Repository navigation
docs(e0): record 5 owner rulings on merge honesty, add 7 work orders - #1121
Conversation
The owner ruled on 2026-11-14 on the 5 decisions of docs/design/2026-11-14-merge-honesty.adoc. This change records them and turns each into a work order: - E0-18: each outcome names its pattern and its kind of match. - E0-19: a category or fuzzy match reaches review at most. - E0-20: the pattern identity, as a design note for a ruling. - E0-21, E0-22, E0-23: the 3 plain defects (fuzzy ranking, fix check by finding, placeholder cluster). - E0-24: an untrained novelty network means novel. The design note is corrected where its review was right: the cost of the label, the fix check as a condition of the measurement, the limit of the label under the pattern id of today, the claim on acceptance A3, and the field that the novelty gate reads. E0-07, E0-08 and E0-10 get a "Theory inputs" part with 4 proved results from echo-types and residual-evidence-types, and their limits. No code changes. A dry run of the launch script accepts all 22 hypatia work orders. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7Zo6gLzkvuBQK3T9P23
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 22 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: df66f59ecb
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| . Carry the pattern id and the kind of match from the router to the outcome | ||
| row through the `metadata` map of `record_outcome/5`. Write them as | ||
| `"pattern_id"` and `"match_kind"`. |
There was a problem hiding this comment.
The production route and outcome write are not connected in-process: DispatchManifest serializes the routed action, then gitbot-fleet's runner later invokes mix hypatia.record_outcome. The manifest and CLI currently have no match_kind handoff, so carrying it only through the router and record_outcome/5 metadata can satisfy a local test while real outcome rows still receive "unknown". E0-18 must include the manifest, CLI, and gitbot-fleet runner transport (or add a dependent fleet work order) so it records the kind chosen for the actual dispatch.
Useful? React with 👍 / 👎.
| * By the owner ruling of 2026-11-14 on | ||
| `docs/design/2026-11-14-merge-honesty.adoc`, the receipt also names the | ||
| pattern and the kind of match (`exact`, `category`, `fuzzy`). See `E0-18`. |
There was a problem hiding this comment.
This newly requires receipts to name the match kind, but E0-09 still depends only on E0-05 and E0-08, while the README schedules E0-18 independently “in any order.” If E0-09 runs first, none of its three learning entry points receives the dispatch match kind, so it must either stop under its unfillable-field condition or emit receipts without the owner-required label; no later work order backfills those receipts. Add E0-18 as a dependency of E0-09 and order it before receipt emission.
Useful? React with 👍 / 👎.
| * Another network of the 8 also returns a fixed value. List it. Do not | ||
| change it here. |
There was a problem hiding this comment.
This stop condition already holds on the measured baseline: SequenceModel.handle_call/3 explicitly implements a placeholder that ignores the finding and returns confidence 0.0 for an empty buffer or 0.3 otherwise, while Coordinator.derive_confidence/1 assigns that result weight 0.05. Consequently every E0-23 run must stop here before removing the VAE's unsupported confidence. Either include the sequence placeholder in the decided remediation or create its follow-up work order and permit E0-23 to finish.
Useful? React with 👍 / 👎.
| . Rank by hit count, then by confidence, then by recipe id, so that the | ||
| result does not depend on the order of the files. | ||
| . Match a keyword as a whole word. |
There was a problem hiding this comment.
The ranking instruction does not require deduplicating keywords extracted from multiple pattern IDs, so repeated words can masquerade as multiple hits. For example, recipe-secret-to-env contains hardcoded in three pattern IDs; a description containing only that word would score three hits and can beat recipe-fix-hardcoded-secrets matching two distinct terms such as hardcoded and credential. Require unique matched keywords before ranking so the new score measures description overlap rather than repetition in recipe metadata.
Useful? React with 👍 / 👎.
What
12 files, documents only. No code changes.
docs/design/2026-11-14-merge-honesty.adocdocs/operations/e0-closeout/work-orders/E0-18toE0-24docs/operations/e0-closeout/README.adocE0-07,E0-08,E0-10Why
The owner ruled on 2026-11-14, through the selector, on the 5 decisions of the design note:
E0-18reviewat mostE0-19E0-20E0-21,E0-22,E0-23E0-24The owner also asked for the theory inputs after the second recon.
Corrections to the design note
The review of the note (hypatia pull request 1114, merged) left 5 findings. I checked each against the code. All 5 were right:
:untrainedas "not novel". The coordinator setsis_noveltofalse(lib/neural/coordinator.ex, line 765), and the guard reads that field.Evidence
scripts/api-agents/run-work-order.shaccepts all 22 hypatia work orders, the 7 new ones included.b660d08, where I read them, and3b8551c(git diff --statprinted nothing).asciidoctor --failure-level WARNbuilds each changed document with no message.bash scripts/check-readme-claims.shpasses../hypatia scan .reports 0 findings on the changed files.dedda75and residual-evidence-types6f94640. I checked that each cited theorem is at its cited line.Not done
E0-18andE0-22first.E0-18toE0-24are not part of the 7 acceptance statements of issue 984. The README says so.Generated by Claude Code