ck:autoresearch

Automated iterative optimization loops run configurable experiments against mechanical metrics, logging results to loop-results.tsv. They apply atomic commits per iteration to git-tracked projects, automatically keeping favorable changes and rolling back regressions.

Autonomous experimentation with git rollback improves mechanical metrics like test coverage, bundle size, performance scores, or ESLint error counts. The process defines Goal, Scope, and Verify requirements to evaluate each iteration and commit only favorable changes.

Autonomous optimization works directly with existing git commit history by applying an atomic-per-iteration commit strategy. It evaluates each automated experiment against defined verification requirements, automatically discarding changes that cause metric regressions.

Automated performance experiments support optional guard steps defined alongside Goal, Scope, and Verify requirements. These guards evaluate each iteration before committing, ensuring only changes that successfully improve the targeted metric are kept in the git history.

Running multiple iterations against a metric without manual reviews requires defining mechanical verification rules and executing autonomous loops. The system automatically logs results to loop-results.tsv and handles keep or discard decisions based on git-tracked experimentation outcomes.

Automated optimization loops discard code changes during iteration when a mechanical metric regresses below its baseline. The rollback mechanism ensures git history retains only atomic commits that meet the predefined Goal, Scope, and Verify requirements for the experiment.