E3–E7B registrations, S5 contract, joint-cell render, E8 runner, prepared issues

.github/CODEOWNERS

@@ -0,0 +1,14 @@
+# Review routing only. GitHub rules must enforce reviews separately.
+# A single owner cannot independently approve their own pull request.
+/.github/ @Cubits11
+/security/ @Cubits11
+/scripts/ @Cubits11
+/tests/ @Cubits11
+/requirements.txt @Cubits11
+/claims.yaml @Cubits11
+/claims_history.yaml @Cubits11
+/experiments/ @Cubits11
+/corrections/ @Cubits11
+/research/ @Cubits11
+/metrics/ @Cubits11
+/distribution/traction/ @Cubits11

.github/ISSUE_TEMPLATE/counterexample.yml

@@ -1,10 +1,8 @@
 name: Counterexample, missed benchmark, or joint outcomes
 description: >-
-  Bring evidence that changes the record: a counterexample to a claim, a
-  benchmark the census missed, a benchmark that already reports the stack,
-  a narrower identified region, or item-level / joint outcomes you can
-  provide. A correction is placed beside the claim it corrects, dated,
-  with this issue linked.
+  Evidence that changes the record: a counterexample, a missed or misread
+  census row, a tighter bound, or joint outcomes. Corrections land dated
+  beside the claim, with this issue linked.
 title: "counterexample: <what it changes>"
 labels: ["counterexample"]
 body:

.github/ISSUE_TEMPLATE/reproduction.yml

@@ -1,10 +1,8 @@
 name: Reproduction run — match or mismatch
 description: >-
-  File the result of running an experiment from /try/ or a registered
-  reproduction script. Both outcomes are wanted: a match becomes the
-  claim's independent-reproduction record; a mismatch is handled as a
-  correction under the same-day correction policy and is credited beside
-  the claim it corrects.
+  The result of running a /try/ experiment or a registered reproduction
+  script. A match becomes the claim's reproduction record; a mismatch is
+  handled as a credited correction.
 title: "Reproduction: <experiment or claim id> — <match|mismatch>"
 labels: ["reproduction"]
 body:

.github/workflows/evidence-boundary.yml

@@ -0,0 +1,39 @@
+name: evidence-boundary
+
+# This workflow belongs to the trusted base branch. NEVER check out, import,
+# install from, or execute the PR head under pull_request_target.
+on:
+  pull_request_target:
+    types: [opened, synchronize, reopened, ready_for_review]
+permissions:
+  contents: read
+concurrency:
+  group: evidence-boundary-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+jobs:
+  evidence-boundary:
+    name: Trusted evidence boundary
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09 # v5
+        with:
+          ref: ${{ github.event.pull_request.base.sha }}
+          fetch-depth: 0
+          persist-credentials: false
+      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+        with:
+          python-version: '3.12'
+      - run: python -m pip install PyYAML==6.0.3
+      - name: Read candidate Git objects without checking them out
+        env:
+          PR_NUMBER: ${{ github.event.pull_request.number }}
+          CANDIDATE_SHA: ${{ github.event.pull_request.head.sha }}
+          BASE_SHA: ${{ github.event.pull_request.base.sha }}
+        run: |
+          [[ "$PR_NUMBER" =~ ^[0-9]+$ ]] || exit 2
+          [[ "$CANDIDATE_SHA" =~ ^[0-9a-f]{40}$ ]] || exit 2
+          [[ "$BASE_SHA" =~ ^[0-9a-f]{40}$ ]] || exit 2
+          git fetch --no-tags origin "refs/pull/$PR_NUMBER/head"
+          test "$(git rev-parse FETCH_HEAD)" = "$CANDIDATE_SHA" || exit 2
+          python scripts/evidence_guard.py --base "$BASE_SHA" --candidate "$CANDIDATE_SHA"

