AutoResearch is a framework for autonomous, recursive self-improvement. By defining a strict objective metric and a sandboxed testing environment, you can let an AI agent continuously run experiments, keep what works, and discard what doesn't.
When implemented correctly, it effectively replaces hundreds of manual A/B tests with an automated loop that can run 24/7. When implemented poorly, it will confidently optimize your system toward a meaningless metric.
Here is a comprehensive guide to bootstrapping an AutoResearch loop for any measurable task.