From 599a0d8f0a5f303c18ba18c05d9276c1c47d919d Mon Sep 17 00:00:00 2001 From: teleworksai <334609622+teleworksai@users.noreply.github.com> Date: Mon, 28 Sep 2026 10:24:52 -0500 Subject: [PATCH] chore: point repository links at moxin-org/aisle The repository moved from teleworksai/aisle to moxin-org/aisle for collaboration. Rewrite the live doc links to the new location. CODEOWNERS stays @teleworksai (an org owner with write access). The hash-bound and historical surfaces left unchanged in #597 stay unchanged; GitHub redirects the old repository URLs. Co-Authored-By: Claude Opus 5.5 --- README.md | 2 +- docs/AISLE-technical-report.md | 26 ++++++++-------- docs/dora-1.0.1-validation.md | 2 +- docs/getting-started.md | 4 +-- docs/paper/aisle-paper.md | 2 +- docs/project-quality-roadmap.md | 54 ++++++++++++++++----------------- 6 files changed, 45 insertions(+), 45 deletions(-) diff --git a/README.md b/README.md index eed0652b..927ae888 100644 --- a/README.md +++ b/README.md @@ -277,7 +277,7 @@ and [getting started](docs/getting-started.md) for manual rollouts and NVIDIA setup. Never install with bare pip/conda. The supported Linux clone and archive quickstarts passed in -[run 34082153283](https://github.com/teleworksai/aisle/actions/runs/34082153283). +[run 34082153283](https://github.com/moxin-org/aisle/actions/runs/34082153283). These are development runs; the benchmark's independent-user, release and physical-evidence gates remain outstanding. diff --git a/docs/AISLE-technical-report.md b/docs/AISLE-technical-report.md index c32fb537..9c20c5d1 100644 --- a/docs/AISLE-technical-report.md +++ b/docs/AISLE-technical-report.md @@ -2840,25 +2840,25 @@ understood. | # | Gap | Why it matters | Section | |---|---|---|---| -| [#264](https://github.com/teleworksai/aisle/issues/264) | Skill manifests record evidence but not **applicability** | As the library grows, undifferentiated same-capability entries make selection harder — the accumulation benefit can invert | §5.5 | -| [#265](https://github.com/teleworksai/aisle/issues/265) | No **sandbox trust tier** | ADR-37's floor leaves an agent no legitimate way to declare an unproven node so its graph validates | §5.5 | -| [#266](https://github.com/teleworksai/aisle/issues/266) | **Idea trees are not retained** | The pre-registrations that make findings falsifiable are gone; "3/4 expectations met" is currently unverifiable prose | §8.5 | -| [#267](https://github.com/teleworksai/aisle/issues/267) | **SFT label undecided**: `joint_cmd` vs `joint_cmd_safe` | Structural safety and clean credit assignment pull against each other; must be settled before any fine-tune | §10.11 | -| [#268](https://github.com/teleworksai/aisle/issues/268) | **No determinism layer for inference** | The staleness floor couples sim time to wall time, so a loaded host can change the trajectory with every seed identical | §10.11 | -| [#269](https://github.com/teleworksai/aisle/issues/269) | **"Safe Learning" is undefined** in the project's own naming gloss | Students reasonably expect a training loop and find none | §3.5 | +| [#264](https://github.com/moxin-org/aisle/issues/264) | Skill manifests record evidence but not **applicability** | As the library grows, undifferentiated same-capability entries make selection harder — the accumulation benefit can invert | §5.5 | +| [#265](https://github.com/moxin-org/aisle/issues/265) | No **sandbox trust tier** | ADR-37's floor leaves an agent no legitimate way to declare an unproven node so its graph validates | §5.5 | +| [#266](https://github.com/moxin-org/aisle/issues/266) | **Idea trees are not retained** | The pre-registrations that make findings falsifiable are gone; "3/4 expectations met" is currently unverifiable prose | §8.5 | +| [#267](https://github.com/moxin-org/aisle/issues/267) | **SFT label undecided**: `joint_cmd` vs `joint_cmd_safe` | Structural safety and clean credit assignment pull against each other; must be settled before any fine-tune | §10.11 | +| [#268](https://github.com/moxin-org/aisle/issues/268) | **No determinism layer for