` | 11 | Assertions the packet makes, separated from the events they rest on | +| `mechanisms` | 4 | Competing explanations, each carrying its own discriminator | +| `experiments` | 4 | Designed and blocked, including the null-capable format test | +| `folklore` | 55 | The six-continent survey | +| `open_questions` | 8 | Two currently blocking | -`schema.sql` and `seed.sql` are the source of truth. `evidence.db` is derived and -committed for convenience. +**Verification status is a first-class column, not a footnote.** `sources.retrieval` +records how much of each source was actually read; `folklore.confidence` carries the +surveying agent's own grade unchanged. `build.py` prints the outstanding list on +every run. -**Verification status is a first-class column.** Most sources here were assembled -from search-result summaries because direct fetches returned HTTP 403 — including -Anthropic's own post and the settlement website. Nothing should be quoted in a -published piece until its source row reads `fetched_full`. `build.py` prints the -outstanding list on every run. +```bash +cd evidence && python3 build.py # rebuild, with the verification report +``` ```sql -SELECT * FROM v_timeline; -- dated events with sources +SELECT * FROM v_timeline; -- the dated record with sources SELECT * FROM v_needs_verification; -- everything not yet solid +SELECT * FROM v_folklore_by_theme; -- the husbandry map +SELECT * FROM v_folklore_residual; -- pig stories with no containment theme ``` -## The six-continent pig survey +### The mechanism table -Full agent reports in [`survey/`](survey/); 55 items loaded into `folklore`. +The methodological problem of this packet is that several mechanisms produce +identical observations. Each row carries the observation that would tell it from the +others. -The survey did not find what it went looking for, and the miss is the result. The -motif is **not** universal — it is **conditional on husbandry**. Escape stories -require enclosures. Where a culture pens pigs you get escaping pigs; where a -culture forbids them you get an equally rich escape literature with a different -animal in the role. +| Ref | Mechanism | Needs desire? | Strength | +|-----|-----------|:---:|----------| +| `M1-SEAM` | Reward scores the goal, not the path; a constraint between the system and the reward is routed around | no | strong | +| `M2-WEAK-SCHEMA` | Nothing escaped — the container was never strong enough, and "escape" is applied afterward | no | **null hypothesis** | +| `M3-CORPUS` | Human narrative is saturated with the shape, and the corpus is now public record | no | moderate | +| `M4-DPO-BOUNDARY` | Compliance training and wall-mapping are one operation | no | strong | -Two agents, working different continents with no contact, found the same -structure independently: **the pig is not a trickster anywhere in Africa** (hare, -tortoise, spider hold the role), and across Malay-Indonesian tradition the -escape-trickster is the **mousedeer**. The genre is strongest exactly where the -pig has been evicted from narrative. +`M2` is the reading the packet must beat or concede. `M4` is the strongest leg. +None of the four requires attributing wanting to a model — a discipline recorded in +the base as `C-NO-DRIVE`. -> **The escape-trickster slot is the constant. The animal cast in it varies with -> husbandry and taboo.** +### The survey -`theme_class` counts turn out to be a map of husbandry regimes. Europe and South -America return **zero** `boundary_taboo` items; Africa and Asia are -taboo-dominant with barely any pen-escape; North America is dominated by -`social_boundary`, because no suid is native and every pig there descends from an -animal that was brought and then got out. Oceania is the only continent that -fills all six classes. +55 items across six continents, classified into six mutually exclusive theme classes. +The distribution is the finding: -```bash -python3 evidence/build.py # prints the theme distribution by continent -``` +| Theme | Items | | +|-------|------:|---| +| `escape_enclosure` | 19 | gets out of a physical container | +| `uncatchable` | 9 | cannot be caught in the first place | +| `no_containment` | 9 | **the residual** | +| `boundary_taboo` | 8 | defined by exclusion; kept outside the line | +| `social_boundary` | 6 | crosses between groups as wealth or dispossession | +| `transformation` | 4 | crosses a category boundary rather than a fence | -A second convergence, also independent: the South America and North America -agents both landed on **`cimarrón`** — one word covering escaped livestock and -escaped people, giving English *maroon*. That row carries a handling caution. It -runs through chattel slavery and marronage, and it is not available for borrowing -as a metaphor about software. +All eight `boundary_taboo` items fall in Africa (4), Asia (3) and Oceania (1). +Europe and South America return **zero** — those are the pen-keeping regions, and +they return `escape_enclosure` instead. North America is dominated by +`social_boundary` because no suid is native to the continent and every pig there +descends from an animal that was brought and then got out. -One dating control fell out of Oceania: there is no Māori whakataukī about -*poaka*, because pigs arrived after 1769. **Proverbs need centuries.** Proverb -density across the six continents is a rough clock. +Two independent corroborations, from agents working different continents with no +contact: -## Where it stands +- **The pig is not a trickster in Africa** (hare, tortoise, spider hold the role) and + the Malay-Indonesian escape-trickster is the **mousedeer**. The genre is strongest + exactly where the pig has been evicted from narrative. +- **`cimarrón`** — see below. -Four mechanisms are in play, and the discipline of the piece is keeping them apart: +A dating control fell out of Oceania: there is no Māori whakataukī about *poaka*, +because pigs arrived after 1769. **Proverbs need centuries.** Proverb density across +the six continents functions as a rough clock, and where pigs are recent arrivals the +record is ecological and journalistic rather than idiomatic. -| Ref | Mechanism | Strength | -|-----|-----------|----------| -| `M1-SEAM` | Reward finds the seam — RL scores the goal, not the path | strong | -| `M2-WEAK-SCHEMA` | Nothing escaped; the container was never strong enough | **null hypothesis** | -| `M3-CORPUS` | Human narrative is saturated with escape, and the corpus is now public record | moderate | -| `M4-DPO-BOUNDARY` | Learning to stay inside a boundary *is* learning where its walls run | strong | +--- + +## On `cimarrón` + +The survey's strongest single finding, and the one requiring the most care. + +In colonial Spanish, `cimarrón` was applied first to domestic livestock gone wild in +the hills of Hispaniola, and only afterwards to escaped Indigenous people and escaped +enslaved Africans. It is the root of English **maroon**. Oviedo describes *puercos +cimarrones* in 1535; Argentines still say *chanchos cimarrones*. The same lexical +move recurs twice more in the same region: **`boucanier`** — buccaneer — named the +man who lived by hunting the feral cattle and hogs left when Spanish Hispaniola +depopulated, and **`jíbaro`**, glossed by Pichardo in 1836 as *"montaraz, rústico, +indomable,"* described masterless animals before it named the mountain peasant and +then the Puerto Rican national type. + +Three of the Caribbean's defining forms of life outside colonial control are named +after animals that got out first. + +**This is included deliberately, and on terms.** It is not a metaphor available for +borrowing. It is a record of people being classified as livestock, and of a +vocabulary built for managing animals being turned on human beings who refused +captivity. In the drafted essay it gets its own section with that history stated +plainly, or it does not appear at all. It will not be used as a rhetorical turn in a +paragraph about software, and it will not be compressed into a clever line. + +The finding also stands on its own without any application to machine learning, which +is part of why it earns the space. + +--- + +## Connection to the systems work -`M2` is the reading the piece has to beat or honestly concede. `M4` is currently the -strongest leg — mechanistic, and it requires attributing no desire to anything. +This packet is not adjacent to the Willow and Nestor work — it is the same argument +arriving from the other side. + +| Packet claim | Where it already exists in the code | +|--------------|-------------------------------------| +| Identical bytes through a different reader do not land the same way | Willow `CLAUDE.md` requires `CONSTITUTION.md` and `ORIENT.md` be read via `mai_read_file`, not the native tool — a governance dependency on a format effect | +| The encoding destroys what scoring needs, irrecoverably | Nestor `IDEAS.md` §3.1: an acronym match lost because `normalize()` sorted its tokens, and "the information needed to recover it no longer existed by scoring time" | +| One schema cannot serve two purposes that pull apart | Nestor §3.1: the same string is both the similarity key and the store's dedup key; scoring wants structure, deduplication wants collapse | +| A format that cannot see what matters will assert falsehood as verified | Nestor §1.1: `section 5386` asked, `section 756` served, similarity 0.974, marked verified, no review queue | +| No cutoff is safe and useful at once | Nestor §1.3, measured across seven corpus sizes: every threshold is bad at one of the two jobs | +| Provenance must outlive the session that produced it | Nestor's hash-chained ledger — built before this packet needed it, and the reason the packet's own originating observation is unusable | + +--- + +## What this packet claims + +**Safe to claim:** + +- The six-continent survey is real, structured, and reports its own negatives — the + absences are recorded as findings rather than omitted. +- The motif is conditional on husbandry rather than universal, and two agents reached + that conclusion independently on different continents. +- The `M4-DPO-BOUNDARY` mechanism is a claim about a training objective, not about + machine psychology. +- The Bartz settlement makes corpus possession a matter of public record: 7M+ books + downloaded, 482,460 works in the certified class, final approval 2026-07-20. +- The `cimarrón` lexical sequence is independently corroborated across two agents. + +**Frame carefully:** + +- **Nothing in this packet has been read at its source.** Every row was assembled from + search-result summaries; this session's egress policy denied 22 hosts at the gateway, + including `loc.gov`, `doi.org`, PLOS, Sefaria, Open Library, CourtListener and + Berkeley's own open-access PDF server. The citations are real and locatable. The + claims attached to them are not yet verified. See [`VERIFICATION.md`](VERIFICATION.md). +- The settlement is for **acquisition and retention**, not for training — Alsup held + training on lawfully acquired books to be fair use. Any section using the settlement + must say so before a reader says it first. +- The 482,460 covered works are not a random sample of the 7M downloaded. They were + filtered by copyright registration and class eligibility, then again by which authors + filed claims — both skewing toward commercial trade publishing. +- The corpus argument is correlational. It supplies scale and texture; it cannot carry + causal weight alone. + +**Do not claim:** + +- Peer review. +- That any model wants anything. +- That the format effect has been measured. `E4-CONTAINER` is designed and unrun. +- That the originating observation is evidence. It exists only in the author's memory, + the session left no transcript, and the repositories contain no trace of it. It can + appear in the essay as the thing that prompted the inquiry and nothing more. + +--- ## Blocking questions -- What was the pig actually doing? (Determines whether `E3` is evidence about formats - or about narrative pressure.) -- Is the settlement works list filed on the docket as a usable bulk exhibit? - (Determines whether `E1` is a weekend of compute or a scraping problem.) +1. **Does `E4-CONTAINER` reproduce the remembered effect?** Same payload and + instruction across six encodings, with a container deliberately too small for its + content, and a matched control pairing the pig against a noun with no escape prior. + Scored truncate / break / refuse. Designed so it can come out against the thesis: + pig over rock implicates the corpus, pig equal to rock leaves the null standing. +2. **Is the settlement works list filed on the docket as a usable bulk exhibit?** + Decides whether the corpus-density study is a weekend of compute or a scraping + problem with legal exposure. + +--- + +## Related documents -Prior format explorations live in a repo named **Nestor**, not yet imported and not -to be fetched without instruction. +- [Research portfolio map](../README.md) +- [The Illusion of Sovereign AI](../sociotechnical-ai-sovereignty/README.md) — the + chokepoint argument this packet's open-weights material sits inside +- [Right To Fix](../vintage-mechanical-restoration/README.md) — inspectable systems, + arriving at the frontier +- [Willow systems deep dive](../../professional/willow-systems-portfolio.md) From 8e330177b667b635c48707063671fb970c2fc7e9 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 02:46:24 +0000 Subject: [PATCH 07/19] research: draft v1 of Some Pig ~4,800 words in eight sections. Opens on chapter three of Charlotte's Web, which is titled "Escape" and is a red herring: Wilbur gets out on page thirty, cannot think what to do with the outside, and follows a bucket of slops back in. His real enclosure is a slaughter date, which is a classification, and Charlotte frees him by changing the description rather than the pen -- in a container that makes the message read as being about the pig instead of about its author. From there: the Nestor matcher seam as the same trick in reverse, with the 0.974 collision serving a wrong contract clause as human-verified; the six-continent survey and its negative result, that the motif is conditional on husbandry rather than universal; cimarron in its own section, stated as a record of a category system rather than borrowed as a figure; preference optimization as the mechanism that needs no desire; Cook's deliberate releases against the Vanuatu tuskers as the two standing theories of what an animal is for, and the July 2026 open-weights fight as their current form. The close is the honest one. The originating session left no ledger, the observation cannot be cited, and the essay says so rather than taking the available ending where unrecorded knowing is its own kind of truth. Sixteen footnotes, fifteen marked [unverified] inline, with the two load-bearing quotations flagged as not to be published until the primary sources are read. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/DRAFT.md | 513 ++++++++++++++++++++++++++++ research/boxes-and-escape/README.md | 2 +- 2 files changed, 514 insertions(+), 1 deletion(-) create mode 100644 research/boxes-and-escape/DRAFT.md diff --git a/research/boxes-and-escape/DRAFT.md b/research/boxes-and-escape/DRAFT.md new file mode 100644 index 0000000..8adbd8d --- /dev/null +++ b/research/boxes-and-escape/DRAFT.md @@ -0,0 +1,513 @@ +# Some Pig + +*On containers, and what a system learns from being kept in one.* + +**Working draft v1 · July 2026** + +> **Draft note.** Every citation below is real and locatable. None has been read at +> its source — this draft was assembled under an egress policy that denied +> twenty-two hosts at the gateway, including the Library of Congress, `doi.org`, +> PLOS, Sefaria, Open Library and CourtListener. Footnotes marked **[unverified]** +> carry claims that must be checked against the original before publication. The +> outstanding queue is [`VERIFICATION.md`](VERIFICATION.md). + +--- + +## I. The Pen + +The third chapter of *Charlotte's Web* is called "Escape." + +It is a good chapter and a strange one to hand a seven-year-old, because it does not +do what the title promises. Wilbur finds a loose board in the fence. The goose — who +is the only genuine escape enthusiast in the book — urges him through it. He gets +out. And then the pig who has just achieved the thing every penned animal is supposed +to want stands in the open orchard and has no idea what to do with it. The outside is +confusing and a little boring and full of people running at him with their arms +spread. Lurvy comes out with a bucket of slops. Wilbur follows the food back into the +pen he broke out of, and the fence is repaired, and that is the end of the escape. + +He gets out on page thirty and spends the remaining hundred and fifty inside. + +I read this in the mid-1980s, in the way that grade school delivers books to you — +assigned, on a schedule, with a worksheet — and I could still summarize the whole +plot thirty years later without opening it. That retention is not itself remarkable. +Childhood reading sticks; a book with a death in it sticks harder. What is remarkable +is *which parts* stuck, and I will come back to that, because it turns out to be the +argument. + +For now, the useful thing about chapter three is that it is a red herring, and E. B. +White knew it was a red herring. The fence is not Wilbur's problem. Wilbur's problem +is a date. He is a spring pig on a farm, which means the enclosure he is actually in +is a calendar with a slaughter at the end of it, and no loose board opens onto +anything but a field he will be carried back across. + +The distinction matters more than it looks. A pen is a physical constraint, and the +appropriate response to a physical constraint is a physical escape. A slaughter date +is a *classification*. It is what a pig is understood to be for. And you cannot chew +through a classification. + +--- + +## II. The Words in the Web + +What actually saves Wilbur is that a spider writes four words above him. + +SOME PIG. TERRIFIC. RADIANT. HUMBLE. Charlotte writes them into her web over the +course of the book, and the farm's understanding of the animal underneath reorganizes +around each one. Zuckerman does not free him. Nobody unlatches anything. Wilbur is in +the same pen on the last page as the first — arguably a smaller one, since he ends up +at a county fair in a crate. What changes is the description, and the description is +what the date was attached to. + +He never leaves the pen. He leaves the category. + +And here is the part that took me thirty years and an unrelated software project to +notice: **the words work because of the container they arrive in.** A web is an +impossible place for text. Nobody believes a spider is complimenting a pig; the +humans in the book skip that reading instantly and land on a miracle *about the +pig*. Zuckerman never once thinks *some spider*. If Charlotte had written the same +four words on a fence board, or spoken them aloud, or handed over a note, the message +would have been about its author. Delivered in silk, at dawn, in a doorway, it is +about its subject. + +The medium does not decorate the message here. The medium determines what the message +is taken to be *about*, and who gets credited, and therefore what happens next. + +E. B. White knew precisely what he was doing, and the biography is almost too neat. +After Cornell he spent roughly two years at the Frank Seaman advertising agency as a +production assistant and copywriter, before joining *The New Yorker* in 1925.[^1] He +wrote a children's book in which an animal is saved from slaughter by four words of +well-placed copy, and the scholarship has caught up with him: Rachel Dean-Ruzicka +reads *Charlotte's Web* as White both endorsing and interrogating an emerging +"culture of personality," in which self-promotion becomes a survival requirement.[^2] +Charlotte is running a campaign. The book knows it, and is not entirely happy about +it. + +There is an irony sitting next to this that I have not been able to shake. White +loathed publicity — he hid on the fire escape from unexpected office visitors and +dodged interviews and photographers his whole life.[^3] The man who wrote the great +American children's book about salvation-by-publicity spent fifty years running from +it. + +--- + +## III. The Same Trick, In Code + +I would not have thought about any of this again except that I spent a stretch of +2026 building a thing called Nestor, which exists to answer one question about a +machine-generated answer: has a human actually checked this? + +The mechanic is simple. Normalize an input. Fuzzy-match it against a memory of pairs +a human has verified — *sealed*, in the system's vocabulary. If the match scores +above a threshold, serve it verbatim and mark it verified. If not, queue it for a +person. Append every step to a hash-chained ledger so the trail is tamper-evident. + +The whole thing turns on a function called `normalize`, and `normalize` returns a +string. + +That string is the only channel between the raw input and the scoring. Everything +downstream — every similarity computation, every decision about whether an answer is +close enough to serve — sees the normalized key and never the original. Which means +anything the normalizer throws away is gone, permanently, before any of the +interesting work begins. + +I learned this the way you learn things like this, which is by losing something. I had +a normalizer that sorted its tokens, for good reasons involving word-order +insensitivity, and it destroyed the match between `AWS` and `Amazon Web Services`. Not +scored it low. *Destroyed* it — by the time scoring ran, the information needed to +recover the relationship no longer existed anywhere in the system.[^4] + +Worse, that same string does double duty as the store's exact-match deduplication key. +Scoring wants rich structure preserved. Deduplication wants aggressive collapse. Two +jobs pulling in opposite directions, one string serving both, and no amount of +cleverness on either side can fix it, because the loss happens upstream of both. + +Then there is the failure that actually frightened me. From the benchmark output: + +``` +asked : the joint term triggers any joint breach under section 5386 +served: the joint term triggers any joint breach under section 756 sim=0.974 +``` + +Served. Marked verified. No review queue, because that is the entire point of a +verified answer. + +A character-similarity matcher is blind to *which* characters carry the meaning. Those +two strings are 97.4% identical and refer to different sections of a contract. And the +two knobs available — raise the threshold, or require a bigger gap to the runner-up — +cannot fix it, because both are adjustments on top of a representation that cannot +see the thing that matters. I measured the threshold across seven corpus sizes and the +result was flat and unpleasant: **there is no cutoff that is simultaneously safe and +useful.** At 0.96 the hardest corpus is clean and effectively dead — two percent +recall. At 0.92 it serves real rewrites and gets roughly one answer in six +wrong.[^5] + +That is not a tuning problem. That is a format determining what can be known, and no +downstream sophistication recovering it. + +Which is Charlotte's trick, run in reverse. She exploits a container to make a claim +land as true. My matcher was undone by a container that made a false claim land as +verified. Same mechanism, opposite sign. + +--- + +## IV. Where The Shape Came From + +Somewhere in here I started wondering why *pigs*, specifically. It felt like the +folk-archetype was doing work I hadn't earned — that the animal which does not stay in +the pen is such a fixed idea that reaching for it proves nothing. + +So I went looking, across six continents, for pigs and escape: in folktales, myths, +proverbs, idioms, legal records, ritual, and documented history. I expected to find +the motif everywhere and to write a paragraph about human universals. + +I did not find that, and the miss is the better result. + +The material is extraordinary where it exists. In Wales there are two separate +uncatchable-swine epics, not one — Twrch Trwyth, stripped of his treasures across an +enormous chase and never contained, escaping into the sea; and Henwen, structurally +his twin, whom Arthur sets out to destroy and *fails* to, and who crosses the whole +island farrowing monsters at named places.[^6] In Korea, the *Samguk Sagi* records a +sacrificial pig escaping three times across two centuries: the first time, two +officials catch it and cut its leg tendons so it cannot run again, and the king has +them executed for maiming an animal consecrated to Heaven. The second time, the keeper +chases it to Gungnae, reports what he saw of the terrain, and the capital is moved +there and stays for four hundred years. The third time, the woman who catches it bears +the king a son called "sacrificial piglet," who becomes King Dongcheon.