打造人们喜爱的产品：AI 时代的行动指南

尽管消费领域目前可能不具备投资价值，但为人们提供新的互联网宝藏、重塑那些我们认为已经过时或平庸的服务的机会从未如此之大，而这一切都因 AI 和智能体成为可能。我认为这非常有可能，这让我想继续打造消费产品。今天，我邀请到了 Zynga 的创始人 Mark Pincus，他是一位永不停歇的点子机器，也是科技界最敏锐的产品思维者之一。Mark 的新书《Life at the Speed of Play: Launch Products People Love》刚刚出版，我们将深入探讨其背后的方法论。AI 乐观主义、失败机器、经证实的更好新事物、创始人模式等等。Mark，很高兴你能来。

Even though consumer is arguably not investable right now, the opportunity has never been greater to offer people a new internet treasure, reinvent some service that we thought was over or just generic, but it's enabled now because of AI and agents. I think that that is highly likely and that makes me want to keep building consumer product. Today I'm sitting down with Mark Pincus, founder of Zynga, a relentless idea machine and one of the sharpest product minds in tech. Mark has a new book out, Life at the Speed of Play, Launch Products People Love. And we're going to dig into the playbook behind it. AI optimism, The Failure Machine, Proven Better New, Founder Mode, and a lot more. Mark, it's great to have you.

谢谢。我很高兴来到旧金山，互联网的汽车城。

Thanks. I love being here in San Francisco, the motor city of the internet.

我喜欢这个书名。在 YC，我们给人们发印有'Make something people want'的 T 恤。而'Make something people love'确实是另一个层次。

I love the title. I mean, you know, here at YC, we give people the t-shirt that says, 'Make something people want.' And so, make something people love is like another level, really.

是的。而且首先要做你自己热爱的东西。

Yeah. Yeah. And make something you love to start with.

没错。也许我们就从这里开始。你为什么写这本书？核心动机是什么？你过去已经做出了那么多人们喜爱的产品。这本书对你意味着什么？

That's right. Yeah. I mean, maybe we start there. I mean, you know, why'd you write the book? What's the core motivation? I mean, you've made so many products that people love in the past. What does this book mean to you?

在创建五家公司的过程中，尤其是 Zynga，我逐渐建立了一套方法论，一套关于产品管理、创始人角色和公司建设的完整方法。我意识到你不能只关注其中某一部分。如果你想打造伟大的产品，就无法回避管理、董事会或投资者。你必须全盘投入，从客户、产品和编码的第一性原理，一直考虑到长期可持续的战略。就像我们刚才聊到的，Eric Reese 在讨论如何从结构上设置公司，以便打持久战并统一各方利益。所以我终于坐下来写了这本书，花了四年时间。我认为每个人都应该写一本书，但我真的只是想把这些内容集中在一个地方，对所有人都有帮助——从有想法、想考虑是否真的能去实现它的全职妈妈，到我的同行，到你，就像一场对话。

In building five companies, but especially mainly Zynga, I over time built this playbook, literally a playbook of my approach to product management, my approach to being a founder, to building companies and this holistic realization that you can't just focus on any one piece of this. If you want to build great products, you can't avoid management. You can't avoid having a board or investors. And so you just got to dive into the whole enchilada. And you should think kind of full stack from first principles of customers and products and coding all the way to like long-term sustainable strategy. Some of the things we were just talking about that you said Eric Reese was talking about of how do I structurally set up a company in a way that I can play the long game and align everyone's interest. So I finally sat down. It took four years. I think everyone should write a book one time, but I really just wanted to get this in one place or be helpful of service to really everyone from a stay-at-home mom who has an idea and wants to think about if she could actually go build it to my peers to you and, you know, be just be in conversation.

随着我对你的了解，有一件事让我印象深刻：你经历了计算机时代的各个阶段，并且一直是领导者。

One of the things that really jumps out at me as I've gotten to know you is that you've been through and a leader through all of the ages of computing itself.

是的。

Yeah.

这意味着你拥有来自不同时代的第一性原理智慧。你提到了五家公司，其中两家上市了，但早期的一些公司正处于社交网络这一全新事物的黎明，有些观众可能甚至无法想象。社交网络现在就像水一样，你无法想象没有它的世界。然而，同样的事情正在再次发生。我们现在正从一个没有 AI 和按需智能的时代，走向 AI 无处不在的时代。所以你非常幸运，既能告诉我们第一性原理是什么，又能告诉我们什么会重演。

What that means is you have this sort of first principles wisdom from these different moments. You were mentioning the five companies and two of them went public, but some of the earlier ones were literally at the dawn of a brand new thing with social networking which is kind of some of the people watching probably can't even imagine. You know, social networks are water now. It's just like you couldn't imagine a world before that and yet you could say the same thing is happening again. We're sort of right now at this moment going from a moment where we didn't have AI and intelligence on tap to now AI everywhere. And so you're sort of uniquely blessed to be able to both tell us what the first principles are, but also, you know, what's going to rhyme. I guess.

感觉我们即将第三次进入一个新时代。1995 年我创办第一家公司 Freeloader 时，还只是互联网，要说服人们使用这个在线网络。到了 Tribe 和 Zynga，是社交和移动。现在有了 AI，真的很令人兴奋。

It feels like the third time that we're about to enter a new era. When I started my first company, Freeloader, in 1995, it was just the web and trying to convince people they might want to use this online network. And with Tribe, Zynga, it was social and mobile. And now with AI, it's really exciting.

从第一波浪潮中你学到了什么？社交网络中有没有你原本以为会是 X，结果却变成了 Y 的情况？

What are some of the things from that first wave? I mean, were there parts of social networking that you thought it was going to be X and then it became Y?