inference** | The staleness floor couples sim time to wall time, so a loaded host can change the trajectory with every seed identical | §10.11 | +| [#269](https://github.com/moxin-org/aisle/issues/269) | **"Safe Learning" is undefined** in the project's own naming gloss | Students reasonably expect a training loop and find none | §3.5 | Refreshed 2026-09-14. All six gaps above have since closed (#264 and #265 through ADR-41 and ADR-40; #267 and #268 on 2026-08-17). The current register: | # | Gap | Why it matters | Where | |---|---|---|---| -| [#347](https://github.com/teleworksai/aisle/issues/347) | The **typed-vs-monolithic causal study is unrun**; pilot v8 is registered with collection pending on the session-bound confinement attestation (#589–#591) | It is the headline claim; every 4xx spec exists to make its result defensible | README status, ADR-66 | -| [#346](https://github.com/teleworksai/aisle/issues/346) | **No perception-eligible task stratum**: the task-band calibration (v13) records BND-7 failed for the only candidate, 22 of 22 strata | Without an eligible non-oracle task, neither the pilot nor the confirmatory study has an instrument | SPEC 490 | -| [#519](https://github.com/teleworksai/aisle/issues/519) | Matched-arm session admission and common evidence (MON-8/MON-12) not complete | Gate on every scored session in both arms | SPEC 420/440 | -| [#483](https://github.com/teleworksai/aisle/issues/483)–[#486](https://github.com/teleworksai/aisle/issues/486) | **External reviews not arranged**: statistical (STA-12), terminology (CLM-12), independent reproduction (RPR-10/11), external benchmark user (BMK-21) | Confirmatory and release gates that no in-house work can satisfy | SPEC 400/410/530/540 | -| [#562](https://github.com/teleworksai/aisle/issues/562) | No declared **halt contract** for motion nodes (bound, semantics, receipt); ADR-64 already makes a halt terminal | Required by SPEC 460/480/520 as an outcome, defined nowhere; parked until SPEC 460 ratifies or Phase 6 opens | SPEC 520 supplement | -| [#566](https://github.com/teleworksai/aisle/issues/566) | Closed idea-tree entries persist but nothing **retrieves a refuted idea** at the moment of choosing | The failure-side half of "does prior work compound"; parked until an H3 rerun | ADR-41, HAR-7 | -| [#356](https://github.com/teleworksai/aisle/issues/356) | **Hardware** entry gates (M1, judge fidelity) are GPU-gated; SO-101 driver is loopback-tested only | Every hardware claim stays `hardware_pending` | SPEC 520 | +| [#347](https://github.com/moxin-org/aisle/issues/347) | The **typed-vs-monolithic causal study is unrun**; pilot v8 is registered with collection pending on the session-bound confinement attestation (#589–#591) | It is the headline claim; every 4xx spec exists to make its result defensible | README status, ADR-66 | +| [#346](https://github.com/moxin-org/aisle/issues/346) | **No perception-eligible task stratum**: the task-band calibration (v13) records BND-7 failed for the only candidate, 22 of 22 strata | Without an eligible non-oracle task, neither the pilot nor the confirmatory study has an instrument | SPEC 490 | +| [#519](https://github.com/moxin-org/aisle/issues/519) | Matched-arm session admission and common evidence (MON-8/MON-12) not complete | Gate on every scored session in both arms | SPEC 420/440 | +| [#483](https://github.com/moxin-org/aisle/issues/483)–[#486](https://github.com/moxin-org/aisle/issues/486) | **External reviews not arranged**: statistical (STA-12), terminology (CLM-12), independent reproduction (RPR-10/11), external benchmark user (BMK-21) | Confirmatory and release gates that no in-house work can satisfy | SPEC 400/410/530/540 | +| [#562](https://github.com/moxin-org/aisle/issues/562) | No declared **halt contract** for motion nodes (bound, semantics, receipt); ADR-64 already makes a halt terminal | Required by SPEC 460/480/520 as an outcome, defined nowhere; parked until SPEC 460 ratifies or Phase 6 opens | SPEC 520 supplement | +| [#566](https://github.com/moxin-org/aisle/issues/566) | Closed