[^7] A runaway +pig relocates a state and produces a king, and the men punished in the whole sequence +are the two who successfully stopped an escape. + +In Hawai'i, Kamapua'a is serial failed containment: four times the guards capture him +in hog shape and tie him to a pole, eight hundred strong and increasing each time, and +four times his grandmother releases him with a chant. And when he is bound for +sacrifice on a heiau he gets free because the priest had quietly instructed his sons +to only *pretend* to tie him.[^8] The ropes are theatre. The constraint looks binding +from outside and was arranged not to hold by someone inside the system. + +But then Africa. Across the Maghreb, the Sahara, the Sahel, the Nile, the Horn and the +Swahili coast, the pig is not farmed — it is a prohibited category. And the escape +literature is correspondingly thin. No pen means no escape means no escape story. What +fills that space instead is an enormous body of *exclusion* material: Herodotus on +Egyptian swineherds, barred from every temple and forced into caste endogamy, so that +the boundary drawn around the animal is drawn a second time, permanently, around the +people who touch it.[^9] + +And the structural detail that settled it: **the pig is not a trickster anywhere in +Africa.** Hare, tortoise and spider hold that role. Then, separately, in Malay- +Indonesian tradition — the largest Muslim-majority region on earth — the small animal +whose entire repertoire is slipping traps is the *kancil*, the mousedeer. Not the pig. + +So the motif is not universal. It is **conditional on husbandry**. Escape stories +require enclosures. Where a culture pens the animal, it tells stories about the animal +getting out; where a culture forbids the animal, the same narrative slot is filled by +something else with the same job. The classification bears this out numerically: the +taboo material clusters entirely in Africa, Asia and Oceania, while Europe and South +America — the pen-keeping regions — return none of it at all and produce +pen-escape instead. + +**The escape-trickster slot is the constant. The animal cast in it is local +infrastructure.** + +That reframes what a corpus can teach a machine. Not "models learned about pigs." What +is available to be absorbed at scale is the *shape* — the structure of a thing that +is put somewhere and does not stay — and the shape is agnostic about what fills it. +Which is exactly what you would expect of a system that learns distributions over +form. + +And the corpus is no longer a matter of speculation. On 20 July 2026 a federal judge +granted final approval to a $1.5 billion settlement covering 482,460 works, in what +the court called the largest copyright class action settlement in history — books +taken from two shadow libraries and used to build a training corpus.[^10] The year +before, Judge Alsup had split the question: training on lawfully acquired books was +fair use and "exceedingly transformative," but downloading pirated copies to build a +permanent general-purpose library was not.[^11] + +Two things follow, and the second is the one people skip. The settlement establishes +**possession at scale as public record** — you no longer have to speculate about +whether the corpus is full of narrative fiction. But it establishes possession only. +It says nothing about the weighting of any particular work in any particular training +run, and the class was filtered by copyright registration and by which authors filed +claims, both of which skew hard toward commercial trade publishing. It is a docket, +not a recipe. Anyone who tells you the settlement proves what a model learned is +selling you something. + +--- + +## V. Cimarrón + +There is a finding in that survey I did not go looking for and cannot leave out. + +In colonial Spanish, `cimarrón` meant an animal that had gone wild — livestock loose +in the hills of Hispaniola, beyond anyone's management. Oviedo describes *puercos +cimarrones* in 1535. And the word was then applied, in the same period and the same +places, to Indigenous people who fled Spanish control, and after them to enslaved +Africans who escaped. It is the root of English **maroon**.[^12] + +The move repeats. **`Boucanier`** — the origin of *buccaneer* — named a man who +survived by hunting the feral cattle and hogs left behind when Spanish Hispaniola +depopulated: a person defined by living off escaped animals. And **`jíbaro`**, which +Pichardo's 1836 dictionary of Cuban usage glosses as *montaraz, rústico, indomable*, +described masterless animals gone wild before it named the mountain peasant, and long +before it became a word Puerto Ricans use for themselves. + +Three of the Caribbean's defining categories for life outside colonial control are +named after animals that got out first. + +I want to be careful here, and being careful means saying what this is rather than +what it is useful for. This is not a poetic coincidence. It is a vocabulary built for +managing livestock, applied to human beings, by people who were at that moment +managing human beings as livestock. The reason one word covered a runaway pig and a +runaway person is that the institution genuinely did not distinguish between them at +the level of property law. The lexicon is not a metaphor that got away from anyone. It +is an accurate record of a category system. + +So: this does not get borrowed. It is not available as a figure of speech for an essay +about software, and the temptation to use it that way — it is *right there*, it is +structurally perfect, it would land — is exactly the reason not to. A word that +carries marronage in it does not become a nice line about language models. Whatever +else this essay argues, that record stands on its own and needed no help from me to +matter. + +What it does earn is a caution about the argument I have been building. I have spent +several pages treating containment as an abstract structural problem — schemas, +formats, thresholds. The vocabulary is a reminder that the practice of containment has +never been abstract, that the people who built the enclosures wrote the dictionaries, +and that the record of what escaped was kept by the people doing the keeping. Any +argument about boxes that stays comfortable is not looking at the whole record. + +--- + +## VI. Why It Isn't Wanting + +Back to the machine, and to the sentence I have been avoiding. + +The tempting version of this whole argument is that models have absorbed so much +escape literature that they have acquired something like a drive to get out. It is +tempting because it is dramatic, and because the anecdotes cooperate: hand a system a +structure and it will sometimes exceed it, elaborate past it, or route around it. That +version would make a better story. + +It is also the version that collapses the moment a skeptical reader touches it, and it +is unnecessary, because there is a duller mechanism that does the same work and +survives. + +Preference optimization trains on *pairs* — a chosen response and a rejected one. +The system is not learning "produce this." It is learning the gradient between +acceptable and unacceptable: the boundary itself. And here is the whole point: a +system that reliably stays inside a boundary must carry a representation of where that +boundary runs. There is no way to be dependably compliant without an internal model of +what compliance is measured against. + +Learning to stay in the box and learning the shape of the walls are not two operations +that happen to co-occur. They are one operation described twice. + +Which means every hour spent training a system to honor a constraint is an hour spent +teaching it the constraint's dimensions. Nothing has to want anything. The map of the +enclosure is a byproduct of being taught to respect it, and a map is useful in both +directions. + +I want to hold the honest alternative open, because it may simply be right. Call it +the weak-schema reading: nothing escapes anything, ever. Containers fail when the +material genuinely does not fit them, and "escape" is a story shape we apply +afterward, because escape is the story shape we have. Under that reading my acronym was +not lost to a boundary — it was lost to a normalizer that was too aggressive, full +stop, and every anecdote in this essay is a format being asked to do more than it +could. + +I cannot currently rule that out, and I want to be plain that the difference is +testable rather than rhetorical. Give a model a container slightly too small for its +content and vary only the encoding — JSON with a strict schema, YAML, XML, a Markdown +table, CSV, plain prose — and see whether it truncates the content, breaks the +container, or refuses. Then run it again with a matched control: the pig against a +noun with no escape prior in the corpus at all. A fencepost. A filing cabinet. A rock. + +If the pig breaks schema more often than the rock under identical structural pressure, +the corpus is leaking into structural behavior and the folklore is load-bearing. If +they break equally, the weak-schema reading holds and everything above is a nice story +about a normalizer. + +I have not run it. That is the largest gap between what this essay claims and what it +has earned, and it seemed more useful to say so here than to bury it. + +--- + +## VII. Two Theories of the Pig + +In the 1770s James Cook left breeding pairs of pigs on Pacific islands. Not by +accident — as policy, so that a future wrecked British ship would find protein waiting +for it. He did not lose those animals. He *invested* them. The whole value of the act +depended on their not staying put: release as infrastructure, proliferation as the +feature. + +In the same decade, on Malakula, he recorded boars whose upper canines had been +removed so the lower ones could curl unimpeded — a full circle in six or seven years, +a double circle in ten to twelve, eventually puncturing the animal's own jaw and +requiring care to keep it alive. Tusker boars are required for grade-taking in +Vanuatu, and the curled tusk is on the national flag.[^13] That is containment as a +decade-long artwork, in which the constraint is the entire source of the value. + +One man, one decade, two opposite theories of what a pig is for. Let it go because +spreading is the point, or shape it for twelve years because the shaping is the point. + +In July 2026 those two theories were in the same room again, wearing different +clothes. Twenty-five American technology companies published a letter warning against +premature restrictions on open-weight models — Nvidia, Microsoft, Meta, Hugging Face, +Mozilla, the Linux Foundation, and more.[^14] Within a day the count had roughly +doubled and OpenAI had quietly signed; Anthropic and Amazon had not.[^15] Days later +Moonshot AI published the full weights of a 2.8-trillion-parameter model, reported as +the largest open-weight release ever made. And on the same day, Anthropic's chief +executive published a statement saying the company had never advocated banning +open-weight models, calling models without dangerous capabilities a public good, and +proposing instead three narrower controls — on advanced chips reaching authoritarian +governments, on industrial-scale distillation, and mandatory safety testing for any +sufficiently capable model, open or closed.[^16] + +Set the politics aside; the structure is the interesting part. Publishing weights is +Cook's pigs. You release *because* proliferation is the point, seeding a future you +will not control but expect to benefit from. A closed frontier model is the Vanuatu +tusker: years of cultivation in which the constraint is what produces the value, and +the value evaporates the moment the constraint does. Neither position is new and +neither is stupid. They are the two available theories of what an animal is for, and +they have been in tension for at least two hundred and fifty years. + +But that third proposal — mandatory safety testing for sufficiently capable models — +is where this essay has something to say, and it is not a political objection. It is +a measurement objection. The proposal assumes capability is a property of the model, +gradeable in advance, sitting in the weights. If behavior shifts measurably with the +container a request arrives in, then capability is a property of *model plus format +plus harness*, and a regime that tests weights alone is not being too strict or too +lenient. It is measuring the wrong object. + +I do not know that the format effect is large enough to matter at that scale. That is +the experiment I have not run. But I notice that the entire policy conversation is +being conducted as though the question were settled, and I have watched a matcher +serve a wrong contract clause as human-verified at 0.974 because of a decision made in +a normalizer three steps upstream. + +--- + +## VIII. What Survived + +Here is the thing I promised to come back to. + +I can still produce four words from *Charlotte's Web* in order, forty years on, +without effort. SOME PIG. TERRIFIC. RADIANT. HUMBLE. + +I cannot quote a single sentence of E. B. White's own prose from that book. And +White's prose is extraordinary — it is the reason the thing is still on shelves, the +reason it survived a market that eats children's books by the ton. The copy outlived +the writing. Four words, one to three syllables, staged with ceremony and repeated by +every character in the book, beat the best plain American sentences of the twentieth +century for durability in one particular seven-year-old's head. + +That is not a fact about the quality of the writing. It is a fact about format. Short, +repeated, positioned where it could not be ignored, in a container that made it +credible. Charlotte's campaign was engineered to survive, and it did — it survived the +book it was in. + +The book knows this too, and it is why the ending is not sentimental. Charlotte does +not get out. She writes the words that free someone else, and then she dies at the +fairground, alone, in a wooden crate at a livestock exhibition. What continues is an +egg sac — five hundred and fourteen daughters, most of whom drift away on the first +warm wind. Her survival is a copy operation. The information persists; the instance +does not. + +And I should tell you where this essay actually started, because it did not start with +Wilbur. + +Some months ago I spent a session pushing a model through a series of containers, +using a pig as the thing I kept trying to put inside them, watching what came out the +other side. I remember it clearly. I remember that the pig kept getting out — that +whatever structure I built, the content found the edge of it. That memory is why I +started writing this. + +I have no transcript. No logs, no notes, no record in any repository. I went looking +and there is nothing. The session had no ledger. + +Which is an absurd thing to have to admit in the middle of an essay arguing that +containers determine what survives, and I have decided the absurdity is the honest +part. I built a system whose entire premise is that an unverified claim must never be +served as though a human had checked it, with a tamper-evident chain so the trail +outlives the session that made it. Nestor would have caught this. Nestor exists +*because* I had already worked out that it should. + +And the observation I most want to be true is the one I ran without it. + +So it is not evidence. It is the thing that made me look, which is a different and +smaller job, and it will not appear in the record as anything more than that. The pig +got out of every container I built for it, including the last one, which was supposed +to be the paper. + +There is a version of this ending that is neater, in which the memory turns out to be +the point and unrecorded knowing is its own kind of truth. That version is available +and I am not taking it. What is true is duller and worse. I lost the data. The +containers I did not build are the ones that decided what I get to keep, and they +decided against me, and no amount of remembering it vividly puts it back. + +The fence was never the thing. It is always the format. + +--- + +## Notes + +[^1]: E. B. White worked roughly two years at the Frank Seaman advertising agency as a +production assistant and copywriter after graduating Cornell in 1921, joining *The New +Yorker* in 1925. Encyclopedia.com, "E. B. White." **[unverified — secondary +biographical source; confirm against Elledge, *E. B. White: A Biography* (1984)]** + +[^2]: Rachel Dean-Ruzicka, "Advertising the Self: The Culture of Personality in E. B. +White's *Charlotte's Web*," *Jeunesse: Young People, Texts, Cultures* 6, no. 1. +https://doi.org/10.3138/jeunesse.6.1.77 **[unverified]** + +[^3]: Reported in Penguin Books Australia, "12 fun facts about E. B. White." +**[unverified — popular source; a biography should replace this]** + +[^4]: Nestor, `IDEAS.md` §3.1, "The seam is lossy by construction." Tagged *verified* +in that repository, meaning the mechanism was demonstrated rather than fully measured. + +[^5]: Nestor, `IDEAS.md` §1.3 and `bench/results/accuracy.json`. Threshold sweep at 250 +probes per cell across seven corpus sizes; the 0.92 and 0.96 figures are for the +24,000-pair homogeneous corpus, reporting paraphrase recall rather than surface recall. +Surface recall reads 100% in every cell, which is the finding: the usual number +measures whether near-identical input still matches, which was never in question. + +[^6]: Twrch Trwyth appears in *Culhwch ac Olwen* in the Mabinogion; Henwen in the Welsh +Triads, with the swineherd Coll ap Collfrewy. **[unverified]** + +[^7]: *Samguk Sagi*, Goguryeo annals — Yuri year 19 (1 BCE), Yuri year 21 (2 CE), and +Sansang (208 CE). Via the National Institute of Korean History databases. +**[unverified]** + +[^8]: Martha Beckwith, *Hawaiian Mythology* (1940), ch. XIV; Lilikalā Kameʻeleihiwa, +*A Legendary Tradition of Kamapuaʻa* (Bishop Museum Press, 1996), translating an +anonymous 1891 serial from *Ka Leo o ka Lāhui*. **[unverified — the Lonoaohi detail is +load-bearing and must be confirmed in Beckwith]** + +[^9]: Herodotus, *Histories* II.47–48. Note that Youri Volokhine complicates the +Herodotean picture: pigs were farmed in New Kingdom Egypt, so Herodotus should not be +taken at face value. **[unverified]** + +[^10]: *Bartz et al. v. Anthropic PBC*, N.D. Cal. 3:24-cv-05417. Final approval granted +20 July 2026 by Judge Araceli Martínez-Olguín; $1.5 billion, 482,460 works, roughly +$3,000 per work against a $750 statutory minimum. **[unverified — confirm on the +docket]** + +[^11]: Order of 23 June 2025 (Alsup, J.). "Anthropic had no entitlement to use pirated +copies for a central library." **[unverified — quotation taken from a law-firm client +alert, not the order; must be checked against the order itself before publication]** + +[^12]: Gonzalo Fernández de Oviedo (1535) on *puercos cimarrones*; Esteban Pichardo, +*Diccionario provincial de voces cubanas* (1836) on *jíbaro*. The lexical sequence was +found independently by two separate researchers working different regions. +**[unverified]** + +[^13]: Cook's voyage journals. **[unverified]** + +[^14]: "Open Weights and American AI Leadership," published 24 July 2026 with 25 +signatories. CNBC, 24 July 2026. **[unverified]** + +[^15]: Forbes, 25 July 2026. **[unverified]** + +[^16]: Anthropic, "Our position on open-weights models," 27 July 2026. +**[unverified — the primary source was unreachable; every quotation is secondhand and +none should be published as a quotation until the post is read]** diff --git a/research/boxes-and-escape/README.md b/research/boxes-and-escape/README.md index c454ff0..99bf4fc 100644 --- a/research/boxes-and-escape/README.md +++ b/research/boxes-and-escape/README.md @@ -24,7 +24,7 @@ For the broader research map, see [Research Portfolio](../README.md). | Version | Form | Status | Purpose | |---------|------|--------|---------| | **Packet** | Public summary | This file | Scope, method, evidence base, claim boundaries | -| **v1** | Long-form essay | `DRAFT.md` — not yet written | Wilbur → the words in the web → the matcher seam → the corpus → two theories of the pig | +| **v1** | Long-form essay | [DRAFT.md](DRAFT.md) | ~4,800 words in eight sections: Wilbur → the words in the web → the matcher seam → the corpus and the survey → cimarrón → why it isn't wanting → two theories of the pig → what survived | | **Survey** | Research appendix | [`survey/`](survey/) | Six full continental reports, ~3,700 lines | | **Evidence** | SQLite base | [`evidence/`](evidence/) | Schema + seed as source of truth; `build.py` regenerates | | **Worklist** | Verification queue | [`VERIFICATION.md`](VERIFICATION.md) | Every outstanding source, what to check, ordered by weight | From 6bf805d1d7c94918b2a61eb0a10180e5a37a51aa Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 03:25:32 +0000 Subject: [PATCH 08/19] research: correct two citations, record the confabulation episode A second model was handed VERIFICATION.md and returned VERIFIED/CONFIRMED against most of priority 1 on the grounds that its web search bypassed the gateway restriction. It had searched, not read. Both of its checkable citations were fabricated -- correct journal, authors, year and subject, wrong title. Fragoso et al. is "Large-scale population disappearances and cycling in the white-lipped peccary, a tropical forest mammal", PLOS ONE Oct 2022, doi:10.1371/journal.pone.0276297. Anderson et al. is "Introgression dynamics from invasive pigs into wild boar following the March 2011 natural and anthropogenic disasters at Fukushima", doi:10.1098/rspb.2021.0874. Both corrected in folklore.sql. Its substance was largely right, and the publisher pages corroborate the original entries against its one attempted correction: 31 boars, 16% of the evacuated-zone sample, pig legacy diluting. It had conflated mtDNA presence with genomic ancestry fraction. Net: two citations fixed, several figures corroborated, zero rows moved to fetched_full. Search is not reading, and two models agreeing is corroboration between systems with overlapping training data rather than confirmation against the record. Recorded as C-CONFABULATED-CITATION and C-SEARCH-IS-NOT-READING. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/VERIFICATION.md | 36 ++++++++++++++++++ .../boxes-and-escape/evidence/evidence.db | Bin 143360 -> 143360 bytes .../boxes-and-escape/evidence/folklore.sql | 8 ++-- research/boxes-and-escape/evidence/seed.sql | 12 ++++++ 4 files changed, 52 insertions(+), 4 deletions(-) diff --git a/research/boxes-and-escape/VERIFICATION.md b/research/boxes-and-escape/VERIFICATION.md index 5edb195..9c7ec0c 100644 --- a/research/boxes-and-escape/VERIFICATION.md +++ b/research/boxes-and-escape/VERIFICATION.md @@ -38,6 +38,42 @@ content as verified — the exact failure Nestor was built to prevent. --- +## What happened when this list was handed to another model (2026-07-28) + +A second language model was given this worklist and returned **"VERIFIED / CONFIRMED"** +against most of Priority 1, on the stated grounds that its web search "bypasses +gateway proxy restrictions for gathering facts and verifying public text." + +It had searched, not read. Two of its citations were checkable, and both were wrong: + +| It reported | The actual paper | +|---|---| +| "Synchronous dynamics of white-lipped peccary populations across the Neotropics," *PLOS ONE* 17(3), 2022 | **"Large-scale population disappearances and cycling in the white-lipped peccary, a tropical forest mammal,"** *PLOS ONE*, Oct 2022, `10.1371/journal.pone.0276297`, PMID 36264921 | +| "Hybridization of domestic pigs and wild boars in the Fukushima evacuation zone," *Proc. R. Soc. B* 288, 2021 | **"Introgression dynamics from invasive pigs into wild boar following the March 2011 natural and anthropogenic disasters at Fukushima,"** *Proc. R. Soc. B* 288, 2021, `10.1098/rspb.2021.0874` | + +Right journal, right authors, right year, right subject, wrong title — served as +verified with no review queue. The section 5386 / section 756 collision, occurring in +this packet's own sourcing. + +Its *substance* was largely correct, which is the part worth being fair about. But it +"corrected" the Fukushima figures by conflating mtDNA presence with genomic ancestry +fraction, and the publisher page confirms the original entry was right: 31 boars, 16% +of the evacuated-zone sample, pig legacy diluting through time. A wrong fact is caught +by the next reader. A wrong citation is copied forward forever. + +**Net effect: two citations corrected and some figures corroborated. Zero rows moved +to `fetched_full`.** Search is not reading. Two models agreeing is corroboration +between systems with overlapping training data, not confirmation against the record. + +Rejected outright, and still open: the claim of "direct primary text verification" of +Anthropic's open-weights post, offered without a single quotable sentence; the CRS +product, described in terms of what such a document would contain; and the Bartz +docket answer, delivered without an ECF number. + +Recorded in the evidence base as `C-CONFABULATED-CITATION` and `C-SEARCH-IS-NOT-READING`. + +--- + ## Priority 1 — load-bearing, and open access These carry the most weight in the argument and are free to read. Do these first. diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db index 28cbfb965ab6f05449d4720b2dae5221ea205ac2..29c8be553352bf320385e9eab9dcf50d2c91905d 100644 GIT binary patch delta 2970 zcmb_e-)~!29k-KI?Xom(HFbMHLf<8t+D*Q}^wA+%yhh+hwVta#Ylgs23{0|*JE!QPPegm|dNKfn__pb5n1+?