docs: refresh project status for 2026-09-14 and add a lab briefing

[heyong4725]

Summary

Brings the public status surfaces up to date for a presentation to a university research lab. No new H-series hypothesis or A-series ablation result, and no claim status changes; the measured record stays dated 2026-08-28 because nothing has been measured since.

• README.md: status date 2026-09-14 (fa339ee); new "Since 2026-08-28" table covering the SPEC 400–540 pre-registration program (all PROPOSED, nine parked under ADR-66), the unrun Pre-register and run the replicated typed-dataflow versus monolithic causal study #347 causal study (confirmatory registration v22 with 7 pending gates; pilot v1 zero sessions and its three blockers; pilot v8 collection pending on fix(harness): bind session audits and reject unsafe loopback authority #589–feat(harness): hidden-access log collected from the macOS sandbox's own denial reports #591; task band v13 BND-7 failed on 22 of 22 strata), the pilot-first order and CON-9 amendment, instrument ablations explicitly labelled synthetic/fixture, M3 v2 (Spearman 0.746, unattested), the local-model floor, ADR-64/spec-change: declared halt contract for motion nodes (SPEC 520 supplement) #562, the PhyAgentOS/TypeGo positioning, and the open external reviews (Arrange the independent statistical review (STA-12) before any confirmatory session #483–Arrange the external benchmark v1 user (BMK-21) #486). Repository map lists the new harness instruments and analysis directories. A pointer to the lab briefing.
• docs/lab-briefing-2026-09.md (new): measured results with their qualifiers, what is not measured and why, how the project is run, positioning, and where graduate students can contribute without breaking the moratorium.
• docs/AISLE-technical-report.md: snapshot note; Appendix D register refreshed (all six original gaps closed; current gaps with issue numbers).
• docs/glossary.md: freeze registry, pilot registration, execution order/moratorium, and the fifteen 4xx/5xx requirement-ID prefixes.
• docs/next-phases.md: 2026-09-14 ledger entry. docs/experiments.md, docs/research-program.md: snapshot headers. docs/contributor-wiki.md: three maturity-map rows.
• docs/claim-evidence.yaml: snapshot dates for the two refreshed overview docs and the new briefing (CLM-8).

Not changed: docs/paper/aisle-paper.md (v1.0, snapshot 2026-08-31; its related-work section already covers PhyAgentOS and TypeGo) and the design doc (frozen pre-registration artifact).

Requirement IDs

Docs only (Class A). CLM-8 snapshot markers; CLM-5 wording.

Gates (CON-9 as amended, docs-only)

• ruff format --check, ruff check
• tools/claim_evidence.py --check ok (release blocker CLM-12 unchanged); tools/docs_inventory.py --check current; tools/trace_check.py ok
• Doc-policing unit tests: test_claim_evidence, test_exposure_wording, test_docs_inventory and neighbours, 133 passed
• All relative links and cited paths resolve; every issue cited as open is open
• Factual review against the repository: every number verified; one stale figure fixed (perception audit now cites the hardened second audit, 22 of 22 strata, not the superseded n=80 run)

Test plan

• CI green
• Owner read of the lab briefing before the presentation

🤖 Generated with Claude Code

https://claude.ai/code/session_019PJqc15AEfswmRJ8zBhXbk

[heyong4725]

[P1] Preserve the post-August measurements in this refreshed snapshot

This says there was "no new measured result since 2026-08-28", but analysis/m3/v2/findings.md is explicitly a measured M3 v2 campaign committed on 2026-08-31 (f002028), and this PR publishes its Spearman 0.746 result in both the README and lab briefing. The new overview is therefore internally contradictory and drops a measurement from the stated delta. Please use the narrower "no new measured hypothesis or ablation result" wording used by the README, or name M3 v2 as the exception.

[heyong4725]

Fixed in #593: the snapshot now distinguishes H/A-series results and explicitly names the post-2026-08-28 M3 v2 measurement.

[heyong4725]

[P1] Keep the wrong-object summary within the catalogued evidence

The table above correctly says this is zero in retained denominators, but "across every campaign" turns it into the project-wide assertion that docs/claim-evidence.yaml explicitly says is not supported: the project-wide session denominator is not represented by one independently audited raw table. For CLM-5 consistency, please use the catalog wording (retained development summaries report zero events; prevention is unproven) instead of claiming complete campaign coverage.

[heyong4725]

Fixed in #593: the briefing now limits the statement to retained development summaries and states that no independently audited project-wide table exists.

[heyong4725]

[P2] Account for explicit traceability waivers

This says CI fails whenever a MUST lacks a citing test, but tools/trace_check.py intentionally honors tools/trace_waivers.toml, and the checker at this head passes with 133 waived MUST IDs. The added glossary repeats the same categorical statement. Please say that every implemented MUST needs a citing test while unimplemented proposed requirements may carry explicit reviewed waivers; otherwise this public briefing overstates current mechanical coverage.

[heyong4725]

Fixed in #593: the briefing and glossary now describe explicit reviewable waivers and the stricter behavior of strict gates.

[heyong4725]

[P1] Do not describe the 2026-08-16 snapshot as the unchanged measured record

The sentence says the measured record is unchanged since the 2026-08-16 snapshot and then names only the pre-registration work as the later delta. The canonical README records later measurements, including H6 on 2026-08-26, the Phase-5 results through 2026-08-27, and M3 v2 on 2026-08-31. If section 9 itself is intentionally frozen, please call it a stale historical section and direct readers to the README; otherwise the refreshed preamble incorrectly implies that those measurements never happened.

[heyong4725]

Fixed in #593: section 9 is now explicitly labeled as the historical 2026-08-16 snapshot, with H6, Phase 5, and M3 v2 directed to the canonical README.

[heyong4725]

Codex cross-review: four actionable documentation-integrity findings posted inline (three P1, one P2). The implementation/checker surface is otherwise clean: both hosted unit jobs pass; ruff format/check, claim-evidence check, docs inventory, trace check, and diff check pass in an isolated worktree. The findings concern the omitted post-snapshot measurements, project-wide wrong-object wording beyond the weakened catalog row, and categorical test-coverage wording that omits trace waivers.

[heyong4725]

P1 — Keep the PR summary aligned with the corrected evidence wording

The opening summary now correctly says there is no new H-series hypothesis or A-series ablation result, but the same sentence still concludes that the measured record stays at 2026-08-28 “because nothing has been measured since.” That remains false: M3 v2 is an unattested development measurement committed on 2026-08-31 (analysis/m3/v2/findings.md, Spearman 0.746), and this PR explicitly adds it to the status surfaces. Please narrow the second clause too—for example, say that the H/A result table remains dated 2026-08-28 while later development measurements are listed separately.

I re-reviewed head e1b9ffa: the four prior file-level findings are fixed, the rest of the final diff is consistent with CLM-5/CLM-8, git diff --check is clean, and both hosted unit jobs pass.