Claude Code Creator Manages Thousands of Agents with Loops
新智元 · wechat · 2026-07-18
Boris Cherny, the creator of Claude Code, shared how he drives large-scale agentic workflows using loop and goal. He now rarely writes code manually, submitting dozens of PRs daily—peaking at 150 in a single day. His company runs hundreds of AI agents during the day and thousands at night.
The article breaks down this methodology:
- loop: Executes fixed tasks on a per-minute, hourly, or daily basis, such as monitoring PRs, fixing CI, and clustering user feedback.
- goal: Given clear completion criteria, the model iterates autonomously until the goal is met.
- The core isn't "better chatting," but an acceptance closed-loop: an independent model verifies the results to prevent agents from churning out garbage.
- He also outlines a 20-minute local context configuration practice, using folders, memory files, and instruction files to ground the model in your business context.
The article concludes that the engineer's role is shifting from "writing code" to "designing systems that autonomously generate code."
Related event: From Prompt Engineering to Loop Engineering(6 posts)→
More from coding & agent
- Agent Field Kit launches $29 security runbooks for production AI agents — alvelda · 2026-10-27
- Dev mixes Claude specs with abliterated GLM to build SS7 threat detector — l4rz · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- GitHub Next's Chopin prototype explores real-time multiplayer agentic planning — max__drake · 2026-10-27
- Veynod: An Open-Source Shared Workspace Where AI Agents Collaborate Like a Team — TKOWNS · 2026-10-27
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27