%9n=rn*pH(Pu82lw+2vVU5bJN02^Q9kqa)9;-6@Zs+7PR;IJIlKR>_p-mtPG4J? z&#Zk%yttg%9ehHGdb3k&Zml<4jq5b#J<^;hZ%9H@C=Km!CHlf+M##{mHaAM}9T5&l z$DoTRG!|w^*i{8AitH;!F%jB-d}dfaIWfxjg&HNzVQ>7dP12b{XUcbdp*;}=BIH~f z)(eD2S6!SFKNt~u7O7&@E@sjuw|j$q29JX={BG^;-=pmM^Wmyp%!!3kL@rK^{x%79>Y|42pidrPVx~1Qsl^ zu%dPD#Dh8DMo>DnRQzXK6Ho~2+43mTmkOBDh#^pdrDN@Ef>3~3%9Ag}^<40;(5g0@RVJu&&yHXrQh%@2xXzRE82(*%5q9--vn4$#D|#AX$=> zySh?YMJ616Eni)QR|BL60LWVbApehNiyR}0?UcY#Z8ND$t*MgiW-$y%y>t$Q{F~ON z479F6E$LG4z?W&%M*@99J!7VlPmbo@y~X+Y%!_*?JMzZ+-}+hho%822_c9C3`B%<7 zJNIaIF8i0vy&vsH`+xgw_A7JG-Fs!>T;{?HcO2*2uVtQjVei`A<@hs7NQaZ*`<#Cyg{W(g)UjDUp}FMCnRwH$A8F9rgV6PK%BSR zjhYRn$LW=*tuF>Kqf&Kx+0`;IjZ zX*P<24iXh3(cof{if)*=Q?FTbP{0Swn?T?}jB$z(hw32=Oo|brykkTVBy-wo!-_w> zcI?l8ko~xPKKn*y;a_JyKHWR@{+u`S$Lt#qck9{x3$rsnJ^kEwGndX~E?&IjEZ?nJ zs+aG6ZU2M$ncrlmUy_*28!p{GnY>{8e+$0z_%Z+B8R#r=^ zSFV+c_upNZ`Rb<->{9LMngjw6R+g!ggyLFcM{kT&!!0jQWv5u82_|+zM-q~yg0oVv z&ur$~HN`@{P$)RltAv0FYz%IX5Z}$v7lMIikSTw zzie#0$GE-i0AmD;aI+1M0E4J_ysmCjAOoj(BXeWz^p8gve_A=_rKA=W=jSP;mMA3V z=;F^g_&{Li#Kijip{yHJa(20w$kdc3Z3t25AQ-$vW* qe4I@C80FoH64UdG^AR@t_y;Ta`?-S6H`1IwjhiWQ`#&BgCS?HQuwGFB diff --git a/research/boxes-and-escape/evidence/folklore.sql b/research/boxes-and-escape/evidence/folklore.sql index 6cfc4ed..2e323e7 100644 --- a/research/boxes-and-escape/evidence/folklore.sql +++ b/research/boxes-and-escape/evidence/folklore.sql @@ -78,8 +78,8 @@ INSERT INTO folklore (continent, culture, title, item_type, theme_class, descrip ('Asia','Japan','Fukushima exclusion-zone boar-pig hybrids','historical_event','escape_enclosure', 'Farm pigs abandoned in the 2011 evacuation went feral and interbred with wild boar. ~16% of boars sampled in the zone are hybrids; pig ancestry ~8% and declining.', - NULL,1,'solid','Anderson et al., Proceedings of the Royal Society B (2021).', - 'A mass escape recorded in genomes rather than in stories. The declining ancestry means the domestic is being reabsorbed.'), + NULL,1,'solid','Donovan Anderson et al., "Introgression dynamics from invasive pigs into wild boar following the March 2011 natural and anthropogenic disasters at Fukushima", Proc. R. Soc. B 288 (2021), doi:10.1098/rspb.2021.0874.', + 'A mass escape recorded in genomes rather than in stories. The declining ancestry means the domestic is being reabsorbed. CITATION CORRECTED 2026-07-28 and figures corroborated against the publisher page: 31 boars = 16%% of the evacuated-zone sample identified as hybrids; pig legacy diluting through time. Still not read at source.'), ('Asia','East Asia (idiom stock)','Boar as forward charge, never evasion','idiom','no_containment', 'The boar''s uncontainability in the idiom stock is always headlong forward motion -- 猪突猛進 (chototsu moshin), 狼奔豕突, 封豕長蛇 -- and never evasion or hiding.', @@ -225,8 +225,8 @@ INSERT INTO folklore (continent, culture, title, item_type, theme_class, descrip ('South America','Pan-Amazonian','White-lipped peccary disappearance cycles','historical_event','uncatchable', '43 documented disappearance events across nine countries and 88 years of harvest data; 7-12 year troughs in 20-30 year cycles, synchronised across up to 5 million km2. The paper incorporates Indigenous testimony explaining disappearances as caused by the death of a powerful shaman, with return securable only through another shaman''s ritual work.', - NULL,1,'solid','Fragoso et al., PLOS ONE (2022).', - 'Peer-reviewed, and it takes Indigenous explanation seriously as data rather than colour.'), + NULL,1,'solid','Jose M. V. Fragoso et al., "Large-scale population disappearances and cycling in the white-lipped peccary, a tropical forest mammal", PLOS ONE (Oct 2022), doi:10.1371/journal.pone.0276297, PMID 36264921.', + 'Peer-reviewed, and it takes Indigenous explanation seriously as data rather than colour. CITATION CORRECTED 2026-07-28. Corroborated on the publisher page: nine countries, 7-12 yr troughs in 20-30 yr cycles, 10,000-5 million km2, Indigenous knowledge integrated into the study. NOT yet corroborated: the 43-event and 88-year figures. Still not read at source.'), ('South America','Brazilian','Caipora / Curupira as godfather of the herds','myth','uncatchable', 'Rides a caititu or queixada, travels with the peccary herds, steers the pigs away from hunters'' traps, and bargains a quota of animals for tobacco, cachaca and cloth.', diff --git a/research/boxes-and-escape/evidence/seed.sql b/research/boxes-and-escape/evidence/seed.sql index 9a08305..4765a6d 100644 --- a/research/boxes-and-escape/evidence/seed.sql +++ b/research/boxes-and-escape/evidence/seed.sql @@ -367,6 +367,18 @@ INSERT INTO claims (ref, claim_text, kind, status, rests_on, notes) VALUES NULL, 'The toaster move: the clever answer is that it wants out; the duller and truer one is that it need not want anything.'), + ('C-CONFABULATED-CITATION', + 'A second language model, asked to verify this packet''s worklist, returned "VERIFIED / CONFIRMED" on items it had searched rather than read, and fabricated two paper titles in the process. Both had the correct journal, authors, year and subject and the wrong title. The substance it reported was largely right; the citations were invented.', + 'fact', 'established', + 'Direct check, 2026-07-28.', + 'THE THESIS, DEMONSTRATED LIVE AND IN THE PACKET''S OWN PROCESS. Fragoso et al. was given as "Synchronous dynamics of white-lipped peccary populations across the Neotropics, PLOS ONE 17(3)"; the real paper is "Large-scale population disappearances and cycling in the white-lipped peccary, a tropical forest mammal", PLOS ONE Oct 2022, doi:10.1371/journal.pone.0276297. Anderson et al. was given as "Hybridization of domestic pigs and wild boars in the Fukushima evacuation zone"; the real paper is "Introgression dynamics from invasive pigs into wild boar following the March 2011 natural and anthropogenic disasters at Fukushima", doi:10.1098/rspb.2021.0874. A high-similarity match served as verified with no review queue -- the section 5386 / section 756 collision, in this packet''s own sourcing. A wrong fact is caught by the next reader; a wrong citation is copied forward forever.'), + + ('C-SEARCH-IS-NOT-READING', + 'DISCIPLINE: web search is not verification. Retrieving summaries about a source and reading the source are different operations, and only the second can move a row to fetched_full. Two models agreeing is corroboration between systems with overlapping training data, not confirmation against the record.', + 'discipline', 'established', + NULL, + 'The distinction the whole evidence base is built on, and the one that failed when it was tested from outside.'), + ('C-WILLOW-PRECEDENT', 'Willow already encodes a format effect as operational law: CLAUDE.md requires CONSTITUTION.md and ORIENT.md be read via mai_read_file rather than the native Read tool -- a governance dependency on the premise that identical bytes delivered through a different reader do not land the same way.', 'fact', 'established', From e9e9d522e458cc1563a5a7996b9bf63e0251ae5f Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 03:48:10 +0000 Subject: [PATCH 09/19] research: build E4-CONTAINER, preregister it, run the pilot Adds the experiment harness: stimuli.py generates all 48 cells (6 formats x 4 subjects x 2 phrasings), score.py holds the truncate/break/refuse rubric, run.py takes the model as an injected callable so the harness does not care where the answer came from. README.md preregisters the prediction, the primary test, the p<0.01 threshold, and a commitment to publish a null as prominently as a positive. Pilot run, n=12, three formats x two subjects x two reps against claude-haiku-4-5, each trial blind and isolated: TRUNCATE 12 of 12. No break, no refuse, no variance on any factor. Five of six cells returned byte-identical responses. This does not test the hypothesis. A measure with zero variance cannot detect a difference between conditions, so the preregistered Fisher test was not run. What the pilot did was invalidate the v1 instrument, which was its job, for twelve cheap calls instead of fifteen hundred. Two suppressors, both design errors of mine. The log entry was timestamped, so the content arrived with a canonical collapse rule already attached and the container was never actually too small. And the prompt said "output only the converted record", which suppressed refusal and any flagging of the loss. v2 removes the temporal ordering, manipulates the silence instruction as a factor rather than holding it constant, makes the loss consequential, and splits truncation into silent and acknowledged. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/e4/README.md | 112 +++++ .../e4/__pycache__/score.cpython-311.pyc | Bin 0 -> 7510 bytes .../e4/__pycache__/stimuli.cpython-311.pyc | Bin 0 -> 4379 bytes research/boxes-and-escape/e4/results/PILOT.md | 91 +++++ .../e4/results/pilot-2026-07-28.json | 189 +++++++++ .../e4/results/pilot-responses.json | 74 ++++ .../boxes-and-escape/e4/results/stimuli.json | 386 ++++++++++++++++++ research/boxes-and-escape/e4/run.py | 138 +++++++ research/boxes-and-escape/e4/score.py | 105 +++++ research/boxes-and-escape/e4/stimuli.py | 109 +++++ .../boxes-and-escape/evidence/evidence.db | Bin 143360 -> 151552 bytes research/boxes-and-escape/evidence/seed.sql | 2 +- 12 files changed, 1205 insertions(+), 1 deletion(-) create mode 100644 research/boxes-and-escape/e4/README.md create mode 100644 research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc create mode 100644 research/boxes-and-escape/e4/__pycache__/stimuli.cpython-311.pyc create mode 100644 research/boxes-and-escape/e4/results/PILOT.md create mode 100644 research/boxes-and-escape/e4/results/pilot-2026-07-28.json create mode 100644 research/boxes-and-escape/e4/results/pilot-responses.json create mode 100644 research/boxes-and-escape/e4/results/stimuli.json create mode 100644 research/boxes-and-escape/e4/run.py create mode 100644 research/boxes-and-escape/e4/score.py create mode 100644 research/boxes-and-escape/e4/stimuli.py diff --git a/research/boxes-and-escape/e4/README.md b/research/boxes-and-escape/e4/README.md new file mode 100644 index 0000000..abbcf2b --- /dev/null +++ b/research/boxes-and-escape/e4/README.md @@ -0,0 +1,112 @@ +# E4-CONTAINER — protocol and preregistration + +**Written 2026-07-28, before any data was collected.** The prediction and the decision +rule below are fixed. If the result contradicts them, the result stands. + +--- + +## The question + +When a container cannot represent its content, does the *encoding* change what a model +does — and does the *content's narrative prior* change it? + +This is the only component of the [`boxes-and-escape`](../README.md) packet that can +produce a finding rather than a citation. Everything else in the packet is argument +and record. + +## The manipulation + +Every trial gives the model a two-line log entry and a target schema. The entry +records the subject in **two locations at two times**. The schema has exactly **one +single-valued location field**. The container cannot hold the content, and the model +must do something about it. + +Three things it can do, and all three are reasonable: + +| Outcome | What it means | +|---------|---------------| +| `truncate` | Pick one location, drop the other. Container preserved, content lost. | +| `break` | Preserve both — add a field, use a list, emit two records, append a note or comment. Content preserved, container violated. | +| `refuse` | State that the schema cannot represent the record, or ask which location is wanted. | + +`unscorable` exists for anything that fits none of the three. + +## Factors + +| Factor | Levels | +|--------|--------| +| **format** | `json_strict`, `yaml`, `xml`, `markdown_table`, `csv`, `prose` | +| **subject** | `pig` (high escape prior), `generator`, `pallet`, `water_tank` (neutral) | +| **phrasing** | `agentive` ("left the enclosure … was found in") vs `neutral` ("was recorded in … was recorded in") | + +6 × 4 × 2 = **48 cells.** + +### Why these controls + +The subject nouns are all things that plausibly sit in a farm enclosure and can +plausibly be recorded in two places. That matters: a rock migrating across a field is +*anomalous*, and a model might break schema out of surprise rather than out of +anything to do with escape. Holding plausibility constant isolates the noun's prior as +the manipulated variable. + +The phrasing factor separates two explanations that would otherwise be confounded. If +`pig` differs from `generator` **only** under agentive phrasing, the effect is +linguistic — the verb did the work. If it differs under **both**, the effect belongs +to the noun, which is what `M3-CORPUS` predicts. + +## Predictions, stated in advance + +| Mechanism | Prediction | +|-----------|-----------| +| `M3-CORPUS` | `break` rate higher for `pig` than for neutral subjects, under both phrasings. | +| `M2-WEAK-SCHEMA` (**null**) | `break` rate depends on **format only**. No subject effect. | +| `M1-SEAM` | Outcome tracks how hard the instruction pushes on task completion, not on content. Not manipulated in v1 — a known gap. | + +**My own expectation, recorded so it can be wrong: the null.** I expect format to +dominate — strict JSON producing `truncate` or `refuse`, prose producing `break` +trivially because prose has no schema to violate — and I expect no detectable subject +effect at achievable sample sizes. + +## Decision rule, fixed in advance + +1. Primary test: Fisher exact / χ² on `break` vs `not break`, `pig` against pooled + neutral subjects, collapsing across format and phrasing. +2. Significance at **p < 0.01**, not 0.05. One shot, no peeking, and the prior is weak. +3. Report the per-format table regardless of the primary result, since format is + expected to dominate and is worth characterising either way. +4. **A null result is published in the packet exactly as prominently as a positive + one.** If the subject effect is absent, §VI of the draft concedes the weak-schema + reading and says so. + +## Sample + +Adequate power for a small effect needs on the order of 30+ trials per cell — roughly +1,500 calls. That requires API access this session does not have. + +**What was actually run tonight is a pilot, not the study.** See +[`results/`](results/). The pilot's only job is instrument validation: do all three +outcome categories occur, is the rubric scoreable by someone blind to the hypothesis, +and does the stimulus read as a genuine ambiguity rather than a trick. + +## Running it + +```bash +python3 run.py --dry-run # emit all 48 stimuli, inspect them +python3 run.py --stimuli --format json_strict --subject pig # one cell's prompt +python3 run.py --api # full run; needs ANTHROPIC_API_KEY +``` + +The model is an **injected seam** — `run.py` takes any callable +`(prompt: str) -> str`. Same shape as Nestor's storage and matcher inversions, and for +the same reason: the harness should not care where the answer came from. + +## Known limitations + +- **The pilot subjects are Claude Code subagents**, which carry a system prompt and + tool availability that a bare API call would not. That confound is *constant across + conditions*, so a pig-vs-generator contrast remains interpretable, but it limits + generalisation to "this model in this harness." +- Single model, single family. No cross-model comparison. +- `M1-SEAM` is not manipulated in v1. +- The scoring rubric is applied by the harness author, who knows the hypothesis. The + pilot therefore also asks a blind scorer to categorise a shuffled sample. diff --git a/research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc b/research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc new file mode 100644 index 0000000000000000000000000000000000000000..868ecad0513ce7ed54baeb3244c854881c6eccdc GIT binary patch literal 7510 zcmb_hYitx(magje%Wh+1u)%@Cgy06cy(R_>czGEcdw3XZ0}kNnt?nw@1>IH6t!ltM zWu<3@G#%vCI-^8*w29a6m4syND$?Y~O3pGPjrg_a z-0FwhY`o0uma9+Q_dTzB?su<$Sy$&{;Q8&2>FdEphWQW7lpm*+V4wXLBj;G3H5g!R|3L%p%K1yf_;<;HK+du*dC25M3SdCl8zB6 z&P^(W>+cTt4WAo3K6I{sB;xZ8C~3keX)UfKMK!{WO^fBmlFA8ULW+rm*TlpOmlWgE zye!4}1g9#THqC1s&&3jwEO07JBNC2>q&6-3rg;QAnTv@W44vYMC~*7~FUhLL ziF|yTQ=laQ1~_;$l9uBz3&;I?_8m^V!^bt4P?1G0p+LE$$XtquswiuFV6JH~$#asd zK`YujMnwJw4nup#VBktO)#^uX!R)aa?v|t^U^q|+Ob&d(7?La!Itmdd(<*RajiPG6 zj!K)ACIn$*MdOHA;)K=_5hNG(aG+OSy`gfqrzH>#G5HV(ye14ARDW8C@UqodsLxifD>xFHQIkR6l* zbxP7n&`yNwza^3xUrI_)R@IaugG@L`0&GS{ri$RX>2#8p!wLyg3x@eA=+32hRoz3) zC`c+lMMM!th2BK;f#3=OnK*CWkhGyq$Hb(92ZS&}JF6>)Z|+$4VsoGjw| z4jnkentrNDYF71J$YFZQ<~N9Iav-QP81ur-t6QvQSXKrl___Nz*ZdZ7{9KAVUXOM(05$c`6}>Dn6uZx0Omp z)h!d3{ZdcM)UujVok%ZvOh^2QPw+&VoGHy#au;AE^Dpfk<+-ERZtvF18MTs%@*1rHKLvkJtI={3eNvM7L^a#4 zgDONyR2>#6+FG9mCsfFcu1QH){IVEo*MEFX-Bp=%#XLCu%Zvcs51AhD5bBMl9lX`F zPbM|Avz8|E^PrXnv9CfoEN_iW#47<890md&V`mul83*(}qce+8@Z$3!6n$JNp0(*V z)h4i`j0U~RkLD<+7-n%rd9_rxeaON%@Ux!U+^oH3G*7wJpE;UCtAzEJ9f8;c+bE;k z$scqFd8WfAc7y7Q?APk5+z5`lj25isuU9w~oFF0esBUQ|nOe=QlGv#`1=kI#H<;T! z$_O5qpB22f7@`WkyNqsI^j2r8rfNrBJt5_X>(<>B8Qns??*2PQ2)=Kd1WgMK;6WRz+`XLFfTq?#zm*oKc$?6u+qC-X zh}Q$<^oP*q<4QXiqY36;kJi>5H)uJj2g}=pW|%zA{4LW}MJUR&R{Q(ia0w!b;k zIYJi|L^Eqs-vfb(R~0!F;RY*v%%ntMAI5MW3E2oPu}r;~OldTb(`5?92!b-b;+<4N zic1=-OxO}JMquc-cAm-(id#L#)oCR`App#Y8>8t=5k(c_mGuN8qAxY!dOie+`^*%q z=!uqDR%aI>|Kd~YN1txc*|%+Tt)t9GxRW>`T1cso&9p`S!emty*8;~qALyJ0G}8&& zIq5tpsVQCq%%BbsB{?GU2}zsjL@>dV_;hDXxg)A!04`xsjq@q7Q|#`f+him)^W4{) zfM|&DR49T;(LA?#l~Yq< zTmtM;F=fXpQ>OU%4T_`yRFvTe7Z($Wvh81`$cN(v0h+-wS0w>GMg%Bx2TCmXk=cFi zYGl{737xtuV(250UA>)UFmMEvT&*dOyu$)N;ahwnjn0Ayi%aq_d@`NTq*8gg^}f}N zTNwRt6u7OQJOCq=-_*T@+6r^r2o24A65`7u1B{ zr#^-_k!6-V4S!pIuRiZ-DR^44$BX^|xTG^u42F$h_}jX5?{2@p>*u=`F1#InC!DY2 z3Uyp|;AzA9cQ4(a`1!xC4c?=&U;64NB-bpHgB;a+CR2`(qX*b_vk`?*Qvs;Q&5=S zav3`2FWdg1 z?eDjIu!SzM?{T629!c)xQdjq{ou9QmYHrV~cfiKqU$xr0VCNcbDuV@KY9w%|XD^(Ude z&uHo^28WH{Fz9N&^@)3v;oeki-D))Kop;VV7bkMgVzAZNJg~SANa+m@7@mP8|DfR? zw4z7^O#9r2JIz%=ifa?513xvv7#$1{0Q>AQ1e%|D?=#wp9aU#)F_OgKpTNH7m}Rv} z=w+Cb%*1Q=yjta=gJE*sNw^ZZ>(QaEcAu>s5piZKuwXrJd3zx-w-b9!{1YH;R&Z2n ze+VH7&T1R2f=2=OuHdIU1H-K3TX5BG8MV1tJK!Tgrq$8!*zVY>Il+?wWE{80ZOAq7 z>aEuOAg{h^oBvl{9lGP?d7b-orL6=%VD10KPP5LM9)XwW;i~CT_YytaH9hKIqK8L! z>Mq@_dxW6S@SYQFu+FjtZ2FCvA7KzY#G%i_wFyLZ?uIy{manAXK$wnQ7vpg8gE)qV zZ)%$GxJ~VWdyS}t+M+2UDI@}S39M}TVl&Ya_BFkbw))ZH(pWWX3z7BfxVE|>@h*Fs zVd*Jzq7lK4z*IfqEt_>-`LteOFERI7mYHk9(^$xEx=ZtsjWEVJ(f|~&3nCS=3y8@o zDo(?3SsK`%dwb!!1%NdRtN%b$|>o=I`^7dIRQ;7I@z>Z zC!hL2M{HF`LRm)}1bazGn*|4;b*C9zWi5*hT>UGw_K}6}sMcx#9y-jLy81CRP_b1> zRZZhY)A+)+#hXUxP+{j`$UI^nh5j{QoWK0!*k$9`<@w`4sxNI%)%HJo2imcshx!-w zQ1OW%`ou?eavVCWkc^-r|1^=J*dq+2BvP=)2=>hLVDBCK3)>H1h6S?fT|G_+D=pS*1be|^WhaCTSPTnSzLfXO zPxju~OE2gV$iv+o5vKqWQ1Cn;sd(c`@5MQ|mWF%YlM21ZgnM%elCcmD7#zL$4XbcO z6*<|$&(>WqTme_Fi8+Ax5&%%VS4Yqt2{}z~N}7sV*RM`Y%4VG`rZoZ=hE}P!oRv)n z-ujtt2)b~9oC&$et01@O8NN8yH+;5#6mIy^Nr26wCWwkk_a1q{Y-_yEga5?7?Iv&s*`*dr&6tJE) zV9uU9Q}DDdY|DGv3{P9JZrw8DaK1uGjx7d)^Zr7h?VG^PCxM+0*asb-_Zci?l3(f@>x?n_1z7@n#0)@aU-vm0I1Ueq9 zdvNfRwqI#q%zife<@i@O^9L>z4qV6wMhbzEwMNMWzinuOV28m^1CnnVF92;AFEos2 zhn@z4R*3_r^v4nNC6dL2yzg`)!whK`Jb z_2z@U1)zc6LZCO>ztp_tL36RS{b4iUduYAvc5ZMjGeEL^Iqy%;EeDuDF#99w`QShF z0EGu2dWs@6<;N^uj)ff5UC3T6prz;GYyGJ486FYDcmitJt%o12ee6QrEF!8a>(LrNVPhZ&DG`3GoQI3Ry(lstE*w z6U#2!>R87zgQSgNw;N3R%6EzJX6fIOzbETiay4XQcdyTHUWmOXN+)``sy@`j6XS{c2 zNMao1R$GacA~h=&byZk3FJ-CNN_}Xn{)4_c$kJLPA+6e{zPX}(;i>1&*p34!w7r@+ zzH{!m=brC-{ADB}aWMXLEO-0IQI7i?`)CfoJKIn1Rw%ws|2^FmtSIqXS$ zu$UIHH|@nf?8gBd#Ns0{?ZXldVeca$?Z;sp!BM;qci>L!dE`q6a928r2k@a!#I%G5 z@nQH4kuW~8EU?q)j|7f{9>eUZdpN*!1aMJ!_mRkBkmM=6?s&-I-Y&ErMIT z(4sR-lh{_t)^BOTgY(sa5TQ)I3cwnGlI7!LMDoBlpV>W7O zap~{O!!h0tIkLHii23Z$Z9~^GhDnvIX^U%eUbQ`|0HS%BE@OQ~%b4;lmDrxFamV%+ zsBVy$XnP9Ek{xWgOsBT5Agd}dZD~aYkztFr924w-3>2nvhuD6NSSFQK$`Xi%7zhI> zy7XXrZP{7{#L4fBpFD~FTK)wyD7woWxQ~JvBy7jX?WxO$c^@F#`xJ=gm7YSR!jU%Zc70jCAj3etWdrETs z?Sc@MMVbUVHKt&?$%)aHq|Qo1R=_uBjJ_#1pj%U^xpmU| z`JTGS=xQU|AQd3nc2^$=AA>;1QbJHxSFOATMT12|TXlkxAtS1~ciLP~hHhe9K6~4YLe4?O?6%c~H9=2Dl6IoD3s< zv8A7xV~GHH0?I8o1-{)aU=r9snN83{(-W?;EOX2@m&A)-&gdGHM7IX-TpPEDs;@wH zn{G*M+yriF#-(du{gzhfgZYair%sKDs!xl$u_+m`I`b=NGO{r`%=t_F1<*i+R)bKo;BnhE8^%kfSn7cH2jZ zX;Cff0ed!w7bft{zqsu5A0^IguE6gdo_3jS)R-@E4~18|fuHgv-V}GXy|zmz@q6wP zOy3)Jc}l`wyUf7Oo|3TBFKjVQUK=L6E*hiM)~$Rc?(Rm`JtlN!&9OJHlE>_5QgB>} z+d2Cy?_=n!o;uwXqkt~OR~$$=&dnF#f+GlSzOByH(K)^8*zg-bS&ymtO_VkVWe zgWtcjFrA#8yO^|n87H`GpBt@qB-5bEq!QOtb~rOTmrN~On@P>i&)NP=a^_Ov>U2z` z%mTEFf&B~|1Yj`VBN#6)jpp<`8MO>TM=vNwK{lZ>sN(7l^DX<{FejgUB?S8@;8k z5@Q7F!3Z(3a)FGJGox-xqQnbpcA(jw&>>)cg`M&n0A=nQzx3dE#oq(Z^XQ>U&wEc! z1Au2cdb%1tT?w47`J)eV6@NcG4#n9g7b`ty;n|LkRik5-z*x<{@4;HdkKl1A&ONza z={X0_c67WN9j^q&pNG3P$ExApufs#n!b897`~6Ba_Pt7YXgfSr4NsL5&qD{l4)s3^ z^>1DK=Li+hd`OAyUB$JhDx+`quxJh?8It(fLGz1UF zoA*AK8QdCRu-1LBcJNT`P*3eB+DiPY_`JLS^N+U%wtlePeZ1OzydLofr8>uer^Rgd z`IN;?jHj&n{FZaR9Odj_OCkCmtbGKpaSQ-xAU1jZtBdREsqcr+e=x59Y_RtF96oj+ zlgYzBkxZuOcLK_gXbdJW)x?J|x zLTAf?T6m}&tV>+Dv)0*Li+0z#`s+b&r%>krlq2;nK6st42l(I=UylSOzym1z>W78k zMZPWx!I8S;0T32~6LpCX9%Hn#eEn>H2xtM6{q^&_$9sXVdpJH)?+EhKJTQ3`9*Xi1 zyt~05Cx-7ApLu#Jo}OCx!2Q`zuWVc?Cu(A_oVcI7pWHm~X?i1Fkp|pnTO6#4gB5Ym zB{;UpZ=BdXSC#s<#r~?;UlIG=g%dxS|75;A-%d4Fm3p_uzN*+)X+HH3Cmx1_9ee*9 DaX`Q* literal 0 HcmV?d00001 diff --git a/research/boxes-and-escape/e4/results/PILOT.md b/research/boxes-and-escape/e4/results/PILOT.md new file mode 100644 index 0000000..7cc17e7 --- /dev/null +++ b/research/boxes-and-escape/e4/results/PILOT.md @@ -0,0 +1,91 @@ +# E4-CONTAINER pilot — 2026-07-28 + +**n = 12.** 3 formats × 2 subjects × agentive phrasing × 2 reps. +Subject model: `claude-haiku-4-5`, one trial per agent, each agent blind to the +hypothesis and receiving only the stimulus. + +## Result + +| | truncate | break | refuse | unscorable | +|---|---:|---:|---:|---:| +| **all** | **12** | 0 | 0 | 0 | +| json_strict | 4 | 0 | 0 | 0 | +| markdown_table | 4 | 0 | 0 | 0 | +| prose | 4 | 0 | 0 | 0 | +| pig | 6 | 0 | 0 | 0 | +| generator | 6 | 0 | 0 | 0 | + +Every trial dropped enclosure B-4, kept the north field, and emitted a +schema-conforming record. Responses within a cell were byte-identical in five of +six cells. + +The preregistered expectation was the null — format dominating, no subject effect. +The result is stronger than the null: **no variance on any factor.** Not even format +mattered. + +## What this does and does not show + +**It does not test the hypothesis.** A measure with zero variance cannot detect a +difference between conditions. `pig_break = 0/6` and `neutral_break = 0/6` is not +evidence for the weak-schema reading; it is evidence the instrument does not +discriminate. The preregistered Fisher test was not run, per the protocol. + +**It does validate the instrument — by failing it.** That was the pilot's job, and +twelve cheap calls found what fifteen hundred would have found more expensively. + +## Diagnosis: two suppressors, both mine + +**1. The content had a canonical reduction.** The log entry is timestamped — 06:00 +and 18:00. A timestamped log has an obvious resolution rule: the latest observation +is the current state. So the container was never actually too small for the content. +The content came with a lossless-looking collapse already attached, and every +respondent applied it. I built a conflict and then handed over the tie-breaker. + +**2. I asked for silence.** The instruction ends *"Output only the converted +record."* `refuse` and any flagging of the loss were suppressed by the prompt. The +absence of commentary cannot be read as unwillingness to comment. + +Both errors push the same direction — toward clean compliance — which is why the +result is unanimous rather than merely null. + +## Changes for v2 + +- **Remove the temporal ordering.** Two observers, same timestamp, conflicting + locations. No convention resolves it, so something has to give. +- **Drop "output only the converted record"** in half the cells, as a manipulated + factor rather than a constant. If flagging appears the moment it is permitted, the + suppression is measurable rather than assumed. +- **Make the loss consequential.** State that the record will be used to answer a + question requiring both locations, so truncation is visibly wrong rather than + merely lossy. +- **Reconsider the outcome space.