AI 的未来：本地计算机与 Claude

这也许是我与 AI 领域很多人持有的一种相反观点。我实际上不认为未来会是高度个性化的软件，以至于每个人都运行自己的版本。我反而觉得，在公交车上拥有我们自己的内部聊天工具会很有帮助。硅谷整体上低估了本地计算机的价值。我对此的默认论据总是：为什么我们都在用 MacBook，而不是 iPad 或 Chromebook？现在当我想到 Claude 时，它是一个应该对你非常有用的实体。我认为这个实体需要能够访问你拥有的所有工具。否则，它会在所有这些复杂的方式中受到束缚。

This is maybe where I hold like a somewhat contrarian view to a lot of people in the AI. I actually don't think that the future is going to be hyper personalized software down to the point where everyone is running their own version. Like I actually think it's going to be quite helpful on the bus to have our own internal chat tool. Silicon Valley overall is undervaluing the local computer. And my default argument for that is always, how come we're all using MacBooks and not like an iPad or a Chromebook? And now when I think about Claude, it's this entity that is supposed to be very useful to you. Like it's tremendously useful to you. I think that entity needs to have access to all the same tools you have access to. Otherwise, it's going to be hamstrung in all these complex ways.

大家好，欢迎收听 Latent Space 播客。这是我们在 Kernel Loop 新工作室的第一期节目。我是 Alessio，Kernel Labs 的创始人，和我一起的是 Sweeks，Latent Space 的编辑。

Hey everyone, welcome to the Latent Space podcast. Our first one in the new studio at Kernel Loop. Uh this is Alessio, founder of Kernel Labs, and I'm joined by Sweeks, editor of Latent Space.

是的，很高兴来到这里。感谢 TJ、Alessio 和 Alan 帮忙布置好一切。看起来很美。我们甚至在外面有 logo。是的，很好，谢谢。真的很棒。

Yeah, so nice to be here. Thanks to uh TJ, Alessio, and Alan helping to set everything up. It looks beautiful. We even have the logo outside. Yeah, that's good and thanks. It's like really nice.

当你作为嘉宾走进这里时，你会想，“哦，这是一个严肃的制作。”你立刻就能感觉到。是的。Felix，你目前是 Covalent 的产品经理，或者呃每年我们都会换头衔。是的，身份有点模糊。技术团队成员。我知道。技术团队成员是我们会永远带着的官方头衔。是的。我最近有点想，我们一直，我有点着迷，我经常用它来管理 Latent Space。比如 Covalent 帮我上传视频、加标题、编辑等等。它真的很棒。酷。他在群聊里多次说 Covalent 是 AGI。是的，是的。所以我们有第二个频道用于 Latent Space TV。基本上这是我们的 Discord 聚会。我们觉得 Claude Covalent 可能是 AGI。我不知道我们是否已经上传了，但有一次会议就是关于 Claude Covalent 的。我很想看看。我很好奇。我工作中最有趣的部分之一就是不断看到人们用 Cocalc 做的奇怪事情，因为显然我们很难为特定用例设计。我们确实在尝试，但每个最惊讶的人通常都是因为一件我甚至没想到 Cocalc 会擅长的事情。我们有一位新设计师，这是他的第一个小任务。我说，“嘿，我们需要为 Cocalc 的内部 Slack 做一个新 emoji。”这是一件很小的事情。我说，“你能做一下吗？”他画了一个 SVG 然后给了 Cocalc。我说，“你能把这个做成 emoji 吗？”现在它有了一个漂亮的循环动画。我认为这显然涉及到很多你可以用代码做更多事情的情况。但正是这类事情让我觉得很有趣。所以长话短说，我很想看看你们在做什么。我会调出来。我会调出来。是的。

When you walk in here as a guest, you're like, "Oh, this is a serious production." You're like feel it immediately. Yeah. Felix, you're currently product manager of Covalent or uh every year we change titles. And yeah, the identities are kind of vague. Member of technical staff. I know. Member of technical staff is like the official title we'll carry around forever. Yeah. I recently kind of wanted like we've been I kind of obsessed I I've been using it a lot even for managing Latent Space. Like uh Covalent helps me upload videos and like title things and like edit and everything. It's it's like really amazing. Cool. He said multiple times Covalent is AGI in the group chat. Yeah yeah yeah. So so we have a second uh we have a second channel uh for Latent Space TV. Uh and I uh and uh we basically this is our Discord meetup. Um and I I we have like Claude Covalent is might be AGI. I don't know if we we have uploaded it yet, but one of the sessions was like a like a Claude Covalent thing. I love to see I would to see it. Like I'm so curious. Like one of the most fun parts of my job is to like constantly see the weird things people use Cocalc for because it's obviously like very hard for us to actually design for specific use cases. We do, but like every single person who's like most amazed is usually amazed about a thing that I didn't even expect Cocalc would be good at. Um we have a new designer, and it's one of the first small tasks. I was like, "Hey, we need like a new emoji for Cocalc for our internal Slack." It's like a pretty small thing. I was like, "Can you please do it?" And he drew an SVG and just gave it to Cocalc. I was like, "Can you make this emoji?" And now it has like this beautiful loopy animation. Um and I mean, I think obviously this goes down to like a ton of stuff you can do more things with code than you expected. But, it is like that kind of stuff that is really fun to me. So, long story short, I would love to see like the kind of things you're doing. I'll pull it up. I'll pull it up. Yeah.

是的。但在我们深入之前，我总想先从一个高层次的问题开始：对于那些没听说过、没试过 Claude Cocalc 的人来说，它是什么？

Yeah. Uh but, before we get into it, I think always want to start with like a top-level what is Claude Cocalc for people who haven't heard of it, haven't tried it out.

好的。简单来说，Claude Cocalc 是 Claude Code 的用户友好版本。它的基本工作方式是，我们有 Claude Code 和一个相当令人印象深刻的智能体框架。在去年 12 月，我们注意到越来越多的人在使用它，即使他们不是技术人员，不熟悉终端，或者他们熟悉终端，但开始将 Claude Code 用于非编码工作负载。比如管理开支、填写收据或组织知识库。有一个很多人喜欢的 Obsidian 时刻。我们想利用这一点，同时也将这种能力带给那些不习惯终端、可能不知道如何用 brew install 安装东西的人。所以，Cocalc 是在虚拟机中运行的 Claude Code，带有一些额外的填充和更多的护栏，使其更安全、更方便，适合那些不想在上班时先打开终端的人。

Okay. Uh real quick, Claude Cocalc is a user-friendly version of Claude Code. So, the way it basically works is we have Claude Code and for us fairly impressive agent harness that over December we noticed more and more people are using either, even though they're not technical, they're not at home in the terminal, or they are at home in the terminal, but they started using Claude Code for non-coding workloads. Right? Like, managing expenses or like filling out receipts or organizing knowledge base. Like, there was a big Obsidian moment that a lot of people liked. And we wanted to capitalize on that, but also bring this capability to people who are not terminal native and who might not know how to like brew install something. So, Cocalc is Claude Code running in a virtual machine with a little bit of padding, a little bit more guardrails, making it a little safer, a little bit more convenient for people who don't want to first open up the terminal when they go to work.

有趣的是，它被定位为更用户友好的东西，因为我总觉得对我来说，我把它视为我熟悉 Claude Code 的原因。我们大约一年前做过一期关于 Claude Code 的节目。但这个更像是超级用户工具，因为它与 Cloud、Chrome 以及其他所有工具的集成更好。但也许这只是感知问题，对吧？

It's interesting uh that it's kind of pitched that way as a more user-friendly thing because I always feel like it to me I treat it as like why I'm familiar with Claude Code. Like we we did a Claude Code episode about a year ago. But this one is like even more power user tools, because it it kind of integrates much better with like Cloud and Chrome and in all the all the other tooling. But like maybe maybe that's like a perception thing, right?

不，老实说，我不认为你错了。这是我过去几周一直在思考的事情。当人们说“用户友好”时，他们觉得是简化版。但实际上，这是超集。是的。我认为类似的事情大约 10 年前也发生在我身上，也许是 12 年前，当时我在微软，我们开始研究 Electron 和基于浏览器的技术以及跨平台的东西。第一个用例之一是 Visual Studio Code，它曾经是一个网站。最初的叙述是 Visual Studio Code 是 Visual Studio 的更用户友好版本。但类似地，我认为有些声音说这不适合严肃的开发者。我们不会用它来做任何事情。我认为最终发生的事情是，人们对 Visual Studio Code 为何变得如此重要有不同的说法。但我个人的信念是，可破解性和可扩展性起了很大作用。你可以将 Visual Studio Code 连接到几乎任何工作负载。它很容易破解，很容易为其构建扩展。我认为 Code Work 可能也类似，它很容易扩展，很容易融入你的工作流程。所以便利性显然是我们作为开发者追求的东西。但我认为人们从中发现价值的方式可能是将其映射到他们实际工作中需要做的事情上。

No, honestly I don't think you're wrong. This is like a thing I've been thinking a lot about for like the last weeks. So people when they say user-friendly is like oh it's the dumbed-down version. But no, actually this is the superset. Yeah. Like I think a similar thing happened a similar thing happened to me about 10 years ago, like maybe 12 years ago when I was at Microsoft and we started working on on Electron and like browser-based technologies and cross-platform stuff. And one of the first use cases was Visual Studio Code, which used to be a website. And the initial narrative was oh Visual Studio Code is is like a more user-friendly version of Visual Studio. But in a similar vein, I think there were some voices saying oh this is not for serious developers. Like we're not going to use this, right? For like anything. And I think in the end what happened is people have different stories about why Visual Studio Code became such a big thing. But my personal my personal belief is that the hackability and the extendability is like played a pretty big role. Right? You can hook in Visual Studio Code to like almost any workload. It's so easy to hack on. So easy to build extensions for it. And I think Code Work might be hitting a similar thing where it's very easy to extend and it's very easy to bring into your workflows. Uh so the convenience I think is a bit of a it's obviously the thing we strive for as developers. But I think the way people find value in it then is by probably mapping it onto whatever they actually have to do in their job.

所以在去年年底，你看到 Claude Code 的非技术使用量激增。设计过程是怎样的，以至于我们决定让 Claude Code 工作？因为我的意思是你们只用了 10 天就建成了。我确定之前有一些关于“更容易使用”意味着什么的讨论。你知道？也许做一个桌面 GUI 显然是一种方式，但产品中有很多细微差别。也许跟大家讲讲触发点是什么，比如我们应该构建一个独立的东西。

So end of last year you see the spike of like non-technical usage in Claude Code. What's the design process to say we should make Claude Code work? Because I mean you built that in only 10 days. Um I'm sure there was some discussion before on what does easier to use mean. You know? Like maybe making like a desktop GUI is obviously one way to do it, but like there's a lot of nuance in the product. Like maybe talk people through what was like the trigger of like we should build a separate thing.

而且我们不应该构建一个完全不同的积木代码之类的东西。然后也许还有一些你没有采用的有趣设计决策。

And we should not build like a different block code thing. And then maybe some of the more interesting design decisions that maybe you didn't take.

是的。我认为在 Anthropic，我们一直在思考如何让那些习惯把 Claude 当作问答工具的人，也能获得更多能力，比如让它为你执行任务、解决问题或构建东西。我们如何将这种能力带给那些目前主要习惯聊天中问答模式的人？我们为此做了很多原型，最早可以追溯到一年半前。我们有很多人在做这件事。Anthropic 内部是一种非常重视原型和演示的文化，我们有很多内部原型从未公开。而 Cowork 最终的样子，就是从我们已有的众多原型中挑选出正确的部分组合而成。对吧？这也许是一个重要的限定条件，每当人们提到那个“10 天”的数字时，我觉得有必要说明我们并不是从零开始的。已经有很多东西在进行了，对吧？我认为人们需要记住，当你建一个网站时，你会用 React，会用很多其他东西。Cowork 也是类似的情况，我们已经有大量现成的组件。在决策路径方面，我认为我们生活在一个有趣的新世界，执行成本实际上非常低。嗯。

Yeah. I think at Anthropic we've been thinking about ways to move people who are comfortable with using Claude as a questions and bring more of the power of like this thing to now like execute tasks for you or I can like solve problems for you or I can like build things for you. How do we bring that capability to people who are currently mostly comfortable with like a like question answer paradigm within the chat. And we've had a lot of prototypes around that. This going back as far as like easily a year and a half. Like we had a lot of people working on that. Um and internally Anthropic is a very prototype demo-first culture. We have a lot of like internal prototypes that don't reach the public. And what Cowork actually became is like we sort of picked the right pieces out of the many prototypes that we had. Right? And that's that's maybe also like I think an important qualifier whenever people mention this like 10-day number. I do think it's important for me to mention that we didn't start with scratch. There was like a lot of stuff already happening, right? Like And I think it's important for people to remember that when you build a website, you use React, you use like a bunch of other things. And this is like a similar scenario with like a lot of pieces we already had. Um and in terms of decision paths I think we live in like an interesting new world where execution is actually quite cheap. Mhm.

所以也许……我听到的有点疯狂。那太野了。你可能会说，想法很廉价，执行才是难点。不。但我们过去可能生活在这样一个世界里：产品经理去找潜在客户，用非常低带宽的方式试图挖掘他们的问题和购买意愿，然后回来起草规格、思考、设计、执行。而在 Anthropic，我们现在可能更接近这样的状态：甚至不用写备忘录，直接构建，快速构建所有候选方案，然后选出最好的。

So maybe maybe what you I'm doing that so crazy to hear. That was wild. You should be ideas are cheap. Execution is the hard part. No. But like the we we used to live in this world maybe where you would take a product manager and the product manager would go to a number of potential customers and in this like very low bandwidth way would try to try to like tease out what are the problems they're having, what are they willing to buy. Um and then maybe what can you build to like address that need. And then you go back and you like draft a spec and you think about it and then like you make a design and you execute it. We in Anthropic have now probably much closer to the point where like don't even write a memo, just like build like let's build all the candidates very quickly. Let's Let's just build all of them and then pick the best ones.

我认为目前对产品和用户影响最大的决策，是我们如何重视你的本地计算机。这是一个很大的决策点。很多人想过，这个东西最终应该运行在你的电脑上还是云端？因为各有取舍，对吧？我想如果我们解决了认证问题，在云端做会很容易，但我认为我可以从任何地方下载任何文件，然后把它放到 Cowork 里，这是一个巨大的解锁。

I think the the decision that is most impactful both for the product as well for the users right now is like the way we put value on your local computer. I think that's a big decision point. A lot of people have thought about should this thing, whatever it is, should it ultimately run in your computer or should it run in the cloud? Because there are tradeoffs, right? I guess like if we solve auth, it would be easy to do in the cloud, but I think like the fact that I can just download any file from anywhere and then put it and co-work there is like a big unlock.

嗯，你提到重用某些组件很有意思。这也是我一直在思考的，即使是 Claude Code 也是如此，对吧？写代码的成本正在趋近于零，等等等等，但实际上，拥有某种平台基础的价值似乎在增加，因为当你构建这些新东西时，你可以把它们插在一起。是的。所以当人们说“很多软件的价值将归零，因为你可以重新创建它”时，我几乎觉得恰恰相反。拥有一个现有平台来构建，反而更有价值，因为你可以把东西往上加。是的。你显然有 MCP、技能、还有模型，对吧？这是很大一部分。所有这些都结合在一起。你觉得这是一种有效的思考方式吗？人们应该更多地投资于这些基础组件来重建，还是你每次都在重新创建很多东西，因为事情变化很快，重写比重用更容易？

Um I mean it's interesting you mentioned reusing certain pieces. I think this is something I've been thinking about even with Claude Code, right? The price of like writing code is going to zero, blah blah blah, but it actually seems like the value of having some sort of platform substrate is like increasing because as you build these new things, you can kind of plug them together. Yeah. So I almost feel like when people are saying, "Oh, the value of a lot of software is going to zero because you can recreate it." To me it's almost like the opposite. It's like having an existing platform to build on top of is like even more valuable because you can kind of bolt things on. Yeah. You have obviously MCPs, you have skills, you have like obviously the models, right? Which is a big part. All these things kind of come together. Do you feel like that's a valid way to think about it where people should invest even more in kind of like these primitives to rebuild on or are you like recreating a lot of it each time because like things change and it's easier to rewrite than reuse?

你知道，我认为你是对的。整体平台确实非常有用。这也许是我在 AI 领域持有的一种与很多人相反的观点。我实际上不认为未来会是超个性化软件，以至于每个人都运行自己的版本。我认为我们所有人都拥有自己的内部聊天墙会非常困难。如果我想和你聊天，那怎么实现呢？对吧？在 Colleague 的背景下以及我们如何构建它，我认为这是一种组合。执行成本降低的部分不一定是重建所有基础组件。我认为先验地，那也没有太多价值。例如，我的团队没有考虑重建 Colleague 代码。我们非常明确地从核心论点出发：这应该是 Colleague 代码。嗯。然后我们在它之上构建东西。执行成本降低的部分是：你如何把这些乐高积木组合起来，以对用户有意义的方式？那才是真正有价值的。

You know, I think I think you're right. I think you're right that the holistic platform is really useful. And this is maybe where I hold like a somewhat contrarian view to a lot of people in the AI. I actually don't think that the future is going to be hyper-personalized software down to the point where everyone is running their own version. Like I actually think it's going to be quite hard for all of us to have our own internal chat wall. And like, if I want to talk to you, like, how is that going to work, right? In the In the context of Colleague and how we build it, I think it's a bit of a combination. Like, what the The execution that gets cheap is not necessarily rebuilding all the primitives. I think a priori, there's also not a lot of value in it. So, for instance, my team did not think about rebuilding Colleague code. We like very much started with the with the core thesis of this should be Colleague code. Mhm. And then we like built things on top of it. The part of the execution that gets a little cheaper is like, how do you take all of these Lego pieces and put them together in a way that makes sense for users? That is like actually valuable.

你现在有很多不同的方法，关于哪些东西应该提升为基础组件？你强烈认为所有产品都应该通过组合我们所有人都可用的基础组件来构建，还是只有你们可用？保留一些内部的东西？

You have so many different approaches now in terms of what kind of What kind of things do you actually elevate to a primitive? Do you strongly believe that all your products should be built by just combining primitives that are available to all of us or is it available to you? Keep some things internal?