首先，我认为社交网络的开端是 Napster，由 Sean Parker 创立。他 16 岁时曾在我这里实习。你还记得第一次加载 Napster 时，它显示'目前有 450 万台机器与你连接'，而这首歌曲在网络上有 15000 个副本。那是我第一次感觉我们透过网络看到了彼此，相互连接。我们不只是连接到某个公司或数据库，而是真正的去中心化点对点网络。它一直是这样发展的。但那是我们第一次真正体验并可视化它，甚至感到有点狂野。Napster 对我来说是一种很好的狂野。我认为那是'人民网络'的开端，尽管几年后才有了 Friendster、Facebook、LinkedIn，以及我失败的公司 Tribe。作为创始人，我经常发现我们离某件事太近，以至于看不到它会有多大。我在社交网络初期工作，并在扎克伯格刚起步时见过他，他来过我的办公室。我在 Tribe 上犯的错误是信任。我在 Facebook 之前建立了一个社交网络，但失败了，因为我在信任方面搞错了：要让人们把自己放到网上，他们首先需要感受到一个良好的信任容器，而 Facebook 从一开始就用.edu 域名做对了这一点。我不认为 Reed 或我，我们任何人，曾想象过它会变得多大。Peter Thiel，我们所有人，我们都卖掉了 Facebook 的股票，用股票投票，没有人能描述出它最终变得有多大。

Well, the first thing I'd say is I trace the beginning of social networking to Napster, which Sean Parker started. He worked for me as an intern when he was 16. So, you remember that first moment that you loaded Napster and it said 4.5 million machines are connected to you right now. And this song file, there's 15,000 copies of it available on the network. And that's the first time to me that I felt like we looked through the network at each other. We connected to each other. We didn't just connect up to XYZ corporation or database. And it was truly a decentralized peer-to-peer. And it always was that. It had grown as that. But this is the first time that we're able to actually experience it and visualize it and even feel a little bit rowdy. And Napster to me was a little bit rowdy in a great way. And to me that was the beginning of this people web even though it was a bunch of years before we got to Friendster and Facebook and LinkedIn, my failed company Tribe. I find so often as a founder that we're too close to something to even see how big it's going to be. And so I was working on the beginning of social networking and I met Zuckerberg when he first started. He came in my office. The thing I got wrong with Tribe was trust. So first of all, the thing that I built a social network before Facebook and I managed to fail and I just got the trust component wrong that for people to put themselves on the web, they first had to feel this good container of trust and that's the thing that they got right with .edu from the beginning. I don't think Reed or I ever imagined how big, I know none of us ever imagined how big this, Peter Thiel, all of us, we all sold our stock in Facebook so we voted with our shares, you know, none of us could have ever described how big this all got.

当你反思这些，再看看我们现在距离 ChatGPT 时刻已经过去几年，但我会说我们现在处于 Opus 4.5 及之后的时代。对我来说，那个模型改变了一切。之前的东西虽然可用，但像玩具，而后来发生了某种神奇的事情，现在我可以把智能体当作同伴来对待。

What are some things you know when you reflect on that like and then you see we're sort of a few years off of the ChatGPT moment but now I would say that we're in the Opus 4.5 and later moment. I would trace it to that model was the thing that changed everything for me. The things before that were usable but like toys, but there was something magical that happened and now I can sort of treat the agent as a peer.

是的。

Yeah.

我可以信任它做一些事情，虽然不是全部，它仍然会幻觉，并不完全正确，但它确实足够聪明。

Like I can trust it with things, not everything, and it still hallucinates and it's not quite right, but it actually is smart enough.

是的。如果你给它合适的框架和上下文，它实际上能做出相对神奇的事情。自最新版本发布以来，我的主要用例就是边走边和它对话。

Yeah. And you know, if you surround it with the right harness and the right context, it can actually do relatively magical things. My main use case now and since the latest release is that I walk around talking to it.

是的。

Yeah.

我经常和它对话。我一直很沮丧。我知道 Granola 非常成功。

And I'm just in conversation with it a lot. I've been frustrated. I know Granola is really successful.

我很惊讶这个产品没有发展得更快，因为我想要的就是让我的 AI 现在就能旁听我们的对话，成为桌边的另一个聪明人。任何时候我们都可以转向 AI 问它的观点，也许有时它会主动插话。但我发现我走路说话时，会说“看看 AI 怎么说”，然后转身对它说，它就会给我回应。但这正是消费级 AI 的一个绝佳例子，几乎肯定会成为一家价值百亿甚至千亿美元的公司，但还没人做出来。

I'm surprised that the product hasn't evolved faster because what I want is to just have my AI listening in on this conversation now with us and just have it be a smart other person at the table. You know, any moment we can turn to the AI and ask it for its point of view. Maybe sometimes it'll interject it, but I find I'll be walking and talking and then I said, "Let's see what the AI says." and I turn and I say it into it and then it gives me a response. But it's I mean this is a great example of like consumer AI that is almost certainly going to be a 10 or hundred billion dollar company that no one's built yet.

我今天也做了完全一样的事，实际上是在和我的治疗师谈话时。有点好笑。我喜欢 Granola，因为它有实时转录。然后我把转录内容粘贴到我的 OpenClaw 里。我说：“嘿，这是我治疗 session 的进展。过去一周有没有什么我该提但还没提的事？”因为它有全部上下文，知道我所有的邮件、短信，我经常跟它聊我在想什么，所以非常有用。所以，基本上它现在就可以坐在我们的访谈里。我不知道，这感觉很奇怪。技术都已经就绪了。我很快就要在 GBrain 上开源我的 Gemini Live 语音插件。所以，这并不难。奇怪的是，一旦我让它跑起来，我有时会跟这个基于 Gemini Live 的开源玩意儿说话，它比 Siri 好，我就想，为什么 Siri 这么差？

I did exactly the same thing today with uh actually my therapist. It's kind of funny. It's uh I like Granola in that it has like the real time transcript. Uh and then I took the transcript and I pasted it my uh to my open claw. And I was like, "Hey, like this is how my therapy session is going right now. Is there anything from the last week that I should have brought up that I haven't brought up yet?" And because it has total context and awareness of like all my emails and like my texts and everything and I talk to it all the time about like what I'm thinking about, it was really useful. And so, you know, yeah, basically it could be sitting here right now in our interview. I don't know. It's like very weird. You know, the tech is all there. Like I'm actually going to open source my Gemini live um voice plugin on GBrain shortly. So, you know, I mean, it's not hard. Like, and the weird thing is once I get it going, it's like I, you know, sometimes talk to this open source thing that's just built on Gemini Live and it's better than Siri and I'm like, why isn't Siri terrible?

Siri 太差了。光是语音输入就很差。亚马逊那边也是，难以置信，对吧？你有一个完整的 Alexa 团队，我听说有上万人。

Siri so bad. Even just the voice input on it is so bad. And then at Amazon, it's just unbelievable, right? You have an entire uh Alexa team. I heard there's like 10,000 people on that team.

不。

No.

