North Star: Self-Improving Agents

Self-improving agents require a feedback loop where both agents and evals evolve together—not just better prompts, but better evaluations too.

Improve Agent

Improve Evals

Collect data

Run evals