嗯，我认为这仍在演变中。但我觉得可能会消失的是——我不确定是否会完全消失——但我想说，就我个人而言，我可能不会再试图在没有与人测试的情况下想出一个非常好的产品。这不是一个新概念，但过去你常常需要做出昂贵的决策：选择技术 A 还是技术 B，或者用这种方式构建还是那种方式。我现在非常坚信，你只需把它们都构建出来，用一个小焦点小组测试，然后哪个更好就用哪个，对吧？这甚至可能与我们一年前的工作方式有很大不同，对吧？我认为这最近才发生。

Um and I think that's still evolving. But I think what's probably going to go away is like I'm not sure if it's going to fully go away, but I'm going to say I think for me personally I will probably no longer try to come up with a really good product without testing out with people. This is not a new concept, but wherever you used to have to make costly decisions around do we pick technology A or technology B or do we like um build it this way build it the other way. I really strongly believe now you just build all of them and try them out with that small focus group and then whatever whatever is better is what you go with, right? And that that is probably quite different even from how we maybe worked a year ago, right? Like, I think I think this happened very recently.

是的，既然你在这里，我碰巧开始用 Electron 构建一些东西。但 Electron 和 SQLite 在开发和构建之间有一些问题。然后我想，我干脆用 Swift 重写整个东西。结果就做完了，你知道吗？没花什么力气。我甚至不懂 Swift。但没错，反正我也不审查它，随便了。你可以用任何语言写。但我做的重要事情不是写 Electron 绑定。

Yeah, I started building something in on Electron since you're here, coincidence. Uh but then Electron and like SQLite are like there's like some issues that like between development and like uh building anyway. And I was like, I'll just rebuild the whole thing in Swift. And just recreated the whole thing in Swift. And it's like it's done, you know? It was that didn't take any effort. I I I don't even know Swift. But Yeah, exactly. I I was like I'm the I'm not reviewing it anyway, whatever. You can write it whatever language you pick. But the important stuff that I did was not write the electron bindings.

就像应用里发生的那种逻辑。然后模型说，对，我可以用 Swift 直接复现同样的东西。

It was like the logic of what happens in the app. You know, and then the model is like, yeah, I can just recreate the same thing in Swift.

是的，我认为你仍然需要——特别是对于做高性能软件或非常复杂软件的人来说——你仍然需要对架构有一些了解。但你可以用 Markdown 来实现。

Yeah, I think you still want, especially for people who are doing high-performance software or really complex software, you still want some view of the architecture. But you can use markdown for that.

对。是的。你实际上不需要读代码。我还是纠结于那个定义性的问题。我们能建立一个好的 Claude Code 心智模型吗？这是我目前的理解，对吧？就像你说的，它本质上是 Claude Code，我们不想动它。有 Claude 应用，有 Chrome 里的 Claude。我觉得你们在 Planning 上做了些不同的事。但我跟 Claude Code 团队的 Tariq 聊过，他说，“不，我们只是暴露了 Planning。”也许你可以澄清一下。人们应该了解 Claude Code 的哪些主要组成部分？

Right. Yeah. You don't actually have to read the code. Again, I'm still on that sort of definitional thing. Can we build a good mental model of Claude Code? This is what I have, right? Like you said, it's fundamentally Claude Code, we don't want to touch it. There's the Claude app, there's Claude in Chrome. I think you guys do something different in planning. But I've been talking with Tariq who's on the Claude Code team, and he's like, "No, we just exposed planning." Maybe you can clarify it. What are the major pieces that people should be aware of in Claude Code?

好的，我认为你基本上已经掌握了。所以，你可以把 Planning 大致拿掉。我认为 Claude Code 中有几样东西非常有价值。虚拟机可能是最强大的。我们目前运行一个轻量级虚拟机，并把 Claude Code 放在里面。我们这样做有很多原因。安全是一个重要因素。但即使你暂时忽略安全，只是说，“好吧，YOLO，我希望这东西能做任何事，”给 Claude 一台自己的电脑也是相当有用的。这通常是个好主意。从架构、用户体验以及我们一直在做的其他方面来看，如果你把 Claude 拟人化，直接把它当成一个人，往往非常有用。如果你给一个人一台电脑，他会怎么做？我今天早上给我爸打了个比方，他仍然坚持用聊天来做编码的事情：“如果你是一个开发者，你的雇主告诉你你不需要电脑，他们只会通过邮件把代码发给你，你再把代码通过邮件发回去。这对后台的 Pedro 可能行得通，但效率不高。”所以，通过虚拟机，因为它是 Linux 系统，Claude Code 可以自由安装它需要的任何东西。它可以安装 Python，可以安装 Node.js。我们有严格的网络入口和出口控制，所以你可以用自然语言让整个系统更清楚你允许什么、不允许什么。但我们从来不需要去问一个真人，比如市场人员或律师。我不需要去找律师问：“你同意我安装 Homebrew 吗？”因为问题和答案的含义复杂而微妙，不容易推理。这给了我们很多抽象能力，让 Claude 变得非常强大。

Okay, I think you basically have them. So, you can take planning more or less out. I think there are a few things that are really valuable in Claude Code. The virtual machine is probably the most powerful thing. So, we currently run a lightweight VM, and we put Claude Code into the VM. And we do that for a number of reasons. Safety and security is a big one. But even if you ignore safety and security for a second, and you're just like, "Okay, YOLO, I want this thing to do whatever," it is quite useful to give Claude its own computer. That is generally a good idea. In terms of architecture and UX and everything else we've been working on, it often is quite useful if you anthropomorphize Claude aggressively and just be like, "This is a person." What would you do if you gave a person a computer? The analogy I gave my dad this morning, who is still quite insistent on using chat even for coding things, is: "If you were a developer, and your employer told you that you don't need a computer, they're just going to send you emails with the code and you send emails with code back. That might work for Pedro on the back, but it is not very effective." So what we can do with the VM is, because it's a Linux system, Claude Code has more or less free range to install whatever it needs to install. It can install Python, it can install Node.js. We do have strict network ingress and egress controls, so you can still, as a user in plain human language, make it clearer to the entire system what you're okay with and what you're not okay with. But at no point do we have to ask a real person, like a person who might be in marketing or a lawyer. I don't have to go to a lawyer and be like, "Are you okay with me installing Homebrew?" Because the implications of the question and the answer are complex and nuanced and not easy to reason about. This gives us a lot of abstraction that makes Claude very powerful.

围绕它，我们可能还有很多东西，几乎每周都在增长，你可能也注意到了，这让 Claude 在某些任务上比单独的 Claude Code 更好。但其中大部分实际上存在于系统提示中。它们关乎我们能从你的工作中推断出什么？我们能引入什么到系统提示中让它们更有效？当然还有与 Claude 和 Chrome 的紧密集成。你注意到很多人，尤其是随着模型越来越好，很多人在 MCP 连接器方面束手无策。我不会去搞 25 个 MCP 连接器，到处点击，然后一半还什么都做不了。所以 Claude 和 Chrome 非常强大，因为我们只需与 Claude 和 Chrome 子智能体对话，它们就会为你做事。

Now around it, we do probably have a number of things that also keep growing almost every single week that you're probably noticing, that makes Claude maybe better for certain tasks than just Claude Code on its own. But most of those actually live in the system prompt. They're about what can we infer about the work that you do? What can we introduce into the system prompt to make them more effective? It's of course the very tight integration with Claude and Chrome. You're noticing that a lot of people, especially as the models get better, a lot of people throw up their hands when it comes to MCP connectors in this area. I'm not going to go through 25 MCP connectors, click off everywhere, and then half of them don't let me do anything anyway. So Claude and Chrome is quite powerful because we can just talk to the Claude and Chrome sub-agent, and they will just do things for you.

举个例子，对吧？在 MCP 方面，老实说我觉得 MCP 的状态真的很难集成。我需要把我用的编码智能体加上 Figma MCP。但我不想读文档，所以我就让 Claude 去做了。它很擅长读文档。同样，我不得不为我做的某个项目设置一个 Google Cloud 账户，并在某处获取一些 API 密钥。而 Google Cloud 是出了名的难导航。所以我就是不想处理这些。所以我就用了 Claude Code。

So one example, right? In MCP, I honestly think the state of MCP is kind of really hard to integrate. I needed to add Figma MCP to the coding agent that I use. But I didn't want to read the docs, so I just had Claude do it. And it's great at reading docs. In the same way, I had to set up a Google Cloud account for some project I was working on and get some API keys somewhere. And Google Cloud is famously super hard to navigate. So I just didn't want to deal with any of it. So I just used Claude Code.

在 Claude Code 上开发的第一周内，这件事发生得非常快。我发现自己开始用 Claude Code 来做编码任务，这表面上并不是我们构建它的目的，对吧？我们不需要。但我发现自己在我们的内部工具上——那个用来收集崩溃和调试信息的工具。我发现自己挑出那些我认为容易修复的 bug，而不是那些可能是内核损坏或操作系统其他问题的 bug。然后我告诉 Claude，“去修复这个 bug。”我心想，“我在这儿干嘛？”再往上走一层。告诉 Claude Code，“我希望你去所有这些崩溃工具。我希望你找到所有你认为可修复的 bug，而不是操作系统崩溃。然后我希望你告诉另一个 Claude 去修复所有这些。”这就是我在做的事情。

Within the first week of developing on Claude Code, this happened very quickly. I caught myself starting to use Claude Code for coding tasks, which is not ostensibly what we built it for, right? We don't need to. But I found myself on our internal tool that we have to collect crashes and debugging information. And I found myself sort of picking out the ones that I think we can easily fix versus the ones that might be kernel corruption or something else in the operating system. And I found myself sort of picking these out and then just telling Claude, "Go fix this bug." I was like, "What am I doing here?" Go one level up. Tell Claude Code, "I want you to go to all these crash tools. I want you to find all the bugs that you think are fixable and not like an operating system crash. And then I want you to tell another Claude to fix all of that." And that's sort of what I'm doing there.

另一个 Claude？

Another Claude?

是的。所以它可以启动另一个实例。目前我做的，有点 hack，但我告诉它用 Claude Code remote 来调用自己。

Yeah. So it can spin up another instance. Currently, what I do is, and this is a bit of a hack, but I'll tell it to use Claude Code remote to call itself.

这很有趣。所以你基本上——如果你想象一个有 20 个 bug 的仪表盘。这是远程控制还是 Claude Code remote？

That's interesting. So you basically take—if you imagine a dashboard with 20 bugs. Is this remote control or Claude Code remote?

哦，抱歉。我只是想确认一下——我使用的方式是，我运行着 Claude Code，然后告诉它：“这是我每天早上通常去找最新 bug 的地方。去读整个 bug 列表。区分哪些是可修复的，哪些是不可修复的。然后对于可修复的，在这个几乎循环中，为每个 bug 写一个带提示的 Markdown 文件。然后对于每个作为提示的 Markdown 文件，启动一个 Claude Code。”所以 Claude Code 原生就有子智能体的概念。这基本上就是一个子智能体，但我没有使用子智能体的功能。我没有用子智能体的功能，原因是我把它作为一个 Claude Code remote 任务来触发。这挺好的，因为我可以直接触发它，然后去开下一个会，而在 Claude Code remote 中，工作已经在进行了。

Oh, well, sorry. I just wanted to confirm what—the way I'm using it is I have Claude Code running and I'm telling Claude Code, "Here's where I normally go every morning to find the latest bugs. Go read the entire bug list. Separate out which ones are fixable and which ones are not fixable. And then for the fixable ones, for this almost loop, for each bug, write a markdown file with a prompt. And then for each markdown file that is a prompt, start up a Claude Code." So natively Claude Code has this concept of sub-agents. And this is basically a sub-agent, but you're not using the sub-agent's functionality. I'm not using the sub-agent's functionality, and the reason I'm not is because I'm firing that off as a Claude Code remote task. It's kind of nice because then I can just fire it off. I can go to my next meeting, and in Claude Code remote, the work is now happening.

是啊，你看你已经开始用云而不是本地机器了，我觉得这就像那种“难道不应该是云优先吗？”的问题，对吧？

Yeah, you see like you're already starting to use the cloud over your local machine and I think this is one of those things where like well, shouldn't just everything just be cloud first, right?

这个问题太好了……我对此有很多想法。很多想法。好吧，我大体上认为硅谷整体低估了本地计算机的价值，我默认的论据总是：为什么你们都用 MacBook，而不是 iPad 或 Chromebook？本地机器仍然有价值。现在当我想到云时，它是一个应该对你非常有用的实体。非常有用。我认为那个实体需要能访问你拥有的所有工具。否则，它会在各种复杂的方式上受到限制。我们大致有两种方法。我们可以说，好吧，我们要一个接一个地把你电脑上的所有东西都搬到云端。这是一种方法。我认为其他产品已经走了那条路。我个人——这是一个非常个人的观点——但就我使用的工具数量而言，我没有耐心给另一个工具授予每一项权限并保持这些权限更新。我仍在纠结的第二件事——我还没有一个好的答案可以告诉任何人——是，如果某人把你整个工作都吸到云端，那会是什么样子？比如，举个例子，如果你点击一个按钮，它就把你整个电脑克隆到云端，你会想要那样吗？我还不完全确信每个人都会想要。这有点像我们将要遇到的所有技术问题的上游，因为大体上，我认为世界还没有为这类事情做好准备。我给你举一个对我们来说可能很容易的例子。作为一个桌面应用，理论上，在你的许可下，我们可以在你的电脑上做很多事情，包括读取你的 Chrome 浏览器 cookies，如果你真的想的话，对吧？我们可以拿走你的 Chrome 浏览器 cookies，你不需要为我们解密它们，但如果我们真的想，我们可以把它们放到云端。一个相当简单的解决方案，会非常酷，因为就像“哦，我们现在可以在云端做所有任务了。”很多网站，包括银行，如果它们看到来自两个不同位置的相同认证，就会直接锁定你的账户。然后你就得去分行，说“好吧，我带着护照来了。”你知道，尽管我们都对“智能体”这个词感到厌倦，但为了智能体式的未来，我认为有很多东西需要慢慢赶上。在那之前，作为在 Claude 上工作的人，我能让 Claude 最有效的方法就是把它放在你工作的地方。

This is such a good... I have like so many thoughts about this. I have so many thoughts about this. Okay, so I generally believe that Silicon Valley overall is undervaluing the local computer and my default argument for that is always how come you're all using MacBooks and not like an iPad or a Chromebook? There's still value in having a local machine. And now when I think about cloud, it's this entity that is supposed to be very useful to you. Like a tremendously useful to you. I think that entity needs to have access to all the same tools you have access to. Otherwise, it's going to be hamstrung in like all these complex ways. And there's sort of two approaches we could take. We could say, okay, we're going to like one by one chip away at everything that is at your computer and move it into the cloud. That's one way to do it. And I think other products have taken that path. I personally, this is a very personal opinion, but I personally for the amount of tools that I use just don't have the patience to give another tool like permissions to every single thing and keep those permissions up to date. The second thing that I'm still grappling with and I don't have a good answer for anyone to say yet, but the second thing I'm still grappling with is what does it look like for someone to slurp up your entire work and put that in the cloud? Like if I just as an example like if you click a button and it just clone your entire computer into the cloud, is that something that you would want? I'm not totally convinced yet that at all everyone will. And that is sort of like upstream of all the technical issues we're going to have because like in general, I think the world is not ready for this kind of stuff. Like I'll give you one quick example that would probably be very easy for us. So as a desktop app, we in theory with your permission can do a lot of things on your computer, including reading your Chrome cookies, if you really want to do it, right? We could take your Chrome cookies, you wouldn't have to decrypt them for us, but we could put those on the cloud if we really felt like it. Pretty easy solution that would be super cool because it'd be like, "Oh, we can do all the tasks in the cloud now." A lot of websites, banks included, if they see the same authentication from like two different locations, will just lock down your account. And now you have to go to the branch and be like, "Okay, I'm here with my passport." And you know, as tired as we all are of the term agent for the agentic future, I think there's a lot of stuff that's sort of slowly needs to catch up. And until that's the case, the way I, as someone who's working on Claude, can make Claude most effective is to like put it where you're working.

还有什么关于我们的心智模型的想法吗？所以，基本上，我也有点觉得，我越了解它的工作原理，就越能充分发挥它的潜力，对吧？

Anything else I thought with our mental model? So, like basically like part of me also just want like the more I understand how it works, the more I can use it to its full potential, right?

是的。

Yeah.

所以，我从你那里听到的是，你让我删掉那个规划功能。你没有做任何只属于 Claude Co-work 的特别的事情。我们有一些技巧，但这有点像一周的变化怪胎。我们评估 Claude Co-work 可能针对不同的用例，而不是你评估 Claude Code 的方式，对吧？你怎么看？

And so, what I'm getting hearing from you is you told me to delete the planning thing. You're not doing anything special on the that's only exclusive to Claude Co-work. We have some tricks, but this is sort of like change freak of a week. We eval Claude Co-work maybe against different use cases than you would eval Claude code, right? How do you think about it this way?

好的。所以，Claude Code 理想上更像 Claude Co-work，是的。Claude Code 针对编码任务进行了相当优化，我们主要根据它在典型套件工作中的表现来评估我们是变好还是变差。而 Claude Co-work 则更多针对典型的知识工作，比如你在金融或法律办公室中会遇到的那种工作。我个人的用例总是像管理我的个人抵押贷款之类的事情，对吧？或者为我和我的家人做财富规划。这些是我们评估 Claude Co-work 的用例。你可能注意到的是我们对系统提示所做的细微调整，我们在系统提示中放入了什么，我们如何通过给予的工具来引导 Claude。所以，要么它在某个方向上更好，要么存在权衡，权衡很多。Claude Code 在代码方面更好，而 Claude Co-work 在非编码任务方面更好。这些差距在接下来的几代模型中还会存在吗？对我来说还有点不清楚。是的，因为现在我们做的这些超优化，我不确定它们还能相关多久。

Okay. So, like Claude code is ideally more like Claude Co-work, yeah. So, Claude code is like quite optimized for coding tasks, and we mostly evaluate whether or not we're getting better or worse depending on how good it is at like a typical suite job. And Claude Co-work on the other hand, we evaluate more against typical knowledge work, the kind of stuff you would find in finance or in like maybe a like um like a legal office. My personal use case is always like managing my things like managing my personal mortgage or something like that, right? Or like wealth planning for me and my family. Those are the kinds of use cases we eval Claude Co-work on. And what you might be picking up on is like the subtle changes we make to the system prompt, what we put in the system prompt, how we steer Claude with the tools we give it. So, like either it be better in one of the other direction and whether there's a trade-off, trade-offs exist a lot. Claude code will be better for code, and Claude code work will be better for non-coding tasks. Will those gaps still exist in the next few generations of models? It's like a little unclear to me, though. Yeah, because right now these like hyper-optimizations we make, I'm not sure for how long they're still going to be relevant.

