71bf04fb44
Squashes the arc's work into one commit; the workflow-file edit it originally carried is excluded (see the end of this message). ## Lane 1 — Workbench recorded a proved answer as ungrounded With deduction_serving_enabled ratified ON (ADR-0256), workbench/api.py's live chat route builds a bare ChatRuntime(), so the deduction composer decides Workbench turns and stamps grounding_source="deduction" — but _coerce_grounding_source carried a hand-copied whitelist of the six pre-arc labels and silently rewrote anything else to "none". The runtime comment reasoned this was inert because "REPL turns do not flow through Workbench's CognitivePipelineRecord path". True, and irrelevant: the traffic flows the other way. Stale since 2026-07-24. Scope is one field. workbench/api.py:818 prefers TurnEvent.epistemic_state, which read epistemic_state_needed — honest. So the UNregistered path degraded honestly while the hand-copied whitelist asserted a falsehood; a second copy of a closed enum was worse than no copy. Hence registration AND derivation: GROUNDING_SOURCES exposes the Literal's members, and the coercion reads it. workbench-ui badges/tokens/snapshot follow; enumCoverage.test.ts forces atomicity. ## Lane 2 — the ratification ceremony The discovery loop was instrumented but not closed. teaching/ratification.py turns a reviewed decision into a chain record, a corpus commit, and a receipt. Its design turns on one observation: _ratified_rows DROPS unadmissible rows silently — correct when serving, a trap when ratifying, because the file grows, the commit lands, and the band count does not move. So the ceremony refuses to call an append a ratification until it has re-read the curriculum through the real loader and seen the chain arrive; a non-admitted append is rolled back. Validation is a pre-flight courtesy, admission is the proof. Arena queue entry and ledger reseal are deliberately NOT performed (bridge rule 1); the receipt names them. Front door: `core proposal-queue ratify`, a sibling of `review` rather than a flag on it. ## Lane 3 — structural closures - ADR-0263 gains rule 5: absence policy is DECLARED in CAPABILITY_LEDGERS, not passed at the call site. An AST-matched test fails if a serving path passes missing_ok again. - Deductive suite added WHOLE to the pre-push gate: 285 tests in 29s against smoke's 216 in 62s, so no coverage trade was needed. - Smoke/CI parity assertion made bidirectional. It was one-directional, and had drifted. - test_prior_surface_deduction_binding.py pins correction binding on the deduction path. The review's diagnosis did NOT reproduce — hash_surface moves in lockstep — so it pins what is there. Mutation-checked. - Domain-keyed ADR index over 312 flat-numbered files, explicitly partial. - Arc-close brief template, plus this arc's own brief filled in against it. ## Lanes 4 and 5 — two premises falsified by measurement, one of them mine Math 4.2: baseline reproduced (correct=5 wrong=0 refused=495); all four named cases traced to one seam with each gap isolated by one-variable probes. Then the number that changes the recommendation: the gap blocking case 0000 affects 1 case in 500, the 'than' gap blocking 0001 affects 2. ADR-0251's prohibition on per-case growth now rests on a count. No reader change made. CGA: versor_condition is 0.22% of a turn, not the "~10x proof latency" I claimed — that multiplied an isolated microbenchmark by a call count and compared it to a single verdict's latency. The real cost is geometric_product at 33,986 calls/turn (~73%) via cga_inner in search paths. The obvious closed form is NOT bit-exact (954/4000 in f32); backend.vault_recall's serial fold IS (3000/3000, worst-rel 0) and is the correct target. cargo test could not run — static.crates.io is denied by the sandbox network policy — so the Rust parity question stays open and the typestate lane is carried forward, not shipped uncompiled. ## Not landed: three lines owed to .github/workflows/smoke.yml The CI smoke gate is narrower than the local one — test_pack_draft_serve_boundary.py (ADR-0253 INV-33) has been local-only, unseen because the parity pin checked one direction. The edit was authored and rejected at push for lacking the `workflow` OAuth scope, so it is recorded as a named, dated PENDING_IN_CI exception rather than dropped: the assertion still fires on any new divergence, and a second guard fires once the three land. [Verification]: pre-push gates all green — smoke 236 passed, warmed_session 10 passed, deductive 285 passed. Ratification 14, ADR index 5, CLI suites 10. Grounding/epistemic sweep 741 passed 1 skipped. workbench-ui 598 passed across 73 files, tsc -b clean. capability index 11 passed, digest unchanged. Math holdout correct=5 wrong=0 refused=495. Committed chain corpora byte-unchanged after the tests that write to them. Environment caveat: the repo pins requires-python ==3.12.13, which uv cannot fetch for linux-x86_64, so `uv sync --locked` fails. All Python runs used a scratch venv on 3.12.11 with declared deps — not the locked universe, not the full ~12k suite. The pin was left untouched. Re-run on a 3.12.13 host before treating this as merge evidence. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FduW6Krm3PPQv3P5iwBYtx
4 KiB
Arc-Close Brief — TEMPLATE
Copy to
docs/handoff/<arc-slug>/BRIEF-CLOSE-<arc-slug>.mdwhen an arc ends. Delete every instruction block (>) as you fill it in.
