Prime Agent: A self-improving RLM agent
Prime Agent is our open-source, self-improving coding harness built around two abstractions: the Recursive Language Model (RLM) and the Continual Harness. With Opus 5, it achieves 95.5% on ARC-AGI-3, surpassing the reported human expert baseline.
Summary
Prime Agent is an open-source coding tool that enhances its own performance using a Recursive Language Model and a Continual Harness. With its latest update, it has outperformed human experts on a key benchmark, achieving a 95.5% score on ARC-AGI-3.
Comments (68 of ) ⌛