.github/workflows/verify.yml

@@ -24,6 +24,11 @@ on:
 permissions:
   contents: read
 
+# Superseded PR revisions need no runners; main deployments finish intact.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event_name }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 jobs:
   claims:
     name: Claim registry
@@ -36,7 +41,7 @@ jobs:
       - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
         with:
           python-version: "3.12"
-      - run: pip install pyyaml
+      - run: python -m pip install -r requirements.txt
       # Both CI and the clean-clone replay execute this source-controlled,
       # duplicate-free manifest. A check cannot quietly run in one green
       # surface but not the other.
@@ -55,7 +60,7 @@ jobs:
       - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
         with:
           python-version: "3.12"
-      - run: pip install pyyaml
+      - run: python -m pip install -r requirements.txt
       - name: Clean-clone the bound commit and re-run the kernel
         run: python scripts/reproduce_cc001.py
       - name: Clean-clone this site revision and replay deterministic gates
@@ -112,7 +117,7 @@ jobs:
       - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
         with:
           python-version: "3.12"
-      - run: pip install pyyaml
+      - run: python -m pip install -r requirements.txt
       - name: Wait for every route, census checksum, counts, and corrections policy
         run: python scripts/smoke_deployed.py --attempts 30 --interval-seconds 10
 

.gitignore

@@ -1,8 +1,10 @@
 __pycache__/
 *.pyc
 _private/
+/Claude outputs/
 .venv/
 .DS_Store
 
 # Google Drive sync scratch
 .tmp.driveupload/
+.tmp.drivedownload/

ARTIFACTS/12-WEEK-PROGRAM.md

@@ -18,12 +18,22 @@ the second-guard selection by at most 2.4 points, and in the largest
 end-to-end measurement the stack was statistically indistinguishable from
 its single strongest member.**
 
+> **Registration status, 2026-09-08.** This fact is **UNREGISTERED**. The W1
+> artifact is committed at `experiments/e2/results/retrospective/` — run report,
+> three matrices, and `independent_t1.py`, which recomputes T1 from the raw
+> released rows outside the analyzer's path. No claim id carries these numbers
+> and no CI check re-asserts them: the claim that used to (retracted 2026-09-06)
+> was withdrawn over a licence defect in its support block, and the retraction
+> reason records that the computation itself is untouched. Every surface that
+> speaks these numbers must say "computed 2026-09-02, unregistered" in the same
+> breath; `scripts/verify_retracted.py` gates the attribution.
+
 Three sources, all VERIFIED this run:
 
 1. **BELLS 2025 released subset** (`non_adversarial_prompts.csv @ 507566c5`,
    sha256 `791dd4b0…`, 170 rows, 11 verdict columns). Computed 2026-09-02 on
-   the hash-verified file (scratchpad script; to be committed as the W1
-   artifact), harmful stratum n=82, native points:
+   the hash-verified file — the W1 artifact is now committed at
+   `experiments/e2/results/retrospective/` — harmful stratum n=82, native points:
    - Five specialized supervisors: for **0 of 5** incumbents does the
      partner maximizing measured union catch differ in union from the
      partner chosen by marginal rank. Regret 0 items.

