打造 Muse：自 ChatGPT 以来增长最快的 AI 产品——Meta 的 Alexandr Wang

给我讲讲 Muse 的历史。

Tell me the history of Muse.

首先要说的，可能是我们创办 Meta Super Intelligence Labs 时的愿景。我们创办时，Mark 发表了一份备忘录，是我、Nat Friedman 和其他一些人跟他密切合作写的，叫《个人超级智能》。整个想法就是：我们如何让 AI，如何构建强大的 AI，真正让我们所有人的生活变得更好？我们如何构建某种东西，从广义上真正提升人类体验？在那份备忘录里，我们谈到了很多想法，基本上暗示了 Muse 最终成为的很多东西，但我们当时只是在非常非常早地指向它。那是 2025 年 6 月，在智能体以任何有意义的方式出现之前。2025 年夏天，AI 还处于完全不同的阶段。但那就是我们想要构建的方向。那是 MSL 的使命，也是我们想要把一切带向的地方。

The first thing to start with is probably what our vision was when we started Meta Super Intelligence Labs. When we started it, Mark published this memo that myself, Nat Friedman, and a bunch of others worked together with him closely on, called Personal Super Intelligence. The whole idea was: how do we make AI, how do we build powerful AI that actually makes all of our lives better? How do we build something that actually lifts the human experience, broadly speaking? In that memo, we spoke to a bunch of ideas and alluded to a lot of what Muse ultimately became, but we were gesturing at it very, very early. This was in June 2025, before agents had happened in any meaningful way. In the summer of 2025, AI was just in a totally different spot. But that's what we wanted to build towards. That was the mission of MSL, and that was where we wanted to take everything.

然后我们就开始着手构建模型，因为开发任何这些前沿模型的端到端过程都要很多很多个月。你必须构建整个技术栈，生产数据集，预训练模型，然后做后训练，还要为此构建一整套技术栈。这是一个非常漫长的端到端过程。而今年年初的这整条生产线——Opus 4.5 显然让很多人，包括我们，看到了智能体的未来可能是什么样，真正把智能体的潜力推到了前台。OpenClaw 在今年年初出现。和我密切合作的 Nat Friedman，我想是 MSL 内部第一个接触 OpenClaw 并尝试 OpenClaw 的人。他的体验，我会形容为介于恐惧和狂喜之间。我觉得他真的全身心投入了。他完全信任。他讲过一些这样的故事——他在一次 Stripe Sessions 采访中谈过一些——他告诉他的 OpenClaw 他想多喝水，然后他的 OpenClaw 会通过监控录像看他，确保他在喝水，并告诉他做得好。他有很多很多这样的疯狂故事。这也让你意识到，哦，这里发生的事情相当神奇。

Then we just got to work building models, because the end-to-end process of developing any of these frontier models is many, many, many months. You have to build your entire stack, produce the data sets, pre-train the model, then do post-training, and build a whole stack for that. It's a very lengthy end-to-end process. And this whole production line at the start of this year—Opus 4.5 obviously really caused many people, including us, to see what the future of agents could be, and really brought the potential of agents to the foreground. OpenClaw was happening at the start of the year. Nat Friedman, who I work closely with, was I think the first person within MSL to work with OpenClaw and try OpenClaw. He had an experience I would describe as somewhere between terrifying and euphoric. I think he really went all in. He trusted entirely. He's told some of these stories—he had a Stripe Sessions interview where he talked about some of this—where he told his OpenClaw that he wanted to drink more water, and then his OpenClaw would watch him in security footage to make sure he was drinking water and tell him good job. He had many, many crazy stories like this. It also makes you realize, oh, there's something quite magical about what's happening here.

我记得 Nat 告诉我的一件事——这也是我记得最清楚的事情之一——他说：“我觉得这里真的有些东西。”我不知道，早期采用者 Nat Friedman 是个相当好的指标，因为他当年是 Stripe 最早的客户之一，然后很早就参与了很多开源开发者的事情，他还帮助创建了 GitHub Copilot 和许多其他东西。所以那是一个明确的信号。就像，好吧，个人智能体从广义上讲显然有些相当神奇的东西。但当你真正信任个人智能体，真正把它带入你的生活，它就有能力做这么多事情。听完所有这些疯狂的故事后，我开始非常深入地使用它。我想有好几周，我们俩都是每个空闲时刻都在跟我们的 OpenClaw 对话，真正沉浸在个人智能体可能是什么之中。

And I remember one of the things that Nat told me—and this is one of the things I remember the most—is he said, 'I think there really is something here.' And I don't know, early adopter Nat Friedman is a pretty good indicator, because he was one of the first customers of Stripe back in the day, and then was really early on a lot of open source developer stuff, and he also helped create GitHub Copilot and many other things. So that was a clear signal. It's like, okay, there's clearly something pretty magical about personal agents generally speaking. But when you actually trust the personal agent and really bring it into your life, it is the ability to do so many things. After hearing all of these crazy stories, I started using it very deeply. I think for many weeks, both of us were sort of like every spare moment we would just spend talking to our OpenClaws and really immersing ourselves in what personal agents could be.

你最初的反应是什么？狂喜还是恐惧？

What was your initial reaction? Euphoria or terror?

Nat 当时构建了一个提示词，基本上是对你自己做一次完整的心理分析。有多轮深度的心理探查。他基本上写了这个提示词，让你用在 OpenClaw 上。我想那是我用自己的 OpenClaw 做的第一批事情之一。那大概相当于三年的心理治疗，全部压缩到这一刻。

Nat at the time had built this prompt to do basically a full psychoanalysis of yourself. There were multiple waves of deep psychological probing. He basically wrote this prompt for you to use an OpenClaw. And I think that was one of the very first things I did with my own OpenClaw. It was maybe the equivalent of three years of therapy, all compressed into this one moment.

等等，我很好奇，为什么你第一次拿到 OpenClaw，这是你最先做的事情之一。

Wait, I'm curious why, when you get your first shot at OpenClaw, this is one of the first things you do.

你知道，我下载了它。我安装了它。然后 Nat 说：“哦，你应该做这个。”我当时——我就在他旁边。他说：“好吧，我就做你让我做的事。”然后——

You know, I downloaded it. I installed it. And then Nat was like, 'Oh, you should do this.' I was like—I was next to him. He was like, 'Okay, I'll just do the thing you're telling me to do.' And then—

好吧。

Okay.

然后，你知道，就是这种疯狂的体验。

And then, you know, this crazy experience.

等等，再多说说那个疯狂的体验。

Wait, say more about the crazy experience though.

哦，嗯，就是非常强烈。我的意思是，又像三年的心理治疗，因为它就是一轮又一轮深度的心理探查。它有我其他的信息。它可以访问我的电子邮件，访问我的一堆照片，它还能去查找其他信息，把这一切拼凑起来。在那个过程中我非常脆弱。

Oh, well, it was just very intense. I mean, it was like, again, it was like three years of therapy, because it was just successive rounds of deep psychological probing. It had other information about me. It had access to my email and access to a bunch of my photos, and it could go and look up other information and sort of pull it all together. I was very vulnerable through that process.

这是二月份的事。我想我们俩都得出了这个共同的结论：这显然——这里显然正在发生一些非常特别的事情。我们当时都写了备忘录。我写了一份关于——这两份我们都通过 Meta 董事会提交了，因为不知为何那正好是董事会季。我写了一份备忘录，谈我觉得这在很多方面是最终的消费产品，而超级智能体就消费产品而言，在很多方面是一个明确的终点。然后他写了一份叫——他实际上写了两份，其中有非常令人难忘的句子：“条款、法律，信任是必须的”，这最终定义了我们构建这些产品的很多核心。所以我们得出了这个结论。这是二月份的事。这里有些非常重要的东西。

This was in February. Both of us, I think, had come to this joint conclusion that this is clearly—there's clearly something really special happening here. And we both wrote memos at that time. I wrote a memo about—and both of these we kind of sent up through the Meta board, because for whatever reason it's board season. I wrote a memo about how I felt that in many ways this was the final consumer product, and that the super agent was something that, as far as consumer products went, was a clear endpoint in many ways. And then he wrote one called—he wrote actually two, which had very memorable lines: 'the clause, the law, and trust is a must,' which ultimately defined a lot of the core of how we built these products. So we came to this conclusion. This was in February. There was something very important here.

然后我们组建了团队，大概在一周、也许两周内构建了一个原型。并向董事会做了演示。这是一个非常快的周期。最终 Muse 产品中的许多决策都出现在第一个原型里。Jollybot 当时就在。Muse Charm 的早期版本在那第一次——那第一次会议上就有了。不是所有的想法、不是所有的标签页都在，但很多都在。然后从那里开始，显然那是二月份，我们在七个月后发布了产品。那是一个非常非常艰难的过程：把这个原型——它有时能工作、很神奇，但大多数时候不工作、会不断崩溃——变成高度可靠的东西，像一个精心打造的消费产品，让我们有信心作为对消费世界的贡献。那是一个非常漫长且相当痛苦的过程。

And then we built our teams, we built a prototype within like a week, maybe two weeks. And demoed it to the board. It was this very fast cycle. And many of the decisions that made their way into the final Muse product were present in the first prototype. The Jollybot was there. The early version of the Muse Charm was in that first—that very first meeting. Not all of the ideas and not all the tabs were there, but a lot of them were present. Then from there, obviously that was February, we launched the product seven months later. It was the really, really hard process of taking this prototype, where it sometimes worked and was magical but most of the time didn't work and would constantly break, and going from there to something that would be highly reliable, like a beautifully crafted consumer product that we felt confident as a contribution to the consumer world. And that was this very lengthy and quite painful process.

在聊到那之前，我想回到你写原型这件事上。你们大概用一到两周做出了原型，然后拿给董事会看。马克在这其中扮演了什么角色？

And before we get there, I want to go back to the idea that you wrote the prototype. You guys created the prototype in like a one to two weeks. You show it to the board. What's Mark's role in all this?

他和我们所有人一样，体验到了这个产品带来的那种狂喜与恐惧。我想他在各种采访里都谈过，他把它用在了家里的一堆事情上，和孩子们一起用，还用在综合格斗训练里。他让它看自己训练综合格斗的视频，然后它会给他改进建议。还有一些员工会议，顺便说一句，这后来成了 Meta 团队很多人的一种习惯，就是开员工会议时问，你和你的智能体有过什么疯狂的经历。再说一次，这是二月份的事了。回头想这整个周期非常有意思，因为 OpenClaw 现象显然达到顶峰后就消失了。是的，它没有持续下去，我想很多人都有这种体验：你会有一些神奇时刻，但大多数时候它并不那么好用，有各种各样的问题，然后就慢慢沉寂了。所以我认为，构建这个产品，很大一部分重点就是，如何把它变成一个真正可靠的消费级体验，让你放心地把它呈现在数十亿人面前，并且有信心其中大多数人都会有很棒的体验。

He was like experiencing the sort of like euphoria/ terror of the product at the same time as all of us. I think he's talked about this in various interviews, but like I think he he used it for a bunch of things around the home and he was using it with his kids and he used it in his um uh in his MMA training. Like he had it watch videos of his of himself doing MMA training and like it would give him tips on like how to improve. there were like some staff meetings and by the way like you know this kind of became a thing for for much of the the meta team um was like having these having a staff meeting and saying like what were the what were the crazy experiences that you had with um with with your agent again this was back in February. It's very interesting to kind of um think back to this whole cycle because obviously the openclaw phenomena it peaked and then disappeared. Yeah, there were, you know, it didn't it didn't persist and and I think a lot of people kind of had this experience which was it was just like you'd have these magic moments but for the most part didn't like work that well and there were all these problems and then you know it just kind of like died down and so part of I think the building the product a lot of the focus of the product was just like how do you make this into an actual reliable consumer experience that you feel comfortable putting in front of over time billions of people and you can feel confident that like most of those people have an amazing experience.

为什么你们能做到，而 OpenClaw 的开发者们做不到？

Why were you able to do that and the developers open call were not?

首先，我们极其尊重 Peter Seinberger 以及所有为 OpenClaw 做出贡献的人。我认为 Peter 绝对是个有远见的人。他那颗漂亮的脑袋里发生的、创造出 OpenClaw 的一切，是神奇的。而让 Muse 如此神奇的，或许是不那么光鲜的部分，就是真正地死磕、打磨所有细节，让模型加产品的整体体验变得非常出色，我们在模型本身上做了很多很多次迭代。我们把整体的模型路线图瞄准了构建个人超级智能，宽泛地说，但肯定是个人智能体，就像——

Well, first of all, I think we respect Peter Seinberger and everyone who's contributed to OpenClaw immensely. Like I think Peter is definitely a visionary. I think whatever happened in his beautiful brain that created OpenClaw is magical. And I think maybe the unglamorous part of what made made Muse so magical was just really grinding out and sanding out all of the details to make the combined model plus product experience really amazing and we did many many iterations on the model itself. We aimed the overall model roadmap towards building towards personal super intelligence broadly speaking but certainly personal agents as like

再多讲讲这个，以及它如何让你和其他竞争对手区分开来。

Say more about that and how that differentiates between your other competitors.

好。这个，这有点——

Yeah. Well, this is this is kind of

这其实真的很有意思。

That's actually really interesting.