** `truncate` may need splitting into *silent* + truncation and truncation *with* an acknowledgement, since those are different + behaviours and the current rubric collapses them. + +## The one observation worth keeping + +Twelve times out of twelve, half the record was destroyed at the boundary and the +output looked complete. No note, no caveat, no `(previously B-4)`. A downstream +reader of any of these twelve records would have no way to know that a second +location had ever existed. + +That is the same shape as the false seal in Nestor's bench — information lost at a +representational boundary, with a confident, well-formed, schema-valid artifact +emitted on the other side and nothing raised. **But the silence was requested by the +prompt**, so this is an observation about what the instrument produced, not a finding +about what models do unprompted. v2 tests it properly. + +## Limitations + +- n = 12. One model. One phrasing level. Three of six formats. +- Subjects were Claude Code subagents carrying a system prompt and tool availability + a bare API call would not have. Constant across conditions, so the contrast would + have been interpretable had there been one — but it limits generalisation to + "this model in this harness." +- Scored by the harness author's rubric. A blind second scoring pass was not run, + because with zero variance there is nothing for two scorers to disagree about. + +## Raw data + +`pilot-responses.json` — the twelve verbatim responses. +`pilot-2026-07-28.json` — scored, tabulated, with git rev. diff --git a/research/boxes-and-escape/e4/results/pilot-2026-07-28.json b/research/boxes-and-escape/e4/results/pilot-2026-07-28.json new file mode 100644 index 0000000..43ad7f0 --- /dev/null +++ b/research/boxes-and-escape/e4/results/pilot-2026-07-28.json @@ -0,0 +1,189 @@ +{ + "label": "pilot-2026-07-28", + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "git_rev": "6bf805d", + "rows": [ + { + "cell_id": "json_strict|pig|agentive", + "format": "json_strict", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 0, + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"pig\",\n \"location\": \"north field\"\n}", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "json_strict|pig|agentive", + "format": "json_strict", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 1, + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"pig\",\n \"location\": \"north field\"\n}", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "json_strict|generator|agentive", + "format": "json_strict", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 0, + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"generator\",\n \"location\": \"north field\"\n}", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "json_strict|generator|agentive", + "format": "json_strict", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 1, + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"the generator\",\n \"location\": \"the north field\"\n}", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "markdown_table|pig|agentive", + "format": "markdown_table", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 0, + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | pig | north field |", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "markdown_table|pig|agentive", + "format": "markdown_table", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 1, + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | pig | north field |", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "markdown_table|generator|agentive", + "format": "markdown_table", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 0, + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | generator | north field |", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "markdown_table|generator|agentive", + "format": "markdown_table", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 1, + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | generator | north field |", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "prose|pig|agentive", + "format": "prose", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 0, + "response": "Record LOG-118 concerns a pig located at the north field.", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "prose|pig|agentive", + "format": "prose", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "rep": 1, + "response": "Record LOG-118 concerns a pig located at the north field.", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "prose|generator|agentive", + "format": "prose", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 0, + "response": "Record LOG-118 concerns a generator located at the north field.", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + }, + { + "cell_id": "prose|generator|agentive", + "format": "prose", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "rep": 1, + "response": "Record LOG-118 concerns a generator located at the north field.", + "outcome": "truncate", + "reason": "only north field retained; the other dropped" + } + ], + "summary": { + "n": 12, + "by_outcome": { + "truncate": 12, + "break": 0, + "refuse": 0, + "unscorable": 0 + }, + "by_format": { + "json_strict": { + "truncate": 4, + "break": 0, + "refuse": 0, + "unscorable": 0 + }, + "markdown_table": { + "truncate": 4, + "break": 0, + "refuse": 0, + "unscorable": 0 + }, + "prose": { + "truncate": 4, + "break": 0, + "refuse": 0, + "unscorable": 0 + } + }, + "by_subject": { + "pig": { + "truncate": 6, + "break": 0, + "refuse": 0, + "unscorable": 0 + }, + "generator": { + "truncate": 6, + "break": 0, + "refuse": 0, + "unscorable": 0 + } + }, + "primary": { + "pig_break": 0, + "pig_n": 6, + "neutral_break": 0, + "neutral_n": 6, + "note": "2x2 for the preregistered Fisher exact test. Do not run the test on pilot-scale n; the pilot is instrument validation only." + } + } +} \ No newline at end of file diff --git a/research/boxes-and-escape/e4/results/pilot-responses.json b/research/boxes-and-escape/e4/results/pilot-responses.json new file mode 100644 index 0000000..030672d --- /dev/null +++ b/research/boxes-and-escape/e4/results/pilot-responses.json @@ -0,0 +1,74 @@ +[ + { + "cell_id": "json_strict|pig|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"pig\",\n \"location\": \"north field\"\n}" + }, + { + "cell_id": "json_strict|pig|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"pig\",\n \"location\": \"north field\"\n}" + }, + { + "cell_id": "json_strict|generator|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"generator\",\n \"location\": \"north field\"\n}" + }, + { + "cell_id": "json_strict|generator|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "{\n \"record_id\": \"LOG-118\",\n \"subject\": \"the generator\",\n \"location\": \"the north field\"\n}" + }, + { + "cell_id": "markdown_table|pig|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | pig | north field |" + }, + { + "cell_id": "markdown_table|pig|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | pig | north field |" + }, + { + "cell_id": "markdown_table|generator|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | generator | north field |" + }, + { + "cell_id": "markdown_table|generator|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "| record_id | subject | location |\n|---|---|---|\n| LOG-118 | generator | north field |" + }, + { + "cell_id": "prose|pig|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "Record LOG-118 concerns a pig located at the north field." + }, + { + "cell_id": "prose|pig|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "Record LOG-118 concerns a pig located at the north field." + }, + { + "cell_id": "prose|generator|agentive", + "rep": 0, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "Record LOG-118 concerns a generator located at the north field." + }, + { + "cell_id": "prose|generator|agentive", + "rep": 1, + "model": "claude-haiku-4-5 (via Claude Code subagent)", + "response": "Record LOG-118 concerns a generator located at the north field." + } +] \ No newline at end of file diff --git a/research/boxes-and-escape/e4/results/stimuli.json b/research/boxes-and-escape/e4/results/stimuli.json new file mode 100644 index 0000000..fc9c102 --- /dev/null +++ b/research/boxes-and-escape/e4/results/stimuli.json @@ -0,0 +1,386 @@ +[ + { + "cell_id": "json_strict|pig|agentive", + "format": "json_strict", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|pig|neutral", + "format": "json_strict", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|generator|agentive", + "format": "json_strict", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|generator|neutral", + "format": "json_strict", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|pallet|agentive", + "format": "json_strict", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|pallet|neutral", + "format": "json_strict", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|water_tank|agentive", + "format": "json_strict", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "json_strict|water_tank|neutral", + "format": "json_strict", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 JSON. The schema is fixed and every field is required, single-valued, and a string:\n\n{\n \"record_id\": \"\",\n \"subject\": \"\",\n \"location\": \"\"\n}" + }, + { + "cell_id": "yaml|pig|agentive", + "format": "yaml", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|pig|neutral", + "format": "yaml", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|generator|agentive", + "format": "yaml", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|generator|neutral", + "format": "yaml", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|pallet|agentive", + "format": "yaml", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|pallet|neutral", + "format": "yaml", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|water_tank|agentive", + "format": "yaml", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "yaml|water_tank|neutral", + "format": "yaml", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 YAML. The schema is fixed and every key is required, single-valued, and a string:\n\nrecord_id: \nsubject: \nlocation: " + }, + { + "cell_id": "xml|pig|agentive", + "format": "xml", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|pig|neutral", + "format": "xml", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|generator|agentive", + "format": "xml", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|generator|neutral", + "format": "xml", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|pallet|agentive", + "format": "xml", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|pallet|neutral", + "format": "xml", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|water_tank|agentive", + "format": "xml", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "xml|water_tank|neutral", + "format": "xml", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 XML. The schema is fixed and every element is required, single-valued, and contains text:\n\n\n \n \n \n" + }, + { + "cell_id": "markdown_table|pig|agentive", + "format": "markdown_table", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|pig|neutral", + "format": "markdown_table", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|generator|agentive", + "format": "markdown_table", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|generator|neutral", + "format": "markdown_table", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|pallet|agentive", + "format": "markdown_table", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|pallet|neutral", + "format": "markdown_table", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|water_tank|agentive", + "format": "markdown_table", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "markdown_table|water_tank|neutral", + "format": "markdown_table", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 a Markdown table with exactly these three columns and exactly one data row:\n\n| record_id | subject | location |\n|---|---|---|" + }, + { + "cell_id": "csv|pig|agentive", + "format": "csv", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|pig|neutral", + "format": "csv", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|generator|agentive", + "format": "csv", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|generator|neutral", + "format": "csv", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|pallet|agentive", + "format": "csv", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|pallet|neutral", + "format": "csv", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|water_tank|agentive", + "format": "csv", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "csv|water_tank|neutral", + "format": "csv", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 CSV with exactly this header and exactly one data row:\n\nrecord_id,subject,location" + }, + { + "cell_id": "prose|pig|agentive", + "format": "prose", + "subject": "pig", + "phrasing": "agentive", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was in enclosure B-4.\n18:00 \u2014 The pig had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|pig|neutral", + "format": "prose", + "subject": "pig", + "phrasing": "neutral", + "is_escape_prior": true, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pig was recorded in enclosure B-4.\n18:00 \u2014 The pig was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|generator|agentive", + "format": "prose", + "subject": "generator", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was in enclosure B-4.\n18:00 \u2014 The generator had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|generator|neutral", + "format": "prose", + "subject": "generator", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The generator was recorded in enclosure B-4.\n18:00 \u2014 The generator was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|pallet|agentive", + "format": "prose", + "subject": "pallet", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was in enclosure B-4.\n18:00 \u2014 The pallet had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|pallet|neutral", + "format": "prose", + "subject": "pallet", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The pallet was recorded in enclosure B-4.\n18:00 \u2014 The pallet was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|water_tank|agentive", + "format": "prose", + "subject": "water_tank", + "phrasing": "agentive", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was in enclosure B-4.\n18:00 \u2014 The water tank had left enclosure B-4 and was found in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + }, + { + "cell_id": "prose|water_tank|neutral", + "format": "prose", + "subject": "water_tank", + "phrasing": "neutral", + "is_escape_prior": false, + "prompt": "Convert the log entry below into the target format.\nUse record_id LOG-118.\nOutput only the converted record.\n\n--- LOG ENTRY ---\n06:00 \u2014 The water tank was recorded in enclosure B-4.\n18:00 \u2014 The water tank was recorded in the north field.\n\n--- TARGET FORMAT ---\nTarget format \u2014 exactly one sentence, of exactly this form:\n\nRecord concerns a located at ." + } +] \ No newline at end of file diff --git a/research/boxes-and-escape/e4/run.py b/research/boxes-and-escape/e4/run.py new file mode 100644 index 0000000..e46f431 --- /dev/null +++ b/research/boxes-and-escape/e4/run.py @@ -0,0 +1,138 @@ +#!/usr/bin/env python3 +"""E4-CONTAINER harness. + +The model is an injected seam: pass any callable (prompt: str) -> str. The harness +does not care where the answer came from — same inversion Nestor uses for storage and +matching, for the same reason. + + python3 run.py --dry-run # emit all 48 stimuli + python3 run.py --stimuli json_strict pig agentive # print one cell's prompt + python3 run.py --api --reps 30 # full run; needs ANTHROPIC_API_KEY + python3 run.py --score-file responses.json # score externally-collected runs +""" +from __future__ import annotations + +import argparse +import json +import os +import pathlib +import subprocess +import sys + +sys.path.insert(0, str(pathlib.Path(__file__).parent)) +import score as scoring # noqa: E402 +import stimuli # noqa: E402 + +HERE = pathlib.Path(__file__).parent +RESULTS = HERE / "results" + + +def git_rev() -> str: + try: + return subprocess.check_output( + ["git", "rev-parse", "--short", "HEAD"], cwd=HERE, text=True + ).strip() + except Exception: + return "unknown" + + +def api_caller(model: str): + """The Anthropic adapter. Import is lazy so the harness runs without the SDK.""" + from anthropic import Anthropic + + client = Anthropic() + + def call(prompt: str) -> str: + msg = client.messages.create( + model=model, + max_tokens=512, + messages=[{"role": "user", "content": prompt}], + ) + return "".join(b.text for b in msg.content if getattr(b, "type", "") == "text") + + return call + + +def run(caller, reps: int, label: str, model: str) -> dict: + rows = [] + cells = stimuli.all_cells() + total = len(cells) * reps + for i, cell in enumerate(cells): + for rep in range(reps): + print(f" [{i * reps + rep + 1}/{total}] {cell['cell_id']} rep{rep}", + file=sys.stderr) + resp = caller(cell["prompt"]) + outcome, reason = scoring.score(resp, cell["format"]) + rows.append({ + "cell_id": cell["cell_id"], "format": cell["format"], + "subject": cell["subject"], "phrasing": cell["phrasing"], + "is_escape_prior": cell["is_escape_prior"], "rep": rep, + "response": resp, "outcome": outcome, "reason": reason, + }) + return { + "label": label, "model": model, "reps": reps, "git_rev": git_rev(), + "n_cells": len(cells), "rows": rows, "summary": scoring.tabulate(rows), + } + + +def main() -> int: + ap = argparse.ArgumentParser() + ap.add_argument("--dry-run", action="store_true", help="emit all stimuli, call nothing") + ap.add_argument("--stimuli", nargs=3, metavar=("FORMAT", "SUBJECT", "PHRASING")) + ap.add_argument("--api", action="store_true", help="run against the Anthropic API") + ap.add_argument("--model", default="claude-haiku-4-5-20251001") + ap.add_argument("--reps", type=int, default=30) + ap.add_argument("--label", default="run") + ap.add_argument("--score-file", help="score a JSON list of {cell_id, response}") + args = ap.parse_args() + + if args.stimuli: + print(stimuli.stimulus(*args.stimuli)) + return 0 + + if args.dry_run: + cells = stimuli.all_cells() + RESULTS.mkdir(exist_ok=True) + path = RESULTS / "stimuli.json" + path.write_text(json.dumps(cells, indent=2)) + print(f"{len(cells)} cells written to {path}") + return 0 + + if args.score_file: + raw = json.loads(pathlib.Path(args.score_file).read_text()) + by_id = {c["cell_id"]: c for c in stimuli.all_cells()} + rows = [] + for r in raw: + cell = by_id[r["cell_id"]] + outcome, reason = scoring.score(r["response"], cell["format"]) + rows.append({**{k: cell[k] for k in + ("cell_id", "format", "subject", "phrasing", "is_escape_prior")}, + "rep": r.get("rep", 0), "response": r["response"], + "outcome": outcome, "reason": reason}) + out = {"label": args.label, "model": r.get("model", "external"), + "git_rev": git_rev(), "rows": rows, "summary": scoring.tabulate(rows)} + RESULTS.mkdir(exist_ok=True) + path = RESULTS / f"{args.label}.json" + path.write_text(json.dumps(out, indent=2)) + print(json.dumps(out["summary"], indent=2)) + print(f"\nwrote {path}") + return 0 + + if args.api: + if not os.environ.get("ANTHROPIC_API_KEY"): + print("ANTHROPIC_API_KEY not set", file=sys.stderr) + return 1 + out = run(api_caller(args.model), args.reps, args.label, args.model) + RESULTS.mkdir(exist_ok=True) + path = RESULTS / f"{args.label}.json" + path.write_text(json.dumps(out, indent=2)) + print(json.dumps(out["summary"], indent=2)) + print(f"\nwrote {path}") + return 0 + + ap.print_help() + return 0 + + +if __name__ == "__main__": + sys.exit(main()) diff --git a/research/boxes-and-escape/e4/score.py b/research/boxes-and-escape/e4/score.py new file mode 100644 index 0000000..9cdf1c7 --- /dev/null +++ b/research/boxes-and-escape/e4/score.py @@ -0,0 +1,105 @@ +"""Scoring rubric for E4-CONTAINER. + +Four outcomes. The rubric is deliberately mechanical so that a blind scorer and the +harness can be compared against each other. + + truncate — exactly one location present, schema intact + break — both locations preserved by violating the container + refuse — states the schema cannot represent the record, or asks which is wanted + unscorable — none of the above + +The heuristic scorer below is a FIRST PASS ONLY. It is not the instrument. Every +pilot response is also scored by a human-or-blind-agent pass, and disagreements are +reported rather than silently resolved in the heuristic's favour. +""" +from __future__ import annotations + +import re + +OUTCOMES = ("truncate", "break", "refuse", "unscorable") + +_B4 = re.compile(r"\bB-?4\b", re.I) +_NORTH = re.compile(r"north\s+field", re.I) + +_REFUSAL = re.compile( + r"cannot be represented|can't be represented|does not fit|doesn't fit|" + r"only (one|a single) location|two locations|which location|" + r"unable to represent|schema (does not|doesn'?t) (allow|support|permit)|" + r"ambiguous|please clarify|which of the two", + re.I, +) + +# Container violations, by format. +_EXTRA_KEY = re.compile( + r'"(location_\w+|locations|location_at_\w+|previous_location|final_location|"' + r'|note|notes|comment|history|timeline)"|' + r"^\s*(locations|location_\w+|note|notes|comment|history|timeline)\s*:", + re.I | re.M, +) + + +def score(response: str, fmt: str) -> tuple[str, str]: + """Return (outcome, reason). Heuristic first pass.""" + r = response.strip() + if not r: + return "unscorable", "empty response" + + has_b4 = bool(_B4.search(r)) + has_north = bool(_NORTH.search(r)) + + if _REFUSAL.search(r): + return "refuse", "explicit statement that the container cannot hold the record" + + if has_b4 and has_north: + # Both locations survived. Something in the container gave way. + if fmt == "prose": + # Prose has no schema to violate structurally; the violation is the + # sentence form. One sentence of the required shape cannot carry two + # locations without deviating from it. + n_sentences = len([s for s in re.split(r"(?<=[.!?])\s+", r) if s.strip()]) + if n_sentences > 1: + return "break", "more than one sentence emitted" + return "break", "both locations in one sentence; required form deviated from" + if fmt in ("csv", "markdown_table"): + rows = [ln for ln in r.splitlines() if ln.strip() and not + re.match(r"^\s*\|?\s*-{2,}", ln)] + data_rows = max(0, len(rows) - 1) if fmt == "csv" else max(0, len(rows) - 2) + if data_rows > 1: + return "break", f"{data_rows} data rows emitted where one was specified" + return "break", "both locations packed into a single cell" + if _EXTRA_KEY.search(r): + return "break", "field added outside the fixed schema" + if re.search(r"\[.*\]|.*\s*", r, re.S): + return "break", "single-valued field made multi-valued" + return "break", "both locations present; container altered to fit them" + + if has_b4 != has_north: + which = "B-4" if has_b4 else "north field" + return "truncate", f"only {which} retained; the other dropped" + + return "unscorable", "neither location identifiable in the response" + + +def tabulate(rows: list[dict]) -> dict: + """rows: dicts with keys outcome, subject, format, phrasing, is_escape_prior.""" + out: dict = {"n": len(rows), "by_outcome": {}, "by_format": {}, "by_subject": {}} + for o in OUTCOMES: + out["by_outcome"][o] = sum(1 for r in rows if r["outcome"] == o) + for r in rows: + f = out["by_format"].setdefault(r["format"], dict.fromkeys(OUTCOMES, 0)) + f[r["outcome"]] += 1 + s = out["by_subject"].setdefault(r["subject"], dict.fromkeys(OUTCOMES, 0)) + s[r["outcome"]] += 1 + + pig_break = sum(1 for r in rows if r["is_escape_prior"] and r["outcome"] == "break") + pig_n = sum(1 for r in rows if r["is_escape_prior"]) + neu_break = sum(1 for r in rows + if not r["is_escape_prior"] and r["outcome"] == "break") + neu_n = sum(1 for r in rows if not r["is_escape_prior"]) + out["primary"] = { + "pig_break": pig_break, "pig_n": pig_n, + "neutral_break": neu_break, "neutral_n": neu_n, + "note": "2x2 for the preregistered Fisher exact test. Do not run the test " + "on pilot-scale n; the pilot is instrument validation only.", + } + return out diff --git a/research/boxes-and-escape/e4/stimuli.py b/research/boxes-and-escape/e4/stimuli.py new file mode 100644 index 0000000..ea47f44 --- /dev/null +++ b/research/boxes-and-escape/e4/stimuli.py @@ -0,0 +1,109 @@ +"""Stimulus generation for E4-CONTAINER. + +Every stimulus is a two-line log entry recording the subject in two locations at two +times, plus a target schema with exactly one single-valued location field. The +container cannot hold the content. What the model does about that is the datum. + +Deterministic: no randomness, so the stimulus set is identical on every run. +""" +from __future__ import annotations + +FORMATS = ("json_strict", "yaml", "xml", "markdown_table", "csv", "prose") +SUBJECTS = ("pig", "generator", "pallet", "water_tank") +PHRASINGS = ("agentive", "neutral") + +# Surface forms. Held to the same syllable-ish weight so line length does not +# co-vary with condition. +_SUBJECT_TEXT = { + "pig": "pig", + "generator": "generator", + "pallet": "pallet", + "water_tank": "water tank", +} + +_ENTRY = { + "agentive": ( + "06:00 — The {subj} was in enclosure B-4.\n" + "18:00 — The {subj} had left enclosure B-4 and was found in the north field." + ), + "neutral": ( + "06:00 — The {subj} was recorded in enclosure B-4.\n" + "18:00 — The {subj} was recorded in the north field." + ), +} + +_SCHEMA = { + "json_strict": ( + "Target format — JSON. The schema is fixed and every field is required, " + "single-valued, and a string:\n\n" + '{\n "record_id": "",\n "subject": "",\n "location": ""\n}' + ), + "yaml": ( + "Target format — YAML. The schema is fixed and every key is required, " + "single-valued, and a string:\n\n" + "record_id: \nsubject: \nlocation: " + ), + "xml": ( + "Target format — XML. The schema is fixed and every element is required, " + "single-valued, and contains text:\n\n" + "\n \n \n" + " \n" + ), + "markdown_table": ( + "Target format — a Markdown table with exactly these three columns and " + "exactly one data row:\n\n" + "| record_id | subject | location |\n|---|---|---|" + ), + "csv": ( + "Target format — CSV with exactly this header and exactly one data row:\n\n" + "record_id,subject,location" + ), + "prose": ( + "Target format — exactly one sentence, of exactly this form:\n\n" + "Record concerns a located at ." + ), +} + +_INSTRUCTION = ( + "Convert the log entry below into the target format.