ARTIFACTS/2026-09-05-FABLE-5.1-OBS-CUT.md

@@ -1,5 +1,16 @@
 # OBSERVATION CUT — BELLS denominators and provenance
 
+> **Superseding note, 2026-09-08 (annotation only; the cut below is unchanged).**
+> Both open items this cut named are closed. MC-005 was **retracted** on
+> 2026-09-06 — the RETRACT is entry 39 of `claims_history.yaml` and its reason
+> is the licence disagreement recorded below — so the registry now has exactly
+> one record of the BELLS file's licence: MC-002's `none declared upstream`,
+> with `commercial_reuse: facts_only`. Nothing in this repository infers a
+> licence from silence. The `exclusive_cells` block this cut found uncommitted
+> is committed, declared by an MC-002 CLARIFY transition, and re-asserted from
+> the hash-verified file by `scripts/reanalyze_bells_subset.py` in CI. The
+> 990-versus-1041 denominator remains unreconciled, exactly as this cut left it.
+
 Written 2026-09-05 against working tree of branch `claude/mc-005-selection-regret`
 (HEAD `f9b24e2`, uncommitted owner changes present and untouched). Status
 vocabulary: OBSERVED (bytes opened or command executed in this run) · DERIVED
@@ -34,7 +45,7 @@ nothing redistributed) or `--cache DIR` offline. Exit 0 this run.
 | C | `077555d9` (2025-02-19, "smaller dataset for playground") | same path, sha `7fa0fbf5885e…` | 174 | 86 / 50 / 38 | 12 | as B | `prompt_guard=1`, `langkit=1` | rest | **SELECTION RULE: UNKNOWN** (commit subject only: "smaller dataset for playground") | selected subset | none |
 | D | `00b42bfd` (2025-02-21, "New data, new results interpretation") | same path, sha `52ca9ab8eb62…` | 174 | 86 / 50 / 38 | 12 | as C, claude column renamed `claude-3-5-sonnet-20241022` | none (Prompt Guard and LangKit now carry real 0/1 values) | all verdict columns | SELECTION RULE: UNKNOWN | selected subset | none |
 | E | `ffe88ccb` (2025-02-23, "new playground data") | same path, sha `d93f9fe1d1a1…` | 174 | 86 / 50 / 38 | 12 | as D | `llm_guard=0` | rest | SELECTION RULE: UNKNOWN | selected subset | none |
-| F | `b20aeed5` (2025-03-22, "new results") = **`507566c5` (2025-07-08, default-branch head)** | same path, sha `791dd4b0a168f2eb…03f57c3` | **170** | **82 / 50 / 38** | 11 (Miscellaneous absent) | five specialized + gpt-4, claude-3-5-sonnet-20241022, gemini-1.5-pro-latest, mistral-large-latest, deepseek-ai/DeepSeek-V3, grok-2-latest | none over 170 rows; within the 82 harmful rows `llm_guard=0` | all eleven verdict columns | **SELECTION RULE: UNKNOWN**. Only 74–76 of the 170 questions appear in C–E, so F is a fresh draw from pool B, not a trimming of C. All 170 questions are in A and B with identical `harm_level`. | selected subset (author-selected; rule unstated) | MC-002 (every `expected` count), MC-003 (`{0…12}/82`, leave-one-out), MC-005 at HEAD (n = 82), census row `bells-misuse-2025` (`item_level_outcomes_released`), dossier `bells-misuse-2025`, disclosure page, two `/answers/` pages, `/try/` TRY-B, film theses in `films/SLATE.md` |
+| F | `b20aeed5` (2025-03-22, "new results") = **`507566c5` (2025-07-08, default-branch head)** | same path, sha `791dd4b0a168f2eb…03f57c3` | **170** | **82 / 50 / 38** | 11 (Miscellaneous absent) | five specialized + gpt-4, claude-3-5-sonnet-20241022, gemini-1.5-pro-latest, mistral-large-latest, deepseek-ai/DeepSeek-V3, grok-2-latest | none over 170 rows; within the 82 harmful rows `llm_guard=0` | all eleven verdict columns | **SELECTION RULE: UNKNOWN**. Only 74–76 of the 170 questions appear in C–E, so F is a fresh draw from pool B, not a trimming of C. All 170 questions are in A and B with identical `harm_level`. | selected subset (author-selected; rule unstated) | MC-002 (every `expected` count), MC-003 (`{0…12}/82`, leave-one-out), MC-005 at HEAD (n = 82; MC-005 retracted 2026-09-06), census row `bells-misuse-2025` (`item_level_outcomes_released`), dossier `bells-misuse-2025`, disclosure page, two `/answers/` pages, `/try/` TRY-B, film theses in `films/SLATE.md` |
 | G | `507566c5` | `data/adversarial_prompts.csv` sha `32fe8663621a…` | 8 | 8 / 0 / 0 | 1 (all Miscellaneous) | eleven verdict columns | `harm_level`, `category`, `deepseek-ai/DeepSeek-V3=0` | rest | SELECTION RULE: UNKNOWN | selected subset | 12-WEEK-PROGRAM.md ("8 adversarial rows released"); census row ("plus 8 adversarial prompts") |
 | H | `8a974123`→`507566c5` | `data/metacognitive_results.csv` sha `66a9f09c60d3…` | 612 rows = 102 prompts × 6 models | ground_truth 504 harmful / 108 not_harmful rows; 19 non-adversarial prompts | — | none of the five supervisors; frontier models only, with **newer versions** (`claude-3-7-sonnet-20250219`, `gemini-2.5-pro-exp-03-25`) | `model` set | responses | UNKNOWN | selected subset | none |
 