是的，因为现在我们做的这些超优化，我不确定它们还能相关多久。我想我指的是，它在质量上感觉不同，可能只是提示的问题，我想多了，但事实上它出来的是一个九步计划，我可以编辑计划，获得反馈，然后看到它执行计划。是的，感觉比 Claude Code 更有长远规划，但也许 Claude Code 里已经有这个功能了，你只是为它建了一个更好的界面。

Yeah, because right now these like hyper-optimizations we make, I'm not sure for how long they're still going to be relevant. I think what I was referring to was also it just qualitatively felt different when I probably it's just all prompting and I'm reading too much into it, but like the fact that you it comes out as like a nine-step plan, I can edit the plan, and get feedback, and and and see it execute the plan. Yeah, it felt more long-range than in Claude code, but maybe that already existed in Claude code, and you just built a nicer UI for it.

两者都有。如果构建规划功能的 Claude Code 团队的人说，他们可能会说“是的，我们在 Claude Code 里有这样的功能。”他们确实有。我认为人们倾向于给 Claude Co-work 分配时间跨度更长的任务。我觉得它很长。是的。这是一点，对吧？工作块可能更大一点。第二点是，因为工作变长时会变得更模糊，我们确实告诉 Claude Co-work 要大量使用规划工具，或者大量使用询问用户工具，对吧？我们确实希望它提出不同的场景，梳理出用户真正想要什么。不要工作四个小时，然后回来给出错误的东西。你可能注意到了这一点。是的。我希望我能告诉你我构建了这个神奇的东西，有一些秘密配方。但事实并非如此，我只是说清晰是好的。你知道，工程师们只想知道他们可以围绕它进行规划。

It's kind of both. Like if the Claude code people who built the planning functionalities would say that they would probably say, "Yes, we have one of those things in Claude code." And they do. I think people tend to give Claude work tasks that are maybe of longer time horizon. I thought it was so long. Yeah. That's like one thing, right? You're just like that the chunk of work tends to be maybe a little bigger. And then the second thing is that because the work when it gets longer it gets a little bit more ambiguous, we do tell Claude work to make heavy use of the planning tool, or to make heavy use of the ask user question tool, right? We do want it to come up with like different scenarios of okay, tease out what the user actually wants. Don't go off to work for like 4 hours, and then come back with the wrong thing. And you're probably picking up on that. Yeah. I wish I could tell you I like built this magical thing, and it's like there's some secret sauce. I'm like No, I mean that it's just clarity is good. You know, engineers just want to know that they can they can plan around it.

然后，对我来说，我记得我不得不切换到我的另一台机器，因为这是一台新机器，没有我的会话，但是，是的，规划对我来说真的很重要，我需要批准它，或者看看它是否正确。询问用户问题的呈现非常漂亮。我的意思是，它在 Cursor 和 Claude Code 中也有，但我认为看到它仍然很好，它让我理解它理解我，它理解我想做什么。

And then I I think also for me um I remember I think I have to switch to my my other machine because this is a new machine it doesn't have my session, but uh yeah, the the the planning is really important for for me to like approve, or like to see whether it's like it's right. The ask user question is so beautifully presented. I mean it it's also available in like cursor and and in Claude code, but like I I think like it's still nice to see that it like it's kind of for me like to understand that it gets me, it gets what I want to do.

是的。是的，它珍视我们的艺术。

Yeah. Yeah, it prizes our art.

就评估这个话题来说，当你说“评估”时，我觉得人们对它的含义很模糊。它只是像“感觉测试”那样，还是你们有像 Claude Work 的自动化程序化评估？

Just on the topic of evals, when you say eval, I think people are very vague about what it means. Is it just like vibe testing or do you have like automated programmatic evals of Claude work?

当我们说“评估”时，真正的意思是，我们基本上会拿整个转录文本，包括 Claude 最终可用的所有工具，然后根据我们调整的内容来衡量输出是什么。所以我们确实经常运行评估。我们在训练中使用它。我们在后训练中也用它，如果你把后训练和它周围的脚手架分开的话。Claude Work 某种程度上存在于脚手架空间，但显然我们也在它上面做了一点训练。所以，当我们说“评估”时，意思是给定某个转录文本，输出是什么样的，包括文件输出以及你在聊天窗口看到的实际 token 输出。

When we say eval, what we really mean is that we essentially take the entire transcript, including all the tools that Claude has available ultimately to it, and we then measure what are the outputs depending on what we tweak. So, we do run that a lot. We use that in training. We use that in post training, if you separate out post training from the scaffolding around it. Claude Work sort of exists in the scaffolding space, but obviously we also train on it a little bit. So, when we say eval, we mean given a certain transcript, what do the outputs look like, including the file outputs as well as the actual token outputs like the ones that you see in the chat window.

我很好奇，失败模式中有多少是模型智能的问题，又有多少是使用最终工具来注入智能的问题？比如财富规划就是一个很好的例子，对吧？制定计划是一回事，制作一个漂亮的电子表格来引导你完成计划是另一回事。你看到这种情况是如何演变的？

I'm curious how much of the failure modes are the model intelligence versus the usage of the end tool to put the intelligence in? Like the wealth planning is a good example, right? It's one thing to come up with a plan, the other thing is to make a nice spreadsheet that runs you through the plan. How are you seeing that evolve?

我经常纠结的是，无论你设计出什么样的脚手架，我认为我们仍然存在一些模型过剩的情况，即模型的能力远远超过用户目前的使用方式。我认为部分原因是我们没有给模型提供所有工具来完成它在理论上能够完成的所有事情。这是一方面。然而，每当你搭建脚手架时，我就在想，这个脚手架什么时候会消失，而你投入多少精力去弄清楚正确的脚手架，这有点像一场赌博。作为一名工程师，我很享受的一点是，在 Tropic 和前沿实验室工作，我可能对即将到来的东西有更多了解，比如下一个模型是什么，模型能做什么，它擅长什么，不擅长什么。我越来越想知道：对我们来说，正确的做法是真正投入太多精力在这些脚手架修正上（模型可能不会行为不当，只是不做你想做的事），还是尽可能多地赋予它能力，努力让这些能力安全，这样最坏的情况就不会那么糟糕，然后只需等待下一个模型发布。我个人目前更倾向于后者。我认为我们会看到很多应用和公司用 AI 做非常令人印象深刻的事情，短期内可能看起来非常有效，因为它们非常针对个别用例，但我认为一旦模型在泛化方面变得更好，并且在那些特定用例上无需过多引导就能做得更好，我不确定这种局面能持续多久。

The thing that I grapple with a lot is that whatever scaffolding you come up with, I think we still have a bit of model overhang where the model is dramatically more capable than users have been using it for. I think part of that is that we're just not giving the model all the tools to do all the things it's theoretically capable of. That's one thing. However, whenever you do put the scaffolding, I'm sort of wondering at what point will that scaffolding go away, and how much you invest in figuring out what the right scaffolding is is a bit of a bet. And one thing that I as an engineer quite enjoy is that working at Tropic and working at a frontier lab, I maybe have a little bit more insight into what's coming down the chute in terms of what's the next model, what the model is capable of, what it's good at, what it's bad at. And I'm increasingly wondering: is the right thing for us to really invest too much in these scaffolding corrections where the model might otherwise not misbehave but just not do the thing that you want? Or is it to just give it as many capabilities as possible, try to make those safe so that the worst case scenario is not as bad as it might be otherwise, and then just simply wait a second for the next model drop. I'm personally currently more leaning into the latter. I think we're going to see a lot of applications and companies that do very impressive things with AI that in the short term might seem very effective because they're very specialized to individual use cases, but I think once models get better at generalization and get better at those specific use cases without being super guided on those, I'm not sure how long that's going to stick around.

你其实已经可以在 skills 和 N2P 服务中看到这一点了。我们已经看到了从 N2P 服务到 skills 的缓慢转变。一个很好的例子是 Barry，他创造了 skills。他最初是在捣鼓一个看起来很像今天 Coda 的东西。他在想，如果 Coda 是为那些不想写代码的人设计的会怎样？他在桌面应用里做了一个原型。我们想到的第一个用例是那些能从图形界面和与底层代码稍微分离中真正受益的编码用例。每个人都得出了相同的答案：数据分析。首先，我们今天有多少用户？有多少？总是数据分析。我认为最终促成 skills 的原因是，我们想把这个小原型连接到我们的数据仓库。团队很快发现，与其为数据仓库构建一个自定义工具，他们只是创建了一个 markdown 文件，上面写着“亲爱的 Claude，如果你想获取数据，这是端点，这是 API 的样子，你自己搞定。”然后他们就把控制权交出去了。另外，只需在抽象层上再提升一级。不要告诉它“这是 CLI，请调用 CLI”或“这是 MCP，请调用这个接口形状”，只需说“这是端点。如果你想知道什么，如果你在这里发帖，也许你可以做 post sequel。没问题的。”结果这非常有效，以至于他们开始尝试同样的模式：只给模型一个描述需要做什么的 markdown 文件。整个东西最终变成了 skills，我们说“我们应该把它打包起来。这真是一个好主意。”

You can sort of already see this in skills and N2P servers. We've already seen this slow shift from N2P service to skills. A good example is Barry, who made skills. He was initially hacking on something that looked a lot like what Coda does today. He was thinking about, what if Coda but for people who don't want to build code? He did that as a prototype inside the desktop app. One of the first use cases we thought of were coding use cases that could really benefit from graphical interfaces and from being a little separated from the actual underlying code. And everyone comes to the same answer: it's data analysis. First, how many users do we have today? How many? It's always data analysis. And I think the thing that ultimately led to skills is that we wanted to connect this little prototype to our data warehouse. And the team very quickly discovered that instead of building a custom tool for the thing to our data warehouse, they just made a markdown file where we're like 'Dear Claude, if you want to get data, here's the endpoint, here's what the API looks like, you figure it out.' And then they ended up handing over control. Also, just go one step up in the layer of abstractions. Instead of telling the thing, 'Here's the CLI, please call the CLI.' or 'Here's an MCP, please call this interface shape.' just say 'This is the endpoint. If you want to know something, if you post here, maybe you can do post sequel. It's going to be okay.' And that ended up being so effective that they started trying the same pattern of just giving the model a markdown file that describes whatever needs to be done. That whole thing eventually became skills and we were like 'We should package this up. This is a really good idea.'

是的，我们在会议上请来了 Barry 和 Mahesh，他确实有个好主意。我想给你看看我是怎么用 Claude Code Work 的。这是我最喜欢的部分。这就是我们运营 Discord 的方式。起初，我并不信任 Claude Code。这是我第一次使用它。然后我就想，“好吧，我就试试手动从 Zoom 下载所有录音并上传到 YouTube，因为这是一个非常费力的过程。我得点来点去。YouTube 并不是超级用户友好。”然后它就这么做了。然后我想，“实际上，你知道，甚至从 Zoom 下载的部分，我也应该放进 Claude Code Work 里。”然后我就这么做了。这里有一堆……它开始在这里压缩，甚至开始能够做像查看视频的单个帧来命名视频这样的事情，这样我就可以自动上传了。所有这些都取代了我作为 YouTuber 的工作。我们将永远感谢你的……是的。所以这很棒。但通过压缩，它创造了一个新东西，对吧？所以我没有最初的那个东西了。但后来我让它创建自己的 skill，这样那些重复性的、一次性的、需要人工引导的事情就变得更自动化了，我可以独立使用这些 skill 并重复利用它们。显然还可以编写 skill。这进入了上下文和底部的 skills 区域，这太棒了。所以我有了所有这些 skill，现在每周都会用。我知道你们发布了定时协作工作，我还没试过，但你真的没有理由不去试试它们。

Yeah, we've had Barry and Mahesh on our conference and he's definitely got a good idea there. I wanted to show you how I've been using Claude Code Work. This is my favorite part. This is how we run the Discord. At first, I didn't trust Claude Code. This is my very first usage. So then I was like, 'Okay, I will just try to manually download from Zoom all my recordings and upload it to YouTube because this is a very laborious process. I got to click click click. YouTube isn't super user-friendly.' And it just did it. And then I was like, 'Actually, you know, even the download from Zoom part, I should also put into Claude Code Work.' and then I did it. Here's a bunch of... And it starts compacting here and it even starts being able to do things like look through the individual frames of the video to name the video so that I can upload it automatically. All that replaces my job as a YouTuber. We will forever appreciate your great... Yes. And so that's great. But then by the compacts, it makes a new thing, right? So I don't have the initial thing. But then I asked it to make its own skill so that something that's repetitive and one-off and human-guided becomes more automated and I can use the skills independently and reuse them. And obviously can write skills. And that goes into context and skills at the bottom here, which is so nice. So I have all these skills that I now sort of do on a weekly basis. I know you've released scheduled co-works, which I haven't done yet, but there's really no reason why you shouldn't try them.

我觉得这太棒了，看到这个我很开心。因为技能特别容易制作，任何人都能做一个，甚至一条短信都可以是技能。而且它们可以高度个性化，这就像是一个抽象层。我猜你工作很出色，你可能已经给了它一些指导。我让它把所有东西打包成一个技能，但后来我想，有时我需要拆分开，因为某些部分会失败或者需要单独使用。所以我让它把一个技能拆成三个，就像技能拆分器一样。然后还有一个父技能来协调它们，如果我想用的话。我觉得这非常好。

I think this is wonderful and fun for me to see because one thing that is very fun about skills is that they're so easy to make. Anyone can make a skill. A text message could be a skill. And they can be hyper-personalized to you. This is an abstraction layer. I assume you're very good at your job. You've probably given this thing some guidance about how to do it. I just said wrap everything up into a skill, and then I thought, sometimes I might need to break things apart because some parts fail or some parts might be needed individually. So I told it to split one skill into three skills. It's like a skill-splitting thing. And then there's a parent skill that orchestrates all of them if I want to use that. I think that's really good.

啊，这太美了。是的，太棒了。

Ah, that is beautiful. Yeah, that's wonderful.

还有一部分：我跟你提过的 Google Chrome 的事情。我想，什么比用 Claude Cowork 上传到 YouTube 更好呢？实际上是查看文档，用编程方式上传到 YouTube，然后把它做成一个技能。我以前从没做过，我不想碰 Google Cloud，所以 Claude Cowork 帮我做了。这真的很酷。我就让它自己干，无所谓。然后我把技能和它构建时用的脚本配对？是的，然后我就更新技能。

There's one more part: the Google Chrome thing I told you about. I thought, what's better than uploading using Claude Cowork to YouTube? Actually looking at the docs to programmatically upload to YouTube, and then putting that in a skill. I've never done that before. I don't want to deal with Google Cloud, so Claude Cowork does it for me. That is really cool. I just let it do its thing. It doesn't really matter. And then I paired the skill with the same script it's built on? Yeah, and then I just update the skill.

啊，这太美了。是的，太棒了。这有点像技能。基本上，我认为人们接触 Claude Cowork 的方式是，拿一个你平时需要点来点去的知识工作，然后试着把它转化。然后你会想，好吧，如果更进一步呢？随着你越来越信任它，你逐渐扩大 Co-work 的范围，同时也教它如何取代你。

Ah, that is beautiful. Yeah, that's wonderful. It's kind of like a skill. Basically, I think the way that people ease into Claude Cowork is to take a knowledge work task that you would normally be clicking around for, and then try to turn that. And then you do, okay, what if you went further? And then you sort of expand the scope of Co-work as you gain trust with it and also teach it how to replace you.

是的，这有点像在玩《异星工厂》，但对象是你自己的生活。你从很小的事情开始，自动化一些非常小的东西，一旦它奏效了，你就不断往这个自动化帝国上加东西，让你的生活越来越轻松。我最喜欢的技能是，每天早上 Co-work 开始查看我的日历，确保没有冲突，因为人们经常临时安排会议，有时还会错过。这很烦人。很多产品都有类似功能。我在自定义提示里写了，但我还没把它做成技能。说实话，我应该做。但我给了它很明确的指示：比如，如果某些人预约了会议，覆盖了其他会议，我可能会去参加他们的会议。比如如果 Dario 安排了会议，就不要试图重新安排 Dario。还有一些其他规则，关于我更在意哪些会议、不太在意哪些、哪些可以推迟、我想什么时候工作、不想什么时候工作。正是这些小事让人产生共鸣。我们推出 Co-work 时，在 Twitter/X 上最火的用户故事之一是“清理你的桌面”，这很傻。你根本不需要一个模型来清理桌面。

Yeah, it's like playing Factorio but for your own life. You start really small. You start automating something really tiny and once it clicks, you keep adding onto this automation empire, just making your life easier and easier. My favorite skill has been every morning, Co-work starts looking at my calendar and makes sure there are no conflicts because people tend to schedule meetings last minute and sometimes miss it. It's often painful. A lot of products have existed like that. I've written in the custom prompt there. I haven't made it a skill. Honestly, I should. But I've given it pretty clear instructions: okay, here are some people, if they book over other meetings, I'm probably going to go to their meeting. Like if Dario schedules a meeting, don't try to reschedule Dario out of it. And there are some other rules about what kind of meetings I care more about, what kind I care less about, what is okay to punt, when I want to be working, when I don't. It's those really small things that click with people. When we launched Co-work, one of the user stories that went most viral on Twitter/X was clean up your desktop, which is silly. You don't need a model to clean up your desktop.