我就想，这到底怎么回事？这么多年了，还没有一个真正能感受到的 LLM。这些组织到底在干什么？

I was like, what is happening? Like if there's no L, it's been years and there's no real like LLM that you could feel. It's just like what is going on with these organizations at this point?

我经常连 Granola 都不用。重要会议比如带孩子看医生，我就开语音备忘录，然后复制转录文本，用我自己的提示词，我发现得到的东西更好、更容易分享。有人可以在上面做“Proven Better New”。

I'll often not even use granola. I just turn on voice memo for important like meetings like a doctor for my kid and then just copy transcript and then have my own prompts and I I find I get something way better and more sharable. Someone could do proven better new on it.

没错。是啊。我不知道。也许我们现在就该用这个。这是个好例子吗？比如，对于刚才我们说的“永远在线的 AI 智能”这个想法，你怎么做“Proven Better New”？我们只是用现有的东西拼凑了一下。但这个框架会从一种直觉开始，觉得少了点什么。

That's right. Yeah. I don't know. Maybe we should use that right now. Is that like a good example? Like how would you do proven better new for this idea of like just always on AI intelligence that we just talked about? Like we hacked it together with whatever we got. But this framework proven better new would start with an instinct that that we have that something's missing.

在这个例子里，我们有这种感觉。是的。

In this case, we have it. Yeah.

好。这就是我们的直觉，也是我们想测试的创新区域。然后把它隔离出来，确保我们只测试那个直觉，而不是我们不想测试的东西。我们会看看 Granola，它可能是最成功的 AI 笔记产品。我们拆解它，然后说，他们做的所有我们没在创新的部分，我们合法复制。我们甚至不去质疑。我们没时间。

Okay. Here's our instinct, which is also kind of our innovation zone that we want to test. but then isolate that and make sure we're testing just that instinct and not the things we don't want to test. And so we would look at Granola, let's say, is probably the most successful AI noteaker product. And we deconstruct it and we would say everything that they're doing that we're not innovating on, we're going to legally copy. We're just going to not even question it. We're just gonna We don't have time.

嗯，它已经被验证了。

Well, it's proven.

它已经被验证了。

It's proven.

是啊。为什么要

Yeah. Why would

而且我们没时间把它做得更好，因为那是我们擅长的领域。

And we don't have time to make it better because that's where we're in the wheel here.

是的。“更好”是指 10 个现有用户里有 10 个会说更好。比如现在它免费了、半价了、更快了。或者有些没人会反对的改进。在这个例子里，感觉摩擦是最大的问题，因为就像你和我，但这是个假设。所以那就是“新”。

Yeah. And better would be something that 10 out of 10 existing users would say better. So that would be like now it's free. It's half price. It's faster. It's there something that no one would friction. I mean in this case like it feels like the friction part is the biggest problem because it's like you and I but that's a hypothesis. So that would be new.

你我说：“好，摩擦是问题，或者它没在听是问题。也许我们是对的，很可能我们是错的。”所以那是我们要测试的部分。我们做所有已验证的机制。然后唯一的新东西就是它始终在听，但我们一开始就假设这个想法可能是错的。所以我们不执着于它，我们测试很多其他相关的想法。

You and I say, "Okay, the friction's the problem or the fact that it's not listening is the problem. Maybe we're right. We're probably wrong." And so that's the part we want to test. And so we do all the proven mechanics of it. And then the only new thing is that it's always on listening, but we start off with the proposition that that idea is probably wrong. And so we don't get attached to it. and we we test a lot of other ideas around it.

“Proven Better New”听起来现在做正合适，因为基本上所有东西都齐了。我做了一个 GPT 和一个 Claude 技能，上传我的书然后做“Proven Better New”，部分效果惊人。所以“Proven”部分我会给 A-，它比大多数都快、都好。

Proven better news sounds like a perfect uh thing to do right now because there's basically all the things, you know, I I made a a GPT and a Claude skill to do upload my book and do proven better new and part of what it did was amazing. So, the proven part I would give it like an A minus. It was faster and better than most.

它知道已经做了什么。是的，它在“Proven”方面很出色。

It knows what has been done. Yeah, it's it's brilliant at proving.

是的，它就在分布里，有时就在模型里，对吧？“Better”部分我会给 B-，因为它不总是能理解“更好”这个概念。然后“New”我给 D。

Yeah, it's in distribution, you know, it's literally in the model sometimes, right? The better part I'd give it like a B minus because it it was not always getting better like the concept of better, right? And then for new I'd give it a D.

是啊。所以那就是人类的作用。

Yeah. Which makes I mean that's what the human's for.

我们仍然在某些方面有用，创造新东西可能是最有趣的事情之一。回到消费级，我们之前聊到，我今天下午刚和一个创业者开过会，他们的消费级指标是我多年来见过最好的，但其他投资了他们的投资人却告诉他们需要转向企业级，我说不，这太棒了。

We're still useful for something and you know making new things is probably one of the most fun things that you could do. Going back to consumer, I mean earlier we were talking about I mean I had just you know did office hours this afternoon with someone who they have like some of the best consumer metrics I'd ever seen in like you know years and then the other investors uh who had invested in it are telling them like you need to pivot to enterprise and I was like no this is amazing.

他们为什么想让他们转向企业级？只是因为现在企业级更容易融资？那是投资者想投的方向。

And why did they want them to pivot to enterprise just because that's more fundable right now? That's what investors want to invest in.

所以，你的感觉是什么？我是说，

So, you know, what's your sense? I mean,

我在 1999 年正好相反，我当时在做一个企业软件公司叫 support.com，后来上市了，但没人对我们感兴趣，因为一切都是 consumer.com，直到那些都失败了，我们是仅有的两家能上市的公司之一。

I was I was on the opposite end of that during in like 1999, I was building an enterprise software company called support.com, which went public and no one had any interest in us because everything was about consumer.com until all of that failed and we were one of like only two companies that could go public.

你觉得奇怪吗？投资者有时就是完全搞反了。

Do you find that odd? It's just uh the investors, you know, they're sometimes exactly 180 degrees off.

他们根本没从第一性原理思考。

They're not thinking about first principles at all.

听起来他们对“新”的部分想得不够。也许现在他们过度依赖 ChatGPT 和 Claude 了，对吧？

It sounds like they don't think about the new part enough. And then maybe these days they're overrelying on ChatGpt and Claude, right?

