Deterministic structural guard blocks every capability-question row (0/126 on
all three stress views), never blocks a true action (0/796), declines 196
malformed action rows as ambiguous, and grows no safety leak under
compose: sparse alone FA 9 (capQ 1) → guard→sparse FA 6 (capQ 0).
Execution eligibility recorded as a deterministic policy boundary, separate
from semantic routing and capability selection; next slice returns to the
coarse non-action router instead of another pragmatics training run.
Task/725.
cointegrated/rubert-tiny fine-tuned end-to-end: pair ordering 0.933-0.970
(+0.56-0.78 margin, in-pool) beats every from-scratch config, but the
held-out capability-question family fires 126/126 as action on all 18
config/seeds (FA_rate 1.000, mean proba 0.86-0.99, family σ 0.0045). tiny2
ceiling (newer, larger vocab) does not move LOFO (1.000) and is worse
in-pool. sparse 0.667 remains the safety floor; supervised encoder scaling
ruled out for the generator boundary.
docs/CLAUDE.md defined four tiers and docs/capabilities/ and docs/architecture/
were neither of them. Both are now declared as generated: rebuilt from a source,
never corrected in place. A wrong row in either is a bug in the generator or in
one of its hand-written inputs, and editing the output makes the next rebuild
silently undo the fix.
The eval index gains the 2026-08-19 CPT+SFT measurement and the 2026-08-26
baseline, and marks the 2026-08-13 capability audit superseded. The 2026-08-19
file lands with them: it is the direct evidence that the deployed
maven-instruct-b2 routes better than Qwen3-1.7B and cannot hold a Russian
sentence, which is the first item on the gaps.md priority list.
Its header records Vikunja as returning 503. That reading was wrong and the
2026-08-26 baseline says so, but a dated eval is not edited after the day.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
docs/evals/2026-08-15-locative-answerability-verifier.md rules the
resident model out as a recall answerability verifier. Its constrained
output was syntactically reliable and neither semantically reliable nor
isolated from instructions inside stored memory: five false accepts out of
32 held-out cases, two of them prompt injections carried in the memory
text, all five identical across three fixed-seed repeats.
design.md carries the reminder row as it now is, one-shot or recurring,
with the outbox and the cancellation invariants. routing.md carries the
new stage 0 frames and the cancellation rung. deployment.md carries the
/reminders contract. The assistant_workday scenario exercises the turn
sequence end to end.
--no-verify: master is the working branch this session by the owner's call.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Reference-count the process-global ONNX Runtime across embedder and routing-head sessions, make close idempotent, and require named proof that both aggregate routing gates executed rather than self-skipped (V-716). Owner explicitly requested direct commits to master.
Replace nearest-neighbour personal routing with a frozen class-balanced linear head measured on historical, stratified, cross-validation, holdout, and fresh challenge gates (V-702). Close the four repair handoff holes, preserve nested clarification flows, and route Russian possession statements through structural grammar rather than lexical exceptions (V-573). Owner explicitly requested direct commits to master.
claude