Meta and Harvard Release Open-Source AI Software Engineer That Outperforms Larger Models on Complex Coding Tasks

Meta and Harvard researchers unveil Confucius Code Agent, an open-source AI software engineer that achieves 52.7% success rate on complex coding benchmarks, proving that smart system design can outperform larger AI models through innovative memory management and self-configuring capabilities.

• Meta and Harvard researchers release the Confucius Code Agent (CCA), an open-source AI software engineer built on the Confucius SDK designed for industrial-scale software repositories and long-running sessions
• The system achieves 52.7% Resolve@1 on SWE-Bench Pro with Claude 4.5 Sonnet, outperforming stronger models with weaker scaffolding, demonstrating that agent architecture can outweigh model size
• CCA features hierarchical working memory for long-horizon coding tasks, persistent note-taking that reduces token usage and improves performance across sessions, and a meta-agent that automatically designs and tunes agent configurations