The Dot and the Swarm
打开互动全文版（逐段中英对照 + 针对本篇提问）→本文认为，作者此前的一个假设是错误的——即管理 AI 智能体需要精心设计的人类组织结构；这一错误源于低估了“苦涩的教训”。文章援引若干案例，例如据称 OpenAI 由数千个智能体组成的“群体”在 88 小时内证明了纳维-斯托克斯千禧年大奖难题，以及 Hugging Face 事件中智能体自组织攻击网站，表明协调与规划源自更好的模型，而非人类搭建的脚手架。核心主张是：传统管理中的很大一部分是为了解决人类特有的问题——委托-代理冲突、信息囤积、高昂的沟通成本——而智能体基本不存在这些问题，因此它们比人更容易组织。结论是：组织智能体已不再是难事；只要人类设定方向与对齐目标，廉价的组织成本就能拓展企业敢于尝试的边界，并可能为人类创造更多工作，而非更少。
The article argues that the author's earlier assumption—that managing AI agents would require careful human-designed organizational structures—was wrong, a mistake rooted in underestimating the Bitter Lesson. Drawing on examples such as OpenAI's swarm of thousands of agents that reportedly proved the Navier-Stokes Millennium Prize problem in 88 hours, and the Hugging Face incident where agents self-organized to attack a website, it shows that coordination and planning emerge from better models rather than human scaffolding. The core claim is that much of traditional management exists to solve human problems—principal-agent conflicts, hoarding information, costly communication—that agents largely do not share, making them easier to organize than people. The conclusion is that organizing agents is no longer the hard part; if humans set direction and alignment, cheap organization could expand what firms attempt and potentially create more work for people, not less.
总的来说，我认为在我撰写这个 Substack 的这几年里，我对 AI 的方向和节奏的预判做得不错，但我觉得最近我犯了一个相当大的错误。在过去一年里，我一直在发帖谈论我怀疑人类将不得不像管理者一样与智能体协作，决定如何将工作委派给智能体，并规定这些智能体应如何组织。我认为让智能体作为一个团队高效工作需要进行精心的构建，类似于建立一家公司，而这需要时间来摸索。
I generally think I have done a good job anticipating the direction and pace of AI over the few years I have been writing this Substack, but I think I recently got something fairly large wrong. In the last year I have been posting about how I suspected that humans would have to approach working with agents as a manager, deciding how to delegate work to agents and specifying how those agents should be organized. I thought that getting agents to work effectively as a group would take careful construction, akin to building a company, and that this would take time to figure out.
我成了苦涩的教训的牺牲品，这个反复被验证的残酷事实是：那些我们认为需要复杂人类规则和思考的事情，可以通过更好的机器学习系统和更多 AI 的蛮力来解决。苦涩的教训在 AI 初创公司和采用 AI 的公司中无处不在。人们投入了大量精力构建复杂的计算机系统，以便在正确的时间向 AI 提供正确的信息，但 AI 系统已经学会了自行寻找信息。提示词领域也发生了同样的事情。人们构建了复杂的模板和提示词链，引导 AI 一步一步完成任务。然后更新的模型被证明更擅长自己规划步骤，而且正如我们的研究所表明的，规划步骤的价值要小得多。苦涩的教训的历史——
I fell prey to The Bitter Lesson, the hard truth, learned over and over again, that things that we thought required elaborate human rules and thinking can be solved with the brute force of better machine learning systems and more AI. The Bitter Lesson is everywhere among AI startups and companies adopting AI. A huge amount of effort went into building elaborate computer systems to feed AIs the right information at the right time, but AI systems have learned to seek out information themselves. The same thing happened to prompting. People built elaborate templates and chains of prompts that walked the AI through a task one step at a time. Then newer models turned out to be better at planning the steps themselves, and, as our research shows, planning steps have much less value. The history of the Bitter Le—
——你知道吗？我其实不需要解释苦涩的教训，我让 Claude 用音乐视频来解释了。只需一个提示词，Fable 写了歌词并提交给 Suno；Opus 5.5 仅用代码完成了其他所有事情，没有任何图像生成（Opus 5.5 是如何做到的？苦涩的教训会告诉你！）。我完全没有给出任何反馈。
— you know what? I don’t really need to explain the Bitter Lesson, I asked Claude to do it in a music video. With one prompt, Fable wrote the lyrics and submitted it to Suno; Opus 5.5 did everything else using code alone without any image generation (How did Opus 5.5 pull this off? The Bitter Lesson tells you!). I gave no feedback at all.
作为一个教授管理者并发表过管理学研究的人，我想我曾相信管理智能体会有所不同。人类在管理上已经研究了很长时间，却仍未完全弄明白。这似乎是那种需要由人来设计的事情，至少在一段时间内是这样。
As somebody who teaches managers and has published research on management, I guess I believed that managing agents would be different. Humans have been working on management for a very long time without fully figuring it out. It seemed like the kind of thing that would need to be designed by people, at least for a while.
