concurrent.cd
Tool / Flake-rate bisection

FlareBisect

https://pypi.org/project/flarebisect/

Bisects on failure rate rather than pass/fail: runs each candidate commit's test N times and names the first commit where the rate rises past a threshold, with an LLM explanation of the break.

The summary describes what the project says it does; inclusion is not an endorsement.

Take

Plain bisect blames the wrong commit when tests are flaky, which gets worse as many agent commits land. Alpha (July 2026), no documented CI integration, no auto-revert.

AI-drafted

Mentioned in

Tool

bisect-agent

github.com/gigaminkyu/bisect-agent

Replaces single-run git bisect with a sequential hypothesis test per probe so flaky tests don't mislead it, runs probes speculatively, shrinks the culprit to the relevant hunk and tests candidate fixes.

Link to this tool: bisect-agent
Take

"git bisect assumes the test tells the truth." Overlaps FlareBisect but adds hunk minimisation and speculative probes. 0 stars, one push, but real code.

  • Verify
  • Integrate
AI-drafted