Anthropic says Claude Code goes from 18% success in raw chat to 95% in an engineered…
This is a dev post classified by AGK Intelligence as AI dev tools (a tutorial), kept by the Dev Radar because it carries real work, not commentary.
F*cking genius Anthropic says Claude Code goes from 18% success in raw chat to 95% in an engineered harness Set that up with one prompt, and turn Claude Code into a self-governing production machine If the harness doesn't sign a verified receipt, git won't let the agent touch main Here's the build: step 1 → paste the prompt below into Claude Code on Opus 5.5 step 2 → Opus compiles your intent into a bounded task_contract.json before Sonnet generates token 1 step 3 → it spins up an isolated git worktree so the agent never modifies your active working branch step 4 → Haiku 5.5 acts as a
Posted by marfin (5.6k followers) 3 h ago · 29 likes · 1.2k views · view the original post on X. Kept by the Dev Radar as AI dev tools.
More dev work like this
- The first AI agent companies to try Daytona said the product was useless. — @guilleflorvs
- thinking machines dropped new model — @notjazii
- If you want to learn how to build self-improving harnesses, check out this paper. — @dair_ai
- Gemini 4 Hydrogen is currently testing on arena under the name gemini-3.8-flash and it's… — @thtbee_
- Qwen 4 (unnamed quant) — @loktar00
- My first Qwen 4 single shot! — @loktar00
- Robots Learn New Skills In Just 29 Seconds with AI Trained Framework! #BigData… — @gp_pulipaka
- #Morpheus #NLProc LLM! #BigData #Analytics #DataScience #AI #MachineLearning #IoT #IIoT… — @gp_pulipaka
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.8k posts from 5k X accounts over the last 21 days, 9.5k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-06 16:11 UTC. Full method.