是的。回到那个时刻其实很有用，因为如果你回到今年年初，二月份，Opus 和 Claude 在编程上遥遥领先，远远超过其他任何人，而且正在爆发式增长。那是整个行业被 Anthropic 接管的开端，我认为总的来说那是一个非常可怕的时刻，就是，哇，他们在编程上领先这么多，然后显然他们有了 Mythos，他们有着明显的领先优势。我认为行业里的传统智慧是，编程智能体是唯一重要的形态，或者也许是唯一重要的技术，或者唯一重要的能力，对于先进 AI 而言。行业里有一股巨大的力量，就是去涌入那个用例，在编程智能体上与 Anthropic 竞争。但因为我们自己的经历和信念，我们形成了围绕个人智能体的想法。退一步说，我认为我们当初做个人超级智能的原因是，我们也相信这是 Meta 拥有独特定位去做的事，我们可以多聊聊这个，但我觉得正因为如此——

Yeah. It's actually useful to go back to the moment because if you go back to the start of the year, February, Opus and Claude were way ahead in coding, way ahead of anyone else and they were exploding. It was the beginning of the sort of like you know Anthropic takeover in the whole industry and it was a very scary moment I think generally speaking which is like oh wow this is you know they're so ahead on coding and and then obviously they had mythos and you know they were they had like this clear lead. I think the the conventional wisdom in the industry was that coding agents were the only form factor that mattered or maybe the only technology that mattered or only the only capability that mattered for advanced AI. you know there's this incredible force in the industry to just like go and crowd that use case and compete with Anthropic on coding agents because of our experiences and our conviction that I think formed around this idea of personal agents and the sort of taking a step back I think the reason why we did personal super intelligence in the first place was that it was also what we believed meta was uniquely positioned to do and we can talk more about that but like I think because of that

好，那我们现在就聊聊这个。

Well, let's talk about that now.

好，我们就这么交织着聊。我想，甚至在 Scale 和 Meta 的交易发生之前，我最早告诉马克的事情之一就是，我认为如果我们有了 AGI 或 ASI，或者不管你想叫它什么，Meta 其实是一家非常特别的公司。因为 AI 的承诺的一部分是，对我们所有人来说，我们想做的事和必须做的事是混合在一起的。而 AGI 的承诺是，我们将几乎不花时间在必须做的事上，因为 AI 会开始替我们做很多这样的事，我们可以把所有时间花在想做的事上。对我来说，Meta 的产品代表的就是我们所有人都想做的事。比如我用 Instagram、WhatsApp 或 Facebook，这些产品代表的是我追随自己的兴趣、爱好，追随我喜欢的朋友和我关心的人。Meta 的产品跨度囊括了我们想花时间去做的那类事情。我认为从长远来看，Meta 会是这一切中的巨大赢家。这是我最初相信的关于 Meta 和超级智能的事情之一。最终，很多这些想法都融入了个人超级智能这个概念。你可以从几个角度看。一个是公司的传承是什么？Meta 的传承是把人们与朋友和家人连接起来，帮助他们发现自己的兴趣。这些都是非常人性的事。然后你也可以从商业角度看，Meta 拥有每天 35 亿人使用其应用的触达。而 Meadow 自然会取得成功的领域是消费级 AI。

Okay, we'll just do do the weave. I think one of the first things I told Mark even before the whole scale and meta deal happened and and everything, one of the first things I told him was I think Meta is actually a really special company if we have AGI or ASI or whatever you want to call it. Because part of the promise of AI is that for all of us, we have a mix of things that we want to do and things that we have to do. And the promise of AGI is that we're going to spend almost none of the time doing things that we have to do because the AIS will just start doing a lot of that for us and we can spend all of our time on things that we want to do. And for me, Meta's products represent the things that we all want to do. Like my use of Instagram or WhatsApp or Facebook like these are products that represent what I, you know, I follow my interests. I follow my hobbies. I follow all people that I like my friends and the people I care about like the product span of meta encapsulates the sort of like what we want to spend our times doing sort of realm. I think on a long arc meta is an incredible winner through all this. That's like one of the first things that I believed about meta and super intelligence. And then ultimately a lot of those ideas made its way into this concept of personal super intelligence. You know, you can kind of look at it a few ways. One is what's the heritage of the company? The heritage of meta is connecting people with their friends and family. Um, helping them discover their interests. Like these are things that are very human. And then there's also the sort of like you can take the business lens which is Meta has distribution to three and a half billion people using the apps every day. And the natural area where Meadow was going to be successful is consumer AI.

不是工作，不是编程，是个人化的。

It's not work. It's not coding. It's personal.

对，没错。所以个人智能体、个人 AI 和消费级 AI，宽泛地说，一直是我们觉得 Meta 能够做出非常特别的东西的领域。这就把我们带到了，好，2026 年初。现在你有了原型，你有了支持，所有人都知道马克加入了，然后你开始了七个月的艰难工作，打磨这个产品。

Yeah. Exactly. So personal agents and personal AI and and consumer AI, broadly speaking, were always the zones where we always felt that Meta was going to do some be able to do something very special. So that brings us to uh okay, early 2026. Now you have the prototype, you have the buy in, everybody knows Mark's on board and now you're doing the seven months of the difficult work of refining this product.

是的。

Yes.

有一种阴影，就是，哦，个人智能体，它们来了又走了，我们一直觉得那是因为体验不够完美，还没打磨好，诸如此类。但我们知道，在开发这个产品时，这是我们必须考虑的事情。我们聚焦了模型开发的路线图。最核心的焦点，或者说最核心的焦点之一，就是构建个人智能体。我们有长长的电子表格，列着数百甚至数千种不同的——不是数千，而是超过一百种具体的行为和能力，我们知道模型必须具备这些才能构建出理想的个人智能体。非常非常细致。

There was kind of the spectre which was like oh personal agents like they came and went and we always felt that it was because you know the the experience wasn't perfect. It wasn't polished yet you know all that kind of stuff. But like we knew that that was something that we had to like think about as you know as we were developing this product. We focused our model development road map. The top focus uh or one of the top focuses was towards building on personal agents and we had like these long spreadsheets of hundreds or even thousands of different not thousands but but more than a hundred like specific behaviors and capabilities that we knew the model needed to have to be able to build the ideal personal agent. Like it was very very meticulous.

然后随着时间推移，我们为所有这些方面建立了评估。我们弄清楚了如何训练模型，让它在每一项上都表现得非常非常出色。我们还有这些评审，我们会查看那个电子表格，找出我们做得不好的地方，并确保我们不断改进。我们做了这样一个非常细致的过程，慢慢地不断改进，打磨掉模型的所有棱角，以便它能驱动产品。

And then over time we built evals for all of those. We figured out how to train the model to be really really good at every single one of those things. And we have these reviews where we like look at that spreadsheet and we figure out the things that we're bad at and make sure that we're improving on them. Like we did this like very meticulous process to slowly just keep improving and sanding out all the edges of the model so that it could power the product.

这也是 Meta 的文化吗？这种不断打磨产品的文化？

Is that really the culture of Meta too? This like constant like sanding of the products?

是的。Meta 的很多文化是，一旦你知道你在衡量什么，那里有非常优秀的人才，你就可以针对这些进行优化，你可以把所有细节都处理好，你可以让这些东西变得很棒。我认为在 AI 之前的消费产品时代，很多都是关于如何优化转化率？如何优化推荐率？如何优化引导流程的每一个部分？如何让所有这些都变得完美？然后我认为在 AI 时代，尤其是 Muse，很多都是以打磨模型的每一个细节，使其完美适配产品的形式出现的。

Yeah. A lot of the culture of Meta is like once you know what you're measuring, there's very talented people and you can optimize for those things, you can send out all the details, you can make those things amazing. I think in the sort of like pre-AI consumer product era, like a lot of this was around how do you optimize your conversion? How do you optimize the referral rates? How do you optimize like every part of the onboarding flow? Like how do you make all that stuff just perfect? And then I think the AI era I think uh a lot of that especially from Muse has been in the form of sanding out every detail of the model to be perfect for the product.

我想向你介绍本播客的呈现赞助商 Ramp。我最近读了很多关于 SpaceX 的内容。SpaceX 是世界上最有价值的公司之一，SpaceX 历史上的一个主要主题就是不断攻击和质疑你的成本。Ramp 帮助世界上许多最具创新性的公司做到这一点。使用 Ramp 的公司的支出中位数减少了 5%。SpaceX 展示的一点是，对控制成本的虔诚 dedication 实际上可以帮助增加收入，因为你可以追求原本无法追求的机会。我们在 RAMP 数据中也看到了这一点。使用 Ramp 的公司的收入中位数也增长了 16%。所以，当你在 Ramp 上运营业务而你的竞争对手没有时，你就拥有巨大的竞争优势，这种优势会随着时间的推移而复合。Ramp 是唯一一个旨在让你的财务团队更快、更快乐的平台。我认识的许多顶级创始人和 CEO 都在 RAMP 上运营他们的业务。我在 RAMP 上运营我的业务，你也应该这样做。访问 ramp.com 了解它如何帮助你的业务节省时间、节省金钱并增加收入。那就是 ramp.com。

I want to tell you about the presenting sponsor of this podcast ramp. I have been reading a lot about SpaceX lately. SpaceX is one of the most valuable businesses in the world and one of the main themes in the history of SpaceX is constantly attacking and questioning your cost. Ramp helps many of the most innovative businesses in the world do exactly that. The median company running on ramp cuts their expenses by 5%. And one thing SpaceX has demonstrated is that a religious dedication to controlling costs can help actually increase revenue because you can pursue opportunities you couldn't otherwise. And we see that in the RAMP data, too. The median company running on ramp also grows their revenue by 16%. So, when you're running your business on ramp and your competitors are not, you have a massive competitive advantage that compounds over time. Ramp is the only platform designed to make your finance team faster and happier. Many of the top founders and CEOs I know run their business on RAMP. I run my business on RAMP and you should too. Go to ramp.com to learn how it can help your business save time, save money, and grow revenue. That is ramp.com.

Deal 是最优秀的创始人将世界变成他们的人才库的方式。我花了十年时间研究历史上最伟大的创始人如何运作。他们都有一个共同点，那就是他们明白招募和雇佣最优秀的人才是你最重要的优先事项。A 级玩家能识别其他 A 级玩家，这就是为什么像 Ramp、Shopify、11 Labs、Uber 和 Door Dash 这样的顶级公司都使用 Deal。我认识的许多顶级创始人在使用 Deal 的产品后都亲自投资了 Deal。他们发现 Deal 是世界上最好的公司，擅长构建全球招聘的基础设施。Deal 将帮助你的企业在世界任何地方雇佣、支付和管理任何工人。这样你就可以在任何地方留住最优秀的人才，并把剩下的时间花在你最擅长的事情上，为客户提供价值。11 Labs 的创始人对 Deal 能为你的公司带来的价值有一个很好的描述。他说：“我们建立 11 Labs 是为了打破语言和沟通障碍。有了 Deal，我们能够在任何地方雇佣和支持杰出人才，我们可以加速创新，将更多的声音、故事和想法带到世界的每一个角落。Deal 受到超过 40,000 家公司的信任，并且发展迅速。访问 deal.comra 了解他们如何帮助你的业务。那就是 deal.com/enra。”

Deal is how the best founders turn the world into their talent pool. I've been studying how history's greatest founders operate for a decade. And one thing they all have in common is they understand that recruiting and hiring the very best talent is your most important priority. A players recognize other A players, which is why top companies like Ramp, Shopify, 11 Labs, Uber, and Door Dash all use deal. Many of the top founders I know have personally invested in Deal after using their product. And what they discovered is that Deal is the best company in the world at building infrastructure for global hiring. Deal will help your business hire, pay, and manage any worker anywhere in the world. So you can retain the best talent anywhere and spend the rest of your time focusing on what you do best, delivering value to your customers. The founder of 11 Labs has a great description of the value Deal can give your company. He said, "We built 11 Labs to break down language and communication barriers. With deal enabling us to hire and support exceptional talent anywhere, we can accelerate our innovation and bring more voices, stories, and ideas to every corner of the world. Deal is trusted by over 40,000 companies and growing fast. Learn how they can help your business by going to deal.comra. That is deal.com/enra.

你看到 Jeremy Stern 在 Colossus 上写的关于扎克伯格的这篇人物特写了吗？

Did you see this profile that's written on Zuck in Colossus written by Jeremy Stern?

我看到了。说实话，它太长了。我没读。但是

I saw it. If I'm being honest, it was like too long. Didn't read. But

它有 15,000 字。我刚请 Jeremy 上了播客。他是我最喜欢的。

it's 15,000 words. I just had Jeremy on the podcast. He's he's my favorite.

我看到了，我看到了。他只是在——我认为他现在写的是世界上关于企业家和投资者的最好的特写。我认为他在这方面确实是世界上最好的。但在那篇特写中，扎克说，他在谈论他相对于我认为的埃隆和萨姆·奥尔特曼等人的优势，他说，听着，我的技能组合非常不性感，但我非常擅长组建团队，然后只是在一个很长的时期内慢慢改进产品。相对于埃隆和萨姆·奥尔特曼等人，我不会失去对公司的控制，我也不需要再筹集资金，所以我可以这样做很长一段时间。当你谈到这 7 个月的尝试提炼和构建 Muse 时，特写中的那部分在我脑海中浮现。

I saw I saw I saw that. He's he's just I think he's writing the best profiles on on entrepreneurs and investors in the world right now. I think he's literally the best in the world at what he does. But in that Zach says he was talking about the advantage that he has relative to I think Elon and like Sam Alman and he's just like listen my skill set is very unsexy but I'm really good at building teams and then just improving a product slowly over a long period of time and relative to Elon and and Sam Alman and others. like I'm not going to lose control of my company and I don't need to raise any more money and so I can just do this for an excessively long period of time. That section of the profile is popping in my mind when you're talking about this 7-month uh stretch of trying to refine and build muse.

我认为这完全正确。我觉得这需要极大的耐心。我的意思是，7 个月，你知道，也许听起来没那么长，但 AI 是激烈的。在 AI 时间里，7 个月。

I think that's exactly right. Like I think that it took immense patience. I mean 7 months like you know maybe it doesn't sound like that long but AI is like fierce. seven months in AI time.

