a recursive self-improving harness designed to help your agents (and future iterations of those agents) succeed on any task
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).
