Description

A framework for building AI agents where the 'harness' (the code managing tool calls, retries, and memory) is autonomously rewritten by a 4B model based on execution reports and scoring sets. Instead of a human writing static logic, the system iterates through rounds of edits, keeping only the code that scores best on held-out test questions.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.