idea-tree entries persist but nothing **retrieves a refuted idea** at the moment of choosing | The failure-side half of "does prior work compound"; parked until an H3 rerun | ADR-41, HAR-7 | +| [#356](https://github.com/moxin-org/aisle/issues/356) | **Hardware** entry gates (M1, judge fidelity) are GPU-gated; SO-101 driver is loopback-tested only | Every hardware claim stays `hardware_pending` | SPEC 520 | Two threads remain owner-facing rather than contributor-facing: the eval's **seed set and episode count remain candidate-chosen** (the same self-grading diff --git a/docs/dora-1.0.1-validation.md b/docs/dora-1.0.1-validation.md index fffae551..12338513 100644 --- a/docs/dora-1.0.1-validation.md +++ b/docs/dora-1.0.1-validation.md @@ -37,7 +37,7 @@ lockstep conformance claim. The legacy fixture runs in free-run mode, while BRG-1 requires lockstep for acceptance; it also asserts TC-4's hardware wall-clock band instead of the specified simulation-rate/liveness rules. Per CON-13, those test changes are paused in -[spec-conflict #497](https://github.com/teleworksai/aisle/issues/497). +[spec-conflict #497](https://github.com/moxin-org/aisle/issues/497). The remaining throughput work is to resolve the fixture migration, then measure cold and warm starts separately with a capture boundary that includes diff --git a/docs/getting-started.md b/docs/getting-started.md index 93bf0dc1..b0b545e1 100644 --- a/docs/getting-started.md +++ b/docs/getting-started.md @@ -11,7 +11,7 @@ Install Git and rustup as well as Python >= 3.11, managed exclusively through ## 1. Install ```bash -git clone https://github.com/teleworksai/aisle && cd aisle +git clone https://github.com/moxin-org/aisle && cd aisle uv sync --extra sim --locked ``` @@ -71,7 +71,7 @@ uv run --extra sim --locked python tools/quickstart.py --runtime-prefix "$AISLE_ This executes graph validation, a public task, bundle validation and reporting. The quickstart uses the `sim` extra; CUDA experiments use the manual rollout path below. The Linux clone and archive paths passed with verified source -receipts in [run 34082153283](https://github.com/teleworksai/aisle/actions/runs/34082153283). +receipts in [run 34082153283](https://github.com/moxin-org/aisle/actions/runs/34082153283). That evidence does not establish independent reproduction or release readiness. For manual harness and graph commands, make the verified binary available to diff --git a/docs/paper/aisle-paper.md b/docs/paper/aisle-paper.md index 81189cbd..3654489d 100644 --- a/docs/paper/aisle-paper.md +++ b/docs/paper/aisle-paper.md @@ -929,7 +929,7 @@ verification.)* ## Reproducibility -Repository: teleworksai/aisle (for double-blind review, this link and +Repository: moxin-org/aisle (for double-blind review, this link and the git metadata are replaced by an anonymized artifact archive; the run dates in committed findings remain there as provenance and are deliberately absent from this text). Every campaign: protocol ADR + analyzer tool + diff --git a/docs/project-quality-roadmap.md b/docs/project-quality-roadmap.md index 19f72bc7..c96bb88d 100644 --- a/docs/project-quality-roadmap.md +++ b/docs/project-quality-roadmap.md @@ -3,7 +3,7 @@ Date: 2026-08-31 Status: ACTIVE planning program Umbrella tracker: [#359 — AISLE 10/10 project-quality -program](https://github.com/teleworksai/aisle/issues/359) +program](https://github.com/moxin-org/aisle/issues/359) ## Objective @@ -24,21 +24,21 @@ simulation, statistics, safety, integrity, and release work can proceed now. | Dimension | Starting review | Objective exit gate | Issues | |---|---:|---|---| -| Novelty and potential | 6/10 | Public versioned benchmark, blind evaluation, external user, and a defensible boundary against adjacent systems | [#357](https://github.com/teleworksai/aisle/issues/357), [#358](https://github.com/teleworksai/aisle/issues/358) | -| Technical system quality | 7/10 | Equal-capability control, sealed evaluator, scoped actuation boundary, and adversarial instrument/safety tests | [#344](https://github.com/teleworksai/aisle/issues/344), [#348](https://github.com/teleworksai/aisle/issues/348), [#350](https://github.com/teleworksai/aisle/issues/350), [#353](https://github.com/teleworksai/aisle/issues/353), [#354](https://github.com/teleworksai/aisle/issues/354) | -| Experimental rigor | 3/10 | Session-level estimands, power, replication, randomized blocks, uncertainty/equivalence, no-fault controls, and retained exclusions | [#345](https://github.com/teleworksai/aisle/issues/345), [#346](https://github.com/teleworksai/aisle/issues/346), [#347](https://github.com/teleworksai/aisle/issues/347), [#349](https://github.com/teleworksai/aisle/issues/349) | -| Support for central claim | 2/10 | Replicated typed-dataflow versus monolithic causal study, with null/negative results accepted without reframing | [#344](https://github.com/teleworksai/aisle/issues/344), [#347](https://github.com/teleworksai/aisle/issues/347) | -| Physical-AI relevance | 3/10 | Positive non-oracle result plus SO-101 physical validation and a live-fault operation cell | [#346](https://github.com/teleworksai/aisle/issues/346), [#356](https://github.com/teleworksai/aisle/issues/356) | -| Reproducibility | 8/10 intent | Independent-machine reproduction, complete raw archive, one-command analysis, and DOI | [#353](https://github.com/teleworksai/aisle/issues/353), [#355](https://github.com/teleworksai/aisle/issues/355) | -| Clarity and focus | 4/10 | Mechanically checked claim/evidence matrix and one canonical external architecture/status narrative | [#358](https://github.com/teleworksai/aisle/issues/358) | +| Novelty and potential | 6/10 | Public versioned benchmark, blind evaluation, external user, and a defensible boundary against adjacent systems | [#357](https://github.com/moxin-org/aisle/issues/357), [#358](https://github.com/moxin-org/aisle/issues/358) | +| Technical system quality | 7/10 | Equal-capability control, sealed evaluator, scoped actuation boundary, and adversarial instrument/safety tests | [#344](https://github.com/moxin-org/aisle/issues/344), [#348](https://github.com/moxin-org/aisle/issues/348), [#350](https://github.com/moxin-org/aisle/issues/350), [#353](https://github.com/moxin-org/aisle/issues/353), [#354](https://github.com/moxin-org/aisle/issues/354) | +| Experimental rigor | 3/10 | Session-level estimands, power, replication, randomized blocks, uncertainty/equivalence, no-fault controls, and retained exclusions | [#345](https://github.com/moxin-org/aisle/issues/345), [#346](https://github.com/moxin-org/aisle/issues/346), [#347](https://github.com/moxin-org/aisle/issues/347), [#349](https://github.com/moxin-org/aisle/issues/349) | +| Support for central claim | 2/10 | Replicated typed-dataflow versus monolithic causal study, with null/negative results accepted without reframing | [#344](https://github.com/moxin-org/aisle/issues/344), [#347](https://github.com/moxin-org/aisle/issues/347) | +| Physical-AI relevance | 3/10 | Positive non-oracle result plus SO-101 physical validation and a live-fault operation cell | [#346](https://github.com/moxin-org/aisle/issues/346), [#356](https://github.com/moxin-org/aisle/issues/356) | +| Reproducibility | 8/10 intent | Independent-machine reproduction, complete raw archive, one-command analysis, and DOI | [#353](https://github.com/moxin-org/aisle/issues/353), [#355](https://github.com/moxin-org/aisle/issues/355) | +| Clarity and focus | 4/10 | Mechanically checked claim/evidence matrix and one canonical external architecture/status narrative | [#358](https://github.com/moxin-org/aisle/issues/358) | Safety is deliberately cross-cutting rather than hidden inside one score: -- [#350](https://github.com/teleworksai/aisle/issues/350) defines and attacks the +- [#350](https://github.com/moxin-org/aisle/issues/350) defines and attacks the agent-to-actuation threat boundary. -- [#351](https://github.com/teleworksai/aisle/issues/351) separates topology, +- [#351](https://github.com/moxin-org/aisle/issues/351) separates topology, kinematic enforcement, and empirical semantic