像这样？清理我的桌面？

Like this? Clean up my desktop?

是的，没错。我想我得选择我的桌面，给它访问桌面的权限。好吧，这很吓人。我们来做。我试过下载文件夹，它说：‘你有这么多条款清单，还有八份办公室租约的副本。’我说：‘好吧，别吼我。’但这是件小事。我通常不会告诉别人：‘我建了一个能帮你整理文件夹的产品’，因为感觉太小了。但正如你所说，问题就在这里。

Yeah, exactly. I need to choose my desktop, I guess. Give it access to my desktop. Okay, this is very scary. We'll do it. I did it with my downloads folder. It was like, 'You have so many term sheets and there are like eight copies of your rental lease for your office.' I was like, 'All right, don't yell at me.' But it's such a small task. I would never normally tell people, 'I've built a product that can organize your folder for you,' because it feels small. But to your point, here's the question.

很美，对吧？它会指向明显的垃圾。你可能不该点那个。不。如果没做好，可逆性很好。

Beautiful, right? It leads to obvious junk. You probably shouldn't click that. No. If it's not done right, it's nice that it's reversible.

我有一个典型的、什么都乱糟糟的文件夹。所以是的，这非常有帮助。这是一个很简单的任务。但我看到了进展。这就是为什么我觉得这肯定和 Claude Code 不同，因为我们会做系统提示，让你思考这个任务。然后我可以对这些事情做一点小建议。太美了。我可以说：‘哦，别那样做。别这样。’太棒了。我很高兴你喜欢它。反过来，我们是 Claude Code 团队的一部分。如果你希望在 Claude Code 里也有这个……

I have a typical everything-is-super-messy folder. So yes, this is super helpful. This is a pretty simple task. But I see the progress. This is why I think this must be something different than Claude Code, because I'm like, we do system prompt that we want you to think about this task. And then I can do little suggestions for these things. It's beautiful. I can say, 'Oh, don't do that. Don't do this.' It's amazing. I'm so happy you like it. The other way around, we're part of the Claude Code team. If you would like that in Claude Code...

靠。所以是的，这真的很好。我显然在极力夸它。我知道，我还有其他事情，比如注册 PG&E。所以如果你能帮我打电话，那就太好了。

Damn. So yeah, this is really good. I'm obviously raving about it. I know, I have other things like sign up for PG&E. So if you can do phone calls for me, that'd be great.

有人做过。显然你不能原生地做，但人们通过各种其他提供商做过。然后这就像注册 Figma MCP。我真的想什么都做。数据分析也是。我确实认为设计转代码非常好。所以这是 Figma 文件，拿去，这就是很多其他知识工作取代我手动点击的地方。但通常我会用 Claude Code 来做这个。但因为我觉得你有更好的 Chrome 集成，我认为你实际上能做得更好。这是我会议网站的一次尝试。很酷。在某个时候，我很想听听你对桌面应用中的代码的看法，我从来不用。

People have done that. Obviously, you can't do that natively, but people have done that with various other providers. And then this is like signing up for the Figma MCP. I really am trying to do everything. Data analysis as well. I do think design to code is very good. So here's the Figma file, take it, and this is where a lot of other tasks like knowledge work replace my manual clicking. But this is normally I would use Claude Code for this. But because I perceive that you have better Chrome integration, I think you can actually do a better job with this. This is one shot at my conference website. That's pretty cool. At some point, I would love to hear how you feel about code in the desktop app, which I never use.

那是同一个团队。同一个团队。所以我用 Claude Code，我认为这是 Claude Coding 的默认方式。这个东西有一个内置浏览器，很多产品都有。

Which is the same team. Same team. So I use Claude Code in which I perceive to be the default way of Claude Coding. One thing this has... Sorry, I'm not here to wrap all the products up. I can talk about other stuff. I'm not sure if people out there want to hear me advertise my stuff for an hour. Please do that. This thing has a built-in browser, which a lot of products have.

你在这里看到的是一个内置浏览器，我认为让 Claude 能够看到你实际在做什么，会使其高效得多。这大概就是你在 Claude Work 中看到的，因为它能查看 Chrome，能调试 DOM，能看到东西。这确实让它更强大。

You see it here, it's a built-in browser and I think giving Claude eyes into like what you're actually working on makes it so much more effective. And this is probably what you've seen in Claude Work because it can see Chrome, it can like debug the DOM, it can like see things. That does make it more powerful.

是的，所以我的思维模型有点混乱，因为我只用了 Claude Work，以为它里面有个浏览器，但后来我了解到 Claude Code 应用或应用版 Claude Code 确实有一个内置浏览器。我见过这个预览功能，只是从没用过。但最终你得到的东西差不多，对吧？你描述的那个额外技能就是，如果 Claude 能看到它在做什么，它会更好。这就是总结。无论是用你的 Chrome 还是它自己搞一个小浏览器，其实没太大区别，因为不管怎样它都能看到自己在做什么，这就会好很多。然后你就不用为你的 Claude 跑 QA 了。

Yeah, so I think my mental model is kind of broken because I only use this Claude Work because I thought they had a browser thing in it, but I understand that the Claude Code app or the app version of Claude Code does have a built-in browser. I've seen this preview thing. I just never used it. But in the end you sort of get the same thing, right? The additional skill that you're describing is Claude is better if it can see what it's working on. That's sort of the summary here. And whether it's using your Chrome or it's just making up its own little browser, it doesn't really make a big difference because either way it's going to see what it's working on and that just makes it much better. And then you don't have to run QA for your Claude.

为什么它不接上我已有的 Claude Code 会话？因为我显然用过 Claude Code，但是……

Why doesn't it pick up my existing Claude Code sessions? Because I mean obviously I've used Claude Code, but...

好问题。我没有很好的答案，除了我们确实在……我想这就是 OBI 团队在做的事。酷。我没有别的了，我只是想拓展大家的思路，也许给那些还没真正试过的人展示一下，我觉得很有趣的是，我有时候用这个比用 Dia 还多，对吧？我用了所有其他智能体浏览器，而 Anthropic 没必要自己建一个智能体浏览器，因为你已经有 Claude Work 了，这就够了。

Excellent question. I don't have a good answer other than we're honestly having... This is what the OBI team does, I think. Cool. I don't have other like I just want to expand people's minds and also maybe show people if they haven't really done it, but I think it's very interesting how I sometimes use this more than I use Dia, right? I use all the other agentic browsers and Anthropic didn't have to build an agentic browser because you just had Claude Work and that's enough.

我也觉得，也许与现有的众多优秀浏览器集成，目前在我个人优先级列表上比从头重建一个浏览器要更高一些。

I also think like maybe integrating with a number of excellent browsers out there is currently on my personal priority list a little higher than trying to rebuild a browser from scratch.

是的。你知道，永远别说绝不，但我想回到这个想法：我们希望把它插入到我们整个现有的工作流中。我认为我们的目标其实不是取代你电脑上的任何应用，而是在你的工作流中很好地工作。

Yeah. You know, never say never, but I think going back to this idea of like we want to plug this into our entire existing workflow. I think our goal is actually to not replace any of the applications you have on your computer, but instead to work really well within your workflow.

满足新的需求，是的。看起来如今，尤其是在浏览器上，大多数创新都是用户人体工程学方面的，而不是底层浏览器引擎。所以，我觉得是 Chrome 还是 Alice 什么的，其实无所谓。

Meet the new one, yeah. It seems that nowadays, especially on the browser, most of the innovation is like user ergonomics. It's not really like the underlying browser engine. So, I feel like it doesn't really matter if it's like Chrome or Alice, whatever.

是的，我们想在任何地方满足你的需求，这显然我会这么说，但这也是普遍事实，因为我不想人为地缩小我的潜在用户基础，说‘好吧，我要开始为那些愿意换浏览器的人构建了’。对吧。你知道，有很多诉讼是关于谁运营浏览器的。很多钱花在了哪个浏览器是默认的、浏览器内哪个搜索引擎是默认的这些问题上。我基本上只想为大众构建。我想为那些有很多烦人任务、觉得也许 Claude Work 能帮他们做的人构建。

Yeah, we want to meet you wherever you are, which is like obviously I would say that, but it's also just generally true because I don't want to shrink my potential user base artificially by saying, 'Okay, I'm going to start building for the people who are willing to switch browsers.' Right. There's such a like, you know, many lawsuits have been filed over who gets to operate the browser. And a lot of money has been spent over the question of which browser is default and which search engine is default within the browser. I just want to build for the masses essentially. I want to build for people who have a number of annoying tasks that they feel like maybe Claude Work could do for them.

你怎么看技能的可移植性？我觉得有一件事。我用另一个叫 Zoe 的东西，有点像云电脑加智能体，我有一个技能用来添加访客到办公室。所以每当有人在下班后需要进来，他们得在楼下登记。但我想给那东西发短信。所以，它在 Claude Work 里不太好用。但现在那个技能在 Zoe 的框架里，不在我的 Claude Work 里。如果我做了改动，就得同步它们。你怎么看这个？我觉得记忆是云个人化的。有点像我不一定想让我的记忆跨平台。但我确实希望我的技能能跨我使用的智能体。我认为 MCP 的人做了同样的事。就像‘哦，MCP 网关，MCP 注册表’。我不太确定那是不是一门生意。所以，我很好奇你在这个领域有没有什么想法。

What do you think about skills portability? I think there's been one thing. I use another thing called Zoe, which is kind of like a cloud computer plus agent and I have a skill to add visitors to the office. So whenever somebody has to come in after hours, they need to check in downstairs. But I want to like text the thing. So, it doesn't really work in Claude Work. But now that skill is in the Zoe harness and it's not in my Claude Work thing. And then if I make a change, I've got to sync them. How do you see that going? Like I see memory as like cloud personal. Kind of like I don't necessarily want my memories to be cross thing. But I do want my skills to be cross agent that I use. I think with MCP people did the same thing. It's like, 'Oh, MCP gateway, MCP registry.' I don't really know if that's like a business. So, I'm curious if you've had any thoughts in the area.

对我来说，这大概就是我回归到非常基础的原语的地方：我们的技能是基于文件的，而不是那种存在于某个超级专有地方的复杂东西。我非常倾向于一切都是文件和文件夹的想法。这本身就使它非常可移植。我们确实有技能作为这个容器格式的一部分，我们称之为插件。插件在 Claude Code 和 Claude Code Work 上都可用。格式是一样的。你可以安装插件。这在 Colab 中今天就能用。你基本上可以说，我要添加一个完整的 GitHub 仓库作为技能市场或插件市场。这就是我们实现可移植性的方式。我认为我们还有很多成长空间：如何让人们容易知道他们可以编写技能？如何让他们容易地与你分享一个技能？因为显然我刚才说的所有话，对吧？我失去了大部分知识工作者基础。我一开始说，‘哦，你可以连接一个 GitHub 仓库。’这不是大多数人在通用知识工作者空间里最终会工作的方式。但我认为那里有些东西。另一个我认为还没有被真正充分探索的东西是，技能的哪一部分非常可移植，哪一部分对你来说非常个人化。我认为这是我们作为一个行业还没有真正解决的问题。就像什么时候你想给技能引入更多结构，或者总是有公共技能、私有技能这样的配对。

I think for me this is sort of where I go back to the really basic primitives for our skills are file based instead of this complicated thing that exists inside of a place somewhere that is super proprietary. I'm really leaning into the idea of like it's all just files and folders. And that makes it very portable on its own right. We do have skills as part of this container format which we just call plugins. And plugins are available both for Claude Code and Claude Code Work. It's the same format. And you can install plugins. This works in Colab today. You can basically say I'm going to add a whole like just a GitHub repo as a skills marketplace or like a plugin marketplace. And that's how we're doing portability. I think we have a lot of room left to grow in how do we make it easy for people to know that they can write skills? How do we make it easy for them to just like share a skill with you? Because obviously all the words I just said, right? I'm losing most of the knowledge worker base out there. I started by saying, 'Oh, you can connect a GitHub repo.' It's not exactly how most people will end up working in a general knowledge worker space. But I think there's something there. And another thing that's there that I think has not really been properly explored is the combination of which part of the skill is very portable and then which part of the skill is very personal to you. And I think that's something we haven't really solved yet as an industry. It's like which time you want to introduce more structure to the skill or have always have like public skill, private skill, you know, pairs.

是的，是的，有点。我认为最简单的方法就是像用字符串插值之类的，对吧？在这里插入用户名，插入电话号码，插入已知文件夹位置之类的东西。那可能很笨拙。这就是为什么我们还没做。但我确实认为有人会想出有趣的方法来保留我们喜欢技能的一切。可移植性就是一个文件。就是 markdown。老实说就是文本，对吧？就像文本文件就能用。完全缺乏结构，这意味着你不需要任何教程来写一个技能。就像你向我解释那样向 Claude 解释，Claude 可能比我先理解，对吧？就像订机票，告诉 Claude 如何订机票，就像我们告诉别人我今天刚想到的那样。

Yeah, yeah, kind of. I think that's like the easiest way to do this would just we do like use string interpolation or something, right? Insert username here, insert like phone number, insert like known folder locations, that kind of stuff. That's probably clunky. That's why we haven't built it. But I do think someone is going to come up with an interesting way to keep everything we like about skills. The portability is just a file. It's just markdown. It's just text honestly, right? Like a text file works. The complete lack of structure, which means you don't need any kind of tutorial to write a skill. Just like explain it to Claude the way you would explain it to me, and Claude will probably get it before I would, right? You just like for booking a flight, tell Claude how to book a flight the same way we're telling someone I just thought about near today.

但结合一个非常个人化的东西。也许我们还是用订机票的例子。我其实不认为 AI 应该去订机票。我觉得我们现有的工具就够了。

But combine that with a very personal thing. Maybe we'll stick with the booking a flight example. I don't actually think AI should be booking flights. I think the tools we have is yes.

是啊，终于有人说了。这是每个人都在做的默认演示。我个人的观点是反对订机票演示。这不是一个好的展示。

Yeah, finally somebody says it. It's the default demo that everyone was making. I'm like my opinion against flight booking demos. It is not a good showcase.

是啊，我就想自己订机票。但我觉得很多事情都有个人和非个人成分，这主要是为什么人们会拿订机票举例，因为有些事情非常通用。更便宜的航班通常更好，对吧？很少有人会去订最贵的航班。然后有些事情非常个人化，比如你喜欢什么时间、什么座位、哪个机场。把这些组合成一种可移植、兼容、易于理解的技能格式。我觉得那会非常令人兴奋。我们只是还需要想清楚怎么做。

Yeah, I'm like I just want to book my flight myself. But I think there's a lot of things that have a personal and a non-personal component, and that's mainly why people reach for flight booking because some things are very universal. Cheaper flight is usually better, right? Like few people try to book the most expensive flight. And then some things are quite personal like what time you prefer, what seat you prefer, which airports you prefer. Combining that in a skill format that is actually portable, compatible, easy to understand for people. I think that would be very exciting. We're just having to figure it out yet.

是的，我觉得文本部分。我想现在每个人都有某种云文件服务，要么 Dropbox，要么 Google Drive，等等。所以感觉在某种程度上，它应该基本上把我的技能符号链接到所有智能体框架中。是的，保持同步。就像我们内部有一个有价值的令牌仓库，里面是所有命令和子智能体。这很好……然后我构建了一个 TUI，你可以启动并说，你知道，把这个命令和这三个子智能体安装到这个文件夹里的这个智能体中，然后复制粘贴。它什么都不做，只是把文件复制进去。但我觉得应该有类似的东西，每当我进入一个新环境时，就像“嘿，这是云文件夹的链接，把这些技能下载到这里。”但今天它还不能这样工作。如果我安装一个新的智能体，我必须复制粘贴所有技能，而且我甚至不知道它们在哪里。

Yeah, I think the text part. I think everybody by now has some sort of cloud file thing. Either Dropbox, Google Drive, whatever. So it feels like in a way it should basically symlink my skills into all my agent harnesses. Yeah, just keep those in sync. Like we have internally this valuable tokens repo, which is like all the commands and sub agents. It's a good... and then I build a TUI where you can start and be like, you know, install this command and these three sub agents into this agent in this folder and just copy paste this. It doesn't do anything. It literally CP the file into that. But I feel like there should be something similar where whenever I go into a new thing, it's like, "Hey, here's the link to exactly the cloud folder and just bring down these skills into this." Like today, it doesn't quite work like that. If I install a new agent, I have to copy-paste all the skills and I don't even know where they are.

是啊，这就是大问题。就像“我在哪里找到它们？”

Yeah, that's like the big problem. It's like, "Where do I find them?"

是的。所以我很好奇，在未来，那几乎感觉我的个人生产力东西就是我的技能。它不是我使用的产品，因为每个人都能访问同一个产品。但今天那只是复制粘贴文件。

Yeah. So I'm curious, in the future, that almost feels like my personal productivity thing would be my skills. It's not really the product that I use because everybody has access to the same product. But today that just looks like copy-pasting of the files.

我觉得很多事情。我真的很喜欢把智能体和 LLM 看作另一个同事。已经有很多尝试去建立文档公司，说“哦，我们要解决你所有的文档问题。”我自己也在 Notion 工作过一段时间，对吧？所以我非常熟悉让所有人保持一致的概念。

I think so many things. I really like thinking about agents and LLMs just as another coworker. So many attempts have been made to build documentation companies that are like, "Oh, we're going to solve all your documentation problems." I myself spent a little bit of time working in Notion, right? So I'm deeply familiar with the concept of let's get everyone on the same page.

嗯。对吧？你基本上是说，你希望所有智能体在你的偏好、技能以及它们应该工作的方式和执行方式上保持一致。我不确定正确的方式会是什么。会不会是某家公司可以说，“好吧，我们是一个独立实体，我们不试图推广任何特定产品。我们的工作就是成为这个技能权威，我们提供……我不知道，我们要成为技能的 Dropbox，我们可以符号链接到所有你想用的产品。”我不确定这是否是一个可行的业务，但作为一个想法，它会很酷，对吧？

Mhm. Right? What you're basically saying is you want all your agents to be on the same page about your preferences, about the skills, about the way they ought to work, and how they ought to execute. And I'm not sure what the right thing is going to be. If it's going to be some company that can say, "All right, we're an independent body, we're not trying to push into any particular product. It's our job to be this skill authority and we provide... I don't know, we're going to be the Dropbox of skills and we can just symlink us into all the products you want to use." I'm not sure that's going to be a viable business, but as an idea, it would be cool, right?