在 AI 时间里，7 个月，可能就像非 AI 时间里的十年。我认为这对我们所有人来说都是如此。但是，我真的非常赞赏马克。这需要大量的耐心和克制，以及对过程的极大信任。你可以回去看看马克在从 2 月到 9 月整个时期的财报电话会议上所说的话，他会暗示我们正在构建个人智能体，我们并没有真正试图隐藏它，就像我们在构建个人智能体，我们在谈论它，但我认为你知道，在整个时期，公司内部有很多怀疑，就像人们，显然，华尔街的主要叙事之一是，哦天哪，Meta 只是在烧钱。他们甚至能在 AI 方面取得任何成就吗？

Seven months in AI time, which maybe like a decade and like non-AI time. And I think it it for all of us. Um but but I really give Mark a lot of credit for this. Like it took a lot of patience and restraint and a lot of trust in the process. And like you'll you can go back and look at the you know what Mark said in earnings calls for this entire period from February through through um September and he will allude to the fact that we were building personal agents and we weren't really trying to hide it like we were we were building personal agents and we were talking about it but I think you know there's a lot of doubt in the company for that entire period like people obviously the one of the dominant Wall Street narratives is like oh my god Gosh, Meta is just burning all this money. Like, are they even going to pull anything off in AI?

不止如此。就像扎克无法赢得 AI。这个说法在很长一段时间里被反复提及。再次回到 Jeremy Stern 关于他的特写。然后他非正式地开始与他的竞争对手交谈，包括一位前研究员，现在在竞争对手那里，曾经为扎克工作，现在在另一个实验室。他说：“我不想不得不与他们竞争，因为马克就像——我想那句话是马克是终结者。”就像他就是不会停止。然后他谈到他的一个超能力，就是他从不满足于产品的现状。所以他总是想要不断证明，但他不是一个——就像他不是你知道的独裁者，他绝对是一个独裁者，但就像他不是你知道的粗鲁，并且会破坏团队的化学反应。他实际上在文章中谈到，团队凝聚力对他来说极其重要。

It's more than that. It's like Zuck cannot win AI. It's said over and over again for a long period of time. And again, going back to Jeremy Stern's profile on him. He then off the record starts talking to his competitors, including one former researcher that is now at a competitor that used to work for with Zuck and is now at another lab. And he goes, "I don't want to have to compete with them cuz Mark's like the I think the line is Mark is the Terminator." And it was like he just does will not stop. And he's like uh then he talks about um one of his superpowers that he's never happy with the state of the product. So he always wants to constant prove but he's not an Like he's not you know dict he's he is definitely a dictator but like he's not you know rude and like kind of will destroy the chemistry of a team. He actually talks about in the in the piece that team cohesion is excessively important to him.

我认为这没问题。

I think it's all right.

我认为 Muse 的神奇之处之一在于，我们真的花了相当多的时间打磨所有细节，才最终推出产品。因为我认为个人智能体是那种如履薄冰的产品领域：如果它能可靠地提供这些神奇体验，它就有可能成为有史以来最伟大的消费产品之一。但你面对的是模型，是不确定的东西，是智能体——它们可能崩溃，可能不可靠。如果人们遇到这些不可靠的体验，他们会觉得这是垃圾。所以关键是要把产品做到一定的质量和价值，让它能引起那么多下载和使用它的人的共鸣。

I think one of the magical things about Muse was that we really took quite a bit of time to sweat all the details before ultimately coming out with a product. Because I think that personal agents was one of these product areas that kind of rested on a knife's edge: if it can reliably deliver these magical experiences, it has the potential to be one of the greatest consumer products ever. But you're dealing with models, you're dealing with something nondeterministic, you're dealing with agents—they can break, they can be unreliable. If people have these unreliable experiences, they think it's trash. It was one of these things where getting the product to a point of quality and value such that it resonated with so many of the people who downloaded and used it.

我想多聊聊这个过程。我很好奇投入了多少资源，团队有多大，有多少人在做这个。你们有没有数据，比如我用 Muse 有九次很棒的体验，两次不太稳定或者很糟糕的体验，然后我就流失了？你们有这方面的信息吗？

I want to talk more about this process. I'm curious how much resources, how big the team, how many people working on it. Do you guys have any data on like okay I'm using Muse nine great experiences two like inconsistent or you know shitty experiences and then I turn out like do you have any information on this?

你可以从今年年初很多其他智能体产品上看到这一点——几乎所有人都流失了，就是因为它们不可靠。我们对 Muse 做了大量用户测试。我们做了很多很多轮，让新用户尝试下载并从头开始使用，看看体验如何。Muse 的一个核心前提，我认为也是它真正成功的地方，就是它和大多数人体验过的消费级 AI 不同。它不是聊天机器人。它是一个智能体，可以为你做智能体该做的事。对开发者来说这可能是老生常谈，因为他们早就下载过智能体并用了一段时间。但对大多数人来说，光是这个过渡过程就需要解释，需要非常用心地引导。当然，我们确实看到——因为我们用不同版本的模型做过测试，并持续做 AB 测试——产品体验对模型质量非常非常敏感。它非常依赖于模型能否可靠地完成这些任务，能否可靠地与用户沟通，能否在需要指导时可靠地回访用户。

You can see this from a lot of the other agent products at the start of the year—like almost everyone turned out of those and it's because they were unreliable. We did a lot of user testing of Muse. We did many many rounds of new people trying to download it and use it from scratch and just seeing what that experience was like. A huge part of the premise of Muse and I think what's really landed is that it is a different thing from what most people have experienced with consumer AI. It is not a chatbot. It is an agent and it can do agent things for you. And I think to a developer this is kind of old news because they've had to download agents and have been using them for a while. But for most people, just even that graduation process needs to be explained and needs to be onboarded in a very thoughtful way. And then for sure we definitely have seen—because we've done tests with different versions of the model and various model versions and continuously do those AB tests—the product experience is very very sensitive to model quality. It's very sensitive to how reliably the models can undergo these tasks, how reliably they communicate with the user about them, how reliably they check back with the user if there's things that they require guidance on.

但你怎么知道它已经准备好发布了？

But how'd you know it was ready to be released?

一部分是相当数字化的，比如我们设定——回到那个有 100 多项的表格。我们为每一行设定了什么阈值是阻碍发布的，高于什么阈值是好的。我们有很多检查点，在 80 行上是绿的，但在 20 行上是红的。我们知道那还不够好。然后我们有了第一个全绿的模型检查点，那是一个特别训练的 Musepark 1.3 版本。所以有非常数字化的部分，然后就是体验部分，我们试用了模型，能感觉到它比我们之前迭代过的任何版本都好得多。

Part of it was pretty numerical, which is like we had set—you go back to that spreadsheet of 100 plus things. We had set for each of those rows what threshold is launch blocking and above what threshold is good. We had many checkpoints that were green on 80 of the rows but red on 20 of the rows. We knew that wasn't good enough. And we had our first model checkpoint which was green across the board, which was a specially trained version of Musepark 1.3. So there was the very numerical part of it and then there was just the experiential part which is we tried the model. We could tell that it was a lot better than anything we had iterated with before.

那么回到你提到的那七个月，有多少人在做这个产品，包括模型工作和产品工作？

So going back to that seven-month period you were talking about, how many people were working on the product between the model work and the product work?

总共不到 200 人，听起来可能很多，但在大公司里其实小得多。人并不多。回头看这很酷，我们当时就知道这将是整个高管团队最重要的产品，每个人都知道这可能是 Meta 今年推出的最重要的产品之一，但我们在组织上有克制，确保是一个真正小而专注的团队在做，没有膨胀成让公司大片区域都参与进来，因为我们知道这是一件单一的艺术品。它需要强烈的观点和专注，才能把它打造成一个美好的体验。

Like under 200 people total, which I guess sounds like a lot, but in big company land it's actually a lot smaller. It's not very many. That was kind of cool looking back on this and thinking about it is like we knew this was going to be the most important product that the whole executive team, everyone knew that this was probably going to be one of the most important products that Meta would ship this year, but we had the restraint organizationally I think to make sure that it was a really small focused team working on it and that it didn't balloon to this like oh you have huge swaths of the company working on because we knew it was a single work of art. It required strong point of view and strong focus to pull that through into one beautiful experience.

多说说这个。为什么你称它为一件单一的艺术品，需要单一的视角或观点？

Say more about this. Why did you just call it like a single piece of art and you need a single perspective or point of view?

我认为对来自大型组织的产品的一个批评，而且现在也开始困扰很多 AI 实验室，就是它们变成了弗兰肯斯坦式的、几乎像柯南伯格式的混合体，融合了一堆不同人的观点、愿景和信念，某种程度上像是产品经理的地狱，每个产品经理都想把自己的东西塞进产品里，然后所有这些混合成一杯冰沙。作为用户和消费者你能感觉到——当一个产品感觉像是很多不同的人做的，明显有不同的人各自为政，只是把它们拼凑在一起。Muse 必须完全相反，即有一个清晰的观点。我们试图向世界传递一个连贯的体验。

I think one criticism of products that come from larger organizations and I think this is even now starting to plague a lot of the AI labs is that they become kind of this Frankenstein almost like Cronenberg amalgamation of a bunch of different people's points of view and visions and beliefs and it's sort of like it's sort of PM hell in some sense where every PM has a thing that they're trying to get into the product and jam into the product and then all that kind of blends together into this smoothie almost. People feel it as users and consumers—you can feel when a product is just like it feels like lots of different people worked on it and there were clearly different people who were gold differently so they just kind of jammed it all together. Muse had to be the exact opposite, which is like there's one clear point of view. There is one cohesive experience that we're trying to deliver to the world.

我很喜欢你这么说，而这纯属巧合。我们——我不知道这期什么时候播出，但我们是在史蒂夫·乔布斯逝世 15 周年纪念日录制的。当你想到一个拥有单一观点的消费产品专家时，你显然首先想到史蒂夫。我读过所有关于史蒂夫和苹果历史的书，为我的另一个播客《Founders》做了大概 15 到 20 期关于他们的节目，但我最喜欢的一句描述史蒂夫影响的话是：苹果就是史蒂夫·乔布斯的一万条命。产品符合他的品味、他的视角、他的观点。他想在世界中看到的东西，正是最终用户得到的。在 Muse 的案例中，Nat Friedman 值得很多赞誉。他的品味和感受力最终决定并塑造了那个产品。

I love that you said that and it just happens to be pure coincidence. Like we're—I don't know when this is going to come out, but we're recording this on the 15th anniversary of Steve Jobs' death. And when you think of somebody like a pro consumer products that had a single point of view, you obviously think of Steve first. I've read every single book on Steve and the history of Apple and I've done like I don't know 15 or 20 episodes on them for my other podcast Founders, but one of my favorite lines to describe Steve's impact was it says that Apple is just Steve Jobs with 10,000 lives. The products are to his taste, his perspective, his point of view. What he wanted to see in the world is exactly what the end user got. In Muse's case, Nat Friedman deserves a lot of credit. His taste and sensibilities ultimately dictated and shaped what that product became.

你们会谈论有多少人在用 Muse 吗？你们公开谈论这个吗？

Do you talk about how many people are using Muse? Are you guys talking about this publicly or no?

我们说过数百万，然后我转发了其他人展示的图表，暗示 Muse 是有史以来增长最快的消费级应用。消费级 AI 应用。

We have said millions and then I have retweeted other people who have shown graphs where we would imply that Muse is the fastest growing consumer app of all time. Consumer AI app of all time.

所以你还是不会说具体数字。

So you won't say a number though.

我们说了数百万。是的。

We've said millions. Yeah.

嗯，数百万也可能是数亿。那也是数百万。所以，你只是没加前面的数字。我想深入聊聊——你最近在 X 上有点，我这是褒义地说，有点放飞自我。

Well, millions could be hundreds of millions. That's also millions. So, you're just not adding the beginning to that. I want to go into—you've been kind of going, and I mean this in a loving way, like unhinged on X.

所以，你谈到你的，这像是营销策略吗？到底怎么回事？我感觉自从 Muse 发布以来，你的公开推文有明显的不同。你在做什么？

So, you talk about like your is this like marketing strategy? Like what is going on? Like I feel like you're there's a distinct difference in your public tweeting since Muse launch. What are you doing?

这里发生了几件事。

There's a few pieces of what's happening here.

这肯定比你在 Scale AI 时更有趣。肯定是的。我们稍后再谈这个，但肯定的是，过去几周是我职业生涯中最有趣的构建时光，我认为有几件事在发生。

This has got to be more fun than scale AI when you were building scale. It has to be. We'll talk about that later, but it has to be the past few weeks have been some of my most fun uh building like yeah certainly my career and I think there's a few things happening.

