autoresearch

Automated code optimization modifies code paths, benchmarks results, and permanently keeps only changes that improve execution time. It targets bottlenecks iteratively, reverting non-improving edits automatically to ensure consistent performance gains without manual intervention.

An autonomous performance optimization loop is a guided process that establishes baseline benchmarks, analyzes bottlenecks, and applies code changes per iteration. It automatically rolls back modifications if performance metrics do not improve, storing all artifacts for traceability.

Autonomous code optimization works across code paths where performance matters, from simple functions to large, data-intensive operations. It targets bottlenecks to reduce execution time and query counts, applying changes selectively based on benchmark results.

Start an iterative benchmark loop by running the /autoresearch command. This initiates the autonomous optimization process, which establishes baseline benchmarks, analyzes bottlenecks, applies changes, and logs progress under the autoresearch/ workspace directory.

When a code change does not improve execution time, the system automatically rolls back the non-improving edit. This ensures only modifications that actively reduce execution time or query counts are kept, maintaining code stability while optimizing performance.