--- format: "aidr-story-markdown/v1" id: "22f676a15f74743af4764036169bb876f6605ff339f2535fa395889308e8945a" canonical_url: "https://aidr.today/22f676a1?lang=en" title: "Google Research Open-Sources RRSI: AI Agents That Improve Their Own Harness Without Overfitting" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-09-29T09:01:36.000Z" category: "Agents" topics: ["google","agent"] source_urls: ["https://www.marktechpost.com/2026/09/29/google-research-open-sources-rrsi-ai-agents-that-improve-their-own-harness-without-overfitting/"] summary: "Google Cloud AI Research has open-sourced RRSI, a framework that lets LLM agents rewrite their own prompts, tools and memory while model weights stay frozen. It adds a leakage critic, a noise floor, a cost rule and pruning so gains carry over to new tasks. With Claude Opus 4.8, Terminal-Bench 2.1 rose from 74.2% to 80.2%, and all 6 held-out splits improved. The post Google Research Open-Sources RRSI: AI Agents That Improve Their Own Harness Without Overfitting appeared first on MarkTechPost ." --- # Google Research Open\-Sources RRSI: AI Agents That Improve Their Own Harness Without Overfitting > [Open the canonical story]() **Published:** 2026-09-29T09:01:36.000Z **Category:** Agents **Topics:** google, agent ## Summary Google Cloud AI Research has open\-sourced RRSI, a framework that lets LLM agents rewrite their own prompts, tools and memory while model weights stay frozen\. It adds a leakage critic, a noise floor, a cost rule and pruning so gains carry over to new tasks\. With Claude Opus 4\.8, Terminal\-Bench 2\.1 rose from 74\.2% to 80\.2%, and all 6 held\-out splits improved\. The post Google Research Open\-Sources RRSI: AI Agents That Improve Their Own Harness Without Overfitting appeared first on MarkTechPost \. ## Sources - [Story source]()