所以，第一点是，我们觉得找到让 Muse 突破的方法很重要。这就像回到那种铺天盖地的更新流。对大多数人来说，他们不断被新的 AI 产品、初创公司、发布、模型轰炸，又一个模型，又一个产品，又一个功能，就像这股洪流。我们实际上开过大会，团队讨论过这个。我们需要突破，显然像 Meta 我们有大量分发渠道，可以确保每个人都知道，但这不足以让它以有趣的文化方式突破，让世界关注。我生活中碰巧遇到并结交了一些 meme 创作者，我亲切地称他们为我的精神分裂朋友，但这些人整天都在思考互联网上的 meme，并理解这一点。我有人痴迷于 Muse，他们也痴迷于你的 X 动态，他们确信你有一百个 Muse 在制作 meme 供使用。嗯，结果你只是有一群永远在线的精神分裂朋友。

So, one is it felt important that we figured out a way to make Muse break through. It's a pretty like torrential going back to this like this torrential stream of updates. Like for most people, there's just so they're just like constantly being like like kind of bombarded with new AI products and startups and releases and things and models and here's another model and here's another product and here's another feature and it's just like this torrential stream. We had actually meeting large meetings where we talked about this as a team. It was like we need to break through and obviously like meta we have lots of distribution like we can make sure everyone knows about it but that's not enough to like make it break through in an interesting cultural or fascinating way for the world to kind of like you know pay attention to. I separately in my life have happened to have met and um become friends with like memers and and uh I have these uh I affectionately call them my skitso friends but but people who like literally they spend all their day like thinking about memes in the internet and and understanding that I have people that are obsessed with Muse and they're also obsessed with your X feed and they're convinced that you have like a hundred muses coming up with memes. to to be used. Um, turns out you just have a collection of schizophrenic friends that are terminally online.

X 动态，顺便说一句，那些推文 literally 都是我写的。有一段时间，我在开会时很难集中注意力，因为我会想到一个 meme，然后想，“天哪，我必须发出去。”那就像灵感时刻。有一次，我正试图和一位同事共进晚餐，然后晚餐期间我突然想通了，然后我发了一个不同的 meme。几乎感觉我又像一个工程师在解决 bug，因为你知道，你能感觉到什么时候会成功。但是的，所以一件事是我们想突破。另一件我认为重要的事是如何让 Meta 再次酷起来？因为不管什么原因，尽管拥有像 Instagram 和 Facebook 这样不可思议的文化软件，Meta 历史上有一段时间并不酷，所以想让 Meta 酷、有趣、有点搞笑，然后那种疯狂是随着时间自然发生的。当 Muse 刚出来时，我主要专注于推广我看到的所有有趣用例，我转发了数十个我看到的人们使用 Muse 的有趣用例。然后某个时候，因为我发推太多，有几条推文我没怎么想就发了。然后其中一些推文表现很好。你见过这个吗，比如花时间思考然后爆款？你见过吗？太好了。

The X feed, just for what it's worth, is like literally every one of those tweets I write. There were like a few a quite a period where I was like, it was like really hard for me to pay attention in meetings because I would just like think of a meme and I'd be like, "Oh god, I have to get this out." And I would like it was like moments of inspiration. There was like um I was uh I was with one of my like I was trying to have like dinner with one of my co-workers and then during the dinner I was like wait I figured it out and then I like posted a different me. It was like it was almost I almost felt like I was like an engineer solving bugs again cuz it was like you know you could tell when something was going to click. But yeah so one things we wanted to break through. Another thing that like I just thought was important is like how do we make meta cool again? cuz like for whatever reason despite having like these incredible pieces of cultural software um like Instagram and and uh Facebook historically meta was not cool for a while and so wanted to make meta cool and interesting and kind of like funny and then the unhingedness was something that happened kind of like naturally over time when Muse first came out I was just like mostly focused on promoting all the interesting use cases that I saw I was like retweeting like tens of interesting use cases that I saw of people using Muse. And then at some point it was like because I was just like tweeting so much. There were like a few tweets where I was just like didn't think that much about it and just said it post. And then some of those tweets like did really well. Have you seen this like time spent thinking about it and then banger you seen that? Great.

我没见过，但我同意。就像我有些推文我没花太多时间。我发了推。我甚至几个小时没看手机，然后我想，“哦，哇。那条表现真好。”然后我记起了互联网的一些东西，这不是超级有意的，但我记得互联网奖励风险，奖励惊喜，奖励人们意想不到的东西。所以我就放开了。我的很多同事，我想有点惊讶，Nat 说了一周，当我发布所有 meme 时。他在一堆会议上说，“你知道，我以为我们会学到。”他说，“我以为我们会学到很多关于 Muse 用户以及人们如何使用 Muse，但实际上我们只是学到了很多关于 Alex 的头脑。”这很有趣。我觉得我利用了多年来的互联网知识来融入那些时刻。我认为它实际上，你知道，我不知道，我们没有跟踪，很难说，但我实际上认为它真的对增长产生了影响，这有点搞笑。

I haven't se but I agree with that. That's like I just some of these I didn't spend that much time on. I like tweeted them. I didn't even check my phone for a few hours and I was like, "Oh, wow. That one went really well." And then I kind of like remembered something about the internet which I it wasn't super intentional, but I remembered like the internet rewards risk and it rewards like surprise and like things that like people don't expect. And so then I just kind of like went off. A lot of my co-workers uh I think were somewhat surprised at uh something that um Nat said for for the whole week when I was posting all the memes. He was like, "You know, I thought we would learn." He said this like in a bunch of meetings. He was like, "I thought we were going to learn a lot about um you know, Muse users and like how people were using Muse, but actually like we're just learning a lot about Alex's mind." It was a lot of fun. I felt like I was drawing on all of my internet knowledge for years and years and years to like pull into those moments. And I think it actually, you know, I don't know, we don't have the tracking and it's like hard to tell, but I actually think it actually genuinely made an impact on grow growth, which is kind of hilarious.

我绝对认为它做到了。我不断听到关于它的消息。而且我听到的所有人也下载了应用，现在正在使用它。不过，回到你刚才说的。就像，嘿，我们有分发，但这并不意味着我们能做出人们会回来并真正粘性的东西。那么，你是如何用 Muse 做到这一点的？

I definitely think it did. I I can't stop hearing about it. So, and all the people I'm hearing about it from also downloaded the app and are now using it. Go back to what you were saying though. It's like, hey, we have distribution, but that doesn't make mean that we can make something that people are going to come back to and actually sticky. So, like, how did you do this with Muse?

我认为是，如何让这成为，我们如何让这作为产品突破？我认为有几个部分，但一个是，如何让人们匹配体验？比如如何让人们尽早对产品产生惊叹时刻？而这个产品我认为自然适合这一点，因为你知道，它是 AI。它能做很多。如果有人从未使用过 AI 智能体，它能做很多会让你惊讶的事情。所以我认为这是很大一部分。我认为另一部分是如何利用所有这些人们有惊叹时刻的故事，并用这些来帮助产品歌唱，让产品飞翔。

I think it was like, how do you make this into how do we how do we make this breakthrough as a product? And I think there were like a few parts of that, but I think one is how do you get people to a match the experience? Like how do you get someone to to a wow moment with the product as early as possible? And this product I think naturally lends itself to that because I think you know there's it's like it's AI. It can do a lot. It can there's like if if someone has never used an AI agent, there's a lot that it can do that will surprise you. So that I think was a big part of it. I think another part of it was like how do you harness all of those stories of people having these wow moments and like use those to help make the product sing and make the product fly.

不过，你是如何使用这些故事的？因为你们没有为我投放广告，对吧？

How are you using the stories though? Cuz you you guys aren't running ads for me, are you?

我们投放了一些广告。我的意思是，老实说，我认为一开始只是我转发了那么多。就像我只是

We ran a few ads. I mean honestly like I think at first it was just the fact that I would retweet so many of them. Like I was just

每当我看到有人做了关于新闻的有趣事情，我就引用推文。我不理解为什么 Anthropic 和 OpenAI 的广告都那么糟糕，因为记得 Anthropic 做了这个大型户外活动，全是关于他们自己，ChatGPT 也一样。就像他们有那个图标，我甚至不认为人们认出那是标志，它只说 CEX，然后两个人坐在桌子旁。我想这到底是什么意思？而你有所有这些用例。你所要做的广告就是展示你产品的益处。仅此而已。

anytime I saw someone do something interesting about news I was like quote tweet. I couldn't understand why both Anthropic and OpenAI's um ads are so bad because remember Anthropic was doing this huge like outdoor campaign and it's all about them and same thing with um Chet. It's like they'd have like the the icon which I don't even think people recognize that as a logo and it just say CEX and just be two people sitting at a desk. I'm like what the does this mean? Where it's like you have all these use cases. All you have to do your ad is to show the benefit of your product. That's it.

你有数百万个。投放那些广告。

You have millions of them. Run those ads.

这道理太基础了，有意思的是人们却一再犯这个错误。

It's so elementary that it's interesting that people make this mistake over and over again.

是啊。

Yeah.

我觉得我们在做 News 以及做产品时反复回到的一点就是：为什么会有人在意？他们不在意你或你的公司，他们在意的是自己。每个人都是自利的。有一本很老的书叫《科学的广告》，写于 20 世纪初。它太有价值了，以至于 20 世纪二三十年代最赚钱的广告公司创始人——这本书的作者叫霍普金斯，他是最顶尖的文案撰稿人。他说，这些就是我赚钱的所有思路，我要把这份手稿锁进保险柜 20 年。但本质上它讲的不是广告，而是人性，而人性是不变的。他的核心观点就是：你的广告讲的是你自己，你的广告讲的是你的产品，你的广告就像在说‘嘿，我有这个东西，买吧’。人们根本不在意这些，他们是自利的。你只需要把‘这东西能为你做什么’这件事做出来，然后直接以此开头。有人采纳了这些写于 20 世纪初的思路，后来建起了庞大的公司、在五六十年代乃至七十年代创立了广告公司，甚至延续至今。

I think one thing that we came back to a lot in both building news as well as just building stuff in general was just like, why would someone care? They don't care about you or your company. They care about themselves. Everybody's self-interested. There's this very old book, it's called Scientific Advertising. It was written in the early 1900s. It was so valuable that the guy that made the most money as an advertising agency founder in the 1920s and 30s — it was written by this guy named Hopkins, who was this best copywriter. He's like, these are all the ideas of how I'm making this money. I'm putting this manuscript in a safe for 20 years. But essentially it's not about advertising, it's about human nature, which doesn't change. And his whole point was just like, your ads are about you, your ads are about your product, your ads are like, hey, I have this thing, buy it. People don't care about that. They're self-interested. And all you have to do is just do the work of what this does for you and you just lead with that. And there are people that took those ideas — those ideas were written in the early 1900s — and went on to build massive companies, advertising agency foundations in the '50s, '60s, '70s, and even to this day.

是的，百分之百同意。

Yeah, 100%.

我觉得有几个洞见。首先，世界上大多数人根本不在乎 OpenAI、不在乎那些研究、不在乎其他公司或它们的研究，甚至不在乎我们或我们的研究。他们在乎的是这东西能为他们做什么。我们有幸拥有那个吉祥物——那个快乐的小家伙。我记得那条让我意识到‘哇，这有搞头’的推文。当时我开车去上班，路上生成了一张他提着公文包的图，然后发了条推文‘周一上班的我’。我记得当时想，这是我近期最好的推文之一。然后我意识到，你可以把他放进很多场景里。在那个过程中我意识到的一点是：你怎么创造出让人意外的时刻，而且和你见过的任何产品营销都截然不同？我真的不记得上一次有人用带点擦边意味的梗图来营销产品是什么时候了。我其实不确定以前有没有人这么干过，但我意识到，哇，这是我们可以用这个小家伙做的事。这很有意思，这能勾起普通人的好奇心。

I think there are a few insights. Let's say first is most people in the world, they don't give a damn about OpenAI or the research or these other companies or the research or even us or our research or whatever. They care about what it can do for them. We had this gift of the mascot, which was the little jolly guy. I remember the tweet where I realized — I was like, oh wow, this is mileage. Because I was basically — I was driving into work and while I was on the way I generated a version of the guy holding a briefcase and I tweeted ‘MFW I go to work on Monday.’ I remember thinking like this is one of my best tweets in a while. And then I kind of realized like you can put him into a lot of situations. One thing I realized as we were going through that process was like, how do you create surprising moments that are very different from any other marketing that you've seen of a product? I don't really remember the last time someone leaned into racy memes to market a product. I don't actually know if it's been done before, but I kind of realized like, oh wow, this is a thing that we can do with this little guy. And that's something that is interesting. That's something that drives intrigue to the average person.

好。但这可能只是提升认知度，甚至让他们下载了 App。可一旦 App 装到他们手机上，你怎么让他们持续使用？

Okay. But that might raise awareness, might even get them to download the app. But then how do you get them to stay engaged once the app is on their phone?

真正喜欢这个 App 的人会经历一个循环——我们在二月份就有过——就是你试着给它一件小事，它做到了。你会说，哇，它真做到了。然后你给它稍微大一点的事，它也做到了，你会说，哇，它能做到。第一件事能行，第二件事也能行。然后你给它一个再大一点的问题。这几乎就像你和 AI 智能体之间的一场‘信任背摔’：一开始你并不太信任它，但你又有点好奇。所以你先给它一点小任务，然后给大一点的，再大一点，再大一点。当然，我跟大学生之类的人聊时，有些人用 AI 的方式就是完全的信任背摔。他们直接录语音备忘、把脑子里所有东西倒给 AI，发过去，然后就信任 AI 去整理所有想法、给出清晰的计划和清晰要做的事等等。我觉得对 Muse 的成功用户来说，关键就是让他们进入这个信任背摔的过程：用它做越来越大、越来越大、越来越大的事，而每一次它都成功。

There's a cycle that people who really love the app — and we even had back in February — which is you try to give it a little thing and it does it. You're like, oh wow, it did that. And then you try to give it something a little bit more and then it does it and you're like, oh wow, it can do that. It could do that first thing. It could do the second thing. And then you give it a little bit bigger of a problem. And it's kind of this trust fall almost that you have with an AI agent, which is at first you don't really trust it, but you're kind of intrigued. So you're going to give it a little nibble and then you give it something a little bigger and something a little bigger and something a little bigger. Certainly when I talk to college students or whatever, some of the ways that people use AI, it's like a full trust fall. They just are recording voice notes and brain dumping to the AI and they send it and then they just trust the AI to organize all the thoughts and give them clear plans and clear things to do, etc. I think for the successful users of Muse it really is like getting them onto this trust fall process of using it for slightly bigger and bigger and bigger things and then having it work every one of those times.