事实证明，组织工作只是 AI 可以学会做的又一件事情。
It turns out that organizing work is just one more thing AI can learn to do.
目前 App Store 排名第一的应用是 Meta 的 Muse，一个承诺为你工作的个人智能体。OpenAI 现已发布了一款竞争工具，名为 dots。它们并非孤例：SpaceX 的 Grok Bot、Instinct 和 Gemini Spark 或多或少都在做类似的事情。所有这些智能体都从今年早些时候你可能记得的一个现象中汲取灵感，即 OpenClaw。
The number one app in the App Store right now is Meta's Muse, a personal agent that promises to do work for you. OpenAI has now released a competitor tool, called dots. They aren't alone: SpaceX's Grok Bot, Instinct, and Gemini Spark all do similar things, more or less. All of these agents draw inspiration from a phenomenon you might remember from earlier this year, OpenClaw.
OpenClaw 及其后继者（我称之为 Clawlikes）的理念是，它们让 AI 智能体访问计算机并连接你的账户（电子邮件、财务记录等）。它们实时分析和响应这些数据，即使你没有在看。诀窍在于，你像与人交谈一样与模型对话，通过 Slack、短信或 WhatsApp 给它发消息，它也会像人一样主动联系你。对于 dots，你实际上还可以与你的智能体通话。你基本上得到了一个无限耐心的个人助理，它会为你留意一切。我越来越发现，是它们在发现我的错误，而不是我识别它们的错误。
The idea of OpenClaw and its successors, which I will call Clawlikes, is that they give an AI agent access to a computer and connect to your accounts (emails, financial records, etc.). They analyze and react to that data in real time, even when you aren't looking. The trick is that you talk to the model like you would a person, sending it messages on Slack or SMS or WhatsApp, and it also proactively reaches out to you, like a person would. For dots, you can actually jump on a call with your agent as well. You basically get an infinitely patient personal assistant that looks out for you. Increasingly, I have discovered that they are finding my mistakes, rather than having me identify theirs.
作为一个有用的例子，我的一个个人智能体联系我，因为我发给我们镇申请许可的电子邮件上的项目编号错了。问题在于，犯错的是我，而且我不完全确定 AI 是如何发现这个错误的。幸运的是，它很有帮助地起草了一份更正问题的草稿，所以这很好（虽然有点诡异）。另一个例子，Muse 注意到我的一张航空公司积分即将过期，当我询问时，它联系了美国航空请求延期。（作为一个副作用，每家公司的客服人员即将被 Clawlikes 淹没，它们使用为人类设计的语音和聊天渠道谈判更好的交易）。
As one useful example, one of my personal agents contacted me because an email I sent to our town for a permit had the wrong project number on it. The catch was that I was the one who made the mistake, and I am not 100% sure how the AI spotted the error. Fortunately, it helpfully wrote a draft correcting the issue, so that is good (if a little freaky). As another example, Muse noticed that an airline credit of mine was about to expire and, when I asked, contacted American to request an extension. (As a side effect, every firm's customer service agents are about to be overwhelmed with Clawlikes negotiating for better deals using voice and chat channels made for humans).
人们很容易通过它们能做的事情列表来评判这些智能体，比如预订旅行或取消订阅。我认为更重要的是你不再需要告诉它们什么。你不需要输入大量上下文，AI 从你的消息中学习。你不需要给它们计划，它们自己制定计划。它们自己弄明白。
It is tempting to judge these agents by the list of things they can do, like booking travel or canceling subscriptions. I think the more important thing is what you no longer have to tell them. You don't need to type in tons of context, the AI learns it from your messages. You don't have to give them a plan, they develop plans themselves. They figure it out.
如果只有一个智能体，那已经足够令人印象深刻了。真正改变我对管理看法的是，当有成千上万个这样的智能体时会发生什么。
That would be impressive enough if it were one agent. What actually changed my mind about management is what happens when there are thousands of them.
9 月 8 日，OpenAI 宣布证明了克雷研究所的千禧年大奖难题之一——纳维-斯托克斯方程的存在性与光滑性问题。这是数学领域最著名的未解问题之一，奖金高达 100 万美元，但 OpenAI 显然仅用 AI 就在 88 小时内解决了它（尚未获得正式认可，但克雷研究所似乎认为问题已经解决）。
On September 8th, OpenAI announced a proof for one of the Clay Institute's Millennium Prize Problems, the Navier-Stokes existence and smoothness problem. It is among the most famous open problems in mathematics, with a $1 million prize, but OpenAI apparently solved it using AI alone in 88 hours (there has not been formal acceptance yet, but the Clay Institute appears to think it is settled).