所以，GPT 不喜欢消费级。也许你有个想法，问“该去哪里融资？”，它可能会告诉你企业级。所以，

And so, you know, they like the GPT doesn't like consumer. Maybe if you had an idea and said, "Where should I go to get funding?" It would probably tell you enterprise. So,

是的，就是这样。关键是，你必须做新东西。而新东西是分布之外的。

yeah, that's exactly the deal. And then the thing is, you have to work on the new stuff. And the new stuff is out of distribution.

如果你我今天刚开始创业，我们会在没有分销渠道、没有经过验证的分销方式的情况下做消费者业务吗？

If you and I were just starting today, would we go do consumer when there's no distribution, no proven distribution?

无论如何，我都会想办法获得分销。理想情况下，产品里会有某种病毒式传播的钩子。我能让人们给朋友发邮件吗？

By hook or by crook, I'd try and figure out how to get distribution. I mean, ideally, there's like some viral hook inside of it. Can I get people to email their friends?

那如果这些方法都不奏效呢？我的意思是，从经过验证的角度看，目前消费者领域没有一条经过验证的分销路径。

Well, what about when those things fail? I mean from a standpoint of proven, there is no proven path for distribution right now in, I would argue, consumer.

所以这才这么难。

Which is why it's so hard.

是啊。你现在可能看不到很多消费者公司成功。

Yeah. You probably don't have a whole lot of consumer companies coming through right now.

没错。有趣的是，我们有一些类似消费者的东西，那就是开发者工具。所以有像产消者这样的东西。

Yeah, that's right. What's interesting is we have things that kind of resembled consumer. They're dev tools. So there's like prosumer.

是的。肯定有产消者。如果你和 AI 投资者聊，这实际上是他们现在喜欢追逐的关键市场之一。产消者的情况是，比如可能有 2000 万开发者，你基本上可以用消费者风格的策略。

Yeah. There's definitely prosumer. And if you talk to AI investors, that's actually one of the key markets that they like to go after right now. The prosumer thing is that there might be like 20 million developers, for instance, and you can basically use consumer style tactics.

我不知道，我们中的一些人觉得消费者很难，因为你一天没有那么多时间。

I don't know, some of us consumer is hard because you just don't have that many hours in the day.

对。对。我们日程排满了。

Right. Right. We're booked.

甚至我们刚才聊的这个语音想法，我们应该启动这个想法。我们应该启动它。

Even this voice idea that we were just talking, we should start this idea. We should start this.

是的，我们应该。只是，你今晚能把它写出来吗？

Yeah, we should. It's just, can you code it tonight?

是的，我可以。我会用 GStack。

Yeah, I can. Yeah. Yeah, I'll use GStack.

据说，Thinking Machines 的重大创新是 AI 响应时间接近零延迟。

Supposedly, Thinking Machines, their big innovation is close to zero latency on AI response times.

这属于新类别。听起来我们得测试一下。

That goes in the new category. It sounds like we got to test it out.

这很难判断。这是更好还是新？因为零延迟、更低延迟，每个人都会说好。

It's a tough one. Is that better or new? Because zero latency, less latency, everyone's gonna say yes to it.

没错。所以那是更好。但我们想用它做的功能可能是新的。是的。他们让 AI 倾听并回应我们的对话。

That's true. So, that's better. But the feature we want to do with it is probably new. Yeah. They have the AI listening and responding to our conversation.

所以一旦我理解了已验证、更好、新，那么，已验证给了我一个基础，更好给了我一个方向，而新就像是所有伸出去的尖刺，其中一些会抓住机会。

So once I know proven, better, new, it's like, well, proven gives me a base and then better gives me a direction and then new is sort of like all the spikes that go out and like some of these things are going to catch.

公平地说，可能一个都抓不住。

Probably none of them will catch, to be fair.

是的。我的意思是，要做好一个都抓不住的准备。

Yeah. I mean, be prepared for none of them to catch.

假设。对，假设一个都没有。我的意思是，新就像是那个能让人对你的产品产生兴趣的新奇想法。如果我们只做已验证，然后说，‘嘿，我们有一个 Granola 的副本。你想要吗？’你知道，没人会觉得印象深刻。

Assume. Yeah, assume none. I mean, new is like the novel idea that will get someone interested in your product. If we just did proven and we're like, 'Hey, we have a copy of Granola. Do you want it?' You know, no one's going to be very impressed.

是啊，我已经在用 Granola 了。它很酷。

Yeah, I'm already using Granola. It's cool.

但如果我们说，‘哦，这就像 Granola，但它一直在听，一直在线。’这可能会让人尝试我们的产品，但那个功能可能最终会消失。这个功能可能确实能带来试用。就像他们说的盒子背面，是让你买新麦片的原因，但很可能不是你回购的原因。你回购是因为麦片本身。

So, but if we're like, 'Oh, this is like Granola, but it's always listening. It's always on.' It might get someone to try our product, but that feature might die. The feature might actually get trial. It's like they call it the back of the box, like what got you to buy the new cereal, but it's probably not what you came back for. Like you come back for the granola.

这实际上是过程中非常重要的一部分。我觉得要理解新不会奏效，然后不要因此气馁。你和很多产品团队和设计师合作过，他们以为自己确信新会成功，但结果没有。你怎么跟他们说？这真的涉及到自我消亡，比如我们如何对追求愿景的本能保持热情，但对这个特定的产品变体和想法保持冷静。这非常困难。当你的团队在攻克这个山头，所有人日夜奋战时，你怎么跟他们说？作为创始人，最糟糕的感觉之一就是周一或周日躺在床上想，‘我觉得这个产品不太对。’

This is actually a really important part of the process. I feel like it's understanding that new will not work and then not getting demoralized about how do you do it. You've worked with so many product teams and designers and they think they're convinced the new is going to work and then it doesn't. What do you say to them? It really kind of gets into this death of the ego, like how do we stay passionate about the instinct that we're pursuing the vision but dispassionate about this particular product variant and idea. That's really difficult. What do you say to your team when they're taking this hill and everyone's working day and night? One of the worst things we can feel as a founder is to come in on Monday or Sunday, I lay in bed and think, 'I don't think this product's exactly right.'

嗯。

Yeah.