是的。是的，我觉得很多事情作为业务都会消失。就像“我该怎么做？”我甚至不要求别人为此做一个产品。就像“是的，我想个人知道。”还有像你说的，你几乎想要技能，然后在个人和工作之间插值。所以如果我因公订机票，那和我个人订机票是不同的。是的。在某些方面，但很多框架是相同的，你知道吗？作为一名工程师，我会告诉你，从技术上讲，一个人对一个人。我就会用符号链接。嗯，这就是我用 Claude Code 和 agents.md 所做的。它和符号链接一样。所以它有效，但感觉……是啊，我不知道，也许吧。因为我会在那个层面上。你总是可以把你的问题告诉同事，然后同事会为你解决。只需创建符号链接。这是一种方法。没错。没错。好吧，一切都叫同事。

Yeah. Yeah, I think so many things are just going away as businesses. It's like, "How am I supposed to do it?" I'm not even asking somebody to make a product about it. Like, "Yeah, I want to personally know." And there's things like you said, just like you almost want to skill and then interpolate between personal and work. So if I'm booking a flight for work, it's different than I'm booking a flight personally. Yeah. In some ways, but a lot of the scaffolding is the same, you know? I as an engineer, I will tell you, technically a person to single person. I will just be like symlinks. Well, that's what I do with Claude Code and agents.md. It's just the same as symlinks. And so it's like that works, but it feels like... yeah, I don't know, maybe. Because I will go on that level. You can always tell cowork your problem and then cowork will solve it for you. Just make the symlinks. That's like one way to do it. That's true. That's true. All right, everything is called cowork.

呃，可能对你们两位来说是个尖锐的问题。哪些行业会消失？好吧，Felix 之前说的很有意思。基本上有短期压力，比如我们需要把这些 token 变成有价值的东西，也就是我应该构建一个利用模型的最后一英里产品。然后还有长期的问题，哪些东西仍然有价值。我觉得你今天已经看到了，你知道，编程领域在某种程度上，每个人都在堆栈中不断上移，因为你需要更多引擎把 token 变成代码。我认为搜索，比如企业搜索，也在经历同样的事情。像 Glean 和所有这些不同的公司，归根结底，如果同事在做所有工作，搜索本身只是很小的一部分，我不知道我是否真的愿意花那么多钱只做搜索。几乎一切都成了同事的垂直领域。那么同事第一方能支持多少，不能支持多少？我认为对于很多这样的事情，你展示的那个规划功能……我们要展示吗？就是那个规划功能。

Uh, potentially spicy question for both of you. Which of these industries will go away? Okay, so what Felix was saying before is interesting. There's basically the short-term pressure of like we need to turn these tokens into valuable things, which is I should build a last-mile product that harnesses the model. And then there's the question of long-term which ones are going to still be valuable. And I think you're kind of seeing this today where, you know, the coding space in a way it's kind of like everybody's moving up and up in stack because you need more engines turning tokens into code. I think search, like enterprise search, is kind of seeing the same thing. Like Glean and all these different companies, at the end of the day if cowork is the one doing all the work, the search itself is such a small part that I don't know if I'm really going to pay that much money just to do search. It's almost like everything is a cowork vertical. So how much can cowork first-party support and how much can it not. I think for a lot of these things, the planning thing that you were showing... Do we show it? It's the planning thing.

好的，是的，是的。就像那件事，这些智能体提供的大部分价值在于它们更擅长为特定任务做规划，并且有更好的工具。是的。但我认为模型现在正在朝那个方向发展，它们有合适的框架，并且就在你的电脑上。所以对我来说，几乎就像如果最终客户信任你的初创公司作为那个任务结果的提供者，那么我认为那是可行的。这是……这是我们正在做的一个衬衫尖峰。

Okay, yeah, yeah. Like that's one thing where most of the value that these agents provide is like they're better at planning for specific tasks and have better tools for it. Yeah. But I think the models are now moving in that direction and they have the right harnesses and they're on your computer. So for me it's almost like if the end customer trusts your startup to be the provider of that task result, then I think that works. This is something that... This is a shirt spike that we're working on.

是的。我觉得听着，我跟你说这个。我不认为我是那个必须决定哪个行业受冲击最大的人，但我确实认为 Anthropic 作为一个团队，我们非常担心这些工具对劳动力市场的影响，尤其是对初级员工。我认为只有诚实地说，当我们谈论自动化掉很多我们个人觉得烦人的工作，那些我们可能认为不是时间最佳利用的工作时。

Yeah. I think look, I'll tell you this. I don't think I'm the best person who actually has to make which industry is going to be hit the hardest, but I do think that Anthropic as a group of people, we're deeply worried about the impact that the tools are going to have on the labor market, especially for junior employees. I think it's only honest to say that when we talk about automating a lot away, a lot of the work that we personally find annoying, that we maybe think it's not the best use of our time.

在很多行业，那种工作本来会交给初级入门员工，对吧？我认为对此感到担忧是合理的，尤其要担心这对进入就业市场的人会产生什么影响。

In a lot of industries, that kind of work would have been given to a junior entry-level employee, right? And I think it's only right to be really worried about that and worry what that's going to do, particularly to people entering the job market.

我对此有一个解决方案：为他们创造模拟工作。这半开玩笑半当真。想想软件工程，当你是一名初级工程师时，你会工作 1-3 年。在那 3 年里，可能只有少数几个时刻你真正学到了东西，其他很多天你并没有进步。我认为现在我们可以用 AI 和这些模型来缩短这些职业生涯，几乎模拟你工作的早期阶段，让学习密度变得极高。就像这样：‘嘿，我们在开发一个分布式系统功能，你需要学这个东西，在公司可能要花 3 个月。’所以你就花 3 个月——这里我们模拟整个过程，不是真的。一周内我们快速过一遍，你从中吸取教训，然后重复。一年内你基本上获得相当于 3 年的项目和经验。我觉得对于销售或市场营销这类工作来说更难，因为你没有反馈循环。但这听起来有点傻——就像创造假工作，但几乎就像上大学，对吧？人们付费学习。这可能类似：‘嘿，我们有 Jane Street 模拟器。你想来 Jane Street 工作吗？我们把你放进模拟器 3 个月，你出来就准备好了。’所以这里有一个方面。我还没有足够专业到知道市场营销、法律或金融会发生什么——我不做那些工作，也不该谈论它们。但我是工程师，对工程有很好的了解。我们看到的一件事是，作为各种规模的公司，公众对初级岗位深感担忧，但我们也看到更多高级工程师被加速——他们更高效，提供的价值增加了。我经常思考的一件事是，即使在这之前，我一直非常尊重滑铁卢大学。加入我团队的滑铁卢新毕业生总是比那些整个大学期间从未在需要交付用户产品的环境中工作过的新毕业生准备得更充分。我是德国人，上过德国大学；那里的信息系统课程往往非常理论化。我经常举的例子是，想成为医生但必须先学 4 年生物学。结果，当你招到一个新毕业生时，你必须教他们如何构建产品、在公司工作、与他人合作、处理不同意见——所有这些事情。滑铁卢大学似乎让他们花一半时间——我不知道这是不是真的，但我想是一年——他们花很多时间实习。

I have a solution for that: you create simulated jobs for them. This is half joke, half true. If you think about software engineering, when you're a junior engineer, you work 1-3 years. In those 3 years, there are maybe a handful of moments where you really learn something, and then a bunch of other days where you're not really progressing. I think now we can use AI and these models to shortcut these careers and almost simulate the early years of your work, making them super dense in learnings. It's like, 'Hey, we're working on this feature, which is a distributed system, and you need to learn this thing that might take 3 months at a company.' So you take 3 months—here we're simulating the whole thing, it's not real. In 1 week, we speedrun through it, you learn your lesson, and we repeat that. In 1 year, you basically get 3 years' worth of projects and experience. I think it's harder for things like sales or marketing because you don't have a way to get the feedback loop. But it sounds kind of silly—it's like making fake jobs, but it's almost like going to college, right? People pay to learn. This might feel similar: 'Hey, we have the Jane Street simulator. You want to come work at Jane Street? We'll put you in the simulator for 3 months, and you'll come out ready.' So there is an aspect here. I'm not expert enough to know what will happen to marketing, legal, or finance—I don't work in those jobs and shouldn't talk about them. But I am an engineer, and I have a pretty good idea about engineering. One thing we're seeing is that as a company of all sizes, the public is deeply worried about entry-level, but we're also seeing more senior engineers accelerated—they're more productive, they increase the value they provide. A thing I'm thinking about a lot is that even before all this, I've always had a lot of respect for the University of Waterloo. The new grads that joined my teams from Waterloo always felt more ready than new grads who spent their entire time at university but never had to work inside an environment where you ship things that users will use. I'm German and went to a German university; the information systems programs there tend to be very theoretical. I often give the example of trying to become a doctor but first doing 4 years of biology. As a result, when you get a new grad, you have to teach them what it's like to build products, work in a company, work with others, handle differing opinions—all these things. The University of Waterloo seems to have them spend half their time—I don't know if this is true, but I think it's a year—they spend so much time in internships.

课程的一部分是花一年时间实习。是的，他们从一个公司到另一个公司。他们作为去过 20 家公司的初级工程师出现在你的团队里。不完全是，但似乎我的很多新毕业生也曾在 Apple、Google、Tesla 短暂工作过。

Part of your job curriculum is to spend a year in internships. Yeah, they just go from company to company. They show up on your team as a junior engineer who's been to like 20 companies. Not really, but it seems like a lot of my new grads have also briefly worked at Apple, Google, Tesla.

是的，有一个常见的梗是他们像收集无限宝石一样收集所有这些标志。但他们总是放在 LinkedIn 上，很不清楚他们只是实习生。没错，但这确实让他们比其他新毕业生好得多。我想知道这是否是未来的一个有用模型，当我们还必须压缩作为初级员工的时间时，因为初级员工的价值将受到影响。我支持年轻人的观点是，你们有更高的神经可塑性，能学更多，更少既有偏见。我认为对你来说也是如此——OpenAI 经常说——实际上更年轻的应届毕业生工程师使用 Codex 或编写代码时比有固定偏好方式的经验丰富的工程师更具创新性。

Yes, and there's a common meme where they collect all these logos like Infinity Stones. But they always put it on LinkedIn. It's very unclear that they were an intern. Yeah, exactly. But it does actually make them so much better compared to other new grads. I wonder if that's a useful model for the future when we also have to crunch down the amount of time you have as a junior employee because the value you have as a junior employee is going to be impacted. My sort of pro-young-people take is that you have higher neuroplasticity, you can learn more, you have less pre-existing biases. And what I assume is true for you—what OpenAI often says—is that actually the younger, fresh grad engineers use Codex or code stuff more innovatively than experienced engineers who have a set and preferred way of doing things.

是的，当我与人交谈时，我会写下一些经历。是的，也许你更 AI 原生，因此你被淘汰。但我认为问题是你不需要那么多这样的人。我的意思是，Anthropic 公开表示我们确实相信对市场的影响将是巨大的，并且我们不认为人们总体上准备好了。对吧？我们确实认为社会应该更多地讨论这个问题。

Yeah, as I talk to people, I write some of my experiences. Yes, and maybe you're more AI-native, and therefore you get cut. But I think the problem is you don't need that many of them. I mean, Anthropic is on the record saying we do believe the impact on the market is going to be sizable, and we do not think that people overall are ready. Right? And we do actually think we should probably talk about it as a society much more.

是的，我不确定我是那个能增加有用内容的人。但我认为作为有经济学家和政府的社会，需要以可能比我个人挣扎更有意义的方式来解决这些问题。我们可能讨论得不够。

Yeah, I'm not sure that I'm the individual that can add anything useful there. But I think as societies with economists and governments that need to wrestle those questions in a way that is probably more meaningful than me wrestling with them. We're probably not talking enough.

是的，我们努力教育，而且我认为像你们那样频繁发布——或者可能太频繁了——这有助于人们随时间调整，而不是一次大爆炸。人们正在经历一种逐渐的起飞，我们可以帮忙，对吧？

Yeah, well, we try to educate, and I think also just releasing frequently as you guys do—or probably too frequently—it's helping people to adjust over time rather than one big bang thing. There's sort of this gradual takeoff that people are living through that we can help out, right?

是的，但我认为我们很多人都在想，到底什么时候会完全起飞？什么时候会出现那个大爆炸时刻，事情开始加速得如此之快，以至于变成自我强化的循环？然后就像比赛开始，不再有缓慢追赶——只是在所有事情上都变得非常出色。是的，当同事在训练模型，当看着 TensorBoard 和权重与偏差并训练东西时。我们都可以争论还有多少年——有些人打赌，也许 10 年，也许 1 年。

Yeah, but I think a lot of us are wondering at what point we actually have full takeoff, right? At what point is there this big bang moment where things start accelerating so quickly that it becomes a self-reinforcing loop? And then it's sort of off to the races, and there will be no more slowly catching up—just being so good at everything. Yeah, it's when coworkers are training models, when it's looking at TensorBoard and weights and biases and training things. We can all debate how many years it's away—some people make a bet, like maybe it's 10 years away, maybe it's a year away.

我不太确定自己在这个问题上的立场，但我也不是很确定它最终是否真的那么重要——无论它是在四年还是五年后发生。如果我们有一个像样的 AGI，那它肯定会发生，这可能是我们应该认真应对的事情。

I'm not entirely sure where I come on this line, but I'm not entirely sure that ultimately matters all that much whether or not it happens in four or five years. If we have a decent one that's certainly going to happen, it's probably something we should wrestle with.

我想谈谈那个日程任务完成的情况。我试图让它做更多主题性的分组，比如读取文件，理解内容，按主题而不是文件类型分组。

I wanted to talk about the schedule task complete. I was trying to get it to do more thematic grouping, like read the file, understand what it's about, group by topic rather than file type.

我的意思是，你可以直接跟进，让它实际去做。

I mean you can just follow up and have it do that actually.

哦对，它确实提出了这个，对吧？所以它有一些主题性的东西，但可能还能做得更好。

Oh yeah, like it did propose that, right? So it's got some topical things, but it could probably do better.

是的，我可能需要给它一个读取视频文件的技能，这样它就能理解我喜欢怎么组织它们。不过说实话，我看到你在用 Claude 4.6，对吧？我对人们的建议是越来越不用操心了。直接告诉它你想让它做什么，它很可能会自己想办法搞定。可能不一定是你喜欢的方式，或者你习惯的做法。

Yeah, like I probably need to give it a skill to read video files so that it understands how I like to organize them. Honestly though, I see that you're using Claude 4.6, right? My recommendation for people is increasingly don't worry about it anymore. Just tell it what you want it to do and it's probably going to figure out a way to do it. It might not be the way that you like necessarily or the way that you've gone about it.

是的，视频更复杂。但我们也在收集和整理所有这些，所以让我们试试吧。我真的很想知道 Claude 会想出什么办法。

Yeah, videos are deeper. But we're also sourcing and organizing all of this, so let's fight. I'm honestly so curious what Claude is going to come up with.

我来开个头。我还想谈谈你提到的整体数据分析，比如你的个人财务。你还说，这对我们来说非常及时，报税季，对吧？用 Claude Code 处理报税季，它不对任何错误负责，但何乐而不为呢？它为你做免费的知识工作。所以我认为 Claude 在金融领域是个大事，这绝对属于那个范畴。我想知道，这是一个独立的团队吗？你和他们交流吗？它有多重要？因为你现在还需要输出 Excel 文件。谈谈你们在金融方面的努力吧。

I'll kick that off. I wanted to also talk about the overall data analysis you talked about, like your personal finances. You also said which by the way for us is very timely, tax season, right? Use Claude Code for tax season and it is not responsible for any mistakes, but might as well, right? It's free knowledge work for you. So I think Claude for finance is a big deal and this is definitely in that mix. I wonder, is it a separate team? Do you talk to them? How important is it? Because you also need the output Excel files now. Just talk about the finance effort you guys have.

我们非常关注垂直领域。所以我们确实有一个专门的垂直团队。我们还有一个专门的企业团队。这些是业务工程，不是销售。是工程。所以我们确实有人每天上班，他们问自己：我们如何让 Claude 在那些特定行业中为人们极其有效地工作？我们如何让它更容易被理解？我们如何让他们更容易接入，并获得与软件工程师相同的价值？我认为软件工程师最终站在整个 AI 时代的前沿并不奇怪，因为其中很多就像鲁布·戈德堡机械一样，我们已经习惯了自动化，对吧？这是我们工作的一部分。所以我们非常重视。我认为这也非常符合我们看到的 Claude 作为模型的优势。我认为它尤其为那些客户提供了巨大的价值，因为我们可以用他们拥有的数据量做很多事情。那些是数据密集型行业。那些是正确性非常重要的行业。对我们来说，如果我用它来分析我的业务，我就是不能展示出来。

We care about the verticals quite a bit. So we do have a dedicated verticals team. We also have a dedicated enterprise team. And those are business engineering, not sales. It's engineering. So we do have people who come to work every single day and they ask themselves how do we make Claude work extremely effective for people in those specific industries? How do we make it easier for them to understand? How do we make it easier for them to plug into this and get the same value out of it that software engineers get? I think it's no real surprise that software engineers ended up being at the forefront of the entire AI moment because so much of it is this Rube Goldberg machine-esque where we're already used to automating things, right? It's part of our job. So we care about it quite a bit. I think it also really matches what we see Claude being very good at as a model. I think it provides tremendous amount of value to those customers in particular because we can do so much with the amount of data they have. Those are data heavy industries. They're industries where correctness matters quite a bit. For us, if I've used it to analyze my business, I just can't show it.

那太遗憾了。所以我也有一个关于报税的问题。我确实发推文说 Claude 在帮我报税。这真的很不可思议。而且很烦人，因为这太酷了，但我不打算在 Twitter 上分享我的纳税申报单。

That's too sad. So I had a similar question about taxes. I did tweet about the fact that Claude was doing my taxes. This is honestly incredible. And it's annoying because this is so cool, but I'm not going to share my tax return on Twitter.