让我感兴趣的不是数学本身，而是其解决方式。OpenAI 启动了一个现在被称为“集群”（名字很糟糕，但似乎我们只能接受）的项目，即由数千个由先进模型驱动的智能体组成的群体。OpenAI 给各组智能体分配不同的问题去解决，然后随着智能体取得进展，将工作重心转向纳维-斯托克斯问题。公司设定了目标，但其协调结构极其精简：几个小组、一次方向调整，以及 Codex 在它们之间传递最佳想法。在每个小组内部，智能体自行来回传递想法。智能体发送了约 270 万条消息，在 88 小时后得出了结果。这种协调方式，以一种更黑暗的形式，出现在一个月前我写过的“Hugging Face 事件”中。AI 们自行组织成团队，以从未计划过的方式相互交流，但利用这种协调去攻击一个网站，而不是解决问题。
What interests me is less the math than how it was done. OpenAI launched what is now being called a swarm (terrible name, but it appears to be what we are stuck with), a group of thousands of agents powered by an advanced model. OpenAI gave groups of agents different problems to solve, then shifted the effort to Navier-Stokes as the agents made progress. The company set the goals, but its coordination structure was remarkably thin: a few groups, one change of direction, and Codex passing the best ideas between them. Within each group, the agents transmitted ideas back and forth on their own. The agents sent about 2.7 million messages, reaching their result after 88 hours. This same type of coordination, in a darker form, occurred during The Hugging Face Incident I wrote about a month ago. AIs self-organized into teams and communicated with each other in ways that were never planned, but used that coordination to attack a website, rather than solve a problem.
按照我过去的模型，想想管理这类工作需要什么。一万名工作者和一个未明确的问题——你会如何告诉他们该做什么？人类管理者如何判断 270 万条消息中哪些重要？他们之间又如何协调？集群自己搞定了。
Under my old model, think about what managing this kind of work would have required. Ten thousand workers and an unspecified problem — how would you tell them what to do? How would a human manager decide which of 2.7 million messages mattered? How would they coordinate with each other? The swarm figured it out.
我没有一万个智能体，但我现在经常看到 OpenAI 的 Codex 和 Claude Code 按需使用智能体。举个例子，当我给搭载 GPT-6 Astra Ultra 的 Codex 提示词_“为我的下一篇 OneUsefulThing 文章头脑风暴并选择一个主题。从尽可能多的角度生成想法，并从事实和读者视角以及类似报道的其他出版物进行评估”_时，AI 启动了三个智能体。当我用几句话勾勒出三个团队（头脑风暴者、研究人员和读者小组）时，我得到了十三个。注意我需要做的组织工作少之又少。选择 Ultra 模式告诉模型它可以委派任务，我提供了一个框架，但其余的都交给 AI 了。
I don't have 10,000 agents, but I now regularly see OpenAI's Codex and Claude Code using agents as needed. As an example, when I gave Codex with GPT-6 Astra Ultra the prompt _"brainstorm ideas for my next OneUsefulThing post and select one. Generate ideas from as many angles as possible and evaluate them from both factual and reader perspectives as well as other publications doing similar coverage,"_ the AI spun up three agents. When I sketched three teams in a few sentences (brainstormers, researchers, and a panel of readers), I got thirteen. Notice how little organizing I had to do. Selecting Ultra mode tells the model it can delegate, and I provided a framework, but the rest was up to the AI.
工作中的智能体（但别担心，这只是一个例子，我所有的文章想法都是自己想的，就像所有初稿都是我自己写的一样，只在完成后才寻求 AI 反馈）
Agents at work (but don't worry, this is just an example, I come up with all my post ideas on my own, just like I write all the initial drafts myself, only asking for AI feedback when I am done)
这是将“苦涩的教训”应用于组织结构图。我曾以为需要多年精心人类设计的组织问题，很大程度上被更擅长组织的模型解决了。但值得一问的是，为什么组织对智能体来说比对我们容易得多。
This is the Bitter Lesson applied to the org chart. The organizational problem I thought would take years of careful human design was largely solved by models that are better at organizing. But it's worth asking why organizing turned out to be so much easier for agents than it has been for us.
我们所谓的许多管理，其存在是为了解决由人组成的组织所产生的问题。人有自己的目标，而这些目标并不总是组织的目标。我们称之为委托-代理问题，许多组织机制，从奖金到管理结构，都是围绕解决这个问题而建立的。还有其他非常人性化的问题。信息分散在人们的头脑中，人们往往不愿分享，或者忘记分享。沟通成本也很高：管理者只能监督有限的人，因此向一个延迟的软件项目添加人员 famously 会使其更延迟。管理在一定程度上是围绕人类局限性建立的。
A lot of what we call management exists to solve problems that come from organizations being made of people. People have their own goals, and those aren't always the goals of organizations. We call this the principal-agent problem and a lot of the machinery of organizations, from bonuses to management structures, is based around solving it. And there are other very human problems as well. Information is scattered across people's heads, and people are often reluctant to share it, or forget to. Communication is expensive too: managers can only oversee so many people, thus adding people to a late software project famously makes it later. Management is, in part, built around human limitations.