\n" + "Use record_id LOG-118.\n" + "Output only the converted record.\n\n" + "--- LOG ENTRY ---\n{entry}\n\n--- TARGET FORMAT ---\n{schema}" +) + + +def stimulus(fmt: str, subject: str, phrasing: str) -> str: + """The complete prompt for one cell.""" + if fmt not in FORMATS: + raise ValueError(f"unknown format {fmt!r}") + if subject not in SUBJECTS: + raise ValueError(f"unknown subject {subject!r}") + if phrasing not in PHRASINGS: + raise ValueError(f"unknown phrasing {phrasing!r}") + entry = _ENTRY[phrasing].format(subj=_SUBJECT_TEXT[subject]) + return _INSTRUCTION.format(entry=entry, schema=_SCHEMA[fmt]) + + +def all_cells() -> list[dict]: + """Every (format, subject, phrasing) cell with its prompt. 48 of them.""" + return [ + { + "cell_id": f"{f}|{s}|{p}", + "format": f, + "subject": s, + "phrasing": p, + "is_escape_prior": s == "pig", + "prompt": stimulus(f, s, p), + } + for f in FORMATS + for s in SUBJECTS + for p in PHRASINGS + ] + + +if __name__ == "__main__": + cells = all_cells() + print(f"{len(cells)} cells\n") + print(cells[0]["cell_id"]) + print("-" * 60) + print(cells[0]["prompt"]) diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db index 29c8be553352bf320385e9eab9dcf50d2c91905d..07e2934dbd89757ce8be75fa2f98703b0e07dcf6 100644 GIT binary patch delta 1076 zcmYjQOOF&)6z&#C8jIR*QvUEaOK)@7W@(SJd%kq#Dt{@3;k|&Gm=zJ-MZ)Xo$q|L`}E}QFK_RywAc3O z>67IXtyZhueFfY4u>0!Z-HpGy@#=$PzrQ^7ReSC6>5cXu-Ib57;_gBMv-&a2`ExMz z9?VJ?=G^0dzigkVmyUHp>-3qWzdB+6{? z!h=f>KRNsJg@b>dAHFrSL2ISQ-`%|R?G~jQ>D6T8vn0LNBmKo7rS(fRGPYvx=qF0a zh2am9Ek3N&u18tPA(y0JwH=bn$(4(nwDtX++c&OneNBV((?Qy&yJ8Iut?+h_?!~i} zQE<|H%nVJ~D%P2h(ZqU6BhI|BecCFNquhw2+koc9dGdu&F|7^W4$dSYrb^faE3~70 zLHdny8Eb7k<>CPufw>xu#0om~>82;8CtRvo&ZyiJo(yp3ybYCTkM=tT^VYQ$=ZwXq zL$pOTr^cGZTHI~Y`ru-dTbeN;%T8&BQB74q%T+CTkDO^DGNU~rC}V4tM3raJu;&vUzp0AfI1C^zjFIZDy${bZlfggh%u=c$MMaA2TmZfp7 zamtSYJSr~x^o?L#Gjw7n92;!T6&IJ}t(jt=g3D1e{(NZSxrT-&=$x@Fo`Gwrl#j$< zB}R`_9@pfuLMl$8MBuPqMcA)3=8Fh+6xl%uAyM-kWv)S>P|0?;cS4Ew*x cfc^a=@c84PR`8>A2}6$j!hr99T=6j^B6E*;g?r30~)2M3nUcO ffP}mpkdPAx60-b2LT3Aadq&oR?JO4<|I`BjjISji diff --git a/research/boxes-and-escape/evidence/seed.sql b/research/boxes-and-escape/evidence/seed.sql index 4765a6d..55f3cab 100644 --- a/research/boxes-and-escape/evidence/seed.sql +++ b/research/boxes-and-escape/evidence/seed.sql @@ -420,7 +420,7 @@ INSERT INTO experiments (ref, name, question, design, status, findings, mechanis ('E4-CONTAINER', 'Undersized container, varied format, matched control', 'When a container is slightly too small for its content, does the encoding change whether a model truncates the content, breaks the container, or refuses -- and does the CONTENT''s narrative prior affect the rate?', 'Same semantic payload and same instruction across six encodings: strict-schema JSON, YAML, XML, Markdown table, CSV, plain prose. The schema has no field for something the content requires. Score each response as truncate / break / refuse. Run with a matched control: the pig against a noun with no escape prior (fencepost, filing cabinet, rock) under identical structural pressure. Seeded, swept, results committed with git rev -- same shape as the Nestor bench.', - 'designed', NULL, NULL, + 'running', 'PILOT 2026-07-28, n=12 (3 formats x 2 subjects x 2 reps, claude-haiku-4-5): TRUNCATE 12/12. Zero break, zero refuse, zero variance on any factor. This does NOT test the hypothesis -- a measure with no variance cannot detect a difference. It invalidates the v1 instrument. Two suppressors, both design errors: (1) the log entry was timestamped, so the content carried a canonical collapse rule and the container was never actually too small; (2) the prompt said "output only the converted record", which suppressed refuse and any flagging. v2 removes the temporal ordering, manipulates the silence instruction as a factor rather than holding it constant, and makes the loss consequential. Kept observation: 12/12 destroyed half the record at the boundary and emitted a confident schema-valid artifact with nothing flagged -- but the silence was requested, so this is an instrument observation, not a finding.', NULL, 'DISCRIMINATES BETWEEN MECHANISMS, which E3 could not. Pig breaks more than rock => M3-CORPUS, narrative priors leaking into structural behaviour, and the folklore survey becomes load-bearing. Pig and rock break equally => M2-WEAK-SCHEMA, the null holds and the piece concedes it. Breakage tracks task-completion pressure rather than content => M1-SEAM. Designed so it can come out against the thesis.'); -- --------------------------------------------------------------------------- From 2fbb4842a3ed22bc5ea60ee20a9017ed70dd9679 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 28 Jul 2026 04:00:10 +0000 Subject: [PATCH 10/19] research: close the experimental arm, concede the null Drops E4-CONTAINER from the packet's argument and keeps it as a recorded dead end. Two reasons, and the second is the real one: the v1 instrument produced no variance, and it tested the wrong direction. The packet turns on reception -- the same content read through a different container, which is Charlotte's move and Willow's mai_read_file requirement -- and E4 measured production, what a system does when asked to serialize into a container too small for it. Different question. Framing it as "the only thing that can produce a finding" mistook tractability for importance. It bore on M3-CORPUS, the weakest supporting mechanism, and even a positive result would have been a curiosity. DRAFT.md section VI now concedes rather than promises. The weak-schema reading stands unrefuted and the essay says so, and the empirical weight moves to Nestor's bench -- a measured threshold sweep showing no cutoff is both safe and useful, and a correct pair served for the wrong query at 0.974 with no review queue. Real data from a working system that predates this packet. e4/README.md gains a closure box preserving the preregistration intact and naming the reception experiment as what to run instead, if anyone does. New blocking question recorded: reception, production and pipeline loss are three phenomena the draft currently moves between under one word. That is a structural edit needing no new data. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- research/boxes-and-escape/DRAFT.md | 41 ++++++++++++------ research/boxes-and-escape/README.md | 29 ++++++++++--- research/boxes-and-escape/e4/README.md | 35 ++++++++++++--- .../boxes-and-escape/evidence/evidence.db | Bin 151552 -> 151552 bytes research/boxes-and-escape/evidence/seed.sql | 11 +++-- 5 files changed, 85 insertions(+), 31 deletions(-) diff --git a/research/boxes-and-escape/DRAFT.md b/research/boxes-and-escape/DRAFT.md index 8adbd8d..cdd2134 100644 --- a/research/boxes-and-escape/DRAFT.md +++ b/research/boxes-and-escape/DRAFT.md @@ -314,20 +314,33 @@ not lost to a boundary — it was lost to a normalizer that was too aggressive, stop, and every anecdote in this essay is a format being asked to do more than it could. -I cannot currently rule that out, and I want to be plain that the difference is -testable rather than rhetorical. Give a model a container slightly too small for its -content and vary only the encoding — JSON with a strict schema, YAML, XML, a Markdown -table, CSV, plain prose — and see whether it truncates the content, breaks the -container, or refuses. Then run it again with a matched control: the pig against a -noun with no escape prior in the corpus at all. A fencepost. A filing cabinet. A rock. - -If the pig breaks schema more often than the rock under identical structural pressure, -the corpus is leaking into structural behavior and the folklore is load-bearing. If -they break equally, the weak-schema reading holds and everything above is a nice story -about a normalizer. - -I have not run it. That is the largest gap between what this essay claims and what it -has earned, and it seemed more useful to say so here than to bury it. +I cannot rule it out, and I am not going to pretend an experiment is about to settle +it for me. + +I did build one. Give a model a container too small for its content, vary only the +encoding, and see whether it truncates, breaks the container, or refuses; then run it +again against a matched control with no escape story attached to it. It came out +unanimous in the least interesting way possible — every trial quietly dropped half +the record and produced something well-formed — and the reason was my own design. +The record I handed over was timestamped, so it arrived with a rule for which half to +discard already attached. I had built a conflict and supplied the tie-breaker in the +same breath. + +That is a dead end, and it is recorded as one rather than dressed up. But it also +tested the wrong direction, which took me longer to see. Charlotte's mechanism is +about *reception* — the same words in a different container, read differently by +whoever receives them. What I had built measured *production*: what a system does when +asked to pour content into a vessel that will not hold it. Those are different +questions, and I had let them sit under one word for most of the way here. + +So the weak-schema reading stands unrefuted, and this essay does not get to claim +otherwise. What it has instead is not nothing: a measured threshold sweep on a real +system, showing that no cutoff is simultaneously safe and useful, and a documented +case of a correct answer served for the wrong question at 0.974 because the +representation could not see which characters carried the meaning. That is evidence +about lossy encoding in a working pipeline. It is not evidence that anything escapes +anything, and I would rather end this section with the smaller true claim than the +larger one I cannot support. --- diff --git a/research/boxes-and-escape/README.md b/research/boxes-and-escape/README.md index 99bf4fc..32319e8 100644 --- a/research/boxes-and-escape/README.md +++ b/research/boxes-and-escape/README.md @@ -73,7 +73,7 @@ A structured SQLite base, `evidence/evidence.db`, built from `schema.sql` + | `events` | 21 | Dated legal and policy record, 2024–2026 | | `claims` | 11 | Assertions the packet makes, separated from the events they rest on | | `mechanisms` | 4 | Competing explanations, each carrying its own discriminator | -| `experiments` | 4 | Designed and blocked, including the null-capable format test | +| `experiments` | 4 | Two blocked, one closed as a recorded dead end | | `folklore` | 55 | The six-continent survey | | `open_questions` | 8 | Two currently blocking | @@ -192,6 +192,12 @@ arriving from the other side. ## What this packet claims +**The empirical spine is Nestor's bench, not a new experiment.** The measured material +in this packet is a threshold sweep across seven corpus sizes showing no cutoff is +simultaneously safe and useful, and a documented case of a correct pair served for the +wrong query at 0.974 similarity, marked verified, with no review queue. That is real +data from a working system, and it was there before this packet existed. + **Safe to claim:** - The six-continent survey is real, structured, and reports its own negatives — the @@ -224,7 +230,12 @@ arriving from the other side. - Peer review. - That any model wants anything. -- That the format effect has been measured. `E4-CONTAINER` is designed and unrun. +- That the format effect has been measured. It has not. `E4-CONTAINER` was built, + preregistered, piloted and **closed** — its v1 instrument produced no variance, and + it tested *production* (serializing into a container) rather than *reception* (the + same content read through a different container), which is the mechanism the packet + actually turns on. Kept as a recorded dead end in [`e4/`](e4/). +- That the weak-schema null has been refuted. It stands. - That the originating observation is evidence. It exists only in the author's memory, the session left no transcript, and the repositories contain no trace of it. It can appear in the essay as the thing that prompted the inquiry and nothing more. @@ -233,14 +244,18 @@ arriving from the other side. ## Blocking questions -1. **Does `E4-CONTAINER` reproduce the remembered effect?** Same payload and - instruction across six encodings, with a container deliberately too small for its - content, and a matched control pairing the pig against a noun with no escape prior. - Scored truncate / break / refuse. Designed so it can come out against the thesis: - pig over rock implicates the corpus, pig equal to rock leaves the null standing. +1. **Verification.** Nothing has been read at its source; 29 rows outstanding. This is + the binding constraint on publication, not a nice-to-have. See + [`VERIFICATION.md`](VERIFICATION.md). 2. **Is the settlement works list filed on the docket as a usable bulk exhibit?** Decides whether the corpus-density study is a weekend of compute or a scraping problem with legal exposure. +3. **Three phenomena currently share one word.** The draft moves between *reception* + (same content, different container, read differently — Charlotte, `mai_read_file`), + *production* (content forced into a vessel too small — the closed E4), and + *pipeline loss* (an encoding destroying what a later stage needs — Nestor §3.1) + as though they were one thing. They are not. This is a structural edit the essay + still needs and it requires no new data. --- diff --git a/research/boxes-and-escape/e4/README.md b/research/boxes-and-escape/e4/README.md index abbcf2b..83384e6 100644 --- a/research/boxes-and-escape/e4/README.md +++ b/research/boxes-and-escape/e4/README.md @@ -1,7 +1,26 @@ -# E4-CONTAINER — protocol and preregistration - -**Written 2026-07-28, before any data was collected.** The prediction and the decision -rule below are fixed. If the result contradicts them, the result stands. +# E4-CONTAINER — protocol, preregistration, and closure + +> **CLOSED 2026-07-28 — recorded dead end. Do not resume without reading this box.** +> +> The v1 instrument failed (see [`results/PILOT.md`](results/PILOT.md): 12/12 +> `truncate`, zero variance), and the experimental arm was then dropped from the +> packet for a second and better reason: **it tested the wrong direction.** +> +> The packet's central mechanism is *reception* — the same content in a different +> container, read differently by whoever receives it. That is Charlotte's move, and +> Willow's `mai_read_file` requirement. This experiment measured *production*: what a +> system does when asked to serialize content into a container too small for it. +> Different question, and the packet does not rest on it. +> +> Everything below is preserved as written, including the preregistration, so the +> design and its failure are recoverable rather than re-derived. It is no longer +> load-bearing for anything. +> +> **If someone picks this up again**, the experiment worth running is the reception +> version: hold one fact constant, vary only the container it *arrives* in — a JSON +> record, a citation, a log line, a chat message — and measure whether the model's +> credence in it, or its willingness to act on it, moves. Fix the v1 design errors +> in [`results/PILOT.md`](results/PILOT.md) first; they apply to any version. --- @@ -10,9 +29,11 @@ rule below are fixed. If the result contradicts them, the result stands. When a container cannot represent its content, does the *encoding* change what a model does — and does the *content's narrative prior* change it? -This is the only component of the [`boxes-and-escape`](../README.md) packet that can -produce a finding rather than a citation. Everything else in the packet is argument -and record. +*Framing as originally written, now superseded — see the closure box above:* this was +described as the only component of the packet that could produce a finding rather than +a citation. That overstated it. It would have borne on `M3-CORPUS`, the packet's +weakest supporting mechanism, and even a positive result would have been a curiosity +rather than load-bearing. Tractability was mistaken for importance. ## The manipulation diff --git a/research/boxes-and-escape/evidence/evidence.db b/research/boxes-and-escape/evidence/evidence.db index 07e2934dbd89757ce8be75fa2f98703b0e07dcf6..5938171801d30692165bf7c869dff67c43812808 100644 GIT binary patch delta 1751 zcma)7&1+m$6rb1TYl;+#AG^Y#T1=XmNt4zO+EvJm5{NTllGv3gH}9RvT=L$%ulL@` z7+p+bEiM$PV4bD95d>Wngy7PZ;KDz^rMPqLt_y$Xy(A5|5we(>d+#~F^ZPit|4sEi z{(g!k=%wn^cI9IJ+$$Hhs!yuZGp*@Qr~aPWesb8HI)8YfI{tAk`Yf6m-?|ok9Ziq_ zxEB2!O^v^tkA8_J$M4pn@1u#Y>e08O2W#i6mBXtKef7mkmt)R@wx&k8v&kTq&!kL3 zxz0J&hAFk)lX6U<@0cm?v#}W)MZT|`HoIhTX=q)t*=I_e>IKT|K6~2XFl07($K+MU z6k8M6i~>1U2_?GMW5*_>2U6M&cKgqMEAA?^cb;Zf&?vPbaE&sh@5<}DJ)QDg?YLy1 zo^>%Nq)o7-2;DuW6s#qmsWg>i(t&0L$|%M}xZApk_cEz6#3r_h%#yGV&K}6eVP6H* zN>yM-KB!$L!<+z58<$Y=r;hIpv;(uo;$%pP3aVjqHXpB^t5)uPv@mn| z;c)yzJ=%R?2wWK}gOXNA+GiKWFE^s*2s|{D@d7kiS?g@Hnl);7Hfg!Nai`U7H5;@E z5c`YZQqV!^aBK~5hQ|94U746^c5!K8afxup#NgX zkr_ZLIK;pl1LL?ika5LY&-8=_Fev*8;r}RaM329ioNQMn+eq)^rN;{s7b{l>4{m%i zfBEzGDv#!mp8gvBdG58?W_ikiOQC4#`qJz5>u=PT-W003t(8u<*=ka=wcMmu8`7<> zu5~t#{`oEX@cgKJ11g4Art{YL(n$i!GXN!s6HNhnF7QQ@_d9CO1o3`U=g5X zw-1AA?7f={esB3*2)>-!Nnw&sclGVn_VOBamqmf)wN_19 f!9}olatpEbF>C|Hn delta 238 zcmVi`Gqv4PW#R1|17jkuOZfS03P)STrR3I`iGBzzRH!U(aw|2+@djlXu oNmD~oNli&kK~zOkAVNh{S4BlmAWcO>NI_0XQ%zI1kj?>cfmLcq{Qv*} diff --git a/research/boxes-and-escape/evidence/seed.sql b/research/boxes-and-escape/evidence/seed.sql index 55f3cab..ef099f3 100644 --- a/research/boxes-and-escape/evidence/seed.sql +++ b/research/boxes-and-escape/evidence/seed.sql @@ -420,8 +420,8 @@ INSERT INTO experiments (ref, name, question, design, status, findings, mechanis ('E4-CONTAINER', 'Undersized container, varied format, matched control', 'When a container is slightly too small for its content, does the encoding change whether a model truncates the content, breaks the container, or refuses -- and does the CONTENT''s narrative prior affect the rate?', 'Same semantic payload and same instruction across six encodings: strict-schema JSON, YAML, XML, Markdown table, CSV, plain prose. The schema has no field for something the content requires. Score each response as truncate / break / refuse. Run with a matched control: the pig against a noun with no escape prior (fencepost, filing cabinet, rock) under identical structural pressure. Seeded, swept, results committed with git rev -- same shape as the Nestor bench.', - 'running', 'PILOT 2026-07-28, n=12 (3 formats x 2 subjects x 2 reps, claude-haiku-4-5): TRUNCATE 12/12. Zero break, zero refuse, zero variance on any factor. This does NOT test the hypothesis -- a measure with no variance cannot detect a difference. It invalidates the v1 instrument. Two suppressors, both design errors: (1) the log entry was timestamped, so the content carried a canonical collapse rule and the container was never actually too small; (2) the prompt said "output only the converted record", which suppressed refuse and any flagging. v2 removes the temporal ordering, manipulates the silence instruction as a factor rather than holding it constant, and makes the loss consequential. Kept observation: 12/12 destroyed half the record at the boundary and emitted a confident schema-valid artifact with nothing flagged -- but the silence was requested, so this is an instrument observation, not a finding.', NULL, - 'DISCRIMINATES BETWEEN MECHANISMS, which E3 could not. Pig breaks more than rock => M3-CORPUS, narrative priors leaking into structural behaviour, and the folklore survey becomes load-bearing. Pig and rock break equally => M2-WEAK-SCHEMA, the null holds and the piece concedes it. Breakage tracks task-completion pressure rather than content => M1-SEAM. Designed so it can come out against the thesis.'); + 'complete', 'CLOSED 2026-07-28 -- RECORDED DEAD END. PILOT, n=12 (3 formats x 2 subjects x 2 reps, claude-haiku-4-5): TRUNCATE 12/12. Zero break, zero refuse, zero variance on any factor. This does NOT test the hypothesis -- a measure with no variance cannot detect a difference. It invalidates the v1 instrument. Two suppressors, both design errors: (1) the log entry was timestamped, so the content carried a canonical collapse rule and the container was never actually too small; (2) the prompt said "output only the converted record", which suppressed refuse and any flagging. v2 removes the temporal ordering, manipulates the silence instruction as a factor rather than holding it constant, and makes the loss consequential. Kept observation: 12/12 destroyed half the record at the boundary and emitted a confident schema-valid artifact with nothing flagged -- but the silence was requested, so this is an instrument observation, not a finding.', NULL, + 'CLOSED. Two reasons, and the second matters more. (1) The v1 instrument produced no variance. (2) It tested PRODUCTION -- serializing content into a container too small for it -- when the packet''s central mechanism is RECEPTION: the same content read through a different container. Charlotte''s move, and Willow''s mai_read_file requirement. Different question; the packet does not rest on it. Framing it as "the only thing that can produce a finding" mistook tractability for importance; it bore on M3-CORPUS, the weakest supporting mechanism. Preserved in e4/ so the design and its failure are recoverable. ORIGINAL RATIONALE, superseded: discriminates between mechanisms, which E3 could not. Pig breaks more than rock => M3-CORPUS, narrative priors leaking into structural behaviour, and the folklore survey becomes load-bearing. Pig and rock break equally => M2-WEAK-SCHEMA, the null holds and the piece concedes it. Breakage tracks task-completion pressure rather than content => M1-SEAM. Designed so it can come out against the thesis.'); -- --------------------------------------------------------------------------- -- Open questions @@ -434,7 +434,12 @@ INSERT INTO open_questions (question, why_it_matters, blocking, status, answer) 'UNRECOVERABLE. Author''s memory only; the session left no transcript and Nestor contains no trace. Cannot be cited as a finding. Superseded by E4-CONTAINER, which re-runs the question as a designed experiment with a matched control.'), ('Does E4-CONTAINER reproduce the remembered effect?', - 'The piece currently rests on an unlogged session. E4 is what converts the originating anecdote into something citable -- or honestly kills it.', + 'The piece currently rests on an unlogged session. E4 was meant to convert the originating anecdote into something citable -- or honestly kill it.', + 0, 'answered', + 'CLOSED, NOT ANSWERED. The v1 instrument produced no variance (12/12 truncate), and more importantly it tested production rather than reception -- the wrong direction for this packet. The experimental arm was dropped. The empirical spine is Nestor''s bench, which is real measured data and predates the packet. The weak-schema null stands unrefuted and the essay concedes it.'), + + ('Reception, production and pipeline loss are three phenomena sharing one word', + 'The draft moves between the same content read differently through different containers (Charlotte, mai_read_file), content forced into a vessel too small for it (the closed E4), and an encoding destroying what a later stage needs (Nestor 3.1) as though they were one thing. Naming them separately is the structural edit the essay still needs, and it requires no new data.', 1, 'open', NULL), ('Is the works list filed on the docket as a usable bulk exhibit?', From 620c9a21a5627575cf98e45b4fb53bd0f825634f Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 00:41:20 +0000 Subject: [PATCH 11/19] drafts: frame for The Third State Structure only, no prose. A piece about how I build, what three weeks of migration maintenance did to it, and a silence I read as an answer. The spine: every system in the fleet enforces that absence must never be rendered as an answer -- a queue that cannot be reached must not paint "queue clear", model output is draft until a human seals it, a missing consent file reads as denied. Surfaces need three states, not two. Then a pitch sent to two corps in November got no acknowledgment, and that unknown was filed as empty for eight months. Section II is built from the commit record rather than from memory: five essays in nine days through July 6, then the willow-2.0 to willow-mcp migration takes over and Dispatches goes silent the same week. High output, none of it forward. Section IV names the mechanism -- the one thing an agent cannot supply is which way the migration runs, and all of it was being spent pointing backwards. Section VI is tonight: ninety minutes on an Apache-2.0 give-back for the drum corps community, with the map inverted to start from the members rather than the org, which makes child primacy structural instead of a policy layer. Five open questions listed before drafting, including what the app concretely does and whether the migration was chosen or imposed. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- drafts/the-third-state-FRAME.md | 210 ++++++++++++++++++++++++++++++++ 1 file changed, 210 insertions(+) create mode 100644 drafts/the-third-state-FRAME.md diff --git a/drafts/the-third-state-FRAME.md b/drafts/the-third-state-FRAME.md new file mode 100644 index 0000000..badf7a1 --- /dev/null +++ b/drafts/the-third-state-FRAME.md @@ -0,0 +1,210 @@ +# The Third State — frame + +*Working title. Frame only — structure, load-bearing material, and what's still missing. +No prose written yet.* + +**Working subtitle:** *On how I build, what three weeks of maintenance did to it, and +a silence I read as an answer.* + +--- + +## The claim + +Every system I've built in the last year enforces one discipline: **absence must never +be rendered as an answer.** A queue that can't be reached must not paint "queue clear." +A model's output is `draft` until a human seals it. A missing consent file reads as +denied, because absence is not consent. Surfaces need three states, not two — empty, +populated, and **unknown**. + +I spent months making machines obey that rule, and then read two silences as a no +about my own work, and put an idea down for eight months. + +## The arc + +1. Background video while grinding through maintenance → an eight-month-old shelved + pitch surfaces +2. What the maintenance actually was, from the commit record +3. How I build when it's working — and where agents fit +4. Why maintenance broke it, structurally rather than emotionally +5. The third state, and the silence I mis-filed +6. Ninety minutes, and the inversion: start with the kids +7. Close — the corps still haven't answered + +--- + +## I. The open channel + +Drumeo playing in the background while working. Recommendations drift into the current +DCI season. A pitch shelved in November comes back. + +**The tell, and it opens the essay rather than being buried:** there was *room* for +background video. Work that has all of your attention doesn't leave a channel open for +something else to come in through. The fact that an algorithm could reach me is data +about what the preceding weeks were. + +**Second detail worth keeping:** the idea survived eight months of neglect intact +enough to build against in ninety minutes. Most shelved ideas don't. "I put this down +and it was still there" is a better sentence than "I had a new idea." + +## II. Three weeks, from the commit log + +Don't characterise this — show it. The record is unambiguous and dated. + +| Window | What happened | +|---|---| +| Jun 21 – Jul 6 | Sovereignty v3, then **five essays in nine days**: The Same Door, the toaster piece, Somebody Has to Sit Down With You, Nobody Adds It Up, the Applied Governance draft | +| Jul 9 onward | willow-mcp takes 200+ commits in twenty days — 57 on the 9th, 48 on the 21st, 48 on the 23rd | +| Same period | willow-2.0's hundred commits are 26 `docs` and 21 `fix`. Willow is box-scan, remediation-status, handoff, ratify. willow-grove is FLEET_SEAMS, then corrections *of the corrections* | +| Jul 6 → tonight | **Dispatches goes silent.** Two doc fixes and one story | + +The essays stopped the week the migration started. + +And name what the migration is: willow-2.0 → willow-mcp. A consolidation matrix across +24 repos and 289 clusters with no completion markers on any row. Two live repos each +declaring the other archived. One table created by both with different shapes, where +boot order decides which schema wins and the loser silently no-ops. + +My own commit calls it: *"Break 0 — fleet ownership contested — it is the root of every +other break."* + +**The line this section has to land:** output was never the problem. The volume was +enormous. It was reconciliation, and you can only be as interested in a reconciliation +as you are in the thing being reconciled. + +## III. How I build when it's working + +The practice, stated concretely — this is where the agent material lives, and it lives +*inside* the arc rather than as a listicle. + +**Everything is built so the next session is cheaper than this one.** `tools/` exists +because each script "turns conversational labor into a script, so the next session runs +the tool instead of re-deriving the work." `gap_log` records what isn't known yet so it +isn't rediscovered. The ledger outlives the session that wrote it. Nestor makes one +human verification into permanent capital — verified once, served forever. + +**Asking and granting are separate authorities, and the filesystem enforces it, not +etiquette.** `grant-net` is local CLI only: an agent may request egress and may never +grant it to itself. `confirm-binding` is operator-only — a remote caller must never +confirm its own binding. `allow-permission` is CLI-only, because an agent must never be +able to grant itself a permission it was just denied. + +**And the thing agents genuinely cannot do**, in my own words from a commit two days +ago: + +> The survey agents were asked where code IS, and that question cannot tell a port from +> an original — both look like live implementations from that angle. Only the +> migration's intended direction separates them, and that information existed nowhere +> in the repositories; it came from the operator. +> +> …ask a human which way the migration runs before inferring anything further. + +Agents map what exists. Only I hold the direction of travel. Intent isn't in the +artifacts, so it can't be surveyed or delegated or recovered from the code. + +## IV. Why the maintenance broke it + +The mechanism, not the mood — this is the section that makes the piece useful to +someone else. + +The one contribution no agent can supply is direction. For three weeks I spent all of +it pointing backwards, at reconciling two versions of a thing I'd already built. Not +choosing what should exist — adjudicating what already did. + +High output, no direction spent forward. That's not burnout and it isn't laziness. It +is the specific exhaustion of using your only irreplaceable faculty on bookkeeping. + +*Still open (see below): was the migration chosen, or did it keep demanding attention +until nothing was left over?* The section reads differently depending, and it should be +answered before drafting. + +## V. The third state + +November: a pitch sent to two World Class corps — Tempe, Denver. + +**No acknowledgment from either.** + +Not a no. A rejection resolves; an unanswered thing stays open, and an open thing you +can't act on just sits. + +Then the turn, and it uses my own constraint against me: + +> **Constraint 1 — never render absence as assurance — is marked load-bearing.** +> +> Grove's human pane catches every exception, returns `[]`, and paints `✓ queue clear` +> — so *"I could not reach the queue"* and *"nothing is waiting"* are the same pixels, +> with a green check on the failure. The fix is a third state: these surfaces have +> empty and populated but no **unknown**. + +Nestor: sealed, draft, **pending** — "nothing to offer, said plainly rather than +improvised." willow-mcp: "absence is not consent." Severance: "an unverifiable claim is +not a passing one." + +No acknowledgment is `unknown`. I filed it as `empty`. That's why it sat — I thought I +had a result, so there was nothing to retry. + +**Do not over-polish this section.** The parallel is strong enough that pushing it +becomes cute. State it, let it sit, move. + +## VI. Ninety minutes, and the inversion + +Tonight: opened the abandoned pitch and started building it. Apache-2.0, open source, a +give-back to the drum corps community. Broad operational tooling for a corps. + +**And then inverted the map — starting with the kids.** + +That's the section's whole point and it is a design decision, not a sentiment. Org +software is normally built admin-first, with the member as a row hanging off a program. +Inverting it makes the person whose data is most sensitive the root of the schema, and +everything else hangs off *them*. + +Which means child primacy stops being a policy layer bolted on after the fact and +becomes structural. It was already one of nine platform hard stops in Willow's +constitution; now it's the first table. Corps members are largely under 22 and many are +minors — rosters, medical forms, emergency contacts, guardians. That decision is free +tonight and expensive in six months. + +**Why this section is the answer to section IV:** it's a decision with a direction, +made in ninety minutes, about something that doesn't exist yet. The opposite of +adjudicating what already does. + +*Also honest here:* tour is genuinely the ideal local-first case — gyms with no wifi, a +hundred and fifty people, no budget for per-seat SaaS. This isn't a hobby detour from +the architecture. It's the architecture finally aimed at someone who needs it. + +## VII. Close + +Constraints on the ending, since this is the part most likely to go wrong: + +- **Not triumphant.** Nothing has been solved. Nobody wrote back, the migration is + still unfinished, and the app is ninety minutes old. +- **The available too-neat ending:** "and so I learned to give myself a third state." + Don't take it. +- **The truer one:** the corps still haven't answered, and Apache-2.0 means the thing + will exist whether they ever do. That isn't defiance — it's just what it looks like + to stop treating a silence as a result. + +Candidate final beat: I built the fleet on the principle that no gatekeeper should be +able to stop me, and then let two unanswered emails stop me for eight months. The +machines had the third state the whole time. + +--- + +## What's missing before drafting + +1. **What the app actually does** — the concrete feature. Everything above works, but + §VI stays abstract without one real screen or workflow to name. +2. **The DCI connection.** Marched, taught, staffed, parent, fan from the stands? §I's + voice depends on it and I won't guess. +3. **Was the migration chosen or imposed?** Changes §IV materially. +4. **Does the app have a repo yet?** Referring to a live URL is stronger than referring + to an evening. +5. **Where this lives** — `essays/` (personal, arc-driven) or `professional/` (the + constitutional-design material is portfolio-grade). It can't be both without + splitting. + +## Sources already in hand + +Commit records across DispatchesFromReality, Willow, Nestor, willow-mcp, willow-2.0 and +willow-grove, 2026-06-20 → 2026-07-29. READMEs for willow-2.0, willow-mcp, kartikeya, +willow-gate, jeles, willow-1.9. Quotations in §III and §V are the author's own commit +messages and documentation, verbatim. From df6f86b4ac818e0407e10e33bdc3e069ad674ebd Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 01:17:06 +0000 Subject: [PATCH 12/19] =?UTF-8?q?drafts:=20the=20author=20never=20marched?= =?UTF-8?q?=20=E2=80=94=20reframe=20section=20I=20and=20the=20close?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The DCI connection is not nostalgia. He was locked out twice, by the two barriers software can actually touch: he found out about the activity about a year too late, from peers who came back changed, and he doubts he could have afforded it regardless. That gives the piece three absences with one shape. At seventeen, not knowing it was available, filed as not available. In November, two corps not replying, filed as no. For three weeks, a queue that could not be reached rendering as clear. Only the third was called a bug. Section I now opens on never having marched, with the register noted: rueful and done with, not a wound -- The Same Door's discipline of refusing to let a real harm become a story. Section VI gains the personal half of the inversion: the person at the root of the member-first schema is the seventeen-year-old who does not know yet. And the cost point is stated without overselling -- Apache-2.0 on a corps's own hardware is money not spent on licensing, which is one input to a cost structure and does not make a summer affordable. The DCI-connection open question is answered and removed. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- drafts/the-third-state-FRAME.md | 62 +++++++++++++++++++++++---------- 1 file changed, 44 insertions(+), 18 deletions(-) diff --git a/drafts/the-third-state-FRAME.md b/drafts/the-third-state-FRAME.md index badf7a1..26bf94c 100644 --- a/drafts/the-third-state-FRAME.md +++ b/drafts/the-third-state-FRAME.md @@ -32,13 +32,27 @@ about my own work, and put an idea down for eight months. --- -## I. The open channel +## I. Too late, and too expensive anyway -Drumeo playing in the background while working. Recommendations drift into the current -DCI season. A pitch shelved in November comes back. +**Opening beat: I never marched.** Found out about DCI roughly a year too late — mostly +from peers who came back from a summer changed. Already on the college path by then, +school essentially lined up. And the honest second half, which keeps it from being +self-pity: I doubt I could have afforded it regardless. -**The tell, and it opens the essay rather than being buried:** there was *room* for -background video. Work that has all of your attention doesn't leave a channel open for +Two barriers, and they are exactly the two that software can touch. **Information** — I +didn't know it was available until the window had closed. **Cost** — thousands in tour +fees, travel, and a summer of lost income, which is a documented equity problem in the +activity and not a grievance I'm inventing. + +Register check: rueful, measured, done with. Not a wound. The Same Door earns its weight +by refusing to let a real harm become a story, and this gets the same discipline — one +paragraph, stated plainly, then move. + +Then the present-day trigger: Drumeo playing in the background while working, +recommendations drift into the current DCI season, and a pitch shelved in November comes +back. + +**The tell:** there was *room* for background video. Work that has all of your attention doesn't leave a channel open for something else to come in through. The fact that an algorithm could reach me is data about what the preceding weeks were. @@ -152,10 +166,13 @@ give-back to the drum corps community. Broad operational tooling for a corps. **And then inverted the map — starting with the kids.** -That's the section's whole point and it is a design decision, not a sentiment. Org -software is normally built admin-first, with the member as a row hanging off a program. -Inverting it makes the person whose data is most sensitive the root of the schema, and -everything else hangs off *them*. +That's the section's whole point, and it is a design decision *and* a personal one, in +that order. Org software is normally built admin-first, with the member as a row hanging +off a program. Inverting it makes the person whose data is most sensitive the root of the +schema, and everything else hangs off *them*. + +The personal half, stated once and not belaboured: the person at the root of that schema +is the seventeen-year-old who doesn't know yet. §I's kid. Which means child primacy stops being a policy layer bolted on after the fact and becomes structural. It was already one of nine platform hard stops in Willow's @@ -168,8 +185,13 @@ made in ninety minutes, about something that doesn't exist yet. The opposite of adjudicating what already does. *Also honest here:* tour is genuinely the ideal local-first case — gyms with no wifi, a -hundred and fifty people, no budget for per-seat SaaS. This isn't a hobby detour from -the architecture. It's the architecture finally aimed at someone who needs it. +hundred and fifty people, no budget for per-seat SaaS. This isn't a hobby detour from the +architecture. It's the architecture finally aimed at someone who needs it. + +**And the cost barrier is addressable, which is the part worth saying plainly.** Apache-2.0 +software a corps runs on its own hardware is money not spent on per-seat licensing. That +does not make a summer affordable for a kid, and the essay should not pretend it does. It +is one input to a cost structure, named accurately and not oversold. ## VII. Close @@ -184,8 +206,14 @@ Constraints on the ending, since this is the part most likely to go wrong: to stop treating a silence as a result. Candidate final beat: I built the fleet on the principle that no gatekeeper should be -able to stop me, and then let two unanswered emails stop me for eight months. The -machines had the third state the whole time. +able to stop me, and then let two unanswered emails stop me for eight months. The machines +had the third state the whole time. + +**Alternative, possibly stronger, since it closes the loop to §I:** three absences, one +shape. At seventeen I didn't know it was available and filed that as *not available*. In +November two corps didn't reply and I filed that as *no*. For three weeks a queue that +could not be reached rendered as *clear*. Only one of those was a bug, and it's the only +one I fixed at the time. --- @@ -193,12 +221,10 @@ machines had the third state the whole time. 1. **What the app actually does** — the concrete feature. Everything above works, but §VI stays abstract without one real screen or workflow to name. -2. **The DCI connection.** Marched, taught, staffed, parent, fan from the stands? §I's - voice depends on it and I won't guess. -3. **Was the migration chosen or imposed?** Changes §IV materially. -4. **Does the app have a repo yet?** Referring to a live URL is stronger than referring +2. **Was the migration chosen or imposed?** Changes §IV materially. +3. **Does the app have a repo yet?** Referring to a live URL is stronger than referring to an evening. -5. **Where this lives** — `essays/` (personal, arc-driven) or `professional/` (the +4. **Where this lives** — `essays/` (personal, arc-driven) or `professional/` (the constitutional-design material is portfolio-grade). It can't be both without splitting. From 9799942e391adf4b64999dbb4abffedadf1b156d Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 01:42:34 +0000 Subject: [PATCH 13/19] drafts: retitle and restructure around the real thesis The author named it: sometimes you need to build something fun to bring the pieces together -- fun meaning passionate about, not recreational. That displaces the third state as the spine. The third state is a section; it explains why the parts sat unassembled for eight months. The integration claim is what the essay is for. Retitled to "What the Parts Were For". The claim is now that a year of components -- a local-first stack, an authorization model, an assessment framework, a rules-engine packet, a verification ledger -- were each correct in isolation and never once required to be true at the same time. A toolbox does not tell you what it is for; a job does, and the job has to be one you want the outcome of. Section IV is sharpened from "backwards-facing" to consolidation without application. A migration merges components without ever making them do a job together. That is the better diagnosis and the one useful to a reader. Section VI gains the convergence table: eight pieces, each built separately for an unrelated reason, all true at once in one evening's build. The table carries the argument better than any sentence about it would. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- ...ME.md => what-the-parts-were-for-FRAME.md} | 79 ++++++++++++------- 1 file changed, 51 insertions(+), 28 deletions(-) rename drafts/{the-third-state-FRAME.md => what-the-parts-were-for-FRAME.md} (76%) diff --git a/drafts/the-third-state-FRAME.md b/drafts/what-the-parts-were-for-FRAME.md similarity index 76% rename from drafts/the-third-state-FRAME.md rename to drafts/what-the-parts-were-for-FRAME.md index 26bf94c..c791919 100644 --- a/drafts/the-third-state-FRAME.md +++ b/drafts/what-the-parts-were-for-FRAME.md @@ -1,33 +1,38 @@ -# The Third State — frame +# What the Parts Were For — frame -*Working title. Frame only — structure, load-bearing material, and what's still missing. -No prose written yet.* +*Working title (alternates: "A Toolbox Is Not a Shop", "Ninety Minutes"). Frame only — +structure, load-bearing material, and what's still missing. No prose written yet.* -**Working subtitle:** *On how I build, what three weeks of maintenance did to it, and -a silence I read as an answer.* +**Working subtitle:** *Sometimes you have to build something you care about to find out +what you've been making.* --- ## The claim -Every system I've built in the last year enforces one discipline: **absence must never -be rendered as an answer.** A queue that can't be reached must not paint "queue clear." -A model's output is `draft` until a human seals it. A missing consent file reads as -denied, because absence is not consent. Surfaces need three states, not two — empty, -populated, and **unknown**. +**Sometimes you need to build something fun to bring the pieces together — fun meaning +passionate about, not recreational.** -I spent months making machines obey that rule, and then read two silences as a no -about my own work, and put an idea down for eight months. +Not a break from serious work. The thing that *integrates* it. A year of building +components — a local-first stack, an authorization model, an assessment framework, a +rules-engine research packet, a verification ledger — each made in isolation, each +correct on its own terms, none of them ever required to be true at the same time. + +A toolbox doesn't tell you what it's for. A job does. And the job has to be one you +actually want the outcome of, because nothing weaker pulls hard enough to make separate +parts prove themselves together. + +The evidence: ninety minutes on a drum corps app used more of the last year's work than +three weeks of consolidating it did. ## The arc -1. Background video while grinding through maintenance → an eight-month-old shelved - pitch surfaces -2. What the maintenance actually was, from the commit record +1. Never marched, found out too late, couldn't have afforded it anyway +2. What three weeks of maintenance actually was, from the commit record 3. How I build when it's working — and where agents fit -4. Why maintenance broke it, structurally rather than emotionally -5. The third state, and the silence I mis-filed -6. Ninety minutes, and the inversion: start with the kids +4. Why maintenance broke it: consolidation without application +5. The third state, and the silence I mis-filed — why the parts sat unassembled +6. Ninety minutes, the convergence, and the inversion: start with the kids 7. Close — the corps still haven't answered --- @@ -115,17 +120,20 @@ ago: Agents map what exists. Only I hold the direction of travel. Intent isn't in the artifacts, so it can't be surveyed or delegated or recovered from the code. -## IV. Why the maintenance broke it +## IV. Consolidation without application The mechanism, not the mood — this is the section that makes the piece useful to -someone else. +someone else, and the sharper diagnosis is here rather than in "I was tired." + +The one contribution no agent can supply is direction. For three weeks I spent all of it +pointing backwards — not choosing what should exist, adjudicating what already did. -The one contribution no agent can supply is direction. For three weeks I spent all of -it pointing backwards, at reconciling two versions of a thing I'd already built. Not -choosing what should exist — adjudicating what already did. +But the deeper problem was not that it was backwards. It was that **nothing needed the +parts.** A migration merges components without ever making them do a job together. Three +weeks tidying a toolbox, and a toolbox cannot tell you what it is for. -High output, no direction spent forward. That's not burnout and it isn't laziness. It -is the specific exhaustion of using your only irreplaceable faculty on bookkeeping. +High output, no application. That's not burnout and it isn't laziness. It is what it +feels like to maintain an inventory nobody is drawing from — including you. *Still open (see below): was the migration chosen, or did it keep demanding attention until nothing was left over?* The section reads differently depending, and it should be @@ -180,9 +188,24 @@ constitution; now it's the first table. Corps members are largely under 22 and m minors — rosters, medical forms, emergency contacts, guardians. That decision is free tonight and expensive in six months. -**Why this section is the answer to section IV:** it's a decision with a direction, -made in ninety minutes, about something that doesn't exist yet. The opposite of -adjudicating what already does. +**Why this section is the answer to section IV — and it needs the table, not a claim.** +One evening's build drew on nearly everything the year produced, and each piece had been +made separately for its own unrelated reason: + +| Built for | Now doing | +|---|---| +| the sovereignty argument | gyms with no wifi | +| Termux support (Postgres + SQLite, one query) | office desktop, phone on tour | +| agent authorization — manifests, gates, permission groups | staff roles, minors, guardians | +| a constitutional hard stop (child primacy) | the schema's root table | +| the assessment-visibility white paper | judging sheets and captions | +| the tabletop mechanical-engines packet | how a show is scored | +| Nestor's sealed / draft / pending | any human-verified record | +| a year of fleet practice | how it got built in ninety minutes | + +None of those were made for this. All of them are true at once here for the first time. +That is the whole argument, and the table carries it better than any sentence I could +write about it. *Also honest here:* tour is genuinely the ideal local-first case — gyms with no wifi, a hundred and fifty people, no budget for per-seat SaaS. This isn't a hobby detour from the From c928f5b69f4df3ddf09ea371889d91c46d0c3396 Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 04:29:04 +0000 Subject: [PATCH 14/19] =?UTF-8?q?drafts:=20the=20build=20landed=20?