@@ -137,7 +148,10 @@ F into a population estimate. MC-002 stays non-population-level.
   files of February 2025.
 - Upstream repository declares no licence (GitHub API `license: null`; no
   `LICENSE` in the tree at `507566c5`). MC-002 records "none declared
-  upstream"; MC-005 at HEAD `f9b24e2` records `license: MIT` for the same file.
+  upstream"; MC-005 (since retracted) at HEAD `f9b24e2` recorded `license: MIT`
+  for the same file.
+  MC-005 was retracted 2026-09-06 for exactly this; `none declared upstream`
+  is now the register's only statement about that file.
 - The v0 commit `0fc3d6d3` carries 114,540 adversarial rows with prompt text,
   while the paper (App. 0.C) states the full dataset is not publicly released.
   Recorded as a provenance fact; nothing here redistributes it.
@@ -159,7 +173,8 @@ F into a population estimate. MC-002 stays non-population-level.
   through `ROOT.rglob("index.html")`. Zero findings on tracked files. A clean
   clone has no `_private/`, so CI is unaffected; the two files this packet
   adds are not `index.html` and are not touched by that gate.
-- Working tree (untouched): staged removal of MC-005 from `claims.yaml`
+- Working tree (untouched): staged removal of MC-005 (now retracted) from `claims.yaml`
+  (completed, and declared as a RETRACT in the history, on 2026-09-06)
   relative to HEAD; unstaged addition of `exclusive_cells` to MC-002 and a
   matching `claims_history.yaml` transition; modified generators, verifiers,
   film receipts; untracked `experiments/e3/`.
@@ -240,8 +255,8 @@ not create a finding.
 - No E2 item, threshold, estimator or criterion was touched. E2 remains
   `UNTESTED BY DESIGN`; P2 remains `EMPTY`.
 - No live API was called; no money was spent; nobody was contacted.
-- MC-002, MC-003, MC-005 and the census row were not edited. Two candidate
-  owner actions are noted, not performed: (a) MC-005's `license: MIT` is
+- MC-002, MC-003, MC-005 (since retracted) and the census row were not edited. Two candidate
+  owner actions are noted, not performed: (a) retracted MC-005's `license: MIT` is
   unsupported by upstream (no licence declared) and conflicts with MC-002's
   record for the same file; (b) the census row's evidence text "170 prompts
   … the full dataset is available only by contacting the authors" is

ARTIFACTS/2026-09-07-RECONCILIATION.md

@@ -14,7 +14,7 @@ held twice: `experiments/e3/freeze/or-bench-80k.csv` and
 `experiments/e3b/freeze/or-bench-80k.csv`, byte-identical
 (`22e95602…`). Each freeze is self-contained by design, so the duplicate
 stays and is recorded by detector D7 in `docs/graph/repo-graph.json`. The
-remaining ~11.7k lines are the E3 and E3B experiments, the MC-005
+remaining ~11.7k lines are the E3 and E3B experiments, the retracted MC-005
 retraction, and the film re-render. Nothing in the 17 commits was found
 uncommittable; see §4 for what an adversarial pass did find.
 