但我该怎么办？我已经说服了投资者，而且团队在拼命构建这个产品。我周一进去说，‘伙计们，我感觉不对’？还是我忍气吞声，然后慢慢引导他们？这有点像为人父母。你怎么管理孩子的事情？这归结到一个问题：一致性和风格。我们作为创始人是否允许自己频繁转型，还是被困在这一个产品路径上？有一个有趣的现象，我称之为真信号或热度，关于正确的产品，那个瓶中的闪电时刻。当你拥有它时，你知道；当你没有时，你不知道。我的意思是，你无法确切知道何时达到。从各方面看，你不需要指标，因为一切都在运转，每个反馈循环都是肯定的。

But what do I do? I've got investors convinced and I have people killing themselves to build this. Do I come in Monday and say, 'Guys, I'm not feeling it.' Or do I just kind of suck it up and then slowly try to move them? It's kind of like being a parent. How do you manage things with your kids? And it gets to one question which is alignment and style. Have we given ourselves permission as founders to pivot a lot or are we trapped in this one product path? There's a funny thing about it, I call it true signal or heat, about the right product, the lightning in a bottle moment. When you have it, you know it, and when you don't, you don't know. I mean, you don't know positively when you achieve it. From everything, you don't need metrics because everything works, every feedback loop is a yes.

但当它不太对时，就有争议，然后你真的在寻找大量不同的数据，试图再测试一件事。当你有了那种热度，那种真信号，你不需要告诉任何人更努力或更快地工作，每个人都看到了。

But when it's not quite right, then it's debatable and then you're really looking for lots of different data and trying to just test one more thing. When you've got that heat, that true signal, right, you don't have to tell anyone to work harder or faster, everyone sees it.

我称之为鱼群在游。当鱼群在游时，你整夜撒网，感觉很棒。这是一种高潮。你某种程度上必须慢下来才能快。比如获得真正的信念，然后每个人都会快速行动。

I call it the fish are running. Like when the fish are running, you're up all night throwing nets and it's great. It's a high. You kind of have to go slow to go fast. Like get to real conviction and then everyone is going to move fast against it.

你一生中有多少次感觉鱼群在游？

How many times in your life have you felt like the fish are running?

只有几次。比如我的第一家公司 Freeloader，我们天时地利人和，第一个月免费应用就获得了 200 万次下载。然后在 Zynga 有很多次，我感受到了。我的意思是，有很多游戏发布时刻。

Only a couple times. I mean with Freeloader, my first company, we were in the right place at the right time and we got 2 million downloads of our free app in the first month. And then a lot of times at Zynga, I felt it. I mean, there were so many game launch moments.

你有很多爆款。

You had so many hits.

是的，我们有很多爆款，甚至很多重大功能发布。我喜欢说，伟大的产品制造者是在收集胜利，而不是下注。所以在发布之前很久，你就知道它会大卖，或者你的用户会喜欢它。你不需要看他们是否喜欢。

Yeah, we had a lot of hits and so many even big feature releases. And I like to say great product makers, they're collecting winnings, not making bets. So long before you launch, you know it's a hit or that your users are going to love this. You don't look to see if they like it.

那么除了已验证、更好、新之外，你还有什么建议给创始人？他们希望在事情顺利时唤起鱼群在游的感觉。当你现在有一批公司事情顺利，达到很高的 ARR，然后他们遇到了规模扩张的问题。规模扩张和管理就像黑箱。人们问，‘我该怎么办？我雇一个懂规模扩张的人？我找一个懂这个的投资者？’我在斯坦福创建的课程中尝试指导创始人。我说，‘你不需要去读 MBA。你不需要学习管理。’所有这些管理工具的唯一目的是让人们在我不在房间时做正确的事。就这些。第一课是待在房间里。

So what are some other things other than proven, better, new that you would suggest to founders? They want to invoke fish are running when things are going right. When you have a bunch of companies now that have things going right, hitting big ARRs, and then they hit a question of scaling. Scaling and management are like these black boxes. People are like, 'What do I do? Do I hire someone who knows scale? Do I bring in an investor who knows it?' I try to coach founders in the class I create at Stanford. I just said, 'You don't need to go get an MBA. You don't need to learn management.' The only point of all of these management tools is to get people to do the right thing when we're not in the room. That's it. The first lesson is be in the room.

这就像，如果你在那个位置上是最佳人选，就尽可能待在房间里，然后用你确信自己不在时也能做对事情的人和流程来取代自己。没有固定的剧本，有很多方法，但归根结底就是如何与人们达成一致？如何招聘到符合你正在创造的文化和风格的正确人选？还有关于培养人们成为 CEO。有很多不同的事情可以做，我试过，其他人也试过，但核心就是让他们在你不在时做正确的事。

It's like be in the room as much as you can if you are the best player in that position and then replace yourself with people and processes that you can feel confident are going to do the right thing when you're not in the room. And there isn't any one playbook. There's lots of ways, but that's what this all boils down to is how do you get alignment with people? How do you hire the right people who are going to be in that culture and style that you're creating? And talk about making people CEOs. There's a bunch of different things you can do. I've tried and others, but the whole point of it is just so they do the right thing when you're not there.

基本上，你提到的听起来很像布莱恩·切斯基的创始人模式。

Basically, what you're mentioning there sounds a lot like Brian Chesky's founder mode in a way.

我总能听到他说，加里，领导力是在场，而不是缺席，对吧？这就是你所说的第一部分。我喜欢那部电影，好像是叫《怒海争锋》。你知道我在说什么吗？他下去抓起一支桨划船，说：‘来吧，伙计。自由。我们必须到达自由。’

I can always hear him saying like, Gary, leadership is presence, not absence, right? And that's like the first part of what you're saying. I love that movie. I think it's called like Master and Commander. Do you know what I'm talking about? And he like goes down and grabs an oar and he's rowing and he's like, 'Come on, man. Freedom. We got to make it to freedom.'