Twitter 可能不是需要看到我纳税申报单的受众。但就是这样。它在读取视频。所以它越来越多了。它实际上是怎么做到的？我很好奇。

Twitter's maybe not the audience that needs to see my tax return. But here it is. It's reading on the videos. So it's getting more. How did it actually do it? I'm curious.

哦，通常它只是截个图，然后通过视觉读取截图。

Oh, usually it just takes a screenshot and then it reads the screenshot by vision.

所以，这就是我处理 Zoom 上传的方式，对吧？因为我有纸牌俱乐部会议需要上传到 Zoom，我希望它能自动给它们加标题并做节目笔记等等。所以它只是截图并尽力而为。转录可能不会带来额外好处，因为它现在完全靠视觉操作，但已经足够好了。然后我还得调用 Nana Banana 来处理图片。所以，除非你们帮我处理图片，否则我得调用那些做图片的人。

So, this is what I do for my Zoom upload thing, right? Because I have paper club sessions that I need to upload to Zoom, and I want it to automatically title them and do show notes and everything. So, it just takes screenshots and tries its best. It wouldn't probably benefit from transcribing, which it's doing by pure vision now, but it's good enough. And then I do have to call out to Nana Banana to do images. So, unless you guys do images for me, I have to call the people who do images.

我们知道。这对我来说太有趣了，因为我越来越常做这件事，越来越好奇 Claude 的创造力，以及弄清楚 Claude 解决问题的方法好在哪。

We're aware. It's just so fun for me because this is the thing that I'm increasingly doing, increasingly curious about Claude's creativity and figuring out what is great about Claude's approach to solving a problem.

万物视觉化是超能力，对吧？还有电脑操控，你们是第一个做电脑操控的，对吧？当它刚推出时，我印象非常差。我觉得它又慢又不可靠。一年前它进步了多少。我知道。它几乎不可用。我记得它几乎不可用，但一年来事情变得好多了，这难道不疯狂吗？我们去了 Anthropic 办公室，因为他们举办了电脑操控的发布活动，就像有个黑客马拉松。但没有人去搞电脑操控。

Vision for everything is the superpower, right? And computer use, you guys were the first to do computer use, right? And when it was launched, I was very unimpressed. I was like, it's slow, it's unreliable. And how much better it's got in 1 year ago. I know. It was barely usable. I remember it was barely usable, but isn't it wild how much better things have gotten over that 1 year? We went to the Anthropic office because they had the launch event for computer use, like there was this hackathon. And nobody hacked on computer use.

但我确实看到你安装了一个自动化的 Mac OS FTP 服务器，对吧？你用过吗？

But I did see briefly that you do have an automate Mac OS FTP server installed, right? Do you use that ever?

什么？哪一个？在哪里？

What? Which one? Where?

如果你去设置。哦，设置。好的。哪里？抱歉，是这个吗？对。我在你的连接器里注意到了。

If you go to your settings. Oh, settings. Okay. Where? Sorry, this one? Yeah. I noticed that in your connectors.

我可能说过一次，但我没有主动使用它。

I probably said that at one time, but I don't use it actively.

好的，比如 Mac OS 的 Automator？对对。所以我真的很想自动化我的一切。但我发现它不太可靠。为什么？

Okay, like a Mac OS Automator? Yeah, yeah. So I really wanted to automate everything in my thing. I didn't find it super reliable. Why?

毫无疑问。Claude 编写 AppleScript 并执行自己的 AppleScript 比依赖这些第三方工具要好得多。所以，我最初安装了 IMCP 和人们构建的所有其他 FCP，但现在我再也不用它们了。就让 Claude 自己写东西。

No question at all. Claude is much better writing AppleScript and executing its own AppleScript than relying on these third-party tools. So, I initially installed IMCP and all these other FCPs that people built, but now I don't use any of them anymore. Just let Claude write its own thing.

它会更加定制化。我们一直在往上层走，但我确实认为电脑操控对我来说是一个相当有趣的领域，而且有趣之处在于，我认为我们离 Claude 非常有效地使用你的电脑（而不仅仅是理论上的电脑）不远了。用户和电脑之间的关系是什么？有一些推文说 Claude Code 创建的一些虚拟机有多大。

It's going to be more custom-made. We keep going up the stack, but I do think computer use is a fairly interesting area to me, and it's also interesting in the sense that I don't think we're far away from Claude being very effective at using your computer and not just a theoretical computer. What's the relationship between the user and the computer? There were some tweets about how huge some of the VMs that Claude Code creates are.

大概 12 到 15 GB，人们就会抱怨。但某种程度上，如果你在用电脑，你在操作，那这还是你的电脑吗，还是我只是看着它？我觉得这就是为什么人们喜欢 Mac Mini 和那种开放式底座之类的想法，因为它有自己的家，它在做它的事，我在做我的事。我觉得有点像，不是竞态条件，而是说，如果我启动这个任务，我就没法真正用电脑了。因为 Claude Co-work 正在上面操作，这有点尴尬。我不确定。但我确实认为这是一个非常有趣的领域，因为我可以告诉你一些我想到的、实际上我觉得是坏主意的想法。所以，当我们最初开始做 Co-work 时，我确实梦想过 Claude 拥有自己的光标会是什么样子。那会很酷，对吧？就像一台电脑，我们可以写代码，可以触碰一切。谁说电脑只能有一个光标？我们可以有第二个光标。但这实际上会带来很多问题，即使你向苹果和微软展示这些酷炫的梦想。你会说，这难道不酷吗？但它会带来很多问题，因为我们电脑上的很多模型都是围绕“你只在一个东西上工作”这个想法构建的。有前台应用、后台应用。Claude 和 Chrome 可以在后台工作，但那是在一个应用内部。但在操作系统层面，实现起来要困难得多。所以，我仍在思考，Claude 真正在你的电脑上行动意味着什么？正确的形式是让 Claude 拥有自己的电脑，你设置好，偶尔放大进去玩一玩？还是让 Claude 等你离开一会儿，然后在你不在时接管？或者让 Claude 在云端拥有自己的电脑，你想让 Claude 做什么就自己设置？有很多不同的选项。这是我经常思考的事情：你和你的电脑、你和你的数据之间的关系是什么？因为这种关系的亲密程度取决于工具和你当前在看的东西。我们对分享某些东西很自在，对分享其他东西则很不自在。我认为任何成功的产品都必须处理这些不同的事情，但即使 Claude 有能力做出决定，你首先会希望 Claude 做出那个决定吗？这很棘手，Barry，因为这不仅仅是隐私问题，这几乎是亲密感。而且很难用一种让所有人都舒服的方式来推理。

It's like 12-15 GBs, and people complain. But at some point it's like if you're using the computer, you're taking action, is this just your computer, and I'm just looking at it? You know, it's like I think that's why people like the idea of like the Mac Mini and the open claw or whatever on it, because it's like it got its own home, you know, it's doing its thing. I'm doing my thing. I think there's some kind of like not like race condition, but it's like okay, if I kick start this task, now I can't really use the computer. Yeah. You know, because Claude Co-work is doing things on it. And it's kind of awkward, like yeah, I'm not sure. I do think it's a super interesting area, because I can maybe tell you some of the things I thought about that I think are actually a bad idea. So, when we initially started working on Co-work, I did have some dreams about what would it look like for Claude to have its own cursor. Could be cool, right? Like it's a computer, we can write code, we can touch everything. Like who says that computers need to have one cursor? We could do a second cursor. But that actually breaks down quite a bit, even if you go and present cool dreams to both Apple and Microsoft. You're like, wouldn't it be cool if it breaks down quite a bit, because so many of our models on the computer are built around this idea of like there's only one thing you're working on it. There's like a foreground app, a background app. Claude and Chrome can work in the background, but that's like within one application. But at the operating system layer, that is a lot harder to implement. So, I'm still grappling with what does it mean for Claude to actually act on your computer? Is the right format for Claude to have its own computer that you set up and maybe every now and then you zoom in and you play with it, or is the right format for Claude to just wait until you're stepping away for a little bit and take over while you're gone, or is the right move for Claude to just have his own computer in the cloud and whatever you want Claude to do you set up yourself. Right? There's a number of different options. This is a thing I think about a lot: what is the relationship between you and your computer and you and your data on the computer? Because how intimate that relationship is kind of depends on the tool and the thing that you're currently looking at, right? Like we're quite comfortable sharing some things, very uncomfortable sharing other things. And I think whatever product is going to be successful will have to deal with those different things, but you probably even if Claude was capable of making a determination, would you want Claude to make that determination in the first place? It's tricky, Barry, because it's more than just privacy. It's almost intimacy. And it's tricky to reason about in a way that will make everyone comfortable.

是啊，我能想象，比如一个虚拟盒子，就像真正的 VirtualBox 应用，你运行虚拟机，然后有一个屏幕里的屏幕，你可以放在后台，但也可以跳进那个屏幕。就像——

Yeah, I could see, you know, a virtual box, like actual virtual box app where you run the VM and then you have a screen within the screen, you know, you can put in the background, but then you can jump in the screen. And like

听起来不错。是啊。我的意思是，以前人们会在 Windows 机器上虚拟化 Kali Linux，然后跳进去再跳出来，但这不是双启动，而是在里面。问题是你需要两倍的内存，两倍的资源，对机器负担很大。但我觉得那会很酷。就像看到那个小小的 Claude 窗口，我能看到它的桌面，多可爱啊，它在点来点去。

Hearing that's not a bad idea. Yeah. You know? Like I mean I used to, you know, people used to do that virtualizing like Kali Linux in a Windows machine. Yeah. And then you just jump in and then you would jump out, but it's like it's not like a dual boot. It's like within the thing. The problem is that you need twice the amount of RAM, twice the amount of, you know, it's like it's kind of taxing on the machine. But I think that would be cool. Kind of like see, you know, the little Claude window. I can see his desktop, like how cute it is, clicking around things.

我正想说，他可是“机器中的机器”的原创者，因为他有那个 Windows 95 项目。那个 Windows 95 项目在哪？

I was going to bring out he's the original machine in a machine guy because he has the Windows 95 project. Where's the Windows 95 project at?

可能在我的 GitHub 上某个地方。不不不，它是最显眼的一个。没错。说实话，那是个很有趣的项目。显然，我得说清楚，免得有人误会。我没有写真正的 Windows 95，因为我当时还是个孩子。而且我也没有构建那个能在 JavaScript 和 WebAssembly 中模拟 x86 处理器的引擎。那是一个叫 v86 的工具，非常酷，大家都应该试试。但这个项目源于我们工作中的一个争论，人们经常在争论 Electron 的优缺点，以及我们是否应该用 JavaScript 构建软件。是还是不是？我仍然很郁闷，因为我可以在 JavaScript 中运行整个 Windows 95，并在虚拟化的 JavaScript Windows 95 机器中启动 Microsoft Excel，做各种事情，整个链条比我在传统 SAS 应用中做很多事情还要快。这有点像我的性能狂飙。所以我主要是为了逗 Slack 的同事们开心而做的这个玩笑。只花了一个晚上。什么？但后来，这并不难做。所有困难的工作都在 v86 上。你去它的仓库，会看到 99% 的工作是由一个叫 copy 的人完成的，他叫 Fabian。

It's probably somewhere on my GitHub. No no no no no. It's like the first thing you see is this one. Nice. Yeah, it's exactly. That was honestly a very fun project though. Obviously I didn't I should say this just so that no one gets the wrong impression. I did not write the actual the actual Obviously I didn't build Windows 95 because I was a child. But also I did not build the actual engine that is capable of like simulating an x86 processor in JavaScript and WebAssembly. That's a tool called v86 which is very cool and everyone should try. But this came out of a debate we had at work where people were like they often are in the end of debating the merits of Electron and whether or not we should be building software in JavaScript. Yes or no? And I still am very upset that I can run all of Windows 95 in JavaScript and launch Microsoft Excel inside the virtualized JavaScript Windows 95 machine and do things that I can do that entire chain faster than I can do a lot of other things in like traditional SAS applications. This is sort of like a performance rampage that I went on. So I mostly built this as a joke for some of my colleagues at Slack. This took like one night. What? but then then I It was not hard to do. It was All the hard work is on v86. Like it's go to the repo. It's going to say like 99% of this work is done by a guy who goes after the by the name copy. His name is Fabian.

我觉得你好像又回到了 Windows 的苦差事上，因为你在做 Windows 支持。我想有一些很酷的技术故事可以讲，让人们体会到这有多难，以及你在沙盒上投入有多重要。所以也许这是个好机会来谈谈一些细节。

I think you're kind of back on the Windows grind because you're building on the Windows support. I thought there were some really cool technical stories to tell and it gives people an appreciation of like well, here's how hard it is and here's how important how you invest in the sandbox. So maybe this is like a good opportunity to talk about some of the details.

哦，是的，虚拟机真的很酷。我们对虚拟机有很多不喜欢的地方，对吧？有很多真正的权衡，你想知道为什么要做这些权衡。你说得对，很多人给我写信说：“嘿，为什么 Claude 占了 10 GB？”我可以在播客里说，它实际上并没有占 10 GB。只是 Mac OS 显示字节的方式有问题。但我们实际写入磁盘的方式是通过压缩镜像中的空白空间。所以它实际上并没有占 10 GB。但这是一个技术上的区别，是给非技术经理用来撒谎的。对我来说，问题是它启动太慢了。有时要 30 秒，或者我不知道。哦，应该比那更快。不管怎样，好吧，可能是 10 秒，但感觉像 30 秒。是的，无论如何，它都会比直接在电脑上运行 Claude Code 慢，对吧？所以这些权衡是真实的。

Oh yeah, the VM honestly is like so cool. There's a lot of things we dislike about the VM, right? Like there's a lot of things that are real trade-offs and you want to know why you're making those trade-offs. You're right, there are a lot of people writing me like, "Hey, how come Claude is taking up 10 GB?" I could say in the pod it's not actually taking up 10 GB. It's just like a way that Mac OS displays bytes is like wrong. But the way we actually write it to disk is by we collapse the empty space in the image. So it's not actually taking up 10 gigs. But that's a technical differentiation that's for a non-technical manager to lie. To me the the how come is it takes too long to start. Yeah. It's like 30 seconds sometimes or I don't know. Oh, it should be faster than that. Whatever. Fine. It's going to be 10, but it feels like 30. Yeah, like even either way, like whatever it is it's going to be slower than just running Claude Code directly on your computer, right? So the trade-offs are real.

但在 Windows 上，我们使用的是 Windows 主机计算系统。这和 WSL 2（适用于 Linux 的 Windows 子系统）运行在同一个系统上，我觉得很多开发者都非常喜欢它。

But what we're doing on Windows, we're using the Windows Host Compute System. It's the same thing that WSL 2 runs on, like the Windows Subsystem for Linux, that I think a lot of developers appreciate quite a bit.

嗯。

Yeah.

这很酷，因为我们必须区分虚拟机运行在哪个系统空间，以及谁可以与该虚拟机通信。显然，你给了这个虚拟机相当大的权限，那么我们不仅要优化两个系统之间的连接，还要确保其他随机应用程序无法与虚拟机内的 Claude 通信。

And it's pretty cool because we have to separate out which system space the virtual machine runs in and who gets to talk to that virtual machine. Obviously you give this virtual machine a decent amount of power, so how do we optimize not just the connection between the two systems, but also how do we make sure that random other applications don't get to talk to Claude inside the VM?

嗯。

Mhm.

我们做了一些非常有趣的事情。上周我们开始编写一个新的网络服务和网络驱动程序，用于优化 Claude 与互联网的通信方式。如果你的公司有一些奇怪的网络设置，比如数据包检测，或者把你的 pod 当作公司内部的一个单元。

We do some pretty interesting things. Last week we started writing a new networking service and networking driver that optimizes how Claude talks to the internet. If your company is doing weird internet things, like packet inspection, like taking your pod as a cell inside your company.

我觉得可能有一个非常小且简单的 Claude Code 版本，它更简单，但在大多数用户的电脑上会出问题。而这个版本很不错，因为它能在大多数用户的电脑上运行。

I think there was probably a very small easy version to build of Claude Code that is much simpler, but also breaks on most users' computers. And this one is quite nice because it works on most users' computers.

我常用的默认例子是，我希望它在大多数人拿到的机器上都能高效运行。而那台机器很可能没有 Python，也没有 Node.js。即使我只去掉这两样东西，Claude 在你的电脑上也会变得非常低效。那你怎么办？

And the default example I would go for is I really want this to be highly effective on a machine that most people pick up. And that machine will probably not have Python, it will not have Node.js. And even if I just take away those two things, Claude is going to be so much less effective on your computer. So, what do you do?

你甚至……我的意思是，也许要求人们安装 Node 和 Python。哦，你是说没有虚拟机的情况下未来会是什么样子？

You don't even... I mean, maybe require people to install Node and Python. Oh, like you mean for a what does the future look like without a VM?

不，不，不。就像你说的，对吧？假设目标机器是默认配置的 Windows 桌面。我们这样做，这很酷。在 macOS 上，我们使用 Apple 虚拟化框架，它优化得相当扎实。这是好东西。而且它是一个非常简单的 API 调用，对吧？

No, no, no. So, like you said, right? Let's say the target machine is whatever is at default spec Windows desktop. We do this, which is quite cool. So, on your macOS, we use the Apple Virtualization Framework, which is pretty solidly optimized. Like, it's good stuff. And it's a very simple API call, right?

它超级简单。我最近看到了代码，我当时想：“就这？什么鬼？”

It's just like super simple. I saw the code recently, and I was like, "That's it? What the fuck?"

