
structure|1️⃣ 三级笔记、思想框架
Simon Willison 提出 designing agentic loops(设计 agent 循环)是 AI 时代一项关键的新技能。Coding agent 的本质是「在循环中运行工具以达成目标」的暴力搜索器——只要你能把问题还原为一个清晰的目标和一组可迭代的工具,agent 就能暴力破解出有效方案。
My preferred definition of an LLM agent is something that runs tools in a loop to achieve a goal. The art of using them well is to carefully design the tools and loop for them to use.
Anthropic 官方建议 Safe YOLO mode:在无网络容器中用
--dangerously-skip-permissions,锁定可访问的 trusted hosts 白名单防止代码外泄。
The value you can get from coding agents and other LLM coding tools is massively amplified by a good, cleanly passing test suite.
Simon 提出的 agentic loop 设计框架可以概括为四步:
贯穿全文的底层逻辑:agent = 暴力搜索器,你的工作不是写代码,而是设计搜索空间(工具 + 循环 + 安全边界),让 agent 在其中高效地暴力求解。
concepts|2️⃣ 关键概念、概念网络
context:
My preferred definition of an LLM agent is something that runs tools in a loop to achieve a goal. The art of using them well is to carefully design the tools and loop for them to use.
费曼一下:Agentic loop 是 LLM agent 的核心运行机制——agent 在一个循环中反复调用工具、观察结果、修正错误,直到达成目标。它不是一次性的“输入-输出”,而是一个持续迭代的闭环过程。Simon 认为“设计这个循环”本身就是一项重要的新技能。
context:
A critical new skill to develop is designing agentic loops.
费曼一下:这是 Simon 在本文中正式命名的技能。它指的不是写代码,而是为 agent 精心设计「搜索空间」——选择哪些工具、如何组织循环、给多少权限、在什么环境中运行——让 agent 能高效地暴力搜索解决方案。
context:
One way to think about coding agents is that they are brute force tools for finding solutions to coding problems. If you can reduce your problem to a clear goal and a set of tools that can iterate towards that goal a coding agent can often brute force its way to an effective solution.
费曼一下:Coding agent(如 Claude Code、Codex CLI)是能直接运行代码、测试、修复错误的 LLM agent。Simon 的核心洞见是把它们理解为「暴力搜索器」,而不是「智能程序员」——它们通过大量试错而非精准计算来找到解决方案。
context:
Each of these tools provides its own version of what I like to call YOLO mode, where everything gets approved by default. This is so dangerous, but it's also key to getting the most productive results!
费曼一下:YOLO mode 是让 agent 自动批准所有操作、不需要人工确认的运行模式。它是释放 agent 暴力搜索能力的关键,但也带来三大风险(文件损坏、数据窃取、被当跳板)。Simon 的关键洞见:安全与效率的平衡点在于「安全地开启 YOLO mode」,而不是不开启。
agentic reading|3️⃣ 费曼 x3