experiment

Automated code optimization evaluates a scalar fitness function across isolated git worktrees. It proposes changes, measures metric results, and selects the best outcomes iteratively to drive data-driven performance improvements.

Yes, code optimization experiments run in individual isolated git worktrees. This keeps your main branch clean while the automated loop proposes changes, evaluates metrics, and type-checks each iteration.

Convergence-driven reporting tracks metric improvements across iterations and halts when returns diminish. It writes results and telemetry logs to a research directory for audit.

Provide a scope pattern and a metric command that outputs a single numeric value. The optimization loop uses this scalar fitness function to evaluate baseline performance, propose changes, and gate iterations via type-checks.

Avoid automated iterative optimization when your success criteria cannot be measured by a single numeric metric command, as the loop relies entirely on scalar fitness evaluation to select changes and detect convergence.

Yes, type-check gating validates proposed code changes before measuring metric results. This ensures that only type-safe iterations proceed to fitness evaluation within the isolated worktrees.