{"version":1,"type":"story","url":"https://digestai.news/story/google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati","json":"https://digestai.news/story/google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati.json","markdown":"https://digestai.news/story/google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati.md","slug":"google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati","headline":"Google releases Harness² framework for AI prompt and tool reorganization","summary":"Google Cloud AI Research and Northwestern University posted the Harness² paper on arXiv on October 9, 2026 and open‑sourced the code on GitHub (≈16,700 stars). The framework lets an AI model keep its pre‑trained weights while a recursive system rewrites eight components – system prompts, guardrails, memory, tool descriptions, sub‑agents, skills, scripts and plugins – to better suit each task. It runs a four‑step cycle of proposing, probing, and composing improvements, then records successful changes in a “recipe notebook” for future jobs.\n\nBenchmarks show up to an 11.5‑point lift per review. In a verification case the score rose from 43.1 to 77.8. On the WorkBuddy‑Bench, GLM‑5 with OpenCode improved from 64.2 to 75.7, overtaking Gemini 3.5 Flash (72.4). With one review Gemini 3.5 Flash and OpenCode moved to 77.5 and reached 80.8 when five candidates were tried in parallel. The authors note that removing the notebook drops performance, confirming the learning effect. Installation requires Linux, Python 3.12, and the uv tool; a demo runs without an API key but the README warns about unrestricted network access, disabled sandboxing for Codex, and API cost accumulation. The code is labeled research‑only and not an officially supported Google product.","keyPoints":["Harness² rewrites prompts, tools and other components without retraining, yielding up to 11.5‑point gains per review.","Reported benchmark lifts include 43.1→77.8 on verification and GLM‑5 64.2→75.7, surpassing Gemini 3.5 Flash.","Open‑source repository (≈16,700 stars) needs Linux, Python 3.12, uv; not an officially supported Google product."],"whyItMatters":"The framework demonstrates a low‑cost way to boost existing model performance by reorganizing the surrounding system, which could reduce the need for expensive retraining across many AI applications.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Google","Northwestern University","TypeSafe AI","Meta"],"models":["Gemini 3.5 Flash","GLM-5","OpenCode","Codex","Jev"],"people":["JosephZhou","Muyang"]},"firstPublishedAt":"2026-10-21T19:05:00Z","updatedAt":"2026-10-21T19:05:00Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"note.com","title":"Google's \"Harness2\" has arrived. A 5-minute breakdown of how AI improves its tools by reorganizing the kitchen without changing the chef's skills","url":"https://note.com/dende2023/n/n5cae3b1096fc?hl=en","publishedAt":"2026-10-21T19:05:00Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"Google Unveils Harness Framework for Self Improving Agents","url":"https://digestai.news/thread/google-research-open-sources-rrsi-for-self-improving-ai-agents-without","storyCount":4},"cite":{"text":"Digest AI, \"Google releases Harness² framework for AI prompt and tool reorganization\", 21 October 2026, https://digestai.news/story/google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati","publisher":"Digest AI","title":"Google releases Harness² framework for AI prompt and tool reorganization","datePublished":"2026-10-21T19:05:00Z","url":"https://digestai.news/story/google-releases-harness2-framework-for-ai-prompt-and-tool-reorganizati"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}