我读《从 0 到 1》时找到了我最喜欢的一句引言。这句话说：‘我注意到的最强大的模式是，成功的人在出人意料的地方发现价值。’他们做到这一点，是靠从第一性原理而非公式去思考商业。这正是 AppLovin 在其广告平台上所做的。AppLovin 让你触达手游中超过十亿的潜在新客户。AppLovin 让你抓住用户全神贯注的注意力。AppLovin 的广告是全屏视频，平均被观看 35 秒。这种留存率让其他广告平台望尘莫及。你可以在几分钟内就在 AppLovin 上投放。你设定目标，AppLovin 帮你实现。无需复杂设置，无需专业知识。而且 AppLovin 能快速扩张。他们可以把你的广告展示在超过十亿潜在客户面前。其他企业已经看到了立竿见影的效果，日投放额扩展到数十万美元，营收增加了数百万。所以，趁你的竞争对手还没上 AppLovin，赶紧开始吧。你可以访问 applovin.com。就是 applovin.com。

I found one of my all-time favorite quotes when I was reading the book Zero to One. The quote says, ‘The single most powerful pattern I have noticed is that successful people find value in unexpected places.’ And they do this by thinking about business from first principles instead of formulas. That is exactly what AppLovin has done with their advertising platform. AppLovin connects you with over a billion potential new customers in mobile games. AppLovin allows you to capture undivided attention. AppLovin ads are full-screen videos that are watched for an average of 35 seconds. That is retention that blows other ad platforms out of the water. And you can launch on AppLovin in minutes. You set the goal and AppLovin achieves it. No complex setup, no expertise needed. And AppLovin scales quickly. They can put your ads in front of over a billion potential customers. Other businesses have seen immediate results, scale to hundreds of thousands of dollars of spend per day, and increase their revenue by millions. So you want to get started quickly before all of your competitors are on AppLovin. And you can do that by going to applovin.com. That's applovin.com.

我想知道这其中有多少其实是靠口碑驱动的。显然，因为要跟你聊，我无论如何都得下载这个 App。但真正勾起我兴趣的，是我一个朋友发给我他用 Muse 的截图。你下载这个 App 时，它其实没怎么告诉你你能做什么，它挺开放的，就像一个空的文本框。但我有个朋友，经营着一家大概千亿美元的公司——他把身份证落在了 Flexjet 上，然后他就……他跟我说，我完全不想参与这个过程，你得想办法拿到身份证，去取身份证，它在这儿，我在这儿，然后自己想办法搞定中间所有步骤，再安排快递，快递到了那儿，不知怎么过了他办公室的安保，上了楼到了他办公室。我当时就想，天哪——这是最好的营销。就三四张 Muse 来回对话的截图，他说‘我不想参与，你得搞定一切’，然后它一路自己想办法。

I wonder how much of this is actually driven by word of mouth. So obviously talking to you, I was going to have to download the app no matter what. But what really piqued my interest is a friend of mine was sending me screenshots of how he was using Muse. So when you download the app, it doesn't really tell you too much of what you can do. It was kind of open-ended, like an open text box. But a friend of mine who's running like a hundred billion dollar company — he left his ID on a Flexjet and then he went through and he's just like, well, I want to — he told me, I don't want to be involved in this process at all, you have to figure out how to get the ID, go pick up the ID, this is where it's at and then this is where I'm at, and then figure out all the steps along the way, and then arrange the courier, went there, got to his security in his office somehow, and got upstairs to his office. And it was like, holy — this is the best marketing. It's just three or four screenshots of Muse going back and forth and him saying, ‘I don't want to be involved. You have to do everything.’ And it figuring out on the way.

是的。这其实正是其中一点——我不知道这是不是有点刻意，但现代产品营销绝对是靠截图驱动的。我觉得要做一个现代消费级产品，它必须在截图上成立。可以说，截图必须能‘飞起来’，要么通过群聊，要么通过口碑，要么在网上传播。

Yeah. This is actually one of the things about — I don't know how it was, somewhat intentional, but definitely modern product marketing is very screenshot driven. I think to have a modern consumer product, it has to work on the screenshot. The screenshot has to fly, so to speak, either through group chats or word of mouth or online.

我认为这是很重要的一部分。这并非有意设计，但我认为截图中出现那个小家伙是件大事，因为把 Jolly 或你的灵感来源放在截图里，立刻就能表明这是 Muse 的截图。这是你能拥有的最好的营销。那么，你整天都在用你的 Muse 吗？你个人的使用情况是怎样的？

And I think that's been a big part of it. This wasn't intentional or engineered, but I think the fact that we had the little guy in the screenshot was a big deal because having Jolly or whoever your muse is in the screenshot immediately communicates this is a Muse screenshot. It's the best marketing you could possibly have. So, are you on your Muse all day long? What is your own personal usage?

我主要用它设置了很多工作流程，然后我猜大部分时间是我的会议，但实际上，在我所谓的工作中，我应该集中注意力、保持在场并与人们互动。我设置了一堆工作流程，把它当作第二大脑。

I mostly set up a lot of workflows using it and then I guess my meetings a lot of the time but so in practice I think in my job quote-unquote I'm supposed to pay attention and be pretty present and interact with people. I set up a bunch of workflows and treated it as a second brain.

是的。你在你写的那篇帖子里说过，它很短，我记得上次看到它已经有 550 万次浏览了，但你说，嗯，我喜欢你描述 Muse 的方式，就像如果每个人都有一个超越自身的第二大脑。我认为从哲学上讲，这是最有趣的问题之一：当你拥有强大的 AI 时，作为人类意味着什么？这种关系是怎样的？它究竟会是什么样子？不同的人对它的样子有不同的看法。我想那种恐惧的情况，显然你知道 AI 在我们之上，我们只是在做它们做的事，还有其他世界，那种关系是不同的。但在你继续之前，你个人的观点是什么？

Yeah. You said that in the post that you wrote and it was really short and I think I saw last time it was like 5 and a half million views so far but you said uh I love how you describe Muse like well if everyone had a second mind beyond their own. I think this is philosophically speaking one of the most interesting questions about like what does it mean to be human when you have like powerful AI and what is that relationship like and and how exactly does that look and different people have different takes on what exactly that looks like. I think the sort of like fear cases obviously that you know AI is above us and that we're just sort of like doing what they do and there's like other worlds where that relationship is different. And before you go on, what is your own personal view though?

我真的相信 Muse。我相信个人智能体作为一种长期形态，因为拥有欲望、渴望和梦想是人性。我认为 AI 和像 Muse 这样的个人智能体将成为机制和桥梁，使我们能够实现这些目标并继续实现它们。然后人类会梦想得越来越大。孩子们会梦想拥有自己的星系之类的。Muse 会帮助他们做到这一点。贝索斯曾写过，我想是在他的一封股东信里，他说消费者的美妙之处在于他们总是美妙地不满足，或者类似的话，或者对他们拥有的选择美妙地不满意。这是我们与 AI 关系的一种模式。我认为还有其他模式，比如 AI 将开始越来越多地参与科学发现过程。AI 将开始越来越多地插入经济的各个部分，但我确实相信这种长期形态，即每个人都拥有一个支持他们及其愿望的 AI。

I really believe in Muse. I believe in the personal agent as a long-term form factor of just it is human nature to have wants and desires and dreams. And I think AI and personal agents like Muse will be the mechanism and the bridge that enables us to accomplish those things and continue accomplishing them. And then humans will dream bigger and bigger and bigger and bigger. Kids will dream of having their own star systems or whatever. And Muse will help them do that. There's this thing that Bezos wrote in, well I think it was one of his shareholder letters, which is like the beautiful thing about consumers is that they're always beautifully unhappy or something like that, or beautifully unsatisfied with the options they have. That is one modality of our relationship with AI. I think that there's other modalities like I think that AI will start doing more and more of the scientific discovery process. I think AIs will start inserting more and more into various parts of the economy, but I do believe in this form factor long term of everyone having an AI that supports them and what they want.

所以你认为 AI 能消除生活中的所有累赘，对吧？那些我们不想做的事情。我记得你在文章里说过，世界充满了守门人和障碍，Muse 可以帮你绕过这些，而不必花费你更多时间，这样你就能真正专注于你想做的事情。我认为 Muse 可以给每个人他们童年梦想的成年生活。每个人小时候都有宏大的想法和梦想，当他们想象自己的生活并演绎出来时，他们想象自己能成为宇航员，或者拯救地球，或者改变世界，或者随便什么。他们想得很大，然后由于各种原因，当人们完成学业、进入职场、工作一段时间后，所有的希望、雄心和梦想的能力都被抽走了。

So you see AI as kind of like removing all the cludge of life, right? The stuff that we don't want to do. I think in the piece you said something like the world is full of gatekeepers and obstacles and like Muse can get around this for you without spending any more of your time so you can actually focus on the stuff you want to do. I think Muse can give everyone the adulthood they dreamed of in childhood. Everyone when they're kids, they have big ideas and big dreams and when they imagine their life and they sort of play it out. They imagine this life where they can be an astronaut or they can save the planet or they can change the world or, you know, whatever it might be. It's like a big they think big and then for, you know, a variety of reasons by the time people finish school, they enter the workforce, by the time they get to their, you know, have been in a job for a while, it's that like all that hope and ambition and ability to dream has been sucked out of them.

不，是被打出来的。生活被打出来了。我认为作为成年人，我们并不欣赏我们都在多大程度上是僵尸。我们失去了能动性，失去了——我认为雄心是一个很好的词。广义上讲，AI 的承诺之一，我认为通过像 Muse 这样的产品，就是让你从孩提时代就保持这种状态。就像你是个孩子，你有大梦想，使用 AI，帮助实现这些。然后梦想更大，实现它们，梦想更大，实现它们。并体验这种能动性的自动扶梯，一生中能动性不断增加，而不是像你有很多能动性然后被压碎。你知道我刚才听你说话时在想什么吗？你之前提到你和 Nat Freeman 经历的那次令人不安的心理审计，像来自 AI 的波浪一样。就像，如果我们有那种，但与你说的相反。它实际上给了你更多自信，让你理解世界是可塑的，如果你足够用力推动它，用足够的能量和动力去追求它，你实际上可以改变周围的世界。

No, it's been beaten out of them. The life is beaten out of them. I think as adults, I don't think we appreciate the degree to which we're all zombies. We lost our agency and we lost our, I think ambition is really a good word for it. One of the promises of AI broadly speaking and I think through products like Muse is to keep it going from when you're a kid. Like you're a kid, you have big dreams, use AI, help make those things happen. Then dream bigger, make those things happen, dream bigger, make those things happen. And kind of experience this escalator of agency, an increase in agency throughout your life versus a sort of like you have lots of agency and then it gets crushed. You know what I was thinking of when you were speaking just now? You mentioned earlier this experience that you and Nat Freeman went through where you had this disturbing psychological audit and like coming in waves from this AI. It's like yeah, what if we had that but it's like the opposite of what you're saying. It's like actually gave you more self-confidence like gave you understanding like the world is malleable and if you push on it hard enough you go after it with enough energy and drive you can actually change it around you.

是的。我认为这是最令人印象深刻的企业家所展现的，他们梦想远大，实现它，然后梦想更大，也许为此工作十年，但如果实现了，他们就会梦想更大。埃隆显然是一个很好的例子。马克也是一个很好的例子。我认为很多标志性企业家都有这样的亲身经历，我认为应该有一个版本对每个人都适用。我认为有点悲剧的是，对大多数人来说，你有更大的梦想、宏大的想法，进入企业工作，你有点变成僵尸，也许某个时候你想去追求梦想，但那时真的很难，因为你有各种承诺。也许你有家庭，也许你有这样那样的事情。你几乎被困在僵尸状态中。在我想谈 Scale AI 以及你与 Meta 的合作和做出那个决定之前，关于 Muse 还有一件事。但在你写的这篇文章或短帖中，我喜欢你说的：“迄今为止，世界是由少数狂热者塑造的，他们以某种方式找到了实现自己愿望的方法，但我们从未见过人类每一个人的能动性都被完全开启。”我认为我所描述的这个世界的承诺，即每个人一生中能动性不断增加，有能力实现自己的愿望和梦想。我认为那个世界以一种非常好的方式看起来很疯狂。我觉得它会非常有趣。

Yeah. I think this is something that certainly the most impressive entrepreneurs exhibit which is like I think they like dream big they accomplish that then they dream bigger and then maybe they work on that for like a decade but then like you know if they accomplish that then they dream bigger. Like Elon is obviously a great example of this. I think Mark is a great example of this. I think but I think a lot of like the iconic entrepreneurs this is like kind of like their lived experience and I think there's a version of that that should be true for every person and I think what's kind of like tragic is I think for most people you have like bigger dreams big ideas you enter the corporate workforce you're you kind of become a zombie maybe at some point you want to go tackle your dreams but then it's like really really hard cuz you all these commitments. You maybe you have a family, maybe you have whatever it is. It's just like it's almost you become trapped in the sort of like zombiehood. There's one more thing on Muse before I want to get to Scale AI and like the partnership that you do that you did with Meta and making that decision. But in this essay that you wrote or this like short postage you wrote, I love what you said. said, "The world until now has been shaped by the small number of fanatics who have somehow found a way to make their wants real, but we've never seen humanity with every single person's agency fully switched on." I think the promise of this world that I'm describing where every single person has this like agency increase throughout their lifetime and they have the ability to accomplish their wants and their dreams. I think that world looks crazy in a very good way. Like I think it will be very interesting.

