autoresearch

Evaluator-guided iterative improvement is automated by defining a single-mission contract and running a bounded loop that applies the evaluator, tracks iterations, and stores per-iteration evaluations. This ensures reproducibility and traceability across runs.

Durable logs for iterative experimentation are maintained by storing per-iteration evaluations and markdown decision logs under a dedicated directory. This approach provides audit trails and supports periodic reruns without losing historical state.

Periodic reruns for an evaluator-driven improvement loop are set up using native cron scheduling. You configure your mission and evaluator, then start the loop to begin bounded, periodic enhancement automatically.

Yes, a predefined evaluator and mission spec are required to start bounded iterative improvement. The system enforces a single-mission contract, meaning both inputs must exist to orchestrate the loop and track iterations reliably.

Iterative experimentation requires a single-mission contract to ensure strict evaluation boundaries, reproducibility, and traceability. This constraint prevents scope drift during the bounded loop and keeps per-iteration logs aligned to one mission.

Yes, cron-based scheduling can be used to trigger periodic enhancement cycles. Native cron support allows the bounded iterative loop to rerun automatically, applying the evaluator and logging results for each iteration.