fix: require new experiments in nightly work #81
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "fix/nightly-new-experiments"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Operator correction: each night must try new experiments, not just inspect yesterday's proposals. Verified previous run stopped after one status-only turn with IMPROVEMENT_DONE.
Prompt now requires active new selection and isolated experiments even without independent adoption evaluator. Structured reported experiment receipts distinguish attempts from status, exact recent hypotheses (200) excluded across nights. Bare done marker without any new receipt does not complete; three unproductive turns stop with explicit failed-night diagnostic. Multiple blocks validated independently; rejected/inconclusive trials count, reports are not independent adoption evidence.
Tests include status-only early stop refusal/failure diagnostic, new experiment completion, cross-night exact repeat rejection, duplicate-first/new-second receipt. Full 485 tests, typecheck/build/diffcheck pass. Fresh review 1eac257b823b77e076f514cf50c5f46a6d8c3c146d3b9d65303e96f2168c861e no findings. No production nightly state reset and no rollout performed.