Era of AI Self-Improvement: Lilian Weng's Deep Dive into Agent Harness Engineering
craigsdennis · x · 2026-08-08
As AI models improve at coding, Recursive Self-Improvement (RSI) for software harnesses is becoming a major industry focus. This thread curates top recent readings on the subject:
- Lilian Weng's Deep Dive: An in-depth look at harness design patterns for self-improvement, covering workflow automation, file systems as persistent memory, and sub-agents, plus the joint optimization of the harness layer and core model intelligence.
- Self-Building IDE: Highlights bb, an agentic IDE built by @ymichael and @sawyerhood that evolves and builds itself.
- METR's Benchmarking: Discusses the difficulty of measuring automated kernel engineering with earlier models (like 4o-level) and the challenges of measuring realistic tasks necessary for RSI feedback loops.
Related event: Lilian Weng Explores Harness Engineering for AI Self-Improvement(3 posts)→
More from AGI Musings
- Beff Jezos declares 'civilizational inflection point' in the acceleration of technological progress — beffjezos · 2026-10-16
- Data scientist: pivoting to AI engineering is the safest career bet for the next 5 years — mdancho84 · 2026-10-16
- AI will be superhuman at science regardless, and Europe can't opt out — teortaxesTex · 2026-10-16
- beffjezos to Aaron Levie: techno-optimism is hyperstitious — basedjensen · 2026-10-16
- AI demos haven't changed since 2023, only gotten better — same scenarios — yangyi · 2026-10-16
- CFR experts: the autonomous weapons debate still has a definition problem — mchorowitz · 2026-10-16