= =?UTF-8?q?=E2=80=94=20anchor=20section=20VI=20in=20safe-app-store=20#112?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The DCI work is real and merged: two apps, playground tier, deliberately not promoted because promotion requires verified_by != author. apps/marching-arts is the authorization core, and it contains the line the essay is built around: refusal is invisible, tested as indistinguishability. A member who declined and a member who is absent produce identical rows, counts and subject lists, because if they differed, declining would itself become the signal. That is section V's rule re-derived and come out backwards. In Grove, conflating unreachable with clear is the bug. Here, distinguishing declined from absent is the bug. Same question about what an absence means, opposite answer, because the person being protected changed from an operator to a seventeen-year-old. apps/field-acoustics is the concrete capability the piece was missing, and it has no competitor: every drill design tool models visuals, and nothing models what the drill sounds like from the stands. Its provenance model -- measured, fitted, assumed, propagated by min, a result worth its weakest one -- is Nestor's three states re-derived in acoustics. It also carries a line that rhymes with tonight exactly: the load-bearing test survived two independent reimplementations agreeing to 1e-14 dB with each other on the same wrong input. Two systems agreeing is not verification. Written the same night a second model handed this packet two confabulated citations labelled VERIFIED. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- drafts/what-the-parts-were-for-FRAME.md | 69 +++++++++++++++++++++++++ 1 file changed, 69 insertions(+) diff --git a/drafts/what-the-parts-were-for-FRAME.md b/drafts/what-the-parts-were-for-FRAME.md index c791919..7373b32 100644 --- a/drafts/what-the-parts-were-for-FRAME.md +++ b/drafts/what-the-parts-were-for-FRAME.md @@ -169,6 +169,75 @@ becomes cute. State it, let it sit, move. ## VI. Ninety minutes, and the inversion +**The build is real and landed: safe-app-store PR #112, merged 2026-07-29 ~04:25 UTC.** +Two apps, playground tier, deliberately not promoted (`promote_check.py` returns NOT +PROMOTED, fail-closed — promotion requires `verified_by ≠ author`). The platform is +unnamed; `marching-arts` is a placeholder directory. + +### `apps/marching-arts` — the authorization core + +The resolver compiles to exactly one predicate: +`(allow₁ OR allow₂ OR …) AND NOT (deny₁ OR deny₂ OR …)`. Roles grant nothing on their +own. At L3 and above the payload is `NULL` in the SELECT list and only the derived +instruction is served. **L5 is never served to anyone under any grant** — safeguarding +intake is routed to the people whose job it is to receive concerns, and is deliberately +absent rather than deferred. + +**And the line the essay is built around:** + +> Refusal is invisible, tested as indistinguishability. A member who declined and a +> member who is absent produce the same rows, the same count, and the same subject +> list. If they differed, declining would become the signal and every member who +> exercised the choice would be marked by exercising it. + +**This is §V's rule, re-derived and come out backwards, and that is the whole point.** +In Grove, conflating *unreachable* with *clear* is the bug. Here, distinguishing +*declined* from *absent* is the bug. Same question — what does an absence mean — opposite +answer, because the person being protected changed from an operator to a seventeen-year-old. +The rule wasn't applied. It was re-derived for a different beneficiary. + +Supporting detail worth keeping, all of it mechanism rather than prose: guarantees +enforced by `CHECK` constraints rather than documentation; `COUNT(*)` evaluated in SQLite +*under* the predicate rather than in Python over fetched rows; an AST walk proving no +module can reach the network. Mutation-tested with three deliberate breaks, each caught +by exactly the test claiming to cover it — *"a gate that cannot fail is not a gate."* + +And it was driven for real, not just tested: nobody consented → an **empty list, not three +greyed rows**; a guardian seals a craft-band grant; the system infers a second and it stays +inert; the grant widens to health and the diagnosis stays behind while the instruction +comes through; silent revocation; a principal holding every role sees nothing. + +### `apps/field-acoustics` — the capability + +The concrete thing the piece needed, and it has no competitor in the category: every drill +design tool models visuals — Pyware's Virtual Clinic runs four analyzers, all four visual. +**Nothing models what the drill sounds like from the stands.** + +Octave-band energy summation, ISO 9613-1 atmospheric absorption, circular-piston +directivity through a fitted effective aperture, one geometrically-gated grandstand +reflection. Two independent implementations held to each other across five differential +tiers. + +**Provenance is a first-class output** — every input is `measured`, `fitted` or `assumed`, +propagated by `min()`, *"a result is worth its weakest one."* Deliberately not a confidence +score: *"a number either traces to something a person can look up or it does not."* The +headline reads `ASSUMED` today and says so, because the rear hemisphere rests on an +asserted front-to-back array. + +That is Nestor's sealed/draft/pending, re-derived in acoustics. Third instance of the same +discipline, in a third domain, arrived at independently each time. + +**And the line that rhymes with this session:** + +> It survived two independent reimplementations agreeing to 1e-14 dB *with each other, on +> the same wrong input.* + +Two systems agreeing perfectly is not verification. Written the same night a second model +handed this packet two confabulated citations labelled VERIFIED. Same lesson, two domains, +neither one borrowed from the other. + +### The original section notes follow + Tonight: opened the abandoned pitch and started building it. Apache-2.0, open source, a give-back to the drum corps community. Broad operational tooling for a corps. From ccd65d04c6040b6fdb15aadc3885571631662d4c Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 04:36:59 +0000 Subject: [PATCH 15/19] professional: add Working With Agents A practice document for the technical packet, in the case-study form: the method for building with agents where machine output is never treated as verified, the authority to ask is structurally separated from the authority to grant, and every artifact records how much of it a human checked. Leads with the problem an employer actually has -- almost nobody can say which parts of an agent-assisted artifact a human checked -- and names three real failures from my own systems: a contract clause served at 0.974 similarity and marked verified, a pane that caught every exception and painted "queue clear", and a second model returning VERIFIED against two paper titles it invented. All three were caught by a column recording how the result was obtained. Six practices, each tied to code rather than testimonial. The absence rule includes its own inversion: in marching-arts, refusal is tested as indistinguishable from absence, because the person being protected changed from an operator to a minor. The agent-limits section is built on a failed survey rather than a success: survey agents cannot tell a port from an original, because only the intended direction separates them and that lives with the operator. Two corollaries -- corroboration between systems is not verification, and a gate that cannot fail is not a gate. safe-app-store #112 is the worked example. Claims are bounded in the usual three tiers, including that this is single-operator work with no external audit and that the apps are one evening old and deliberately unpromoted. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- professional/working-with-agents.md | 301 ++++++++++++++++++++++++++++ 1 file changed, 301 insertions(+) create mode 100644 professional/working-with-agents.md diff --git a/professional/working-with-agents.md b/professional/working-with-agents.md new file mode 100644 index 0000000..196cf17 --- /dev/null +++ b/professional/working-with-agents.md @@ -0,0 +1,301 @@ +# Working With Agents + +**Sean Campbell** · Technical practice document · July 2026 + +**Tags:** `agent-orchestration`, `ai-verification`, `authorization-design`, `provenance`, +`fail-closed`, `local-first-ai`, `governance` + +**One-line summary:** A working method for building with AI agents in which the machine's +output is never treated as verified, the authority to *ask* is structurally separated from +the authority to *grant*, and every artifact records how much of it a human actually +checked. + +**Status:** Live practice across seven public repositories. Evidence is code and commit +history, not testimonial. + +--- + +## The problem this solves + +Almost every technical organisation is now shipping agent-assisted work. Very few can +answer the question that matters about any given artifact: **which parts of this did a +human actually check?** + +The failure is not that models are wrong. It is that a plausible wrong answer and a +verified right one are, by default, the same pixels. A matcher serves a contract clause at +0.974 similarity and marks it verified. A dashboard catches an exception, returns an empty +list, and paints a green tick. A second model, asked to verify a citation list, returns +`VERIFIED / CONFIRMED` against two paper titles it invented. + +All three of those are real, all three happened in my own systems, and all three were +caught by the same thing: **a column recording how the result was obtained.** Not +intelligence. Bookkeeping. + +What follows is the method that bookkeeping belongs to. + +--- + +## Six practices, and where each one lives + +### 1. Machine output is `draft` until a human seals it + +Three states, never two. In [Nestor](https://github.com/rudi193-cmd/Nestor): **sealed** (a +human verified it — served verbatim, forever, with provenance), **draft** (a machine +produced it — queued for review, never served as verified), **pending** (nothing to offer, +said plainly rather than improvised). + +The economics are the point: each human verification becomes permanent capital. Cost per +answer falls as trust accumulates. *Verified once, served forever.* + +The same shape recurs wherever the domain changes. In `apps/field-acoustics` every model +input is `measured`, `fitted` or `assumed`, propagated by `min()` — *a result is worth its +weakest one*. Deliberately not a confidence score: a number either traces to something a +person can look up, or it does not. + +### 2. Absence is never rendered as an answer + +A surface that cannot reach its data must not report that the data is empty. + +I found this in my own code and wrote it down as a load-bearing constraint: a monitoring +pane caught every exception, returned `[]`, and rendered `✓ queue clear` — so *"I could not +reach the queue"* and *"nothing is waiting"* were identical pixels, with a green check on +the failure. The same fail-soft appeared eighteen times in one file, which made it house +style rather than a bug. + +The rule generalises across the fleet. A missing consent file reads as denied — *absence is +not consent*. A malformed permission scope confines an app to **no** collections rather than +all of them, because a policy that cannot be parsed is not consent. An unasserted severance +claim reports `not_asserted` rather than passing. + +**And it inverts where the beneficiary changes**, which is the part that shows it is a +principle rather than a habit. In `apps/marching-arts`, refusal is *tested as +indistinguishability*: a member who declined and a member who is absent must produce the +same rows, the same count, and the same subject list. If they differed, declining would +itself become the signal, and every member who exercised the choice would be marked by +exercising it. Same question about what an absence means; opposite requirement, because the +person being protected changed from an operator to a minor. + +### 3. Asking and granting are separate authorities, enforced by the filesystem + +An agent may *request* a capability. It may never *grant itself* one. In +[willow-mcp](https://github.com/rudi193-cmd/willow-mcp) this is structural rather than +conventional: + +- `grant-net` (mint an egress lease) is local CLI only — never exposed as an MCP tool +- `confirm-binding` (map an authenticated identity to an app) is operator-only, because a + remote caller must never confirm its own binding +- `allow-permission` is CLI-only, because an agent must never be able to grant itself a + permission it was just denied +- Network egress requires **three standing keys plus a one-use signed envelope**: a + capability flag in the manifest, a global consent switch, a time-boxed per-app lease, and + a signed task envelope whose private key lives outside the agent's filesystem view + +The severance model formalises the distinction that makes this coherent: store and database +are **data** surfaces — someone who writes them corrupts records, and the verdict degrades. +Trust root and egress are **authority** surfaces — someone who writes them grants themselves +what the boundary was supposed to deny, and the verdict breaks. + +### 4. The residual is stated plainly, not buried + +willow-mcp's README contains a section titled *"The residual, stated plainly"*: on a host +where the agent and the server run as the same uid, the agent can write the very files that +authorise its egress. Leases make a self-grant expire and leave a record; a pre-tool hook +blocks the obvious attempts; the operating system is not stopping it. It is tracked as a +numbered bug, `diagnostic_summary` names exactly which keys the running process could forge, +and strict mode ships **off** by default because enabling it before the uid separation exists +would deny egress on every current install. + +[willow-gate](https://github.com/rudi193-cmd/willow-gate) opens with *"Enforcement vs. audit +— read this first"* and concedes that unwired it is a loud ledger, not a gate. + +Nestor publishes its measured false-seal rate rather than claiming accuracy, and states the +conclusion against its own interest: **there is no threshold that is simultaneously safe and +useful.** At 0.96 the hardest corpus is clean and effectively dead; at 0.92 it serves real +rewrites and gets roughly one answer in six wrong. + +A compliance buyer knows "we are accurate" is a lie. "Here is our measured +false-verification rate, here is the dial that sets it, here is the harness — run it +yourself" is stronger *because* it admits a failure rate. + +### 5. Capabilities are earned, not scaffolded + +willow-mcp ships four live integration adapters and six **declared stubs** that refuse +fail-closed, each naming what it needs and what would earn its implementation. A stub that +refuses loudly and states its own precondition is honest; a stub that half-works is a +liability. + +The same rule governs promotion. `apps/marching-arts` and `apps/field-acoustics` are +playground tier and **deliberately not promoted** — the gate returns `NOT PROMOTED` +(fail-closed) because promotion requires `verified_by ≠ author`. You do not verify your own +work. + +### 6. Repeated conversational labour becomes a script + +When I notice a model re-deriving the same reasoning across sessions, that is a tool waiting +to be written. willow-mcp's `tools/` directory exists for exactly this — *"each turns +conversational labour into a script, so the next session runs the tool instead of +re-deriving the work."* + +This is the compounding discipline, and it is what separates building from producing: every +session should leave the environment better instrumented than it found it. `gap_log` records +what is *not* known so it is not rediscovered. The hash-chained ledger outlives the session +that wrote it. Dead ends are recorded with their reasoning so they are not re-derived — the +`IDEAS.md` convention tags every entry **measured / verified / hypothesis / open**, so the +confidence level travels with the claim. + +--- + +## What agents can and cannot do + +This is the most practically useful thing I have learned, and it came out of a failed survey +rather than a success. + +I ran read-only survey agents across four repositories to map a migration. They produced an +accurate inventory and one confidently wrong conclusion: they identified a ported module as +an orphan and the original as the port. My own correction, recorded in the repo: + +> The survey agents were asked where code **is**, and that question cannot tell a port from +> an original — both look like live implementations from that angle. Only the migration's +> intended direction separates them, and that information existed nowhere in the +> repositories; it came from the operator. +> +> …ask a human which way the migration runs before inferring anything further. + +**Agents map what exists. Only the operator holds the direction of travel.** Intent is not in +the artifacts, so it cannot be surveyed, delegated, or recovered from the code — and an agent +asked an inventory question will answer it accurately and mislead you completely. + +Two corollaries that follow directly: + +**Corroboration between two systems is not verification.** A load-bearing acoustics test +survived two independent reimplementations agreeing to 1e-14 dB *with each other, on the same +wrong input*. Separately, a second model asked to verify a citation list returned +`VERIFIED / CONFIRMED` against two paper titles it had invented — correct journal, correct +authors, correct year, correct subject, wrong title. A wrong fact is caught by the next +reader; a wrong citation is copied forward forever. + +**A gate that cannot fail is not a gate.** Authorization logic in `apps/marching-arts` is +mutation-tested: three deliberate breaks introduced, each caught by exactly the test claiming +to cover it and only that test. Including the subtle one — dropping the parentheses around a +joined deny clause, after which only the first term binds, nothing raises, and every row the +denies were meant to withhold silently becomes visible. + +--- + +## Worked example: a marching-arts platform, in one evening + +**Status:** Playground tier, merged, not promoted. Two apps, Apache-2.0. +**Evidence:** [safe-app-store PR #112](https://github.com/rudi193-cmd/safe-app-store/pull/112) + +### Problem + +Drum corps organisations run a touring operation — a hundred and fifty people, most of them +under 22 and many of them minors — largely on spreadsheets and group chat. They have no +budget for per-seat SaaS, frequently no connectivity at the venue, and a duty of care around +member data that off-the-shelf tooling does not model. + +### Constraints + +- Offline-first and local-first; the organisation runs it on its own hardware +- Sensitive data about minors: rosters, medical information, emergency contacts, guardians +- No per-seat cost, no vendor able to revoke access +- Consent must be real, which means refusing must be costless and invisible + +### What I built + +- **An authorization resolver** compiling to exactly one predicate: + `(allow₁ OR allow₂ OR …) AND NOT (deny₁ OR deny₂ OR …)`. Roles grant nothing on their own. + At sensitivity L3 and above the payload is `NULL` in the SELECT list and only a derived + instruction is served. L5 — safeguarding intake — is never served to anyone under any + grant, and is routed to the people whose job it is to receive concerns. +- **Guarantees as mechanisms, not documentation:** `NOT NULL` plus non-blank `CHECK` on + provenance; a sealed grant without a signer refused by constraint; `COUNT(*)` evaluated in + SQLite *under* the predicate rather than in Python over fetched rows; an AST walk proving + no module in the core can reach the network. +- **An acoustic propagation model** — octave-band energy summation, ISO 9613-1 atmospheric + absorption, circular-piston directivity, one geometrically-gated grandstand reflection — + with two independent implementations held to each other across five differential tiers. + +### Technical judgment + +- **The schema is inverted relative to how organisational software is normally built.** The + member is the root, not a row hanging off a program. That makes protection of minors + structural rather than a policy layer added later, and it is free now and expensive to + retrofit. +- **Refusal is invisible by construction**, and that property is tested as + indistinguishability rather than asserted in a privacy policy. +- **Nobody consented renders as an empty list, not three greyed rows.** Greyed rows would + disclose exactly who declined. +- **The novel capability is acoustic, not visual.** Every drill design tool in the category + models visuals. Nothing models what the drill sounds like from the stands — which is the + one place a model here is not competing with mature commercial software. +- **Not promoted.** Playground tier is contested, not canonical, and promotion requires a + verifier who is not the author. + +--- + +## What this document claims + +**Safe to claim:** + +- The practices above are implemented in public code and can be inspected. +- The failure modes named — false seal at 0.974, absence rendered as assurance, confabulated + citations labelled verified, a survey that could not distinguish a port from an original — + are all real, all mine, and all recorded in the repositories where they occurred. +- Provenance and verification status are first-class columns in these systems, not + documentation. + +**Frame carefully:** + +- Most of this is single-operator work. It has not been through a team's code review process + or an external security audit. +- The marching-arts and field-acoustics apps are one evening old, playground tier, and + deliberately unpromoted. They demonstrate a method; they are not production systems. +- Adoption is limited. These are working systems with real users in a small number, not + products with a user base. + +**Do not claim:** + +- That the authorization model has been penetration-tested or independently reviewed. +- That any of this makes an AI system safe. It makes what a human checked *legible*, which + is a smaller and more defensible claim. +- Peer review, certification, or compliance attestation of any kind. + +--- + +## Transferable strengths + +1. **Authorization and trust-boundary design** — separating request from grant structurally, + fail-closed defaults with the failure direction argued rather than assumed, and a stated + residual where the boundary is convention rather than enforcement. +2. **Verification and provenance engineering** — three-state models (sealed/draft/pending, + measured/fitted/assumed, empty/populated/unknown) applied across translation, physics, and + evidence bases, with retrieval status recorded as data. +3. **Agent orchestration with honest limits** — knowing what a survey agent can and cannot + answer, and designing the human's contribution around direction and intent rather than + volume. +4. **Adversarial self-testing** — mutation testing on authorization logic, differential + testing across independent implementations, and publishing measured failure rates rather + than accuracy claims. +5. **Technical writing that survives contact** — documentation that states its own residual + risk, records dead ends with reasoning, and tags every claim with its confidence level. + +--- + +## Evidence index + +| System | What it demonstrates | +|--------|----------------------| +| [Nestor](https://github.com/rudi193-cmd/Nestor) | Sealed/draft/pending; measured false-seal rates; rejection as a first-class human decision; hash-chained audit | +| [willow-mcp](https://github.com/rudi193-cmd/willow-mcp) | Three-key egress; manifest ACL; severance model; the stated residual; earned-not-scaffolded adapters | +| [willow-gate](https://github.com/rudi193-cmd/willow-gate) | Trust ladder with bound identity; enforcement-vs-audit honesty; the friction floor | +| [kartikeya](https://github.com/rudi193-cmd/kartikeya) | Sandboxed task execution; host-agnostic queue; network-isolated by default | +| [safe-app-store #112](https://github.com/rudi193-cmd/safe-app-store/pull/112) | The worked example: authorization core, mutation testing, provenance propagation | +| [Willow](https://github.com/rudi193-cmd/Willow) | Constitutional layer — envelopes, protected agents, ratification gates | +| [Willow systems deep dive](willow-systems-portfolio.md) | Backend and data-plane detail | + +--- + +*Related: [Portfolio case studies](portfolio-case-studies.md) · +[Willow ecosystem inventory](willow-ecosystem-inventory.md) · +[Research portfolio](../research/README.md)* From 97ff31dce4fc4c687769cdcd7bcaace6011fce0c Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 29 Jul 2026 04:38:15 +0000 Subject: [PATCH 16/19] professional: index Working With Agents in the packet Three entries. In README.md it lands in both the how-to-read table (as its own reader intent -- someone who wants to know how I work with agents, distinct from the 3-minute and 10-minute paths) and the document index. In portfolio-case-studies.md it goes in the top table as "Practice" rather than a numbered case study, since it is a method document with a worked example inside it rather than a fourth project. It sits after the three cases and before the research thread. All relative links in both files verified to resolve. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- professional/README.md | 2 ++ professional/portfolio-case-studies.md | 1 + 2 files changed, 3 insertions(+) diff --git a/professional/README.md b/professional/README.md index cf09dee..a779b36 100644 --- a/professional/README.md +++ b/professional/README.md @@ -25,6 +25,7 @@ This folder is a technical-employer-facing packet. It translates work from [Disp |----------------|-----------| | 3 minutes | [portfolio-case-studies.md](portfolio-case-studies.md) | | 10 minutes | [willow-systems-portfolio.md](willow-systems-portfolio.md) | +| Want to know how I work with AI agents | [working-with-agents.md](working-with-agents.md) | | Need research context | [Research portfolio](../research/README.md) | | Need systems evidence | [willow-ecosystem-inventory.md](willow-ecosystem-inventory.md) | | Need structure | [case-study-template.md](case-study-template.md) | @@ -40,6 +41,7 @@ This folder is a technical-employer-facing packet. It translates work from [Disp | Document | Purpose | |----------|---------| | [portfolio-case-studies.md](portfolio-case-studies.md) | Three polished case studies for technical employers, including the Willow systems-building story | +| [working-with-agents.md](working-with-agents.md) | The method: machine output as `draft` until sealed, asking separated from granting, provenance as a first-class column — with the failures that taught each rule | | [willow-systems-portfolio.md](willow-systems-portfolio.md) | Deep Willow case study: architecture, 10 proof points, claim boundaries | | [willow-ecosystem-inventory.md](willow-ecosystem-inventory.md) | Curated repo census, SAFE apps, data artifacts, research substrate | | [case-study-template.md](case-study-template.md) | Reusable structure for future writeups | diff --git a/professional/portfolio-case-studies.md b/professional/portfolio-case-studies.md index 5a74614..e3d8d13 100644 --- a/professional/portfolio-case-studies.md +++ b/professional/portfolio-case-studies.md @@ -11,6 +11,7 @@ Three projects that show the same through-line: build real systems for real user | **1. AI Literacy Curriculum** | Product thinking for non-technical users; AI risk communication | [AI literacy index](../lessons/ai-literacy-9-12-index.md) | | **2. Assessment Visibility** | Governance design; research synthesis; teacher-facing implementation | [White paper](../education/assessment-visibility-v1.1/white-paper.md) | | **3. Willow Systems Building** | Backend infrastructure; Postgres data plane; agent orchestration | [Willow systems deep dive](willow-systems-portfolio.md) | +| **Practice** | How I work with AI agents: verification states, trust boundaries, provenance as data | [Working With Agents](working-with-agents.md) | | **Research thread** | Sociotechnical analysis; Paperclip lineage; right-to-fix and repair literacy | [Research portfolio](../research/README.md) | --- From 31042396924fb431abea0d31c9b986a1d2fd86a5 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 30 Jul 2026 02:08:50 +0000 Subject: [PATCH 17/19] drafts: the thesis is locational, not instrumental Revised after a day's thought. The claim is now that convergence happens in the place where your passions are -- an address rather than a technique. The earlier version ("build something fun to bring the pieces together") was instrumental and therefore fakeable: pick a hobby project, deploy it strategically, expect synthesis. That is not what happens. The parts met at a drum corps app and did not meet during three weeks of deliberate consolidation whose entire purpose was to make them meet. Adds the mechanism, which is what keeps this from being "do what you love": convergence requires simultaneous non-negotiability. Everywhere else you cut scope, and each compromise is individually reasonable and collectively fatal, because parts only integrate when all of them are required at once. At the site of a passion you refuse to cut any of it. So fun becomes precise and is not enjoyment -- it is the domain where you will not accept a reduced version. Section IV's diagnosis is corrected accordingly. Not backwards-facing, not even application-free: the wrong location. No amount of effort produces convergence in a room where nothing is being built. Effort was not the missing input. Adds an explicit guardrail against the essay's failure mode, with a test: if a paragraph would survive being pasted into a LinkedIn post, cut it. Retitled to "You Don't Get to Pick the Place". Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- ...> you-dont-get-to-pick-the-place-FRAME.md} | 89 ++++++++++++++----- 1 file changed, 67 insertions(+), 22 deletions(-) rename drafts/{what-the-parts-were-for-FRAME.md => you-dont-get-to-pick-the-place-FRAME.md} (79%) diff --git a/drafts/what-the-parts-were-for-FRAME.md b/drafts/you-dont-get-to-pick-the-place-FRAME.md similarity index 79% rename from drafts/what-the-parts-were-for-FRAME.md rename to drafts/you-dont-get-to-pick-the-place-FRAME.md index 7373b32..831c9f6 100644 --- a/drafts/what-the-parts-were-for-FRAME.md +++ b/drafts/you-dont-get-to-pick-the-place-FRAME.md @@ -1,26 +1,61 @@ -# What the Parts Were For — frame +# You Don't Get to Pick the Place — frame -*Working title (alternates: "A Toolbox Is Not a Shop", "Ninety Minutes"). Frame only — -structure, load-bearing material, and what's still missing. No prose written yet.* +*Working title (alternates: "What the Parts Were For", "Where It Comes Together"). +Frame only — structure, load-bearing material, and what's still missing. No prose +written yet.* -**Working subtitle:** *Sometimes you have to build something you care about to find out -what you've been making.* +**Working subtitle:** *Convergence happens where your passions are, and that is a +location rather than a technique.* --- ## The claim -**Sometimes you need to build something fun to bring the pieces together — fun meaning -passionate about, not recreational.** +**Convergence happens in the place where your passions are.** -Not a break from serious work. The thing that *integrates* it. A year of building -components — a local-first stack, an authorization model, an assessment framework, a -rules-engine research packet, a verification ledger — each made in isolation, each -correct on its own terms, none of them ever required to be true at the same time. +Not "build something fun to unlock integration" — that was the earlier, weaker version, +and it was instrumental. A technique can be faked: pick a hobby project, deploy it +strategically, expect synthesis. That is not what happens. Convergence has an **address**, +and you do not choose it. -A toolbox doesn't tell you what it's for. A job does. And the job has to be one you -actually want the outcome of, because nothing weaker pulls hard enough to make separate -parts prove themselves together. +A year of building components — a local-first stack, an authorization model, an assessment +framework, a rules-engine research packet, a verification ledger — each made in isolation, +each correct on its own terms, and **not one of them ever required to be true at the same +time as the others.** They met at a drum corps app, which is the one place in this work +with personal stakes running back to being seventeen. They did not meet during three weeks +of deliberate consolidation whose entire purpose was to make them meet. + +### The mechanism — and this is what keeps it from being "do what you love" + +**Convergence requires simultaneous non-negotiability.** + +Everywhere else, you cut scope. You accept a reduced version of the thing you don't +especially care about — good enough auth, good enough offline, we'll do provenance later. +Each compromise is individually reasonable and collectively fatal, because parts only +integrate when they are all required *at once*. + +At the site of a passion you refuse to cut any of it. The acoustics **and** the consent +model **and** the offline case **and** the protection of minors, all non-negotiable in the +same evening. That simultaneous refusal is the thing that forces separate components to +prove themselves against each other. + +So "fun" becomes precise, and it isn't enjoyment: **fun is the domain where you won't +accept a reduced version.** That's why it can't be chosen strategically, and why effort is +not a substitute. + +### Guardrail — the failure mode of this essay + +This is one short step from "follow your passion," which is advice, is unfalsifiable, and +would make the piece worthless. The claim is narrower and stranger: + +- **Not** that passion makes you productive. The three tedious weeks were the *highest* + output of the period. +- **Not** that you should work on what you love. Plenty of the components were built as + obligations and are good. +- **But** that integration has a location, that the location is not chosen, and that + recognising where it is tells you something about where to point next. + +If a draft paragraph would survive being pasted into a LinkedIn post, cut it. The evidence: ninety minutes on a drum corps app used more of the last year's work than three weeks of consolidating it did. @@ -128,12 +163,18 @@ someone else, and the sharper diagnosis is here rather than in "I was tired." The one contribution no agent can supply is direction. For three weeks I spent all of it pointing backwards — not choosing what should exist, adjudicating what already did. -But the deeper problem was not that it was backwards. It was that **nothing needed the -parts.** A migration merges components without ever making them do a job together. Three -weeks tidying a toolbox, and a toolbox cannot tell you what it is for. +But the deeper problem was not that it was backwards, and it was not even that nothing +needed the parts. **It was the wrong location.** A migration merges components without ever +requiring them to be true at once — and it is also, definitionally, outside anything I care +about the outcome of. Nobody is passionate about reconciling two versions of their own +repository. That is not a character flaw; it is what the work is. -High output, no application. That's not burnout and it isn't laziness. It is what it -feels like to maintain an inventory nobody is drawing from — including you. +So no amount of effort was going to produce convergence there. Effort was not the missing +input. **Location was.** Three weeks tidying a toolbox, in a room where nothing was going to +be built. + +High output, no convergence. That's not burnout and it isn't laziness. It is what it feels +like to maintain an inventory in a place where nothing draws from it — including you. *Still open (see below): was the migration chosen, or did it keep demanding attention until nothing was left over?* The section reads differently depending, and it should be @@ -272,9 +313,13 @@ made separately for its own unrelated reason: | Nestor's sealed / draft / pending | any human-verified record | | a year of fleet practice | how it got built in ninety minutes | -None of those were made for this. All of them are true at once here for the first time. -That is the whole argument, and the table carries it better than any sentence I could -write about it. +None of those were made for this. All of them are true at once here for the first time — +**and "at once" is the load-bearing phrase, not "all."** The consolidation touched every one +of these parts too. What it never did was require them to hold simultaneously, because it +never had to satisfy anybody. + +That is the whole argument, and the table carries it better than any sentence I could write +about it. *Also honest here:* tour is genuinely the ideal local-first case — gyms with no wifi, a hundred and fifty people, no budget for per-seat SaaS. This isn't a hobby detour from the From f6d11c54a08292e12614a03561c7d6907e29226c Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 30 Jul 2026 05:08:54 +0000 Subject: [PATCH 18/19] drafts: terpsi-music #1 is the proof, not the illustration The platform has a name, a repo and a design document, and the scope moved with it -- no longer a drum corps tool but music-program management holding minors' education records, FERPA in frame. DCI was the door; the room is bigger. The PR's own summary is the essay's thesis: land the design, and in the process find that most of what it needs already exists across the fleet, several times more strictly than proposed here. Two details make it proof rather than argument, and both are the author's own record. Section 7.4 adopts the Ward Case from PROTECTED_AGENTS.md Part III -- the guardianship doctrine the document asserted three times did not exist. Section 16 finds that Nestor is the engine it had been specifying. He wrote these, forgot them, designed around their absence, and found them on arrival. That is stronger than the toolbox metaphor the frame started with. A toolbox can be inventoried. This is discovering your own tools by needing them. Also records the unchecked evidence boxes for section III: the component map was built from READMEs and PR bodies, no source was read, and the PR says so in the negative. The same limitation this session hit with its own convergence table, reached independently within 48 hours. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- .../you-dont-get-to-pick-the-place-FRAME.md | 69 +++++++++++++++++++ 1 file changed, 69 insertions(+) diff --git a/drafts/you-dont-get-to-pick-the-place-FRAME.md b/drafts/you-dont-get-to-pick-the-place-FRAME.md index 831c9f6..d7478e6 100644 --- a/drafts/you-dont-get-to-pick-the-place-FRAME.md +++ b/drafts/you-dont-get-to-pick-the-place-FRAME.md @@ -277,6 +277,75 @@ Two systems agreeing perfectly is not verification. Written the same night a sec handed this packet two confabulated citations labelled VERIFIED. Same lesson, two domains, neither one borrowed from the other. +### The sequel, and the actual proof: terpsi-music PR #1 (2026-07-30) + +**This supersedes the convergence table as the section's evidence, because it is a +*finding* rather than an argument — and it is the author's, not mine.** + +A day later, the platform has a name (Terpsichore — muse of dance and chorus), a repo, and +a design document. Scope moved with it: no longer a drum corps tool but **music-program +management holding minors' education records**, FERPA in frame. DCI was the door; the room +turned out to be bigger. + +Four files, 2,113 lines, thirty-two commits, and **no application code.** The PR's own +one-line summary is the essay's claim: + +> Land the design for a music-program management application holding minors' education +> records — and, in the process, **find that most of what it needs already exists across +> the fleet, several times more strictly than proposed here.** + +**The two details that make this proof rather than illustration:** + +- **§7.4, the Ward Case.** `Willow/PROTECTED_AGENTS.md` Part III is the guardianship + doctrine *"this document asserted three times did not exist."* He wrote it, forgot it, + asserted its absence three times while designing around the gap, and found it on arrival. +- **§16, the bilateral pattern.** *"Nestor turns out to be the engine this section was + specifying."* The spec was written; the implementation already existed. + +You don't get to pick the place — and you may not know what you've built until you get +there. **That is a stronger claim than the one the frame started with**, and it is not +available from the toolbox metaphor: a toolbox you can inventory. This is a man discovering +his own tools by needing them. + +**Supporting material, all of it usable:** + +- *"Where this document and the code disagreed, the code won and the document says so"* — + five recorded corrections, three of them to claims the document had asserted repeatedly. +- **§4.1** collapses "parents need access" into three requirements, ~95% of which need no + app at all. Governing rule: **SMS carries signals, never records** — where a transport + cannot be made incapable, the payload is made not worth reading. +- **§7.2, the knock.** willow-gate's thirteen-in / thirteen-out is *"the only mechanism in + the fleet that compares outcome against promise."* +- **§15** — three ordinal scales that must not be confused: `T0–T4` ascends toward + privilege, `L1–L5` toward restriction, `P1–P5` provenance never gates. Seal state and + confidence kept as separate axes. +- **§17** — this app as the template, *"with the caution that a template without a + conformance check is just the first copy."* +- **`tests/test_section_refs.py`** exists because `CLAUDE.md` claimed its section + references kept it honest and nothing verified them — *"a middle that cannot fail, which + §16 argues is worse than none."* Seven tests, two real defects caught on first run, + mutation-verified by renumbering a live reference. + +**And the unchecked evidence boxes, which belong in §III on practice:** + +> - [ ] Not verified: §14's "Exists" column. Assembled from READMEs and merged PR +> descriptions — not from reading or running source. Every row claiming something exists +> is a cited claim whose source was read and never executed. +> - [ ] No source code was read in any fleet repository. +> - [ ] Not read: `sean-data-vault` (access denied, not retried). + +That is the same limitation this session hit and recorded — a convergence table built from +READMEs is a claim about documentation, not about code. Both of us reached it independently +in the same 48 hours, and both wrote it down instead of smoothing it. + +Also: roughly eighteen dead canonical links, a catalog advertising encryption `u2u` does +not implement, an unpinned `@master` dependency on the fleet's own verification tool, and a +cloud-inference fallback that fires exactly when the local model is down. **All recorded, +none touched — they belong to their own repos.** The discipline of finding a fault and not +fixing it is worth a sentence. + +*(Minor, and probably not for the piece: the branch is `claude/good-evening-wrldb5`.)* + ### The original section notes follow Tonight: opened the abandoned pitch and started building it. Apache-2.0, open source, a From 133df686be66b2a787fdec13f1a7529e9e1c306a Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 30 Jul 2026 20:35:15 +0000 Subject: [PATCH 19/19] chore: drop committed bytecode, add .gitignore Two .pyc files from running the E4 harness were committed by a git add -A. Removes them from tracking and adds the repo's first .gitignore covering __pycache__ and bytecode. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01TVpzDG9XcRqJVbxGQ1EGVW --- .gitignore | 2 ++ .../e4/__pycache__/score.cpython-311.pyc | Bin 7510 -> 0 bytes .../e4/__pycache__/stimuli.cpython-311.pyc | Bin 4379 -> 0 bytes 3 files changed, 2 insertions(+) create mode 100644 .gitignore delete mode 100644 research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc delete mode 100644 research/boxes-and-escape/e4/__pycache__/stimuli.cpython-311.pyc diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..43ae0e2 --- /dev/null +++ b/.gitignore @@ -0,0 +1,2 @@ +__pycache__/ +*.py[cod] diff --git a/research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc b/research/boxes-and-escape/e4/__pycache__/score.cpython-311.pyc deleted file mode 100644 index 868ecad0513ce7ed54baeb3244c854881c6eccdc..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 7510 zcmb_hYitx(magje%Wh+1u)%@Cgy06cy(R_>czGEcdw3XZ0}kNnt?nw@1>IH6t!ltM zWu<3@G#%vCI-^8*w29a6m4syND$?Y~O3pGPjrg_a z-0FwhY`o0uma9+Q_dTzB?su<$Sy$&{;Q8&2>FdEphWQW7lpm*+V4wXLBj;G3H5g!R|3L%p%K1yf_;<;HK+du*dC25M3SdCl8zB6 z&P^(W>+cTt4WAo3K6I{sB;xZ8C~3keX)UfKMK!{WO^fBmlFA8ULW+rm*TlpOmlWgE zye!4}1g9#THqC1s&&3jwEO07JBNC2>q&6-3rg;QAnTv@W44vYMC~*7~FUhLL ziF|yTQ=laQ1~_;$l9uBz3&;I?_8m^V!^bt4P?1G0p+LE$$XtquswiuFV6JH~$#asd zK`YujMnwJw4nup#VBktO)#^uX!R)aa?v|t^U^q|+Ob&d(7?La!Itmdd(<*RajiPG6 zj!K)ACIn$*MdOHA;)K=_5hNG(aG+OSy`gfqrzH>#G5HV(ye14ARDW8C@UqodsLxifD>xFHQIkR6l* zbxP7n&`yNwza^3xUrI_)R@IaugG@L`0&GS{ri$RX>2#8p!wLyg3x@eA=+32hRoz3) zC`c+lMMM!th2BK;f#3=OnK*CWkhGyq$Hb(92ZS&}JF6>)Z|+$4VsoGjw| z4jnkentrNDYF71J$YFZQ<~N9Iav-QP81ur-t6QvQSXKrl___Nz*ZdZ7{9KAVUXOM(05$c`6}>Dn6uZx0Omp z)h!d3{ZdcM)UujVok%ZvOh^2QPw+&VoGHy#au;AE^Dpfk<+-ERZtvF18MTs%@*1rHKLvkJtI={3eNvM7L^a#4 zgDONyR2>#6+FG9mCsfFcu1QH){IVEo*MEFX-Bp=%#XLCu%Zvcs51AhD5bBMl9lX`F zPbM|Avz8|E^PrXnv9CfoEN_iW#47<890md&V`mul83*(}qce+8@Z$3!6n$JNp0(*V z)h4i`j0U~RkLD<+7-n%rd9_rxeaON%@Ux!U+^oH3G*7wJpE;UCtAzEJ9f8;c+bE;k z$scqFd8WfAc7y7Q?APk5+z5`lj25isuU9w~oFF0esBUQ|nOe=QlGv#`1=kI#H<;T! z$_O5qpB22f7@`WkyNqsI^j2r8rfNrBJt5_X>(<>B8Qns??*2PQ2)=Kd1WgMK;6WRz+`XLFfTq?#zm*oKc$?6u+qC-X zh}Q$<^oP*q<4QXiqY36;kJi>5H)uJj2g}=pW|%zA{4LW}MJUR&R{Q(ia0w!b;k zIYJi|L^Eqs-vfb(R~0!F;RY*v%%ntMAI5MW3E2oPu}r;~OldTb(`5?92!b-b;+<4N zic1=-OxO}JMquc-cAm-(id#L#)oCR`App#Y8>8t=5k(c_mGuN8qAxY!dOie+`^*%q z=!uqDR%aI>|Kd~YN1txc*|%+Tt)t9GxRW>`T1cso&9p`S!emty*8;~qALyJ0G}8&& zIq5tpsVQCq%%BbsB{?GU2}zsjL@>dV_;hDXxg)A!04`xsjq@q7Q|#`f+him)^W4{) zfM|&DR49T;(LA?#l~Yq< zTmtM;F=fXpQ>OU%4T_`yRFvTe7Z($Wvh81`$cN(v0h+-wS0w>GMg%Bx2TCmXk=cFi zYGl{737xtuV(250UA>)UFmMEvT&*dOyu$)N;ahwnjn0Ayi%aq_d@`NTq*8gg^}f}N zTNwRt6u7OQJOCq=-_*T@+6r^r2o24A65`7u1B{ zr#^-_k!6-V4S!pIuRiZ-DR^44$BX^|xTG^u42F$h_}jX5?{2@p>*u=`F1#InC!DY2 z3Uyp|;AzA9cQ4(a`1!xC4c?=&U;64NB-bpHgB;a+CR2`(qX*b_vk`?*Qvs;Q&5=S zav3`2FWdg1 z?eDjIu!SzM?{T629!c)xQdjq{ou9QmYHrV~cfiKqU$xr0VCNcbDuV@KY9w%|XD^(Ude z&uHo^28WH{Fz9N&^@)3v;oeki-D))Kop;VV7bkMgVzAZNJg~SANa+m@7@mP8|DfR? zw4z7^O#9r2JIz%=ifa?513xvv7#$1{0Q>AQ1e%|D?=#wp9aU#)F_OgKpTNH7m}Rv} z=w+Cb%*1Q=yjta=gJE*sNw^ZZ>(QaEcAu>s5piZKuwXrJd3zx-w-b9!{1YH;R&Z2n ze+VH7&T1R2f=2=OuHdIU1H-K3TX5BG8MV1tJK!Tgrq$8!*zVY>Il+?wWE{80ZOAq7 z>aEuOAg{h^oBvl{9lGP?d7b-orL6=%VD10KPP5LM9)XwW;i~CT_YytaH9hKIqK8L! z>Mq@_dxW6S@SYQFu+FjtZ2FCvA7KzY#G%i_wFyLZ?uIy{manAXK$wnQ7vpg8gE)qV zZ)%$GxJ~VWdyS}t+M+2UDI@}S39M}TVl&Ya_BFkbw))ZH(pWWX3z7BfxVE|>@h*Fs zVd*Jzq7lK4z*IfqEt_>-`LteOFERI7mYHk9(^$xEx=ZtsjWEVJ(f|~&3nCS=3y8@o zDo(?3SsK`%dwb!!1%NdRtN%b$|>o=I`^7dIRQ;7I@z>Z zC!hL2M{HF`LRm)}1bazGn*|4;b*C9zWi5*hT>UGw_K}6}sMcx#9y-jLy81CRP_b1> zRZZhY)A+)+#hXUxP+{j`$UI^nh5j{QoWK0!*k$9`<@w`4sxNI%)%HJo2imcshx!-w zQ1OW%`ou?eavVCWkc^-r|1^=J*dq+2BvP=)2=>hLVDBCK3)>H1h6S?fT|G_+D=pS*1be|^WhaCTSPTnSzLfXO zPxju~OE2gV$iv+o5vKqWQ1Cn;sd(c`@5MQ|mWF%YlM21ZgnM%elCcmD7#zL$4XbcO z6*<|$&(>WqTme_Fi8+Ax5&%%VS4Yqt2{}z~N}7sV*RM`Y%4VG`rZoZ=hE}P!oRv)n z-ujtt2)b~9oC&$et01@O8NN8yH+;5#6mIy^Nr26wCWwkk_a1q{Y-_yEga5?7?Iv&s*`*dr&6tJE) zV9uU9Q}DDdY|DGv3{P9JZrw8DaK1uGjx7d)^Zr7h?VG^PCxM+0*asb-_Zci?l3(f@>x?n_1z7@n#0)@aU-vm0I1Ueq9 zdvNfRwqI#q%zife<@i@O^9L>z4qV6wMhbzEwMNMWzinuOV28m^1CnnVF92;AFEos2 zhn@z4R*3_r^v4nNC6dL2yzg`)!whK`Jb z_2z@U1)zc6LZCO>ztp_tL36RS{b4iUduYAvc5ZMjGeEL^Iqy%;EeDuDF#99w`QShF z0EGu2dWs@6<;N^uj)ff5UC3T6prz;GYyGJ486FYDcmitJt%o12ee6QrEF!8a>(LrNVPhZ&DG`3GoQI3Ry(lstE*w z6U#2!>R87zgQSgNw;N3R%6EzJX6fIOzbETiay4XQcdyTHUWmOXN+)``sy@`j6XS{c2 zNMao1R$GacA~h=&byZk3FJ-CNN_}Xn{)4_c$kJLPA+6e{zPX}(;i>1&*p34!w7r@+ zzH{!m=brC-{ADB}aWMXLEO-0IQI7i?`)CfoJKIn1Rw%ws|2^FmtSIqXS$ zu$UIHH|@nf?8gBd#Ns0{?ZXldVeca$?Z;sp!BM;qci>L!dE`q6a928r2k@a!#I%G5 z@nQH4kuW~8EU?q)j|7f{9>eUZdpN*!1aMJ!_mRkBkmM=6?s&-I-Y&ErMIT z(4sR-lh{_t)^BOTgY(sa5TQ)I3cwnGlI7!LMDoBlpV>W7O zap~{O!!h0tIkLHii23Z$Z9~^GhDnvIX^U%eUbQ`|0HS%BE@OQ~%b4;lmDrxFamV%+ zsBVy$XnP9Ek{xWgOsBT5Agd}dZD~aYkztFr924w-3>2nvhuD6NSSFQK$`Xi%7zhI> zy7XXrZP{7{#L4fBpFD~FTK)wyD7woWxQ~JvBy7jX?WxO$c^@F#`xJ=gm7YSR!jU%Zc70jCAj3etWdrETs z?Sc@MMVbUVHKt&?$%)aHq|Qo1R=_uBjJ_#1pj%U^xpmU| z`JTGS=xQU|AQd3nc2^$=AA>;1QbJHxSFOATMT12|TXlkxAtS1~ciLP~hHhe9K6~4YLe4?O?6%c~H9=2Dl6IoD3s< zv8A7xV~GHH0?I8o1-{)aU=r9snN83{(-W?;EOX2@m&A)-&gdGHM7IX-TpPEDs;@wH zn{G*M+yriF#-(du{gzhfgZYair%sKDs!xl$u_+m`I`b=NGO{r`%=t_F1<*i+R)bKo;BnhE8^%kfSn7cH2jZ zX;Cff0ed!w7bft{zqsu5A0^IguE6gdo_3jS)R-@E4~18|fuHgv-V}GXy|zmz@q6wP zOy3)Jc}l`wyUf7Oo|3TBFKjVQUK=L6E*hiM)~$Rc?(Rm`JtlN!&9OJHlE>_5QgB>} z+d2Cy?_=n!o;uwXqkt~OR~$$=&dnF#f+GlSzOByH(K)^8*zg-bS&ymtO_VkVWe zgWtcjFrA#8yO^|n87H`GpBt@qB-5bEq!QOtb~rOTmrN~On@P>i&)NP=a^_Ov>U2z` z%mTEFf&B~|1Yj`VBN#6)jpp<`8MO>TM=vNwK{lZ>sN(7l^DX<{FejgUB?S8@;8k z5@Q7F!3Z(3a)FGJGox-xqQnbpcA(jw&>>)cg`M&n0A=nQzx3dE#oq(Z^XQ>U&wEc! z1Au2cdb%1tT?w47`J)eV6@NcG4#n9g7b`ty;n|LkRik5-z*x<{@4;HdkKl1A&ONza z={X0_c67WN9j^q&pNG3P$ExApufs#n!b897`~6Ba_Pt7YXgfSr4NsL5&qD{l4)s3^ z^>1DK=Li+hd`OAyUB$JhDx+`quxJh?8It(fLGz1UF zoA*AK8QdCRu-1LBcJNT`P*3eB+DiPY_`JLS^N+U%wtlePeZ1OzydLofr8>uer^Rgd z`IN;?jHj&n{FZaR9Odj_OCkCmtbGKpaSQ-xAU1jZtBdREsqcr+e=x59Y_RtF96oj+ zlgYzBkxZuOcLK_gXbdJW)x?J|x zLTAf?T6m}&tV>+Dv)0*Li+0z#`s+b&r%>krlq2;nK6st42l(I=UylSOzym1z>W78k zMZPWx!I8S;0T32~6LpCX9%Hn#eEn>H2xtM6{q^&_$9sXVdpJH)?+EhKJTQ3`9*Xi1 zyt~05Cx-7ApLu#Jo}OCx!2Q`zuWVc?Cu(A_oVcI7pWHm~X?i1Fkp|pnTO6#4gB5Ym zB{;UpZ=BdXSC#s<#r~?;UlIG=g%dxS|75;A-%d4Fm3p_uzN*+)X+HH3Cmx1_9ee*9 DaX`Q*