Why this template exists
The handoff system is mature in one direction only. BRIEF-O / BRIEF-S are
downward briefs: a planning tier telling an execution tier what to build.
Nothing goes back up. When an arc closes, the tier that actually executed it
holds the corrected ground truth — which assumptions were falsified, what the
flag state now is, which "next steps" the measurements killed — and that
knowledge currently survives only as commit messages and research docs, which
the next arc's opener must reconstruct by re-reading every ADR from the arc's
first number onward.
The generalization arc minted eight ADRs in roughly 48 hours. That is the cadence this template is sized for: an opener should be able to start scoping from this one file and reach for ADRs on demand, not as a prerequisite.
The single most important section is Falsified assumptions. A downward brief states a plan; an arc-close brief states which parts of that plan turned out to be wrong. That is the part which cannot be recovered from the diff.
Arc-Close Brief —
Arc: · Closed: · Closing tier: <Fable/Opus/Sonnet + risk tier>
Plan of record: docs/plans/<plan>.md
ADRs minted: <ADR-NNNN … ADR-NNNN>
1. What shipped
One line per landed capability, each naming its enforcement site. Not commit messages — capabilities. A reader should be able to tell what the system can now do that it could not before.
| Capability | ADR | Enforcement site | Evidence |
|---|---|---|---|
2. Flag state at close
The single highest-value table here. Every flag the arc touched, its default now, and what ratified it. A flag flipped ON is a live capability whose test suite is a regression suite — say so.
| Flag | Default at close | Ratified by | Notes |
|---|---|---|---|
3. Falsified assumptions
Do not skip this section. For each: what the plan assumed, what the measurement showed, and where the evidence lives. An assumption that was merely not reached is not falsified — say "untested" and mean it.
| Plan assumed | Measurement showed | Evidence |
|---|---|---|
4. Binding constraint going into the next arc
One paragraph. Not a list. If the next arc has one thing standing between it and progress, name it and quantify it. "Curriculum volume: every band is 24×–73× short of the entailed-bucket floor" is a binding constraint; "we should improve coverage" is not.
5. Trigger state
For any phase the plan declared trigger-gated: is the trigger now met? Cite what met it. If met, say what the next tier must design before writing code — the design work that should land in a higher tier than the implementation.
6. What is NOT next, and why
The most expensive thing an opener can do is resume a thread the closing tier already killed. List the plausible-looking next steps that the arc's own evidence rules out, each with its reason. Be specific enough that someone cannot re-propose them without engaging the evidence.
7. Open items carried forward
| Item | Kind | Where it is recorded |
|---|---|---|
| deferred / blocked / untested |
8. Gates and their state at close
Local-first is the merge bar (
AGENTS.md), so record what was actually run, with counts. "Green" without a count is not evidence.
| Gate | Command | Result at close |
|---|---|---|
| smoke | uv run core test --suite smoke -q |
|
| deductive | uv run core test --suite deductive -q |
|
| lane pins |
9. Ground truth corrections for the next opener
Free text. The things you would say out loud if you could brief the next session in person — including anything in the plan of record that is now stale. If a comment in the code reasons from a premise that has since changed, name the file and line: that is exactly the class of staleness nobody greps for.