feat(harness): route code and fix to sonnet by default (POC) - #877
Closed
guyoron1 wants to merge 2 commits into
Closed
feat(harness): route code and fix to sonnet by default (POC)#877guyoron1 wants to merge 2 commits into
guyoron1 wants to merge 2 commits into
Conversation
July-12 30-issue benchmark on the full pipeline: triage and review are where Opus is least needed (triage 52.6% cheaper with RTK+Ponytail alone; review is diff-bound). Code and fix keep opus. See fullsend ADR 0090. Signed-off-by: guy oron <goron@redhat.com>
…t code/fix Previous commit inverted the routing: triage needs opus for planning judgment, review needs opus to catch real issues, not just diff-bound mechanical checks. Code and fix are the phases with headroom — sonnet is cheaper implementation without the planning/judgment load. triage: haiku -> opus review: sonnet -> opus code: opus -> sonnet fix: opus -> sonnet Signed-off-by: guy oron <goron@redhat.com>
Functional tests did not runFunctional tests run automatically for org/repo members and collaborators on pull requests. For other contributors, a maintainer must add the |
Author
|
🤖 Finished Retro · ❌ Failure (creating sandbox: sandbox creation failed after 3 attempts: sandbox create failed: exit status 1 (output: Error: × code: 'The system is not in a state required for the operation's │ execution', messag…) · Started 11:49 AM UTC · Completed 11:49 AM UTC Commit: |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Draft / POC for the 2026-08-19 contributors meeting — not ready for review.
Cost win: code →
sonnet, fix →sonnet(harness/*.yaml+agents/*.md, 8 lines). Triage and review keepopus.Why: triage needs opus for planning judgment, review needs opus to catch real issues rather than rubber-stamp diffs — those two stay put. Code and fix are implementation-following-a-plan work, where sonnet is the cheaper tier without giving up the judgment calls that matter. See fullsend ADR 0090 (draft PR in fullsend core) for the full cost-aware routing plan; this is its "now" step.
hack/lint-agent-docspasses. Open question for the meeting: whether code/fix on sonnet holds implementation quality — worth an A/B on the next 30 issues before un-drafting.