Last released Aug 11, 2026
Run agent optimization loops against an evaluator the agent cannot change
Supported by