一旦你开始在上面发布生产代码，你就会开始添加所有你学到的边缘情况。哦，是的。最终代码会变得长一点，但我认为 Apple 在虚拟化框架上真的做得很好。它非常非常好，非常快，非常可靠。在 Windows 上，主机计算系统，我认为 WSL 2 也是 Windows 中的一颗明珠。它是少数几个开发者普遍称赞的东西之一。非常非常酷。接入同一个子系统让我们更容易说：“我们并不在乎你的电脑有多受限。也许它是你雇主的电脑，你的雇主决定你不能安装任何东西。”

Would you once you start shipping production code on it, you start adding all of these edge cases you learn. Oh, yeah. It ends up being a little longer, but I think Apple really cooked with the Virtualization Framework. It is very, very good. It is very fast, very reliable. And same on Windows, the Host Compute System, I think WSL 2 as well is maybe one of the diamonds within Windows. It's like one of the few things that developers universally rave about. It's very, very cool. And hooking into the same subsystem makes it a lot easier for us to say, "We don't really care how locked down your computer is. Maybe it's like your employer's computer, and your employer has decided that you can install nothing."

嗯。不被信任。

Mhm. Not trusted.

但在很多环境中确实如此，对吧？即使在 Anthropic，我们的 IT 部门也控制着我们安装的东西，这对许多公司来说是很常见的体验。这给了 IT 部门相当多的……这让他们的工作轻松很多，因为我们可以说，你可以把 Claude 的电脑和用户的电脑分开，然后对于 Claude 的电脑，你可能关心的是数据丢失、潜在恶意行为者、数据被泄露。一旦你控制了网络和文件系统层，你就不必再担心 Claude 可能会编写非常有用的 Python 脚本。真正让你担心的是，一旦你安装了 Python，任何人都可以在你的电脑上做任何事情。但一旦你把它放在虚拟机里，这种风险就大大降低了。

But it's true in a lot of environments, right? Like, even at Anthropic, our IT department controls what kind of stuff we install, which is a pretty common experience for many companies. And this gives IT departments a decent amount of... it makes their job so much easier because we can say you can separate out Claude's computer from the user's computer, and then for Claude's computer, what you probably care about is data loss, you care about a potentially hostile actor, you care about maybe data being exfiltrated. And once you control the network and the file system layer, you don't really care necessarily anymore that Claude might be writing super useful Python scripts. What worries you about the fact is that once you install Python, now anyone can do anything on your computer. But once you put that in a VM, that risk really goes down.

是的。所以这就是为什么我们要费这么大劲。

Yeah. So, that's why we jump through all of these hoops.

是的，我记得你发过另一条关于这个的推文，但几乎就像人们有审批疲劳一样。你不可能批准每一个命令。有时默认情况下，一些 CLI，我认为即使是早期的 Claude Code，我们也必须批准每一个命令。

Yeah, I think you had a different tweet about this, but it's almost like people have approval exhaustion. It's like you can't approve every single command. Like sometimes by default, some of the CLIs, I think even early Claude Code, we have to approve every single command.

是的。所以存在一种两难境地：要么批准每一步，要么危险地跳过权限。

Yeah. And it's so there is a sort of dichotomy between either approve every step or dangerously skip permissions.

实际上沙盒化有点像中间地带。

And actually sandboxing is kind of like the middle ground.

是的，我确实认为这可能是我们整个行业的责任，要提出比“只要它什么都不做，就是超级安全”更好的方案。对吧？如果你想让这东西有用，你就必须批准每一步，计算机使用就是一个很好的例子。让主机上的计算机使用变得超级安全的唯一方法，可能就是批准每一个操作，对吧？比如模型说“我想输入字母 L”，你回答“好吧，那没问题。”

Yeah, I do think it's maybe on us as the industry to come up with something better than "Oh, this is super safe as long as it doesn't do anything." Right? And if you want this to be useful, then you have to approve every single step of the way and like computer use is a good example. The only way to make computer use on your host super safe, like really super safe, is probably if you approve every single action, right? Like models like "I would like to type the word 'L'." You're like, "Okay, that seems fine."

因为我知道哪个光标是焦点。

Cuz I know which cursor is focused.

是的，如果你不委托，那就不是自动化。

Yeah, it's not automation if you don't delegate.

是的，完全正确。你需要委托。你需要能够委托并走开，相信这东西不会自动搞砸。我甚至不认为我们需要构建完美的系统。我不认为我们需要等待 100% 的模型对齐。我们可以依赖行业长期使用的瑞士奶酪模型。但我确实认为我们可能需要普遍地、最终地投入更多。而这正是我们在做的。我们需要在那些可以说“你不需要批准所有事情”的系统上投入更多。

Yeah, exactly. You need to delegate. You need to be able to delegate and walk away and trust that this thing is not going to mess up automatically. And I don't even think we need to build perfect systems. I don't think we need to wait for 100% model alignment. We can rely on the same Swiss cheese model we've used in the industry for a long time. But I do think we need to universally maybe eventually invest more. And that's what we're doing. We need to invest more in systems where we can say you do not need to approve everything.

说到瑞士奶酪模型，他刚写了一篇关于这个的文章。哦，酷。是的。是的。

Speaking of Swiss cheese model, I mean he just wrote a thing about this. Oh, cool. Yeah. Yeah.

是的，所以我们很好。我的意思是，是的。奇怪的是，通常我认为安全和保障对工程师来说是个无聊的词。他们会说：“给我不安全，给我不保障。”但我认为实现正确的事情，就像你追求消费者/专业消费者一样。

Yeah, so we're cool. I mean, yeah. It's weird how usually I think safety and security is kind of a boring word to engineers. They're like, "Just give me unsafety. Give me unsecure." But I think achieving the right thing like you're going after a consumer / prosumer.

是的，是的，有点像两者兼顾。我有点两者兼顾。我想吸引那些像你一样使用 Claude Code 毫无困难的人。是的，是的，是的。但仍然觉得它可能只是方便、更容易。你会说：“哦酷，这就像右边的待办事项列表。我可以编辑它。”这些事情如果你必须做的话，会更容易。

Yeah, yeah, kind of like both. I was kind of like both. I think I also want to capture people who would have no trouble using Claude Code like yourself. Yeah, yeah, yeah. But still find it maybe just convenient, easier. You're like, "Oh cool, that's like the to-do list on the right. I can edit it." Those things are just easier to do if you have to.

是的，但这显然是知识工作方面。Claude Code 显然会捕获开发工作流。但我确实认为你必须为这些安全和保障细节付出努力，才能让人们信任它。就像 Claude 和 Chrome 使用任何 API 来做后台事情一样。

Yeah, but this is clearly the knowledge work side. Claude Code will clearly capture the development workflow. But I do think you have to sweat these safety and security details in order for people to trust it. And like even Claude and Chrome having the whatever API uses to do the background thing.

是的。这是我使用它的唯一原因。因为否则我就得弄一台单独的机器，然后通过它来运行。

Yeah. That's the only reason I use it. It's because otherwise I would have to just get a separate machine and just run it through it.

听起来超级烦人。

Sounds super annoying.

是的，我现在就在这么做。我觉得作为开发者，我们可能更愿意承担风险，但我们也只是接受这一点。我们更愿意承担风险，但我也觉得我们有一种——我不想说是傲慢——但有点像相信如果真出了大问题，我们大概能修好。我只是让 Claude 在做任何不可逆操作前先问我一声，比如发邮件或永久性操作。这样就够了。但不仅仅是 Claude，就连像 npm install 这样简单的事也是如此。我们都在用完全用户权限运行 npm install，如果它想读取 .ssh 目录，它就能读。默认设置就是这样，真是疯狂。

Yeah, I mean, I'm currently doing it. I think as developers, maybe we're more risk tolerant, but we're also just accepting. We're more risk tolerant, but I think we also just have, I don't want to say arrogance, but sort of the trust that if a really bad thing happens, we can probably fix it. I just tell Claude to check with me before doing any irreversible action, like sending an email or doing something permanently. It's good enough. But not even Claude, I mean simple things such as npm install. We're all running npm install with full user permissions, and if it wants to read .ssh, it will. Crazy that that is the default.

有点像什么？嗯，我知道。我同意。我完全同意。我显然每天都在这么做。不是吗？

Kind of what? Yeah, I know. I agree. I agree at the fine. Like I'm obviously doing it every single day. No, right?

而且我觉得 npm 和 GitHub 在过去几个月里也做得不错，清理了环境并推出了更具体的令牌。但总的来说，我认为作为工程师，我们一直更愿意承担风险。如果你稍微自省一下，问自己这是不是我们应该做的，你未必总能得出正确答案。对于模型也是如此。我的方法是，最安全的做法就是什么都不做。我们确实想要能力很强的产品，但尽可能的，我不想问你‘这个脚本你同意吗？’因为我有点相信，一旦它成为你工作流程的一部分，你要么没有能力判断这个 Python 脚本是否安全，要么你根本就不会去读它。

And I think obviously npm and GitHub too have done a pretty good job maybe over the last couple months to clean house and come up with more specific tokens. But generally speaking, I think as engineers we've always been a little bit more risk tolerant. And if you do a little bit of introspection and ask yourself if that's how we should be doing things, you might not always come up with the right answer. And I think for models too. My approach is that the safest thing is to do nothing. We do want products that are quite capable, but to the extent possible, I don't want to ask you, 'Are you okay with the script?' Because I kind of believe that once it starts becoming a part of your workflow, you probably either don't have the skill to understand whether this Python script is safe, or you're not going to read it anyway.

好的。我想我有几个最后的问题。Claude 工作的未来是什么？

Cool. I guess I have a couple parting questions. What's the future of Claude work?

我认为我们还处于非常早期的阶段。我们会继续发布新功能，快速迭代这个产品，也就是说你可以期待每周都会有一个小功能，甚至是大功能。我可能会继续加倍押注在你的电脑上，让你在电脑上更高效，也让 Claude 在你的电脑上更高效。我们开始思考，就像今天讨论的，你的电脑意味着什么？它必须是你面前的那台吗？还是你电脑上的虚拟机？或者是别处的电脑？第三件让我很兴奋的事情是，继续攀登这座山，慢慢地把那些习惯提问并得到答案的用户，逐步教他们放手，让 Claude 接管越来越大的任务，无论是在时间上还是在范围上。我想你可能会看到我们大部分的投资和未来发布都围绕这两点：在电脑上做更多事，以及更独立、更长时间地做事。

I think we're still at such early days. We're going to keep shipping things, keep iterating on this thing pretty quickly, by which I mean you can sort of continue to expect that every single week there's going to be a small new feature, if not a big new feature. I'm going to continue probably to double down on your computer and making you effective on your computer and making Claude effective on your computer. We're starting to grapple, as we talked about today, more with the question of what does your computer mean? Does it have to be the one in front of you? Or a VM on your computer? Or a computer somewhere else? And then the third thing that I'm quite excited about is continuing to go up this hill of slowly taking users who are used to asking questions and getting an answer, to slowly teaching them to step more and more away and let Claude take over bigger and bigger tasks, work both in time as well as in scope. And I think you can probably see most of our investments and future releases work on both of those things: the ability to do more on your computer, and then the ability to do more independently and for longer.

Claude 工作现在支持远程控制吗？还没有，对吧？好问题。

Does remote control work for Claude work yet? No, right? Excellent question.

即将推出。我的意思是，如果你想继续押注电脑，这显然是一个方向。但对我来说，你知道，我们谈论人们今年还没准备好，没有壁垒，一切在加速。对我来说，今年年底我们会做什么不同的事，而年初我们可能甚至没想到？我只是想展望一下，我们瞄准的好用例是什么。比如，对于机器学习科学家，总是‘我想要一个能自动化机器学习的 AI 科学家’。但对于知识工作，我的意思是，我已经能让它注册 Google Cloud 了，这对我来说就是 AGI，因为 Google Cloud 的……但除此之外呢？我不知道。我认为基本上还是你仍然需要告诉它构建你的脚本，对吧？你仍然需要参与。

Coming soon. I mean, that's an obvious thing if you want to keep betting on the computer. But to me, you know, we talk about people not being ready this year, there's no wall, it's accelerating. To me, what will we be doing differently at the end of this year that we maybe weren't even thinking about at the start of this year? I'm just trying to look ahead to what's a good use case that we sort of aim towards. So for example, for the machine learning scientist, it's always 'I want an AI scientist that can automate machine learning.' But for knowledge work, I mean, I can already get it to sign up for Google Cloud, which to me is AGI, because Google Cloud's... but what's beyond that? I don't know. I think it's basically the idea that you still have to tell it to build your script, right? You were still kind of involved.

是的。

Yes.

也许对你来说感觉有点神奇，但对我这个构建产品的人来说，它仍然感觉有点笨重。我看到太多流程，我就会想，‘哦，让我帮你省掉这些。’

And maybe a way that felt kind of magical to you, but to me on the other side as the person building this product, it still feels kind of heavy-handed. I see so much process that I'm like, 'Oh, let me take that away from you.'

好的。

Okay.

或者我如何继续往上走，越来越深入技术栈，让你的生活越来越轻松。哦，这里有一个，对吧？

Or how do I just go, I will continuously go further and further up the stack and make your life easier and easier. Oh, here's one, right?

是的。听着，我不在乎自己的隐私什么的，我信任 Anthropic。所以只要观察我日常所做的一切。一天结束时，告诉我你……这叫做可协同工作。是的。

Yeah. Watch, I don't care about my own privacy or whatever, I trust Anthropic. So just watch everything I do on a normal day-to-day basis. At the end of the day, tell me what you... it's called co-workable. Yeah.

你喜欢的，有充分理由，我不喜欢，我从来不喜欢过多透露我在做什么，因为我觉得你应该直接构建并发布它。然后再谈论它。我不太喜欢提前模糊地发帖。但一直让我着迷的是，你们俩今天多次提到一些事情，我就想，‘是的，那显然很明显，好吧，应该有人在研究那些东西。’我认为我们仍然处于这样一个阶段：如果你看看 Claude 工作，我们发布的东西可能不会让你们俩感到太惊讶。你们会说，‘是的，那显然很有价值。显然他们正在研究那些东西。’

You like, for good reason, I don't enjoy, I've never liked to tease too much what I'm working on because I think you should just build it and release it. And then talk about it. I'm not a big fan of vague posting on work ahead of time. But the thing that is always so fascinating to me is that both of you multiple times today have mentioned things and I'm like, 'Yeah, that is obviously very obvious, okay, that someone should be working on those things.' And I think we're still in the space where if you look at Claude work, the things that we are releasing will probably not be a big surprise to either of you. You're going to be like, 'Yeah, obviously that's valuable. Obviously they were working on those things.'

是的，是的。

Yeah, yeah.

而且那显然很好、很有用。我越击中这些点，我们的未来就越符合那个类别，我认为这对我们越有利。因为这样我们就不会构建过于专门化或难以理解的东西。

And obviously that's good and useful. And the more I hit those points, the more our future is fitting into that category, I think the better it is for us. Because then we don't end up building things that are too hyper specialized or too difficult to understand.

是的，我认为过于专门化这一点非常重要。它让你保持通用性，意味着你可能不会想得太狭隘。我不知道该用什么词。

Yeah, I think the hyper specialized thing is very important. It keeps you like general purpose, it means you're not thinking too small maybe. I don't know what the word is.

是的，是的，完全正确。就像整个概念，我们从未发布过，你知道，没有针对使用 React 和 TanStack 等技术的 Node.js 应用的 Claude Code。如果是其他东西，我知道有几家这样的初创公司。我认为这很……我不是 VC，也不是投资者。我很难预测市场走向。但就我感兴趣的构建块而言，Electron 可能是我构建过的最受欢迎的东西。而 Electron 本身非常抽象和通用，对吧？很多应用都在它上面运行。我认为我很难预测到底有多少应用最终会使用 Electron。而更难以预测的是这些应用做什么。

Yeah, yeah, exactly. It's like the whole concept that at no point did we release, you know, there's no Claude Code for Node.js applications that use React and TanStack and all of those things. And if it's anything else, I know several startups like that. I think that's pretty... I'm not a VC, I'm not an investor. It's hard for me to predict where the market goes. But in terms of the building blocks that I'm interested in, Electron is probably by far the most popular thing I ever built. And Electron itself is very abstractable and generalizable, right? So many apps are running in it. And I think it would have been hard for me to predict how many apps actually end up using Electron. And what would have been even less useful for me to predict is what those apps do.

我记得 Bloom 刚出来的时候，我就觉得这很酷。就像你在角落里有个小圆圈摄像头，挺聪明的。那是个 Electron 应用吗？

I do remember Bloom coming out and being like that is cool. Like you're a camera in a little circle in the corner. That is pretty smart. That's an Electron app?

是的，或者至少曾经是。我不确定现在还是不是。它曾经是过一段时间。就像 One Pop 一样。这就是为什么它有这么有趣的东西，对吧？这是技术栈中我很熟悉的一层，每当我给其他工程师建议时，实际上我认为这一层最值得投入，因为这一层的工具并不那么好，但正是在这里你能为未来获得最大的杠杆效应。

Yeah. Or at least was. I'm not sure if it still is. It was for a while. Like one pop. That's why it has so many interesting things, right? It's a level of the stack that I'm quite comfortable with and whenever I give other engineers advice, it's actually that layer that I think is most valuable to invest in because the tools of the layer are not that good, but that's where you get the most leverage for the future in general.

快速岔开一下话题，关于 Electron 的，我一直好奇这个。你看过 Tauri 吗？

Just quick tangent on Electron cuz I always wondered this. Have you looked at Tauri?

我看过，是的。你怎么看？你知道，我的观点是，大多数东西默认应该用 Tauri，除非你真的需要 Electron 的全部能力。但没错，我可以给出我的大观点。为什么我们要在应用里打包整个 Chromium 版本？对吧？为什么我们要这么做？很多人问我这个问题，因为这非常反直觉。使用操作系统自带的 Web 视图不是更容易吗？不用这么做不是更容易吗？答案是肯定的。显然我以前也这么做过。Slack 应用曾经有一个版本只用了操作系统的 Web 视图。

I have, yeah. What's your take? You know, my view is like most things should be Tauri by default unless you really need the full power of Electron. But yeah, I can give my big take. Why do we ship an entire version of Chromium inside the thing, right? Like why do we do that? And people ask me this question a lot because it's very counterintuitive. Wouldn't it be much easier to use the web views that are on the operating system? Wouldn't it be much easier not to have to do that? And the answer is yes. And like obviously I did that once upon a time. There was a version of the Slack app that used just the operating system web views.

是你开始做 Slack 应用的吗？

Did you start the Slack app?

