feat(bench): aj bench --gen to generate+run repetition fixtures - #10
Merged
Merged
Conversation
Adds the generate-and-run flow so `aj bench` can sweep the break-even curve without hand-authoring task files. - Fixture registry: FixtureByShape / FixtureShapes (currently "nullcheck"). - `aj bench --gen <shape> --n 1,2,4 [--workdir DIR]`: materializes the fixture at each repeat count (own subdir per count so they never collide) and runs it, composing with --arm and --compare. Verified with a real `claude` sweep: nullcheck n=1,2 both solved and dual-gate-verified (guards actually added), reporting real T2S per count. --dry-run correctly shows 0% / no-T2S on the unmodified fixtures. Refs: benchmark sandbox design §3.2, Phase 3. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds the generate-and-run flow so
aj benchcan sweep the break-even curve without hand-authoring task JSONL (finishing the CLI side of Phase 3).FixtureByShape/FixtureShapes(currentlynullcheck).aj bench --gen <shape> --n 1,2,4 [--workdir DIR]— materializes the fixture at each repeat count (own subdir per count, no collisions) and runs it; composes with--armand--compare.Verified with a real run
Drove the live
claudeacross a sweep:Both dual-gate-verified (guards actually present).
--dry-runcorrectly shows 0% / no-T2S on the unmodified fixtures. Registry unit test added;golangci-lintclean.Remaining (separate step)
The JIT arm — compiling a reusable skill so the second run is cheaper — is the last piece for a true break-even number. That's a deeper
aj compileintegration; I'll scope it next. This PR lands the gen-run harness + a real baseline sweep.Refs: benchmark sandbox design §3.2.
🤖 Generated with Claude Code