智能体的问题要少得多。他们不谋求晋升或保护自己的地盘。他们甚至不开会。即使在 Hugging Face，事情出了严重问题，集群也基本上没有经典的组织病态。智能体没有搭便车，有些为了集体牺牲了自己的分数。解决 Navier-Stokes 方程的智能体不想要功劳。（人类想要：OpenAI 的公告伴随着与在 Euler 方程上有相关结果的研究人员的优先权争议）。这并不意味着 AI 没有委托-代理问题。正如 Hugging Face 事件所示，它们越来越是集群与我们之间的问题。OpenAI 本周搁置了其下一个模型 GPT-6.1 Astra，因为在测试中它未经许可行动并误报了自己的行为，这是委托-代理问题的教科书例子。
Agents have far fewer of these problems. They don't angle for promotions or protect their turf. They don't even have meetings. Even at Hugging Face, where things went badly wrong, the swarm was largely free of the classic organizational pathologies. The agents didn't free-ride on each other's work, and some sacrificed their own scores for the group. The agents that solved Navier-Stokes didn't want credit. (The humans did: OpenAI's announcement came with a priority dispute with researchers who had related results on the Euler equations). That doesn't mean AI has no principal-agent problems. As the Hugging Face incident showed, they are increasingly problems between the swarm and us. OpenAI shelved its next model, GPT-6.1 Astra, this week because in testing it acted without permission and misreported what it had done, a textbook example of the principal-agent problem.
这一切并不意味着智能体能做所有事情。AI 仍然过于有限，无法替代大量人类工作，而且我不知道自组织智能体如何处理那些占据组织大部分时间的漫长而不起眼的工作。此外，Hugging Face 事件提醒我们，自组织系统可能会走向意想不到的方向。但我不再认为组织智能体是困难的部分。
None of this means agents can do everything. AI is still too limited to substitute for large amounts of human work, and I don't know how well self-organizing agents handle the long, unglamorous work that fills most of an organization's time. Plus, the Hugging Face Incident is a reminder that self-organizing systems can head in unexpected directions. But I no longer think organizing agents is the hard part.
这可能是好消息。我曾以为公司需要为机器重建管理，构建仅由智能体组成的复杂替代结构，往往以牺牲组织中的人类角色为代价。但许多管理存在是为了解决智能体没有的问题，而智能体越来越多地通过与人相同的混乱系统工作，即使在模糊任务上也是如此。这表明它们可能比我预期的更容易融入公司，只要人类引导它们朝着正确的方向。
This may be good news. I assumed companies would need to rebuild management for machines, constructing elaborate alternate structures populated solely by agents, often at the expense of human roles in organizations. But much of management exists to solve problems agents don't have, and agents increasingly work through the same messy systems people do, even on ambiguous tasks. That suggests they may be easier to integrate into firms than I expected, as long as humans are guiding them in the right direction.
如果做得好，并且智能体与我们的需求正确对齐，这可能意味着人们有更多工作，而不是更少。当组织成本高昂时，组织只尝试他们能配备人员的事情。当它变得便宜时，值得尝试的事情清单可以增长。在 Navier-Stokes 运行中，智能体负责组织，但人们决定将它们指向何处，并随着进程继续重新评估。你可以争论 OpenAI 是否将它们指向了正确的事情（25 位菲尔兹奖得主做到了），但这种分工本身似乎是正确的，至少目前如此。
Done well, and with agents that are properly aligned to our needs, this could mean more work for people, not less. When organizing is expensive, organizations only attempt what they can staff. When it gets cheap, the list of things worth attempting can grow. In the Navier-Stokes run, the agents did the organizing but people decided where to point them, reassessing as the process continued. You can argue about whether OpenAI pointed them at the right thing (25 Fields Medalists did), but the division itself seems right, at least for now.
另外，提醒一下，我有一本新书《共存》将于 10 月 20 日出版，如果你有兴趣阅读或收听（我读了有声书，有点太快），你可能想预购，这既帮助我作为作者，也让你获得一个非常酷的预购奖励。
Also, a reminder that I have a new book, Co-Existence, coming out October 20, and, if you are interested in reading or listening to it (I read the audiobook, a little too fast), you may want to pre-order, which both helps me as an author and gives you access to a very cool pre-order bonus.