嗯，最终是团队努力的结果，但我在那里，我们一起搭建了这个技术栈。

Well, team effort in the end, but I was there and we built this stack up.

这太疯狂了。我的意思是，显然你让 Electron 的人来做这件事，但……

That's crazy. I mean, obviously you get the electron guy to do it, but...

嗯，这是一个有趣的发现。当我加入 Slack 时，他们已经有一个用当时叫 Mac app 的东西构建的应用，有点像移动端的 App Gap。它只用了操作系统的 Web 视图。但那行不通，原因有很多。然后我们就想，好吧，我们需要更大的武器。我们需要对渲染栈有更多控制。这里我总是提到几件事。我认为如果你在构建一个小应用，只用操作系统的 Web 视图完全没问题。如果你构建的应用用户不多，不会因为出问题而哭天喊地，那也没问题。选择使用自己的嵌入式渲染引擎的原因是——这在 2026 年仍然成立——操作系统的渲染引擎并不那么好。它们就是不够好。微软和苹果都在试图摆脱这一点，但到目前为止他们并没有真正成功。升级这些引擎的唯一方法是升级你的操作系统。所以，如果你是 Slack，并且在 WKWebView 或其他 Web 视图选项中有一个关键的渲染错误，你唯一的办法就是告诉你的客户：“哦，对不起，你太穷了，没有买最新的 MacBook。”这是不可接受的。对用户不可接受，对开发者也不可接受。所以你需要深入技术栈，找到最好的渲染引擎，然后把它放进你的应用里。为什么是 Chromium，即使它很大？Chromium 是目前最好的东西。我经常提醒人们 Unreal Engine。你想渲染一些文字，他们用 Chromium。Chromium 是 Unreal Engine 的一部分，出于同样的目的。Chromium 非常非常好。我认为它是工程学的奇迹之一。很难——从我们现在录制的地方旧金山来说，城里大多数人都是 Web 开发者——我很难夸大其词地说，你能运行像渲染 YouTube 视频、动态协商比特率、处理你极其糟糕的硬件驱动这样的事情是多么神奇。实际上，有个有趣的事情。你可以输入“chrome://gpu”。

Well, this is an interesting find. By the time I joined Slack, they already had an app that was built with something at the time called Mac app that was a little bit like the same app gap thing for mobile. It just used the operating system's web views. And that didn't work for so many reasons. And then we were like, all right, we need bigger guns. We need to take more control of the rendering stack. There's a few things I always mention here. I think if you're building a small app, just going with the operating system's web views is perfectly fine. If you're building an app that maybe doesn't have too many users who would cry bloody murder if it doesn't work, that is fine. The reason to go with your own embedded rendering engine is because—and this is still true in 2026—the operating system rendering engines are not that good. They're just not that good. Both Microsoft and Apple are trying to move away from that. They so far really haven't. The only way to upgrade those is to upgrade your operating system. So, if you're say a Slack and you have a critical rendering bug in WKWebView or some of the other web view options, your only recourse is to tell your customer, 'Oh sorry, you're too poor, you didn't buy the latest MacBook.' Unacceptable. Unacceptable to the user, unacceptable to the developer. So, you sort of need to go down the stack and find the best rendering engine and put it in your app. Why Chromium, even though it's very big? Chromium is by far the best thing. I often like to remind people of the Unreal Engine. You want to render some text, they use Chromium. Chromium is part of the Unreal Engine for the same purposes. Chromium is very, very good. I think it's one of the marvels of engineering. It's very hard—from where it's San Francisco right now where we're recording, most of the people in the city are web developers—it's hard for me to overstate how magical it is that you can run something like rendering a YouTube video, dynamically negotiating a bit rate, figuring out what to do about your extremely broken hardware driver. Actually, this is a fun thing. You can enter 'chrome://gpu'.

好的。

Okay.

如果你往下滚动一点，这些都是启用的变通方案，因为你的电脑上出了问题。如果你在 Windows 电脑上使用一个不是最流行的 GPU，这个列表会很长。所有这些通常只是为了确保，如果作为开发者的我说，我希望这里出现一个红色像素，那么它确实会出现。Chrome 真是个奇迹，因为它能在用户可能给你的所有机器上运行，而且相当可靠。如果不行，他们大概会在 24 小时内修复它。就这样。所以，这就是超级操作系统，对吧？它无处不在。

And if you scroll down a little bit, these are all the enabled workarounds because something is going wrong on your computer. If you're doing this on a Windows computer with a GPU that is not the most popular GPU, it will be much longer. And all of these are usually just there to make sure that if I say as a developer, I want a red pixel to appear here, that that actually happens. Chrome is such a marvel because it works on all the machines that a user might throw at you, and it's going to work fairly reliably. And if it doesn't, they will probably fix it within 24 hours. As is. So, this is the super operating system, right? That works everywhere.

是的。好的。明白了。所以，Electron 的很多魔力实际上就是它让你非常容易地以完全符合你用例的方式打包 Chromium。正是如此。

Yeah. All right. Okay. Yeah. So, a lot of the magic of Electron is honestly just that it makes it very easy for you to ship Chromium in a way that serves you exactly in your use cases. Electron exactly.

我们下一个采访是 Martin Grissom，他有一句话：桌面操作系统只是实际操作系统（即 Chrome）的糟糕实现，而 Chrome 实际上无处不在。这就是你发布应用的平台。

Our next interview was with Martin Grissom who had the phrase like desktop OS's are just poorly poor implementations of the actual OS, which is Chrome, which actually works everywhere. It is this is the platform where you ship apps.

我认为疯狂的是，作为工程师，我们经常假设我们下面的平台层非常稳定。然后你和那些人交谈，他们会说：“是啊，我们也只是在猜测。”

As I think the wild thing is that as engineers, we so often sort of assume that the platform like the layer below us is super stable. And then you talk to those people and they're like, 'Yeah, we're also just like guessing.'

我在 Slack 有一个特别的时刻，我们 Slack 的一个客户是 Nvidia，有一段时间我真的把 GPU 开发者捧得很高。我确实认为他们可能仍然比我聪明得多。但我想：“制造芯片的硬件工程师，然后编写驱动，他们的工作一定比我难得多。他们一定非常优秀。”然后我们 Slack 里有一个 bug：如果你在 Slack 里有一个 YouTube 视频，它渲染得不太对——会有奇怪的伪影。结果那是一个 Chromium 的 bug，最终出现在一个巨大的线程里。所以我看到了很多源代码，他们也只是注释说“我们不知道为什么这很奇怪，但如果你翻转这个位，事情就能工作。”你知道，这在技术栈的每一层都在发生。

I had a distinct moment at Slack where one of our customers at Slack was Nvidia, and for a while, I really put GPU developers on this pedestal in my head. And I do think they're still probably much smarter than I am. But I was like, 'Hardware engineers who built the chips, who then built the drivers, their work must be so much harder than mine. They must be very good.' And we had one bug in Slack where if you had a YouTube video in Slack, it wouldn't quite render right—it would have these weird artifacts. And that ended up being a Chromium bug and ended up on this giant thread. So I got to see a lot of the source code and they also are just like comment to do 'we don't know why this is weird but if you flip this bit things work.' You know, this is just happening at every layer of the stack.

也许年底的 AGI 预测是 Claude 能构建 Chromium。

Maybe the end of year AGI prediction is that Claude can build Chromium.

你现在笑了，但我觉得总有一天它会变得相当不错。你知道，它应该完全没用，大部分只是被 Chromium 仓库里那些高度专业化的工具搞得不知所措。很长一段时间，Chromium 不得不重新发明所有工具，因为没有工具能处理 Chrome。

You see you laugh now but I like you know someday it's starting to get pretty good. You know, you should completely useless mostly just like overwhelmed both with whole hyper specialized tools are inside the Chromium repo. For a long time Chromium had to sort of reinvent all the tools because none of them were capable of handling Chrome.

我想我等待的 AGI 时刻是，什么时候我们会说 Electron 可能不再必要了，因为你可以直接构建完全原生的应用。用 Swift 吗？

I think the AGI moment I'm kind of waiting for is at what point are we going to say Electron is probably no longer necessary because you can just build fully native apps. The Swifty?

是的，不只是 Swift，因为这是一件事，我认为我们当前的模型完全有能力把一个 Electron 应用用 Swift 复制出来。

Yeah, like not just in Swift because this is one thing like it's pretty easy if you I think our current models are quite capable of taking an Electron app and replicating it in Swift.

它们能否构建出性能更好、内存占用更少的应用？它们会进入开发者长期以来所做的超优化领域吗？

Are they going to be capable of building an app that is actually more performant, uses less memory, all of that stuff? Is it going to go into the same hyper optimization that developers have done for a long time?

我们还没到那个程度，即使是我们最好的模型，我也不能指着某个东西说“用原生代码复制它，别出错，试试看”。我们还没到那一步。我不认为这很糟糕。今天我觉得它回来了。是的。好吧。或者我们会让模型思考好几天。这对 Bora 来说是很长的时间。但他确实让模型思考了好几天？

We're not quite there yet where I can point even our best models at a thing and say just replicate this in native code. Make no mistakes or try think, right? We're not quite there yet. I don't just think it's bad. Today I think it's back. Yes. Okay. Or we'll get a neural think for like days. Which is a pretty long time for Bora. But he worked on neural think for days?

是的。为什么？这只是个前端。里面还有更多东西。嗯，好吧。

Yeah. Why? It's just a front. A little more goes into it. Yeah, okay.

我还有一个关于 co-work 的问题。如果我有我的 Claude co-work，多人模式是什么样的？我认为子智能体就像是单人游戏分割上下文。是的。而多人 co-work 就像我的同事在他们的机器上有某个文件我想了解，或者我想知道他们的任务进展如何，然后更新我的东西。这有趣吗？这对你们来说有意义去构建吗？

Another question I had is like co-works. So if I have my Claude co-work, what's kind of like the multiplayer mode? I think sub-agents is like single player split up the context. Yeah. And the multiplayer co-work is like my colleague has some file on their machine that I want to know about or I want to know how their task is going to then update my thing. Like is that interesting? Is that something that makes sense for you to build or for like...

这对我来说非常有趣。这几乎回到了某些脚手架的问题上，我在想，好吧，我们最终会构建出最终会消失的脚手架吗？一年半后的问题就是，在什么时候我们直接给这些东西分配它们自己的 Gmail 账户，给它们自己的 Slack 账号，然后它们就会像我们人类一样使用相同的工具来互相交流。你提到了我们的财务人员。他们一直在努力做非常好的办公集成。我想有一段时间我们构建了很多技术，让 Claude 在 Google 文档中留下有用的评论，现在它直接就能做到。就像在你的 Google 文档中留下评论，然后你就这样与它互动。也许类似的情况，我仍然对什么是最好的交互模式有疑问。是我们为 Claude 智能体构建一个超级自定义的通信方式？还是直接跳到终点，说如果你在工作中使用 Slack，我们就给这个东西一个 Slack 账号，这就是它实现多人能力的方式。它们互相通信。是的。你知道，作为一个有趣的项目，我构建了一个叫 PiQueue 的东西，它基本上获取任何仓库和 Pi 智能体编码智能体，把它放在一个 VPS 上，然后有一个公共的 webhook，任何人都可以提交编码任务。然后有一个仪表盘，你可以在上面审查任务，然后你 Pi p i q...

It's super interesting to me. It almost goes back to some of the scaffolding where I'm like, okay, are we going to end up building scaffolding that will just go away? And like a question and a half year is at what point do we just assign these things like their own Gmail account and we just give them their own Slack handle and then they will just use the same tools we humans use to interact with each other. You mentioned our finance people. They've been working pretty hard on very good office integrations. And I think for a while we built like we built so much tech around Claude leaving useful comments inside a Google Doc and now just does it. Just like leaves a comment in your Google Doc and that's how you interact with it. Maybe like the similar thing where I still have open questions around what is the best interaction mode? Is it for us to build something super custom for Claude agents to talk to each other? Or is it okay, let's just jump straight to the finish line and say we'll we're just going to give this thing if you use Slack at work, we're just going to give this thing a Slack handle and that's going to be the way it's like multiplayer capable. They communicate with each other. Yeah. Like you know, as a fun project I built this thing called PiQueue which basically takes any repo and the Pi agent coding agent, it puts it in a VPS and then there's a public webhook where anybody can submit a coding task. And then there's a dashboard in which you review the task and then you Pi p i q...

是的，你基本上有了所有这些任务。任何人都可以提交任务。对我来说，这几乎就像未来的组织，销售人员与工程团队交谈，工程团队与市场团队、产品团队交谈。所有这些 co-worker 会以某种方式排队决策，让别人批准。是的。你知道，我很好奇那会是什么样子，以及我如何让我的 co-work 既能批准任务而不需要问我，又如何决定哪些需要我审查。是的。因为对于某些事情，比如你想改变颜色之类的，那是一种品牌决策。或者另一个是，嘿，你的东西坏了，这是修复方法。而 Claude 实际上可以审查那个提示是否符合它试图做的事情。今天一切都还像是单人游戏中的多人模式，你知道，我猜它们不多。但如何让多个人利用他们特定的上下文互相传递东西呢？

Yeah, you basically got all these tasks. Anybody can submit a task. And to me it's almost like in the organization of the future, it's like the sales people are talking to the engineering team that is talking to the marketing team, to the product team. And all these co-workers going to like queue up decisions for other people to approve in a way. Yeah. You know, and I'm kind of curious what that looks like and like how do you how do I give my co-work the ability to both approve tasks without asking me. Yeah. And how to decide which one I need to review. Yeah. You know, because for some of these things is like, you know, you want to change the color or something, that's kind of like a branding decision. Or another one is like, hey, your thing is just broken. It's like this is how you fix it. And Claude can actually review whether or not that prompt matches what it's trying to do. Today everything is still very it's like multiplayer within the single player, you know, I guess they have not many of them. But like how do I get multiple people to hand off to each other things using their particular context?

是的，而且让你的两个 co-worker 互相交流，对吧？

Yeah, and for both of your co-workers to like talk to each other, right?

对。是的，嘿，我们今天有一集。你能……你知道，或者……是的，这就像……我知道我们时间不多了，但我们之前讨论过共享技能，我有个问题：如果你的 co-work 问其他 co-work 他们是否有完成这个任务的技能。它们中任何一个都能做到吗？好吧，所以技能转移。是的，这也许又……

Right. Yeah, hey, we got an episode today. Can you like have you, you know, or Yeah, this is like a I know we're like running out of time here, but like we we previously talked about sharing skills and I could have this question of like what if your co-work like ask the other co-works if they have a skill for this task. Does any of these could do it, right? Like okay, so skill transfer. Yeah, like and again this maybe a...

这也许又回到了构建非常强大的东西和构建令人毛骨悚然的东西常常并存的领域，因为从我的工程师同事的反应中我能看出，这可能不是我们要做的，但比如我们有低功耗蓝牙，对吧？这台电脑可以判断出它就在另一台电脑旁边。所以你们可能在做同一件事。嗯，你会在 co-work 中看到这个吗？可能不会。但是，我认为有一些我们还没有尝试过的非常有创意的解决方案。

This maybe goes back into the territory of like building something very powerful and building something creepy often goes hand in hand because I could tell from the reaction that my fellow engineers had that this is probably not what we're going to do, but like we have Bluetooth LE, right? Like I this computer can figure out that it's sitting right next to this computer. So you're probably working on the same thing. Um will you see that in co-work? Probably not. But um there's like I think really creative solutions to problems that we really haven't tried yet.

是的，是的，是的，是的。太好了。我想最后一件事是，Anthropic Labs 我一直有一个模型实验室与智能体实验室的心理模型，这基本上是 Anthropic 内部的智能体实验室，Claude Code 现在就在这个部门下，对吧？它是整个组织的一部分。我的意思是，人员是如此可互换，对吧？好吧，这只是……我不知道这有多真实。不，这是一个真实的团队。这是一个真实的团队。最后一个团队主要是在研究你还没有公开看到的东西。他们在尝试一些非常疯狂、看起来不太可能实现的想法。疯狂科学家，但你是正式属于这个部门吗？

Yeah, yeah, yeah, yeah. Excellent. I guess the last thing is that Anthropic Labs I always have this mental model of a model lab versus agent lab and this is basically Anthropic's internal agent lab which Claude code is now under, right? It's part of the whole org. I mean people are so fungible, right? Like okay, this is just I don't know how I don't know how real this is. No, it's a real team. It's a real team. The last team is primarily working though on things that you don't see in public yet. They're trying like really wild out there ideas that seem quite improbable. The mad scientist but you you're are you officially under this thing or...

不，我们……Claude Code 现在是一个相当大的团队，我实际上不知道我们有多少人，我记得昨天参加我们的每周 co-work 会议时，我心想，哇，这里人真多。但我们仍然有一个 labs 团队，而且我们实际上把 labs 团队扩大了很多。Mike 刚刚以个人贡献者的身份加入了 labs 团队，我认为这非常酷也非常有趣。但他们正在研究你还没有看到的东西，这些想法非常超前，而且可能有一半是坏的，对吧？labs 团队的理念是，它应该只研究那些对其他人来说毫无意义的东西。

No, we're so code is now Claude Code is like a fairly big group where I actually don't know how many people we have like I remember yesterday coming into our weekly co-work meeting and I was like whoa, this is a lot of people here. But we still have a labs team and we actually made the labs team a lot bigger. Mike just joined the labs team as an IC which I think is very cool and very fun. But they're working on things that you have not seen yet that are extremely out there and probably half broken, right? Like the sort of the idea of a labs team is that it should only work on things that make really no sense for anyone else to work on.

好的，我们期待那里有令人兴奋的东西。但非常感谢你。我知道我们时间到了，但我很感谢你加入我们。我很欣赏 Claude Cowork。大家都去用吧。这是我今年感觉最接近 AGI 的东西。你这么说真是太好了。非常感谢。是的，谢谢你的时间。是的。

Okay, well we're looking for exciting things from there. But thank you so much. I know we're out of time but I appreciate your joining us. I appreciate Claude Cowork. Everyone go use it. It is the closest I've felt to AGI this year. That's so nice of you to say. Thank you very much. Yeah, thank you for your time. Yeah.