autoresearch

To automate a research and verification loop, you need a tracked issue, a measurable goal, and a defined iteration cap to manage the convergence process. The system then handles planning and execution procedures to reach the target.

Bounded convergence is a controlled loop of research, implementation, and testing that executes iteratively to reach a defined goal. It ensures deterministic proof and consistent revision history before handing off the final artifact.

To ensure automated code fixes are independently verified, use an evidence-based validation process that records every iteration with deterministic proof and reviewer dispositions. This guarantees the agent checks its own work across multiple iterations.

To start an automated review and execution loop, you must provide a specific target issue and your desired outcome. This initiates the bounded cycle of research, implementation, and testing to resolve the problem.

You should use an automated fix and verification loop when manual iterative development becomes inefficient. It is ideal for resolving specific issues or implementing features where the agent must independently verify its own work through multiple iterations.

The automated research process requires an iteration cap to strictly manage the convergence process. This bounded approach prevents infinite loops, ensuring the research, fix, and verification cycle reaches a deterministic conclusion within set limits.