Terminal-RSI: self-improving agent system lifts 27B model success rate from 20.68%
Dr_Singularity · x · 2026-11-09
Researchers introduced Terminal-RSI, a system that improves AI agents through recursive cycles of rewriting successful problem-solving strategies and refining decision-making policies. The resulting 27B-parameter model reportedly saw its average success rate jump from 20.68% significantly higher. The post frames it as a contrast to US 'ban AI' discourse — AI that recursively improves itself and gets dramatically better.
More from coding & agent
- GitManual turns any GitHub repo into a step-by-step setup guide — adnan_hashmi · 2026-11-09
- DeepSeek Harness hits 1.54M daily users in under 3 months, burning 92T tokens/day — rohanpaul_ai · 2026-11-09
- Hermes Agent ships Security plugin: parallel subagents find and verify code vulnerabilities — NousResearch · 2026-11-09
- Warp fires back at 'factories don't ship software' jab: 400 factory PRs in one day — vikvang1 · 2026-11-09
- The fence metaphor for agent trust: draw boundaries, then let agents roam free inside — manosaie · 2026-11-09
- Claude Code project structure explained: when every config file loads — alex_verem · 2026-11-09