智能体、循环、图。你需要知道的一切尽在此处。
[
[
Agents, Loops, Graphs. Everything You Need to Know in One Place.
智能体、循环、图。你需要知道的一切都在这里。
Most people use AI at one speed: ask, read, fix, ask again. It works. It's also the slowest way to do it.
大多数人使用 AI 的速度都是一样的：提问、阅读、修改、再提问。这行得通。但这也是最慢的方式。
There is a faster setup. It has three parts: an agent that plans and acts without being walked through each step, a loop that runs until the work actually clears the bar, and a graph of parallel workers that handles what a single prompt never could. Claude runs all three. Most people never find out.
有一种更快的设置。它包含三个部分：一个无需逐步引导就能规划和执行的 agent，一个一直运行直到工作真正达到标准的 loop，以及一个由并行 workers 组成的 graph，用来处理单个 prompt 永远无法处理的任务。Claude 运行了所有这三个部分。大多数人从未发现这一点。
Not because it's complicated. Because nobody put all three parts in one place.
不是因为它复杂。而是因为没有人把这三个部分放在一个地方。
By the end of this article you will understand what Claude is actually capable of. What agents, loops, and graphs are, how each one builds on the last, when they are worth using and when they are a trap, and how to build all three yourself with prompts and code you can use today.
读完本文，你将了解 Claude 的实际能力。什么是智能体、循环和图，它们如何层层递进，何时值得使用、何时是个陷阱，以及如何用今天就能用的提示词和代码自己构建这三者。
Agents
Agents
1 - What an agent actually is
1 - agent 实际上是什么
Asking Claude "summarize this article" is not an agent. That is a chat.
让 Claude "总结这篇文章"并不是智能体。那只是聊天。
Telling Claude "find the three most cited papers on this topic, pull the key claims from each, cross-check them against each other, and write me a one-page briefing", and having it do all of that without you touching anything in between, that is an agent.
告诉 Claude "找出关于这个主题被引用最多的三篇论文，提取每篇的核心观点，相互交叉核对，然后给我写一份一页纸的简报"，并让它在你不干预任何中间过程的情况下完成所有这些，这就是智能体。
The difference is not the model. It is the structure around it.
An agent has three things a regular chat does not.
区别不在于模型，而在于围绕它的结构。
智能体拥有普通聊天所不具备的三个要素。
Tools it can call on its own: search, file systems, code execution, external APIs. It does not wait for you to go get the informati...