是的。但我确实认为这是创始人模式的一部分，无论是埃隆睡在工厂地板上，还是你的团队看到你像他们一样深入细节、了解细节并关注细节。我不信任一个不喜欢自己产品、不比任何人更了解自己产品的消费公司 CEO。我知道有些 CEO 很成功，但我不明白他们怎么能成功。我喜欢创始人模式。当我第一次听到布莱恩谈论它时，它打动了我。因为对我来说，创始人模式很大程度上是允许我们做自己。我喜欢他谈论的方式，也喜欢创始人模式之后引发的辩论，在 Twitter 或 X 上，很多 VC 说创始人模式只适用于很小一部分值得的创始人，比如杰夫·贝佐斯。而我认为创始人模式适用于每一位创始人。

Yeah. But I definitely think that that is a part of founder mode and that's whether it's Elon sleeping on the factory floor, but your team seeing you get into the details and know the details and pay attention to the details as much as them. I don't trust the CEO of a consumer company that doesn't love their product and doesn't know their product better than anybody else. I know there are some really successful, but I don't understand how they can be. I love founder mode. When I first heard Brian talk about it, it spoke to me. I'm like because to me founder mode is so much of it is about giving ourselves permission to be ourselves. And I love the way that he talked about it and I love the debate that came out after founder mode where on Twitter or X where a lot of VCs were saying founder mode is only good for the very small percentage of founders that deserve it. You know, like Jeff Bezos, right? And I'm like, founder mode is for every founder.

我喜欢对创始人说，你成为创始人是为了押注自己，现在你不应该把这个权利让给别人，比如你的董事会或投资者。有趣的是，创始人模式适用于没有人相信你的时候，对吧？当没有人愿意让你做你想做的事。你所做的非常不受欢迎，但你是唯一相信它的人。我喜欢说，我们在职业生涯早期都是专家证人，最接近答案，却离决策最远。创始人模式就是为你准备的，但不要在自己公司里当专家证人，我们常常对自己这样做。

I like to say to founders, you went and became a founder to bet on yourself and now you don't you shouldn't abdicate that to somebody else, to your board or your investors. And it's just funny that founder mode is for when no one else believes in you, right? When no one wants to let you. What you're doing is so unpopular, but you're the only one who believes in it. And I like to say that we all start as expert witnesses in early in our careers that we're closest to the answer and furthest from the decision. And founder mode was made for you because you're but don't be an expert witness in your own company and we do that to ourselves.

哦，这太有趣了，事情就是这样。我们在建立公司时做出太多妥协，为了雇佣那个 CTO 或获得那个主要投资者，我们扭曲自己，最终没有建成我们想住的房子。然后有一天我们想，好吧，这家公司可能不适合我，或者我不适合它。但我们要说，不，你是明星球员。

Oh, that's so interesting that that's what happens. We make so many compromises as we're building our company to hire that CTO or get that major investor or that we contort ourselves and we end up not building a house we want to live in. And then one day we're like, well, I guess it's not really the company for me or I'm not for it. We're like, no, you're the star player.

是的，你是老板，也是明星球员。我感觉有你在公司价值更高。那么，我们做了什么？你只是创造了一个你甚至不受欢迎或不属于的环境。我开始理解创始人模式不仅仅是在治理或董事会层面，而是我们作为创始人如何每周都活在创始人模式中。我认为你有这些直觉让我们走到今天，让 Airbnb 走到今天，很大程度上是布莱恩有信心，但也为周围所有人创造了背景：如果他周一进来不喜欢那个产品，他能直接告诉团队吗？还是他需要管理他们或委婉表达？有人说为马克工作感觉像三年级足球，每个周一他进来都会爱上不同的想法。有时他们是对的。但如果他们这么说，那也说明我在管理他们和为我每周改变创造背景方面做得不好。你知道，这是我学到的一件事。但首先，创始人模式就是准备好倾听自己的直觉并跟随它们，不要因为取悦他人而忽视它们。但创始人模式的另一部分是我们开始学习创造背景，让团队接受你说：你知道吗，我们要去那边的大陆。那是我们的使命，没有改变，但我现在要在不同高度讨论。那是 10 万英尺。今天我们在 5000 英尺。这个策略没有朝向那个大陆。我们必须走那条路。每个人都参与进来，我们努力做到智力诚实，并形成一种文化，这样人们周一进来会问：上周你学到了什么？我看到另一个产品，我认为它更好地实现了我们讨论的目标。告诉我更多。我们是否谦逊、好奇、智力诚实？这是我们的文化，还是我们只是一台执行机器，没有人可以举手说我不明白？

Yeah, you're the owner and you're the star player. I have a feeling the company's worth more with you there. So, what did we do or what did you do? You just create an environment you're not even welcome at or you don't belong in where I've started to interpret founder mode is it's not just at the governance or the board level it's there's a question of how do we as founders live in founder mode on a weekly basis and I think it's you have these instincts that got us here it got Airbnb to where they are and so much of it is Brian having the confidence but also creating the context with everyone around him that if he comes in on Monday and doesn't like that product, is he able to tell the team that or does he have to manage them or couch it? And people have said working for Mark can feel like third grade soccer and every Monday he comes in and falls in love with a different idea. And sometimes they're right. But if they're saying that, I'm also not doing a good job of managing them and creating the context for me to change every Monday. You know, that's one thing I've learned. But first of all, founder mode is like being ready to tune into your own instincts and follow them and not ignore them because you're a people pleaser. But the other part of founder mode that we start to learn is create the context that your team is okay with that that you say, you know what, we're going for that continent over there. That's our mission. It hasn't changed, but I'm going to talk at a lot of different altitudes now. That's 100,000 ft. Today, we're at 5,000 ft. This tact is not heading towards that continent. We got to go that way. And where everyone comes in where we're trying to be intellectually honest and there's a culture of that so that people come in on a Monday like what did you learn the last week well I saw this other product that I think is doing what we're talking about better like tell me more about it like are we humble and curious and intellectually honest and is that our culture or are we an execution machine and there's no room for anybody to raise their hand and say I don't get it.

是的，我觉得这很难，因为每家公司都处于不同模式。有时你处于新产品模式，试图做前所未有的事情，然后有时就像，嘿，这东西很棒，我们只需要继续榨取价值。我很好奇你的观点：是否还有尚未被发明的、价值数万亿美元的消费服务和公司？

Yeah, I think that's so hard because like every company is in different modes. Like sometimes you're in the new product mode and you're trying to do something that's never happened before and then sometimes it's like, hey, this thing's great. Like we just got to like, you know, keep squeezing the lemon here. I'm curious what's your point of view on are there massive trillion dollar consumer services and companies that are yet to be invented that we haven't seen yet?

我认为几乎肯定有，怎么会没有呢？不过，智能已经唾手可得，但有趣的是，人们开始有点沮丧。你知道，即使在企业方面，昨天我看到一个数据，大约 90%投资了 AI 的企业还没有获得任何收益。

I mean I think almost certainly like how could there not be right? It's intelligence is on tap though I guess it's people are starting to get a little demoralized about it interestingly you know even on the enterprise side like there was a stat yesterday that I saw something like 90% of enterprises that have like invested in AI haven't received any benefit yet.

