The Mythical Man-Month in the Age of AI

Last weekend I watched four people desperately trying to get a laptop to connect to a screen. I was happily observing their efforts, with zero intent to join in.

上周末，我看着四个人拼命想把笔记本电脑连接到屏幕上。我愉快地观察着他们的努力，完全没有参与的打算。

But someone nudged me: can’t you have a look at it?

但有人推了我一下：你不能帮忙看看吗？

My reply: “Haha sorry no! My experience is that adding more people will not solve this problem any quicker”.

我的回答是：“哈哈，抱歉不行！根据我的经验，增加人手并不会让这个问题解决得更快。”

I was of course — indirectly — referring to Brooks’s Law:

我当然是在间接引用布鲁克斯定律（Brooks’s Law）：

“Adding manpower to a late software project makes it later.”

“向进度落后的软件项目中增加人手，只会使进度更加落后。”

It was formulated by computer scientist Frederick P. Brooks Jr. in his 1975 book The Mythical Man-Month, based on his experience managing the IBM OS/360 project.

这是计算机科学家小弗雷德里克·P·布鲁克斯（Frederick P. Brooks Jr.）在他 1975 年的著作《人月神话》中提出的，基于他管理 IBM OS/360 项目的经验。

However I got to thinking. Does this still hold up in the age of AI agents and LLMs?

然而我开始思考。在人工智能体（AI agents）和大语言模型（LLMs）的时代，这一理论还成立吗？

At first glance, you’d think no. The traditional bottlenecks — onboarding time, endless stand-ups, and office politics — disappear overnight. An AI agent is up to speed in milliseconds and doesn’t demand a coffee break.

乍一看，你可能会觉得不成立。传统的瓶颈——入职时间、无休止的站会和办公室政治——一夜之间消失了。AI 智能体可以在几毫秒内进入状态，而且不需要休息喝咖啡。

But when you look closer, and the law hasn’t vanished. It just mutated.

但当你仔细观察时，会发现这条定律并没有消失。它只是发生了变异。

The famous “Baby Problem” remains: nine agents still can’t deliver a baby in one month. Sequential tasks are still sequential.

著名的“生孩子问题”依然存在：九个智能体仍然无法在一个月内生出一个孩子。顺序任务依然是顺序的。

Overhead hasn’t disappeared; it shifted: instead of managing people, you are now managing agents. Someone still needs to define “done”, review the code, catch hallucinations, and stitch the pieces together.

管理成本并没有消失，只是发生了转移：你不再是管理人员，而是管理智能体。仍然需要有人来定义“完成”的标准、审查代码、捕捉幻觉，并将各个部分拼接在一起。

Throw nine autonomous agents at a single codebase without crystal-clear orchestration, and they will happily rewrite each other’s work in a high-speed loop.

如果将九个自主智能体投入到一个没有清晰编排的代码库中，它们会乐此不疲地在高速循环中互相重写对方的工作。

Maybe the 2026 update to Brooks’s Law should be:

也许 2026 年版的布鲁克斯定律更新应该是：

“Adding agents to a project with unclear requirements gets you the wrong answer faster.”

“向需求不明确的项目中增加智能体，只会让你更快地得到错误的答案。”

⬅ Previous post Interview with Ondřej Surý in the Atlantic

⬅ 上一篇 《大西洋月刊》对 Ondřej Surý 的采访