Repository navigation
The review cannot tell a criterion that failed from one the worker never mentioned - #179
Merged
Merged
Conversation
The six dispatches parked at the retry ceiling on 2026-09-04 all had an unmet count that was exactly their omitted count: criteria that were never judged at all, recorded indistinguishably from criteria that were judged and failed. Three reached maximumRealAttempts and escalated to a person without the work ever being assessed; the oldest waited 91 hours. - unjudgedCriteria names the promised criteria no verdict line answered, matching through quote drift (punctuation, case, the parser's 300-character slice) with a containment guard so one word cannot judge the whole contract. - The round records judged and unjudged per dispatch (FR-001) and the send-back names the two lists under distinct headings (FR-002), so a worker can tell 'you did not do this' from 'you did not say whether you did this'. - Once a bee has judged any criteria, the unjudged ones are added to the question as unmet, so the policy decides sendBack-versus-escalate on the complete picture rather than waiting for lines that will never come. A block missing entirely is still a wait: that is the torn- transcript signature (browseros-ai#1335), not a wall of omissions. - An attempt whose unmet criteria are ALL unjudged does not increment send_backs (FR-003): the retry budget is spent on work that was attempted and found wanting, never on silence. reviewFinishedDispatches is exported for the suite, as runRound is. The new suite drives the real policy binary and pins both acceptance scenarios: 2-of-5 judged leaves the counter alone; 5 covered with 2 failings advances it.
❌ Tests failed — 7/1615 failed
Failed tests
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Work the Queen accepted on , closed as Trinity-S-AI-DNA/trios#1420, and never merged.\n\nMeasured 2026-09-04: 174 branches on the remote and FIVE contained in . The other 169 all carry a real diff. A closed issue whose code is not in the branch is a false statement about the repository, and the loop had been making 169 of them.\n\nMerges cleanly against the current base, checked with before this PR was opened.