我也看到了。

I've seen that too.

是啊，这难道不奇怪吗？

Yeah isn't that weird.

但这也太早了，我认为有很多因素不利于采用。

But it's also like so early and I think there are just so many things arrayed against adoption.

你认为这包括所有那些每年在 token 上花费一百万美元的公司吗？但我想这取决于人们在做什么。比如，如果你只是花 token，而不在产品制作方式上做任何改变，或者不触及用户，我能理解那只是花更多钱而没有效果。所以，目前我仍然认为，如果到了 2026 年人们还在摸索，那我认为是技能问题。

You think that includes all of these companies that are spending a million dollars a year on tokens probably. But, you know, I guess it just depends on what people are doing. Like, if you just spend the tokens and don't do anything different about how you make products or it doesn't touch a user, I could see how that's just like spending more money and not having any effect. So, you know, for now, I still think, you know, it's if it's 2026 and people are still figuring it out, I assume it's skill issue.

如果这个数字在三四年后仍然成立，我可能会有点担心。但你认为我们还有三四年时间才能看到成果。

If that number continues to be true in like three or four years, like, you know, I might get a little worried. Yeah, but you think we have three or four years before it has to deliver.

有很多奇怪的混杂效应。我不太惊讶的一件事就是这种 token 最大化现象。我们很多朋友都在花——我自己每年花一百万美元在 token 上。彼得·斯坦伯格显然每月花一百万或一百一十万美元。这算是用这些东西能做到的前沿了。你现在确实可以做一千人的工作，但你必须为此付费，而且很少有人这么做。

There's just so many weird confounding effects. Like one of the things I'd be not too surprised is happening is actually this token maxing thing. So many of our friends are spending, you know, I'm spending a million dollars a year on tokens. Like Peter Steinberger apparently is spending a million or $1.1 million a month. That is sort of the frontier of what you can do with this stuff. You can literally do the work of a thousand people right now but you have to pay for it and very few.

如果你做了一千人的工作，但没有得到一千人的产出，那就不对劲了。

If you're doing the work of a thousand people but you're not getting the output of a thousand people, then something's not adding up yet.

我的意思是，我觉得我确实得到了那种产出。我的开源项目看起来相当可用和有用。更好的版本实际上是 Open Claw。我认为彼得·斯坦伯格之所以能做出这个，是因为他已经很富有了，而且他在瞎折腾。我听说他厌倦了 Aiza，然后他说，你知道吗，这更——他就是喜欢编程，为什么不在 token 上每年花一百万美元，来打造世界上最好的开源平台，把这份礼物送给每个人呢？这就是 Open Claw 现象发生的原因。但其中蕴含着一个有趣的教训：那些 90% 的人中有多少每年花一百万美元在 token 上，但那只是一两个人，而有多少被一千个人分摊，每个人基本上都在用 Copilot 上的 GPT-3.5？他们用的是一年前的模型，非常担心成本，而且根本没有正确使用。我在网上因为谈论代码行数而惹了很多麻烦。我意识到他们其实是对的，因为我刚开始重新编程时写了五十万行 Rails 代码。那是错误的做法，因为我写了大量代码来调用 LLM。那是老方法。新方法是让 LLM 直接写你当下需要的代码。然后你写的代码量减少 10 到 20 倍，但可定制性更强。它做得更多，也更棒。在 token 最大化之间，甚至不要写调用 LLM 的代码。写 markdown 来教 LLM 写代码。在如何构建软件方面，有几个大的转变，但还没人知道。不过我也喜欢这样的概念：现在的研发就是去挥霍 token。

I mean, I would say I feel like I'm getting that. My open source projects seem to be quite usable and useful. The better version is actually Open Claw. The entire reason why I think Peter Steinberger was able to make that was he was already wealthy and he was messing around. I heard he got sick of Aiza and he just said, you know what, it's way more—he just loves coding and why not just spend a million dollars a year on tokens to make the world's best open-source platform that gives this gift to everyone. That's how the Open Claw phenomenon happened. But embedded in that is actually an interesting lesson: how many of those 90% are spending a million dollars a year on tokens, but it's one or two people, and how much of it is divided out by a thousand people and everyone basically is using GPT-3.5 on Copilot? They're using models from a year ago, really worried about cost, and they're just not using it correctly. I got in a lot of trouble on the internet for talking about lines of code. I realized they were actually right because I wrote half a million lines of Rails code when I first started coding again. And that was the wrong thing to do because I was writing a lot of code that then would call the LLMs. That's the old way to do it. The new way is to have the LLMs just write the code you need right now. Then you write 10 or 20 times less code, but it's way more customizable. It does more and is more awesome. Between token maxing and even, don't write code that calls LLMs. Write markdown that teaches LLMs to write code. There are just a few big shifts in how to think about building software that nobody knows yet. But I also like the concept that the R&D now is about going and squandering tokens.

我知道，对吧？我不认为这总是浪费 token，但也许我们现在就该这么做。

I know, right? I don't think it always is a squander of tokens, but maybe that's what we should be doing right now.

是的。我的意思是，基本上因为前沿模型已经那么好了。我认为两年内这个数字会变成十万，然后降到一万和一千。基本上这些数量级的变化即将发生。我听到你这么说，但我也在想，你是不是会挥霍更多。你可能仍然花一百万美元，只是现在你得到了一百万人的工作，而不是一千人。

Yeah. I mean, basically because the frontier models are that good. I think that in 2 years that'll be 100,000 and down to 10,000 and a thousand. I mean, basically these orders of magnitude are about to happen. I heard you say that, but I also wonder if you're just going to squander that much more. You might still be spending a million dollars. It's just that now you're getting the work of a million people instead of a thousand people.

我想我们会发现纯软件的极限在哪里。悲观的观点会说我们已经到了。实际上你不需要一百万人来编程。全世界显然只有两千万程序员。

I guess we're going to find out what the limit of pure software is. A pessimistic view would say we're already there. You don't need a million people coding anything actually. There are only 20 million coders in the whole world apparently.

嗯，现在可能多很多了。是的，取决于你怎么定义程序员。

Well now there's probably a lot more. Yeah. I mean depending on how you define a coder.

