← 回總覽

傅盛实践分享:7 个 Agent 并行工作流

📅 2026-03-21 21:59 傅盛 人工智能 1 分鐘 1046 字 評分: 82
Karpathy Agent工作流 多Agent协作 AI编程实践 傅盛
📌 一句话摘要 傅盛引用 Karpathy 观点并分享个人实践:同时运行 7 个 Agent 各司其职,只负责给方向和验收结果。 📝 详细摘要 该推文是 Thread 的后续内容,傅盛进一步解读 Karpathy 的观点:AI 未能给出理想结果,不是 AI 能力问题,而是用户未找到将多个 AI 能力串连起来的方法。Karpathy 羡慕的人是 Peter Stemberg——同时管理十几个 Agent,每个运行 20 分钟,人负责切换、验收和任务分配。傅盛表示自己目前也是采用类似工作方式:7 个 Agent 各司其职,他只负责给方向、看结果。这条推文比前一条更具实操性,是个人实践经验的分享

Warning: This page contains shadow DOM that are currently hidden, consider enabling shadow DOM processing.

One Sentence Summary

Fu Sheng quotes Karpathy's insight and shares his personal practice: running 7 Agents in parallel, each with its own role, while only responsible for providing direction and reviewing results.

Summary

This tweet is a follow-up in the thread where Fu Sheng further interprets Karpathy's viewpoint: when AI doesn't give the desired results, it's not an AI capability issue, but rather the user hasn't found a way to chain multiple AI capabilities together. The person Karpathy admires is Peter Stemberg—managing more than a dozen Agents at the same time, each running for 20 minutes, with humans responsible for switching, reviewing, and assigning tasks. Fu Sheng shares that he currently uses a similar working method: 7 Agents each with their own responsibilities, and he only provides direction and reviews results. This tweet is more practical than the previous one, sharing personal hands-on experience.

查看原文 → 發佈: 2026-03-21 21:59:00 收錄: 2026-03-22 00:00:28

🤖 問 AI

針對這篇文章提問,AI 會根據文章內容回答。按 Ctrl+Enter 送出。