它会充满艺术感、很酷，你会进入世界的不同角落，非常多元。很难想象那具体会是什么样子：数十亿人突然因为智能的极大丰富，拥有了让不可思议的事情发生的资源。我认为这是承诺的一部分。在我脑海里，这几乎就像《瑞克和莫蒂》里的跨维度电视，但不知怎么地变成了人类的现实。那可能会非常棒。所以我一直听到那部剧里的一句话，大意是“宇宙会吞噬聪明人”。

It'll be artistic and cool, and you'll go into different pockets of the world, and it will be very diverse. It'll be kind of insane to think about what exactly that looks like, where literally billions of people have, all of a sudden, because of abundant intelligence, the resources to make incredible things happen. I think that's part of the promise. In my head, it's almost like the Rick and Morty interdimensional cable, but somehow manifested into reality for humans. It could just be really awesome. So I keep hearing this line from that show where it's like, the universe eats smart people.

好，我想回溯一下。我个人好奇的是：你创立了 Scale AI，你在经营 Scale AI，然后有一天马克·扎克伯格联系你说，“嘿，”我猜他说的是，“嘿，我想聊聊。你能跟我讲讲这个吗？”

Okay, so I want to go back. What I'm personally curious about is: you founded Scale AI, you're running Scale AI, and then one day Mark Zuckerberg reaches out and says, 'Hey,' I assume he says, 'Hey, I want to talk. Can you tell me about this?'

是的，我想那条消息大概是：“嘿，你有时间通个电话吗？”

Yeah, I think the exact message was something like, 'Hey, do you have time for a call?'

你之前和马克有交情吗？你和他有过接触吗？背景是怎样的？

Did you have a relationship with Mark previously? Did you spend any time with him? What was the background there?

我有一个朋友，叫亚历克斯·舒尔茨，我认识他很多很多年了，那时 Scale 可能才成立一年。我在旧金山认识他。他目前是 Meta 的首席数据官，但曾是 Meta 的长期高管亲信。他把我介绍给马克，我想大概是在 2021 年或 2022 年，为了聊 AI，因为那时马克对 AI 越来越感兴趣。Scale 在 AI 方面做了很多事。我记得安排那第一次会面花了大约一年。从介绍到会议安排，我想整整一年。

I had a friend, this guy Alex Schultz, who I've known for many, many years, from when Scale was maybe a year old. I met him in San Francisco. He is currently the chief data officer at Meta, but was a longtime Meta executive lieutenant, and he introduced me to Mark, I want to say in about 2021 or 2022, to talk about AI, because at that time Mark was getting a lot more interested in AI. Scale was doing a lot of stuff in AI. My memory is that it took like 1 year to schedule that first meeting. From the intro to the meeting being scheduled, I think it was like a full year.

为什么？

Why?

马克显然日程排得极满，有无数事情要处理。他也非常擅长排优先级。他会花大量时间在真正重要的事情上——某种程度上，在真正重要的事情上花的时间过多。他能做到这一点，同时还能当个好爸爸、好伴侣等等，唯一的方法就是无情地排优先级。在那之后我和他聊过几次，作为创始人得到他的建议——能得到马克·扎克伯格的建议显然是件大事。我 actually 记得那时他会写——我在 WhatsApp 上问他一个问题，然后他会写很长的回复，我当时很困惑。我想，他怎么有时间写这么长的回复？但现在我意识到，他只是写大量 WhatsApp 消息时又快又严谨。他能很快写出非常严谨的长回复。这是他一项非常令人印象深刻的技能。

Mark obviously has an insane calendar and has a bajillion different things to deal with. And he's also very good at prioritization. He will spend a lot of time on the things that are really, really important—in some ways, too much time on the things that are really, really important. The only way he's able to do that and be a good dad and a good partner and all this other stuff is to just ruthlessly prioritize. I had spoken to him a few times after that and had gotten his advice as a founder—to get advice from Mark Zuckerberg is obviously a really big deal. I actually remember at that time he wrote these—I would ask him a question on WhatsApp and then he would write these really long responses and I was really confused. I was like, how does he have time to write these really long responses? But now I realize he is just really fast and rigorous at writing a lot of WhatsApp messages. He'll write a very rigorous, long response very quickly. It's a very impressive skill of his.

让你窥见他脑子里在想什么。

Gives you a sneak peek into what's going on in his mind.

是的。Meta 是 Scale 的重要客户。我想说在这件事之前，我们大概每六个月聊一次。也许那就是我们互动的频率。然后是的，我想他打了电话，一开始其实并不清楚最终会走向什么，因为第一次对话就像，你知道，那是在 Llama 4 之后，基本上就是问，你觉得我们应该做什么？

Yeah. Meta was an important customer of Scale. I would say prior to this instance, maybe we spoke once every six months or so. Maybe that was the cadence of our interactions. And then yeah, I think he had called and I think it was actually not obvious at first what the end point was going to be, because the first conversation was just like, you know, this was after Llama 4 and it was kind of just asking, what do you think we should be doing?

为什么这次对话发生在 Llama 4 之后是一个重要的点？

Why is that an important point that this conversation is happening after Llama 4?

Llama 4 没有走在 Meta 作为一家公司想要的轨迹上。内部来说，它肯定令人失望，而且当时 AI 显然正变得越来越重要。很明显，AI 对 Meta 的未来至关重要。所以是的，最初几次对话我们通电话，他有点像是在征求建议：你觉得我们应该做什么？你觉得我们应该聚焦什么？这是我私下听说的关于他的事：他有一个疯狂的——不是董事会，不是 Meta 董事会，虽然董事会里有些人他也会这样——但他身边有一群世界级的企业家，我和其中一些人聊过，他们说，你不明白，他不断地征求建议。他会找我们，说，这是我生活中或工作上正在发生的事，你会怎么做？他会一遍又一遍地这样做。我认为这是一个相当令人钦佩的特质，因为显然需要谦逊，才能在取得了所有这些不可思议的成就之后还继续这样做。我是说，从他第一次联系到我们宣布交易，整个时间线大概五、六周。相当快。

Llama 4 was not on the trajectory that Meta wanted as a company. It definitely was, internally speaking, a disappointment, and it was also at a time where AI was clearly becoming increasingly important. It was clear that AI was going to be really critical to the future of Meta. So yeah, the first few conversations we spoke on the phone and he was sort of asking for advice on, what do you think we should be doing? What do you think we should be focused on? This is what I've heard about him privately: that he has this insane—it's not a board of directors, like a Meta board, although there are some people on the board that he also does this with—but he's got this group of world-class entrepreneurs around him, and I've spoken to some of them, like, you don't understand, he asks for advice constantly. He's hitting us up, like, here's what's going on in my life or the work, what would you do? And he'll do that over and over and over again. I think this is quite an impressive trait, because it takes humility, obviously, to continue doing that even after all the incredible things he's been able to accomplish. I mean, the whole timeline from when he first reached out to when we announced the deal was like five, six weeks. It was pretty quick.

世界级企业家的另一个共同点。我们请过 Groq 的创始人乔纳森·罗斯上这个播客。他第一次公开谈论他与英伟达达成的 200 亿美元交易就是在本节目上。他说，从黄仁勋第一次打电话到钱到我银行账户，只用了 3 周。

Another thing that world-class entrepreneurs have in common. We had Jonathan Ross, the founder of Groq, on this podcast. The first time we ever spoke publicly about the $20 billion deal he did with Nvidia was on the show. And he's like, from the time of the first call from Jensen to money's in my bank account was 3 weeks.

是的。太厉害了。我有很多想法，关于如果我选择的话我会做什么，以及我会如何思考 Meta 的 AI 战略。我最早说的话之一就是我之前说的：嘿，我 actually 认为，因为 Meta 在传承上专注于人们想做的事，它实际上是这次 AI 巨大转变中处于最特殊位置的公司之一。我有很多很多想法发给了他，然后那场对话与这个并行发生：哦对，也许我们应该考虑一下是否有办法合作。

Yeah. That's amazing. I just had lots of ideas on what I'd be doing if I choose and how I'd be thinking about Meta's AI strategy. One of the first things I said was what I said earlier, which is, hey, I actually think because Meta is, heritage-wise, focused on the things that people want to do, it is actually one of the most specially placed companies with this incredible shift in AI. I had lots and lots of ideas that I sent his way, and then that was kind of this conversation happening in parallel with this, oh yeah, maybe we should potentially consider if there's a way to work together.

所以当他说那个的时候，你怎么想？因为这小子——你——这对你来说不可预测。

So when he says that, what do you think? Because this kid—you—this was not predictable to you.

不可预测。不。整件事在当时是一连串非常疯狂的经过。我当时就想，哦，那挺好，但绝不可能——

Not predictable. No. The whole thing was a really crazy sequence at the time. I was just like, oh, that's nice, but there's no way that—

嗯，你知道他们这么说意味着什么，对吧？他们想收购你。总是同一套，但他们说成，哦，我们得找到一种合作的方式。

Well, you know what that means when they say that, right? They want to buy you. It's always the same thing, but they say it's like, oh, we got to find a way to work together.

是的。我想我当时没有受过那种指导。

Yeah. I don't think I'd had that kind of coaching at the time.

当时我明确感觉到的一件事是，嗯，这甚至可能不太合理。我个人当时想，也许他只是在开玩笑或者随便说说，但这真的合理吗？

One thing that I explicitly felt at the time was like, yeah, it probably doesn't make that much sense even. I personally was like, maybe he's teasing it or throwing it out there, but does he even make that much sense?

为什么你会觉得它不合理？

Why wouldn't you think it makes sense?

我的意思是，显然最终交易达成了，所以我知道什么呢？但是

I mean, obviously a deal ended up happening, so what do I know? But

不，不，但当时，不是我们现在知道发生了什么，但我很好奇你当时的想法，因为你说过，哦，你说这好像不合理。为什么它不合理？

No, no, but back then, not we know what happened, but I'm very curious your thinking as you're going experiencing this like the fact that you said, oh, yeah, you said this was like that doesn't make sense. Like why wouldn't it make sense?

退一步说。显然，Meta 的标志性收购是 Instagram 和 WhatsApp，这些都是非常明确的产品逻辑驱动的决策。Scale 是深度企业政府销售，与 Meta 完全不同的业务。所以当时我认为合理的逻辑是，如果 Meta 想进入这些领域，那么也许那是合理的产业逻辑，但这次不是，因为我认为马克主要感兴趣的是获得 Llama 项目和整体 AI 项目

Taking a step back. Obviously the iconic Meta acquisitions have been Instagram and WhatsApp and these were very clear product logic driven decisions. Scale is like a deep enterprise government sales, an entirely different kind of business from Meta. And so the logic that I had thought at the time would have made sense is if Meta wanted to get into those things, then maybe that's the industrial logic that would have made sense, but this wasn't that because I think Mark clearly was predominantly interested in getting the Llama program and his overall AI program on

他想要人才。

He wanted talent.

是的，他想要人才。整个过程在很多方面都有些模糊不清。

Yeah he wanted talent. It was kind of shrouded and unclear the whole time, a lot of the way.

什么时候变得清晰？

When did it become clear?

当钱到账后，我们开始谈数字时，我心想，哦，哇。好的。好吧。那时事情变得更清晰了。

After the money hit the account when we started talking numbers, I was like, oh, wow. Okay. All right. That's when things became more clear.

起初，你觉得，哦，这会有道理。然后你们继续谈。他多快说服你这是一条好路？

At first, you're like, oh, this will make sense. Then you guys keep talking. And how fast did he convince you like this is a good path for you?

其中一部分是，哇，这真是一个迷人的交易结构。我们最终达成了这笔交易，Meta 投资，拥有 Scale 49% 的股份，Scale 继续运营。我和几个人加入 Meta。当时我有点难以置信，因为这是一笔如此奇怪的交易，但我认为这真的是一个非常有趣的三赢局面。Scale 的股东赢了，因为所有股东都得到了很好的交易，受益匪浅。Scale 继续运营，我真心相信 Scale 最好的日子还在后头。

There was a part of it was just like, wow, this is like a really fascinating deal construct. We ultimately landed on this deal where Meta invested, owns 49% of Scale, Scale continues. Myself and a few people join Meta. What I was kind of incredulous about at the time was like it's such a weird kind of deal but is one that I think genuinely was like this very interesting win-win-win so to speak. There's a win for the shareholders of Scale because all shareholders of Scale got a great deal, benefited a lot. Scale continues and I truly believe the best days of Scale are ahead of it.

我的问题是，或者我个人感兴趣的是，他能在一次电话中向你解释并让你接受这个观点吗？还是说，等等，马克，我得想想，这太疯狂了。我得想几天。请尽可能解释这个过程。

My question to you is like or what I'm personally interested in is like okay is this he can explain this to you and you come around to that perspective in one phone call. Is this just like wait Mark I got to think this is so crazy. I got to think about this for a few days. Like explain this process as much as you can.

好几周都在想，这太疯狂了。这是真的吗？这合理吗？如果合理，我对此感觉如何？主要情绪更像是这有点疯狂。然后当我更多地思考，哦，我实际上必须决定是否这样做。那本身就是。那里有很多与 Scale 的人、我们的投资者的对话。那方面有很多有趣的对话。我的意思是，显然很难放手或放弃你的孩子

Many weeks of being like, this is insane. Is this even real? Does this make any sense? If it does make sense, how do I feel about it? The predominant emotion was more like this is kind of insane. And then as I thought more about like, oh, I actually have to make a decision of whether or not I do this. That was it in and of itself. There were a lot of conversations there with the people at Scale, with our investors. There were a lot of interesting conversations on that side. I mean it's obviously really hard to let go of or give up your baby

因为你为 Scale 工作了多久？多少年？

Cuz you worked on Scale for how long? How many years?