outcomes. -- [#352](https://github.com/teleworksai/aisle/issues/352) decides whether AISLE +- [#352](https://github.com/moxin-org/aisle/issues/352) decides whether AISLE can support an identity-aware semantic authorization claim; otherwise H5 is permanently narrowed. @@ -47,13 +47,13 @@ Safety is deliberately cross-cutting rather than hidden inside one score: ### Wave 0 — rules before results - [#345 — session-level statistics and power - analysis](https://github.com/teleworksai/aisle/issues/345) + analysis](https://github.com/moxin-org/aisle/issues/345) - [#353 — treatment integrity - v3](https://github.com/teleworksai/aisle/issues/353) + v3](https://github.com/moxin-org/aisle/issues/353) - [#354 — independent instrument audit and mutation - benchmark](https://github.com/teleworksai/aisle/issues/354) + benchmark](https://github.com/moxin-org/aisle/issues/354) - [#358 — claim-to-evidence matrix and architecture - narrative](https://github.com/teleworksai/aisle/issues/358) + narrative](https://github.com/moxin-org/aisle/issues/358) Exit: experimental units, claim scopes, treatment identity, audit expectations, and analysis decisions are mechanically explicit. @@ -61,17 +61,17 @@ and analysis decisions are mechanically explicit. ### Wave 1 — build fair instruments - [#344 — equal-capability monolithic - control](https://github.com/teleworksai/aisle/issues/344) + control](https://github.com/moxin-org/aisle/issues/344) - [#346 — non-oracle reachable task - band](https://github.com/teleworksai/aisle/issues/346) + band](https://github.com/moxin-org/aisle/issues/346) - [#348 — sealed hidden fault bank and - injector](https://github.com/teleworksai/aisle/issues/348) + injector](https://github.com/moxin-org/aisle/issues/348) - [#350 — actuation threat model and bypass - validation](https://github.com/teleworksai/aisle/issues/350) + validation](https://github.com/moxin-org/aisle/issues/350) - [#351 — safety exposure and held-command - ablation](https://github.com/teleworksai/aisle/issues/351) + ablation](https://github.com/moxin-org/aisle/issues/351) - [#352 — semantic authorization - boundary](https://github.com/teleworksai/aisle/issues/352) + boundary](https://github.com/moxin-org/aisle/issues/352) Exit: expert parity, task reachability, blinding, confinement, and safety evidence gates pass before confirmatory data collection. @@ -79,9 +79,9 @@ evidence gates pass before confirmatory data collection. ### Wave 2 — confirmatory campaigns - [#347 — replicated typed versus monolithic causal - study](https://github.com/teleworksai/aisle/issues/347) + study](https://github.com/moxin-org/aisle/issues/347) - [#349 — replicated typed-evidence versus logs-only fault - study](https://github.com/teleworksai/aisle/issues/349) + study](https://github.com/moxin-org/aisle/issues/349) Exit: analyzer-derived session-level treatment estimates, uncertainty, no-fault false-alarm measurements, and all exclusions are retained. A null or negative @@ -90,11 +90,11 @@ result is a valid exit. ### Wave 3 — independent and physical closure - [#355 — independent-machine reproduction and DOI - artifact](https://github.com/teleworksai/aisle/issues/355) + artifact](https://github.com/moxin-org/aisle/issues/355) - [#356 — SO-101 physical benchmark and live-fault - operation](https://github.com/teleworksai/aisle/issues/356) + operation](https://github.com/moxin-org/aisle/issues/356) - [#357 — versioned public benchmark and blind evaluation - path](https://github.com/teleworksai/aisle/issues/357) + path](https://github.com/moxin-org/aisle/issues/357) Exit: an external operator can run and verify the benchmark without the original campaign machine, and physical claims rest on retained hardware @@ -117,7 +117,7 @@ evidence. ## Completion -Close [#359](https://github.com/teleworksai/aisle/issues/359) only after every +Close [#359](https://github.com/moxin-org/aisle/issues/359) only after every child issue is complete or explicitly dispositioned with a narrower claim, the claim/evidence matrix has no unsupported headline claim, and an independent reviewer audits the completed program.