Harness Engineering for Self-Improvement
lilianweng.github.ioAbout
The concept of recursive self-improvement (RSI) dates back to I. J. Good (1965) , where he defined an “ultraintelligent machine” as a system that can surpass humans in all intellectual activities and design better machines to improve itself. Yudkowsky (2008) used the phrase “recursive self-improvement” for a specific feedback loop: an AI uses its current intelligence to improve the cognitive machinery that produces its intelligence. This feedback loop in modern AI may ind
Why it made the leaderboard
Weng reframes the AI coding conversation from prompting tricks to the harness itself—the scaffolding that lets a system improve the machinery producing its own intelligence, tracing the idea from Good's 1965 ultraintelligent machine through Yudkowsky's recursive feedback loop. Read this if you're building agentic tooling and want to think structurally about where self-improvement loops actually close, not just how to squeeze one more good completion out of a model.
Comments
No comments yet.