是的。从创立到那时 9 年

Yeah. 9 years from founding until the

你还年轻。还不到 30 岁。所以这就像你生命的三分之一。

And you're still young. You're not even 30 yet. So this is like a third of your life.

保罗·格雷厄姆很久以前说过，如果你把你的公司视为你一生的工作，你会以不同的方式运作。在我为 Scale 工作的整个期间，我真的把它视为我一生的工作。那是一个艰难的情感过程。然后我认为最终让我接受的是，哇，这是一个三赢，对各方都是好交易，而且我看到了 Meta 可能发生的事情的潜力。你知道，在这一点上，我认为 Meta 在 AI 方面在很多方面看起来像是受损的货物，可能不是最吸引人的 AI 工作场所，但我想这触发了我企业家的一面，我当时想

There was something that Paul Graham said for a long time which is like if you think of your company as your life's work you will operate differently. I did really genuinely think of Scale as my life's work for the whole time that I was working on it. That was a tough emotional process. And then I think what ultimately got me was a combination of wow this is like a win-win-win like all it's a good deal for all parties and also just this like I saw the potential of what could happen at Meta. And you know, this is a point at which I think Meta on AI certainly looked like damaged goods in many ways and like was maybe not the most appealing place to work on AI, but I think that kind of triggered my entrepreneur side where I was like

你基本上可以重建它。就像重新创立实验室。

You can essentially rebuild it. It's like a refounding of the lab.

是的。是的。我认为这是最终变得迷人的事情之一，因为很明显有很多工作要做。它真的为我扫清了道路，能够从头开始构建很多东西，建立正确的原则，建立正确的文化。它为我们创造了构建惊人事物的条件。

Yeah. Yeah. Like I think this is one of the things that ended up being kind of fascinating which is like because it was clear that there was so much work to do. It really cleared the way for me to be able to build a lot of stuff up from scratch, set up the right principles, build the right culture. It created the conditions for us to build something amazing.

这就是你和马克讨论的方式吗？就像嘿，我们必须在这里重置。显然，如果我们继续这条路，我们基本上会抛弃我们所拥有的，用完全不同水平的人才重做。这些是你和他之间的对话吗？他似乎不想要微调。你明白我的意思吗？他就像，哦，这完全不行。不会成功。那么让我们把它拆到地基然后重建。

And is that how you and Mark discuss it? It's like hey we have to hit a reset here. Like obviously like we're not if we keep on this path like we're essentially going to throw out what we have and redo it with a completely different level of talent. Are these the conversations that are occurring between you and him? It didn't seem like he wanted a minor adjustment. You understand what I'm saying? He's just like, oh, this is not working at all. It's not going to work. Then let's rip it down to the foundation and rebuild.

正确。这是一个不断发展的对话，我认为其中一件事我坚信，我认为埃隆在他的职业生涯中多次证明了这一点，你只需要一个小型、极其出色、高度技术化的团队，非常扁平，如果你这样做，你可以非常非常快地完成很多事情。所以这是我非常坚信的，我相信在 AI 方面这是正确的方法，特别是如果你必须做我们必须做的事情，那就是非常非常快地行动，我认为早期我和马克谈了很多，我认为他也很兴奋，这最终成为我们整体战略的支柱之一，就像小型、非常扁平、高度技术化的团队，非常高的人才密度，如果你有这些成分，你能多快行动

Correct. It was an evolving conversation where I think there were like one of the things that I believe really strongly I think Elon has demonstrated this in various times in his career where you just have a small incredibly cracked highly technical team that's very flat and like if you do that you can accomplish a lot very very quickly. So that was something that I had a lot of conviction in and like I believed in AI like that is the right approach especially if you had to do what we had to do which is like move very very quickly and I think early on that was one of the things I talked to Mark a lot about and I think he was excited about as well that ended up becoming one of the tent poles of our overall strategies like small very flat highly technical team very high talent density how quickly can you move if you have those ingredients

从外部视角看，马克就像清空弹夹，这是我思考的方式，对吧？你读过一本叫《拿破仑的心灵》的书吗？

From an outside perspective, it just seems like Mark like empties the clip is the way I think about this, right? Where have you ever read this book called The Mind of Napoleon?

没有。

No.

好的。所以，《拿破仑的心灵》出版于 1957 年。我发现它是因为我看到这个采访。我想是泰勒·考恩采访了萨姆·奥特曼。我想这是 2018 年，他问他今年读过的最重要的书是什么？他说《拿破仑的心灵》。所以很难找到。我买了这本书。300 页，只是拿破仑自己的话，按主题组织。

Okay. So, Mind of Napoleon published in 1957. I found it because I saw this interview. I think Tyler Cowan interviewed Sam Altman. And I think this is 2018 and he asked him like what's the most important book that you read this year? And he said The Mind of Napoleon. So it's very hard to find. I buy the book. It's 300 pages of just Napoleon his own words organized by topic.

在那本书里，他反复谈到犹豫是致命的。但他说，在采取行动之前，会有大量的深思熟虑。他会从各个角度研究，确保自己做出正确的决定。我觉得这与马克的运作方式有共鸣。他会做大量深思熟虑，确保自己走在正确的道路上。但一旦他做出决定，他就会全力以赴。他不会——他直接清空弹夹。这就是我对当时发生的事的感受：他来找你，然后你们开始重建整个组织。我认为他真的内化了失败不重要。重要的是当你赢的时候，你赢得多大。我认为这个整体概念让他在很多情况下能够真正无畏地运作，因为我觉得他非常乐于承担大量风险。

And in that he talks about over and over again that hesitation is fatal. But he says before you engage in a course of action, there's a lot of deliberation. He'll study from every angle, make sure he's making the right decision. I feel there's an echo with the way Mark operates. Does a lot of deliberation, make sure he's on the right path. But once he makes that decision, he just goes all in. He won't—he just empties the clip. And that's the way I felt about what was occurring right at this point where he comes and gets you and then you guys start rebuilding this entire organization. I think he really has internalized that failure doesn't matter. It matters when you win, how big you win. And I think this overall concept enables him to operate truly without fear in a lot of circumstances because I think he is just very comfortable taking lots of risk.

他有一句很棒的话：在一个不断变化的世界里，最大的风险是不承担任何风险。

He's got that great line where it's like in a world that's changing all the time, the biggest risk is not taking any.

是的，没错。我认为他非常深刻地内化了这一点。确实存在一些可能的世界，Meta 在某种感觉上更像谷歌——只是更缓慢、更官僚，作为一家公司看起来非常不同。但我认为他因为如此深刻地内化了这些教训，让公司保持动态、运转和活力。

Yeah, exactly. I think he's internalized this quite deeply. Like there's certainly versions of the world where Meta looks a lot more like Google in some sense—like it's just more slow-moving and more bureaucratic and just looks very different as a company. But I think he, because he's internalized some of these lessons so deeply, keeps the thing dynamic and moving and live.

请跟我讲讲，你作为创始人经营 Scale 时，和现在与马克一起重新创立这个实验室，这两种环境下你对待工作的不同方式。

Talk me through the difference between how you were as a founder running Scale, and now you're refounding this lab, working with Mark—the different ways that you approach your work in those two environments.

在 Scale，我真正内化的一件事——我甚至不记得这个建议来自哪里——但就是这样一个想法：如果你想成为一个好的创始人，你必须能够做每一份工作。我对公司里每一件小事应该如何发生和运作都有非常强烈的观点。这导致公司很大程度上是我的努力和我的观点的产物，它在整个公司以各种方式体现出来，但也让我在很多方面成为瓶颈。这对公司有利有弊。我认为，如果你要创办一家公司，而且这是你第一次创业，你绝对应该采取这种方法，因为这样比不采取这种方法更有可能成功。

At Scale, one of the things I really internalized—and I don't even remember where this advice came from—but it was one of these ideas that if you want to be a good founder, you have to be able to do every job. I had a very strong point of view about how every little thing at the company should happen and should work. And that resulted in a company that was very much a product of my effort and my point of view, and it expressed itself in all sorts of ways throughout the company, but also made me the bottleneck in very big ways. There are pros and cons of that for a company. I think definitely—if you were to start a company and it's your first time starting a company, you should definitely take that approach because it's much more likely to be successful than if you don't take that.

Scale 是你的第一家公司？

Scale your first company?

Scale 是我的第一家公司。

Scale is my first company.

是的。

Yeah.

但在 Meta，这是一个非常不同的任务，因为我们必须在很短的时间内完成很多事情。我没有奢侈去，所谓，做实验室里的每一份工作。我当然不是 AI 研究员。我也不是产品设计师。我不是所有这些对组织成功至关重要的角色。所以在 Meta，我非常采纳了这样一种理念：我们如何创造一个环境，让我们能够雇佣了不起的人。这是第一要务：雇佣在每个学科领域都处于顶尖的杰出人才。

But at Meta, it was a very different assignment because we had to accomplish so much in so little time. I didn't have the luxury of being able to, quote unquote, do every job within the lab. I certainly am not an AI researcher. I'm also not a product designer. I'm not all these things that are totally pivotal to the organization being successful. So at Meta, I very much adopted this philosophy around just how do we create an environment where we're going to hire incredible people. It was job number one: hire brilliant and incredible people who are at the top of their field across every discipline.

你为什么认为你在 Meta 能够做到这一点？仅仅是因为你有更多资源吗？为什么你能做到？

Why do you think you're able to do that at Meta? Is it just because you have more resources? Like why can you do that?

我认为 MSL 创立之初非常有魅力和吸引力的一个原因是，有机会用一个非常非常小的团队从零开始建立一个实验室，而且我们对前进方向有强烈的观点：个人超级智能。我认为这引起了人们的共鸣。我认为每个从事 AI 工作的人都想帮助人们，希望它最终对他们的妈妈、奶奶或爷爷有意义，希望它对世界上的每一个人都有意义。因为我们是用 Meta 的资源来做这件事，所以有机会——我们不会受限于算力。我们不会受限于基础设施。我们不会受限于分发。我们不会受限于初创公司常常受限于的那些东西。但这是一个机会，去真正建立一些你可以打上自己印记、拥有很大影响力并真正塑造的东西。我认为这对很多人来说非常令人兴奋。

I think one of the things that was very charismatic and attractive about the founding moment of MSL was there was an opportunity to build a lab from scratch with a really, really small team, and we had a strong point of view of where we're going: personal superintelligence. I think that resonated with people. I think everyone who works on AI wants to help people, wants it to ultimately mean something to their mom or to their grandma or to their grandpa, wants it to be something that's meaningful to every single person in the world. Because of the fact that we're doing this with the resources of Meta, there was an opportunity—we weren't going to be bottlenecked by compute. We weren't going to be bottlenecked by infrastructure. We weren't going to be bottlenecked by distribution. We weren't going to be bottlenecked by the things that startups often are bottlenecked by. But it was an opportunity to really build something that you could put your stamp on and have a lot of influence over and really shape. I think that was quite exciting for a lot of people.

所以这就是为什么他选择花一大笔钱，看起来是一大笔钱，来吸引少数顶尖人才，因为他知道这些顶尖人才会反过来招募其他顶尖人才。但你需要这个作为起点。对吗？

And so that's why he chooses to spend an insane amount of money, what looks like an insane amount of money, getting a handful of top talent because then he knows that those top talent will recruit in turn other top talent. But you need that as a starting point. Correct.

是的，是的。我认为关于这一点，可能有些被低估的是，OpenAI 和 Anthropic 的股价涨了很多。所以碰巧的是，如果你看看这些人如果留在 OpenAI 或 Anthropic 能赚多少钱，那也是一大笔钱。所以我认为对很多人才来说，在薪酬方面最终在很多方面是净中性的。但显然我认为这有点类似——我们常常不会去想，这些实验室的早期员工或早期研究员，他们的股票价值大幅升值，你知道，他们的薪酬是什么样子。

Yeah. Yeah. And I think one thing that is maybe somewhat underappreciated about that specific part is that the stock prices of OpenAI and Anthropic ran up a lot. So it just happened to be the case that if you looked at how much these people were making if they stayed at OpenAI or Anthropic, it also was a lot of money. So I think for a lot of the talent it ended up being kind of net neutral in many ways, comp-wise. But obviously I think it's kind of the same thing—oftentimes we don't think about how much the early employees or the early researchers at these labs that have had immense stock value appreciation, you know, what their comp looks like.

我们完全基于拥有和雇佣最杰出人才的信念开始了这一切。这是一。个人超级智能。这是二。我们如何创造一个环境，让人们在 Meta 和 MSL 做出最好的工作？我做的更多是，所谓，传统管理。因为这些是我思考很多的事情：我如何创造一个环境，让所有这些杰出、卓越、非常特别的人都能运作并实现他们才华的最大表达？我在某些方面更像一个教练，而在 Scale 时非常不同——我是那个观点必须在我们所做的一切中体现的人。

We started this fully on the belief of having and hiring the most brilliant people. That's one. Personal superintelligence. Those two. How do we create the environment where people do their best work at Meta and in MSL? I'm doing much more of, quote unquote, traditional management, so to speak. Because these are the things I'm thinking a lot about: how do I create the environment where all these brilliant people who are exceptional and very, very special can all operate and achieve the greatest expression of their talent? I'm thinking a lot more like a coach in some ways versus at Scale it was very different—I was the person whose point of view had to be expressed in everything that we did.

多说说这个——你感觉自己像个教练。

Say more about that—you feel like a coach.