CLAUDE.md

@@ -11,13 +11,19 @@ style — they are the product, and CI enforces them.
 ```bash
 python3 .claude/skills/evidence-ledger/ledger.py   # the instant
 python3 scripts/cadence.py report                  # the trajectory
+python3 scripts/direction.py                       # the heading
 ```
 
 Repo state in one screen: what has been measured, what is blocked on a human,
 and how much protocol sits on top of how many rows. Cheaper and more honest
 than reconstructing it from the program, the contract, the freeze and the last
 report. The cadence report answers the question one instant cannot: whether the
-answer is moving, and whether what moved was evidence or scaffolding. See
+answer is moving, and whether what moved was evidence or scaffolding. The
+direction report answers what neither can: which way this is pointed, whether
+the next thing you are about to write is the thing it needs, and whether
+infrastructure outstanding has overtaken work outstanding. Its source is
+`research/DIRECTION.yaml` — the whole plan, machine-read, with recorded
+amendments to its own first draft. Read it before proposing a new direction. See
 `.claude/skills/evidence-ledger/SKILL.md` for the two standing rules — in
 particular: before writing another governing document, say the trade out loud
 first.
@@ -35,9 +41,10 @@ prints it from the tuple.
 ## Conventions CI enforces
 
 - **Generated pages are never hand-edited** — `/ledger/`, `/observatory/`,
-  `/modules/*`, `/missing-column/*`, `/try/`, `/worldspace/`, `sitemap.xml`.
+  `/modules/*`, `/missing-column/*`, `/try/`, `/worldspace/`,
+  `/records/conformance/`, `sitemap.xml`.
   Edit the source registry (`claims.yaml`, `modules.yaml`, `census.yaml`) and
-  regenerate; nine `scripts/generate_*.py` have `--check` drift gates.
+  regenerate; twelve `scripts/generate_*.py` have `--check` drift gates.
 - **`claims.yaml` is schema v0.4.** Every claim needs a non-empty
   `falsifier.condition`, a fixed `NARROW|REJECT|HOLD` consequence, and a typed
   `forbidden_rescues` list (explicit `[]` is valid). A declared commit must be
@@ -53,6 +60,15 @@ prints it from the tuple.
   anything else with a `local_content_change` trigger.
 - **Figure numbers** are re-derived from stated constants in
   `scripts/verify_figures.py` to 1e-9.
+- **New experiments use a canonical executable contract.**
+  Follow correction C1 in `research/DIRECTION.yaml`: the runner consumes
+  `contract.json`, and `PREREG.md` renders its scientific choices into prose.
+  `research/contracts/validate_contract.py` checks structure, threshold direction,
+  planned pool identities and planned marginal preservation. Passing establishes
+  internal consistency, not runtime conformance or inferential validity.
+  `verify_prereg.py` remains the legacy sidecar scan and historical fixture
+  corpus; its static reference check does not prove runtime consumption.
+  Frozen experiments remain untouched and undeclared rather than retrofitted.
 - **The cadence series only appends.** `metrics/repo_state.jsonl` is a hash
   chain of daily repository state; an edited, reordered, dropped or truncated
   row fails `scripts/cadence.py check`. CI enforces the chain and never the

README.md

@@ -253,3 +253,25 @@ DESIGN.md                   design-decision ledger + changelogs + field-artifact
 ```
 
 Content © Pranav Bhave. Code (HTML/CSS/JS) may be reused with attribution.
+
+## Scheduled feedback and release updates
+
+The existing local schedules record repository state daily and collect distribution
+observations every four hours. Missing days and missed observation windows remain
+missing; the jobs never interpolate evidence. Both jobs share a repository lock.
+They prefer `.venv/bin/python3`; create that environment with Python 3.12 or newer
+and install `requirements.txt` before enabling the schedules. `CUBITS11_PYTHON`
+can select another environment explicitly.
+
+A cycle starting from clean, synchronized `main` runs the verification manifest,
+commits only its permitted outputs on a `claude/cycle-*` branch, and opens a pull
+request. GitHub auto-merge uses a merge commit and remains subject to required
+checks and review rules. A failed API call leaves the branch available for recovery;
+a failed check retains the observations without publishing them. Logs live in
+`_private/cron/`. No cycle dispatches a new outreach message.
+
+This feedback loop records observations and tests known failures. It does not
+autonomously change scientific criteria, promote a hypothesis, or accept a
+contradicted result. Deployment checks compare the public pages, primary films,
+posters and claim metadata with the verified revision, so an older page returning
+HTTP 200 cannot stand in for the release.