Bingchen is working on recursively self-improving AI

Previously built

• AIDE² /
• Hyperagents /
• LLM Speedrunner /
• Semanticist /
• OOD-CV /
• LLaMA-3 Recaption /
• LayerNorm Tuning

Currently

Building systems that improve themselves—AI that rewrites its own training to get better at getting better.