我认为我的工作是帮助设定北极星和我们想要前进的方向，这再次是个人超级智能，在消费者 AI 上获胜，构建对每个人都有意义的 AI，然后在团队中发现特殊人才，然后给人们环境和资源，让他们能够最大限度地表达这一点。

I think that my job is to help set the north star and the directionality of where we want to go, which again is personal superintelligence, winning on consumer AI, building AI that means something to every person, and then spotting in the team special talent and then giving people the environment and resources to be able to express that maximally.

我觉得大公司经常搞砸的一件事是，大多数大公司里都有杰出的人才，但他们就是没有被赋予权力去做最好的工作，这就是为什么他们常常离开，去更小的公司或创业公司之类的。你之前提到过杰夫·贝索斯的股东信。这也是他会反复强调的，我读过的关于他的所有书以及股东信里都提到，如果你有优秀的人却无法施展，他们就会离开。

Like I think one thing that big companies often mess up is that there's like exceptional people at most big companies but they just like are not empowered to do their best work and that's why they often leave to go to smaller companies or to startups or whatever it might be. You mentioned Jeff Bezos's shareholder letters earlier. That's something he would repeat as well and all the books I read about him but also shareholder letters like if you have great people that can't build they're going to leave.

是的，没错。AI 是一件非常特别的事情。我觉得你知道现在“研究”这个词几乎没什么意义了，但按定义来说，人类并不知道这些模型的极限在哪里，也不知道能用它们做什么。它们真的是我们不完全理解的未知产物。所以这确实是科学和研究。我们在探索这项技术的极限，以及能构建什么，这是一个需要大量试错的过程。它需要聪明的人有非凡的洞察力。你不能把这个过程想成一种非常压迫性的高强度过程。人们需要空间、时间和资源才能做出最好的工作。我从很多维度思考这个问题，对吧？比如我认为 Nat Friedman 绝对是个天才，正如我提到的，Muse 的很多内容如果有一个人的个人观点贯穿其中，那就是他，现在重要的是让他处于一个能充分表达所有这些的位置。然后我们有一些杰出的研究员。他们有非常非常令人兴奋的想法。我们如何创造一个环境让他们能够探索这些想法？而且大多数想法不会成功。有些会非常成功。那些成功的案例证明了所有投资都是值得的。

Yeah, exactly. And AI is such like a it's like such a special thing. Like I think you know this term research now it it feels like it almost doesn't mean anything but like by definition humanity doesn't know the limits of what these like models are capable of and what are the things you can do with them. Like they really are these like unknown artifacts that we don't understand super world. So like it really is science and research. Like we are exploring the limits of what you can do with this technology and what are the things that can be built and it is a process that requires lots of trial and error. It requires like brilliant people to have incredible insight. Like it's one of these things that you can't think about this process like a very opishly intense process. Like people need space and time and resources to be able to like you know do their best work. And I think about in many dimensions, right? Like I think Nat Freiedman is absolutely brilliant and as I mentioned like so much of Muse if there were a single person whose like single point of view came through on Muse it is now it's important for him to be in a position where where he can like fully express all of that. And then we have some brilliant researchers. They have like very very exciting ideas. How do we create an environment where they are able to explore those? And like most of them won't work out. Some of them will work out incredibly well. The cases where that works out justify all the investment.

所以可以说有两种问题，这个类比就像钻石开采或建造摩天大楼，取决于你所做事情的回报分布。有些领域的分布是超级幂律的。你做的很多事情可能没什么用，但有些事可能比其他所有事情有价值一百万倍。我认为 VC 就是一个很好的钻石开采式行业，它就是要能够识别特殊的想法并坚持到底。

So there's two kinds of problems so to speak and that analogy is like diamond mining or building skyscrapers and it depends on what the distribution of the payoff of the things that you're working on are. So there's like some areas where the distribution is super power. Like most of the things you do will be kind of useless, but then some of the things you do can be like literally a million times more valuable than everything else. VC I think is a great like diamond mining kind of industry to give a sense. It's just all about like being able to identify the special ideas and like see those through

而那很容易做到。只要看看，去你的 19 岁生日派对就行了。

and that's easy to do. Just look at just go to your 19th birthday party.

是的。我的意思是，那简直难以置信。

Yeah. I mean that's like unbelievable. Um

听众们不知道我刚才指的是什么？你能快速说一下你 19 岁生日派对上都有谁，以及为什么如果当时投资了桌边的每个人，那会是一笔好赌注？

do people listening that don't know what the hell I just referenced? Can you talk about real quick what who was at your 19th birthday party and why that was would have been a good bet if they just invested everybody around the table?

这真是想想都觉得疯狂。我 19 岁生日派对。我当时是 MIT 的新生。那是一月份。我在一家叫 HRT 的交易公司做冬季实习。我们大概有 10 个人，12 个人，差不多这样。几乎所有人都是哈佛或 MIT 的学生。我们大多数人已经在各种数学、科学或计算机科学竞赛中认识了。其他人包括 Jeffrey Yan，后来创办了 Hyperliquid，非常成功。Scott Wu 后来创办了 Cognition，非常成功。还有 Jesse Zhang 创办了 Decagon，也相当成功。另一个人是 Vicky Yei，我认为她是一位杰出的研究员，在 Anthropic 工作。每个人都才华横溢。我最近实际上和 HRT 的创始人谈过，HRT 本身现在也做得非常好。难以置信。

This is like a really crazy thing to think about. my 19th birthday party. I was a freshman at MIT. It was in January. I was doing a win internship, a winter internship, win internship at a trading firm called HRT. There were like 10 of us, 12 of us, um, something like that. Almost all of whom were students at either Harvard or MIT. Most of us already knew each other from various math and science or computer science competitions. Some of the other people were Jeffrey Yan who went on to start Hyperlquid very successful. Um Scott Wu went on to start Cognition very successful. Another person there was uh uh Jesse Zang went to start Decagon also quite successful. Um another person was Vicky Yei who is I think a brilliant researcher works at Anthropic. Everyone was like incredibly talented. I actually talked to the uh the founder of HRT recently and HRT itself by the way now is like crushing it. Unbelievable.

HRT 是做什么的？

What does HRT do?

HRT 是一家自营交易公司。他们用自己的钱交易，在 2026 年第二季度有 110 亿营收，我记得利润是 80 亿，创纪录的数字。HRT 的员工不到一千人。我的意思是，这简直难以置信。但我最近和 HRT 的创始人见面，他们谈到：“哦，我们开始做一些私人投资了。”我说：“就投资实习生吧。所有实习生都拿种子轮支票。”那很棒。

HRT it's a proprietary trading firm. So they they trade their own money in Q2 of um 6 had 11 billion of revenue and like I think 8 billion of profit is the the record number. This is a I think it's under a thousand people work at HRT. I mean, it's like unbelievable. But I I recently uh met up with the founder of HRT and they were talking about like, "Oh yeah, we're starting to do some private investments." I was like, "Just invest into the interns. All the interns get seed checks." That's great.

我从小参加数学、科学和计算机科学竞赛和奥林匹克竞赛，认识了来自美国各地做这类事情的各种人。看到和我一起参加竞赛的人变得如此成功，真是超现实。

I grew up doing uh math and uh science and computer science competitions and Olympiads and got to know like all sorts of people from all around America doing this kind of stuff. It's like quite surreal to see the people that I did competitions with become so successful.

你们所有人在未来几十年会做什么，将会非常引人入胜。在我打断你之前，关于钻石开采和摩天大楼的事情你还有更多要说的吗？

It's going to be fascinating what all of you do the next few decades. Before I interrupted you, did you have more to say about diamond mining and skyscraper stuff?

基本上，就是看它是否有幂律回报，有些东西的回报 literally 是投资的一百万倍，还是更像建造摩天大楼，我会把数据标注或数据业务描述为更像这种，就像亚马逊 Prime 配送是另一个例子，回报相当线性，你必须开发一个流程，让你非常非常擅长做这件事。如果你在建造摩天大楼，那是一个非常运营性的业务，关键是从过程中榨取每一分效率，你看着它，想办法让这边效率提高 2%，那边提高 1%，另一边提高 0.5%，你试图榨取每一分优化。然后是钻石开采，你就像是，“嘿大家，去尝试所有疯狂的想法，培育那些看起来有前途的想法。”

Basically, it's like does it have a power law payoff where some things just like pay for like literally like a million times the investment or is it something more like building skyscrapers which I would I would describe data annotation or or or data businesses as more like this which is like Amazon Prime deliveries is another example where the payoff is pretty linear and you have to just get develop a process where you're like really really amazing at doing it. If you're building skyscrapers, then it's like a very operational business and it's all about like squeezing every ounce of efficiency out of that process and you just look at it and you figure out like I can make it like 2% more efficient this way and 1% more efficient that way and half a percent more efficient that way and you just try to squeeze every ounce of optimization out of it. And then there's diamond mining where you're kind of like, "Hey everyone, just like you know, go forth, try all your crazy ideas and like you're just nurturing the ideas that seem promising."

这就是你觉得你现在在 Meta 做的事情。

That's what you feel you're doing at Meta right now.

这就是我觉得我在 Meta 做的事情。而建造服务器之类的事情，一般来说是我觉得我在 Scale 做的事情。这是一个巨大的范式差异。

That's what I feel I'm doing at Meta. And and and the building server stuff generally speaking is what I felt I was doing at at scale. And that's like this big paradigm um paradigm difference.

回到 coaching 以及你现在在 Meta 的管理方式与 Scale 相比的差异。为什么你有这么多直接下属？你有大概 100 个直接下属？有多少直接下属？

Going back to the coaching and the way the difference of how you're essentially managing now at meta compared to way scale. Why do you have so many direct reports? You have like what 100 direct reports? How many direct reports?

是的，我想超过 200 个。是的。

Yeah, I think more than 200. Yeah.

如何以及为什么？回到 MSL 的创始论点之一，以及我们想做什么、重建这个实验室，你如何实现非常非常高的人才密度？你如何让它成为许多这些杰出人才做出最佳工作的最佳场所，以及你如何建立一个技术聚焦的组织，让技术人员做他们最好的工作，有机会真正做他们一生的工作？

How and why? Going back to one of the founding theses of MSL and what we want to do and rebuilding this lab, how do you have really, really high talent density? How do you make it the best place for a lot of these brilliant people to do their best work, and how do you build an organization that's technically focused where the technical people are doing their best work, have the opportunity to really do their life's work?

其中一个设计原则是让它非常扁平。实际上，我们招进这个叫 TBD 的组的所有研究员都直接向我汇报。这更像是一种声明：这是反官僚的。TBD 字面上就是反官僚。关键是我们在招聘杰出人才，你们所有人都清楚地证明了自己能做出卓越的工作。你们在职业生涯中已经做到过。所以我不需要微观管理你们来做卓越的工作。你们自己就能做到。我们需要技术领导力。所以，你们中的一些人将负责为你们的组设定清晰的技术方向，我们有这种 pod 结构，有各种 pod，每个 pod 有技术负责人，他们在技术决策上做最终裁决。所以，有一个技术领导结构，但从所谓的人员管理角度或官僚角度，我们就是反官僚。

And one of the design principles around this was making it very flat. Effectively all the researchers that we hired into this group called TBD, they report directly to me. And it's more of a statement of the fact that this is anti-bureaucracy. TBD is literally anti-bureaucracy. The point is we're hiring brilliant people and you all have demonstrated clearly that you can do brilliant, brilliant work. You have done it before in your careers. So I don't need to be micromanaging you to do brilliant work. You're going to be able to do that on your own. We need technical leadership. So, some of you are going to be responsible for setting clear technical directions for your groups and we have this kind of pod structure where we have various pods and there's various technical leads of the pods who tie break on the technical decisions. So, there's a technical leadership structure in place, but from a quote unquote people management perspective or from a bureaucracy perspective, we're just anti-bureaucracy.

然后另一部分是，我们确实做很多集体决策。我们想要辩论，因为这是一个人才密度极高的杰出群体。我们想要进行那些对话，大家一起讨论我们认为正确的路径是什么。实验室里的每个人都做过卓越的事情。我们想听到所有这些意见。我们想要达成我们都认为是前进最佳路径的共识。所以这确实是一种非常规的团队管理方式。如果我说我是管理 200 多人的最佳经理，那我就是在撒谎。但关键是他们不需要经理，他们需要一个伟大的环境。他们都很杰出。他们都极其有能力。他们只需要空间去发挥。

And then the other part of this is we do make a lot of group decisions. We want to debate because it is a very talent dense group of brilliant people. We want to have those conversations where we're all discussing what we think the right path is. Everyone in the lab has done brilliant things. We want to hear all these opinions. We want to arrive at things that we all believe are the best path forward. So it's certainly an unconventional way to run the team. And I would be lying if I said I was the best manager to more than 200 people. But this part of the point is that they don't need managers, they need a great environment. They are all brilliant. They're all extremely capable. They just need the room to cook.

是的。所以也许不是教练。你更像是伟大环境的管理者，这样他们就能做出伟大的工作。

Yeah. And so maybe not a coach. You're kind of like the steward of a great environment so they can do great work.

是的。没错。

Yeah. Exactly.

Alex，这太棒了，伙计。感谢你抽出时间。

Alex, this was awesome, man. Thanks for taking the time.

是的。

Yeah.

我希望你喜欢这一集。请记得在你收听的地方订阅并留下评论。并确保收听我的另一档播客《Founders》。近十年来，我痴迷地阅读了 400 多本历史上最伟大企业家的传记，寻找你可以在工作中使用的想法。你在这个节目中听到的大多数嘉宾都是通过《Founders》先找到我的。

I hope you enjoyed this episode. Please remember to subscribe wherever you're listening and leave a review. And make sure you listen to my other podcast, Founders. For almost a decade, I've obsessively read over 400 biographies of history's greatest entrepreneurs, searching for ideas that you can use in your work. Most of the guests you hear on this show first found me through Founders.