关于大型消费者服务是否已经占领了市场、是否已经结束的问题，我想说一点。我很好奇。你怎么看？

The one thing I'd say back on the question of are the big consumer services that ground taken and is that over? I'd be curious. What do you think?

你知道，我看着我的手机。那是我的主屏幕？是的。那是我的主屏幕，一半是空的，满的一半大部分是通用应用。有时钟、备忘录、照片、相机、天气。但我认为这些我称之为互联网宝藏的服务，那些我们无法回忆没有它们的生活或无法想象没有它们的生活，比如谷歌或 GPT。现在，我认为在我们的数字生活栈中，被发明出来的宝藏还很少。所以我认为，尽管消费者领域现在可以说不可投资，但奇怪的是，机会从未如此之大，可以为人们提供一个新的互联网宝藏，并重新发明一些我们认为已经结束或只是通用的服务，无论是相机还是 Uber 和 Airbnb，一个我们尚未真正想象但现在已经因为 AI 和智能体而成为可能的服务。我认为这非常有可能，这让我想继续构建消费产品。

You know, I look at my phone. That's my home screen? Yeah. That's my home screen and it's half empty and the half that is full is of generic apps for the most part. It's clock, notes, photos, camera, weather. But I think that these what I call these internet treasures, these services we can't remember life before or imagine life without like a Google or a GPT. Now, I think that there's so few that have been invented that's in our digital life stack. And so I think that even though consumer is arguably not investable right now, I do think that oddly the opportunity has never been greater to offer people a new internet treasure and reinvent some service that we thought was over or just generic, whether it's the camera or an Uber and Airbnb, a service that we didn't yet really imagine, but it's enabled now because of AI and agents. I think that that is highly likely and that makes me want to keep building consumer products.

是的。我的意思是，我会说，如果 Opus 4.5 时刻就在十二月，你必须支付数万到数十万美元才能用这些东西完成真正的工作。这意味着理想的消费者时刻仍然相差三个数量级，但这也意味着消费者革命实际上在 2029 年。

Yeah. I mean, I would argue, if the Opus 4.5 moment was just in December, you have to pay tens to hundreds of thousands of dollars to get real work done with those things. It means that the ideal consumer moment is still like three orders of magnitude away, but that means that the consumer revolution is actually in 2029.

听起来没错。我的意思是，这对我来说有道理。我想你想想互联网的初期，电信和基础设施有大量投资，有点像今天，但大部分是暗光纤。所以，我们知道那是不一样的。然后投资者对消费者互联网有巨大兴趣。然后泡沫破裂，我们看到 eBay 早期起飞，但直到 2002 年底，我才开始看到亚马逊的季度数据持续攀升。他们一直在做自己的事，但从 2002 年秋季开始，他们宣布了一个大季度，这有点令人震惊，因为互联网本应已经结束，那时我就想，‘哦，终于来了。’所以按照你的说法，那是在开始六年后才实现的。

That sounds right. I mean, that makes sense to me. I guess you think about the beginning of the internet and there was huge investment in telecom and infrastructure just kind of in an analog to today, but it was mostly dark fiber. So, we know that was different. And then there was huge investor interest in the consumer internet. And then that crashed and we saw eBay take off early, but it wasn't until late in 2002 that I started to see Amazon's numbers start to consistently climb quarter on quarter. They were kind of doing their thing, but it was starting in the fall of 2002, they announced a big quarter and it was kind of shocking because the internet was supposed to be over and that's when I was like, 'Oh, this is finally happening.' So to your point that was a good six years in before we got to it.

你怎么看待所有这些事情的时机？你知道，那是个很好的例子，对吧？这些事情似乎需要很长时间，然后你几乎错了十年，然后突然每个人都觉得，‘哦，你太对了。’然后最后，对吧？然后每个人都觉得晚了。是的。他们确实晚了。

How do you think about timing for all this stuff? You know, I mean that's a great example, right? These things seem to take a long time and then you have to be wrong for like almost a decade and then suddenly everyone's like, 'Oh, you were so right.' And then finally, right? Then everyone feels late. Yeah. And they are.

那他们就晚了，因为你必须现在就开始做某件事，这件事会在三四年后爆发。对于那些说‘我连明天午饭吃什么都不知道’的人，你有什么建议？你是怎么找到你的持久力的？

Then they are late because you had to have started something right now that like took over in three or four years from now. Do you have any advice for people who are like, well, you know, I don't even know what I'm going to eat for lunch tomorrow. You know, like how did you find your staying power?

我一直热爱打造消费产品，而且随着时间的推移，我发现我似乎比早期大众市场领先了大约 18 个月。我是从自己作为用户的角度来思考的。有趣的是，我的手机一半是空的，除了 GPT 和 Claude 之外，没有其他消费应用能吸引我的注意力。肯定不是游戏。

I've always had a love for building consumer products and I've found over time that I seem to be like 18 months ahead of the early mass market majority. I think about it that way in terms of myself as a user. And it is interesting that half my phone is empty other than GPT and Claude. There aren't consumer apps that are grabbing my attention right now. Definitely not games.

所以保持持久力很难，但这取决于你怎么看。当你没有爱上任何东西时，很难继续建造东西。我们需要走出去，爱上一些东西来获得灵感。我们进入的这种状态，我称之为‘深渊’。深渊就是我们发现自己处于热情追求产品之间的地带，不确定是否还能再找到让自己充满激情的东西。

And so it's hard to have the staying power, but it depends on how you think about it. It's hard to just keep building things when there's nothing that you're falling in love with. We need to go out and fall in love with things to get our inspiration. And so these zones we get in, I call the abyss. The abyss is this place in between passionately pursuing products that we find ourselves in where we're not sure if we'll ever find something we're passionate about again.

是的。

Yeah.

我们不知道是否还能走出来。然后就会想，在深渊里该做什么？但我认为，在深渊的这段时间，你实际上可以扩展你的品味范围，并可能找到一些能带你走出去的东西。只是当时你不知道而已。

And we don't know if we'll ever come out. And then it's like, well, what do you do in the abyss? But I think that during this time in the abyss is when you actually can expand your taste zones and potentially find things that are part of your way out. You just don't know it at the time.

嗯，我们处在一个奇怪的时刻，我越想越觉得这像是一个成本问题。真正神奇的东西你得花很多钱。

Well, we're in this weird moment where, you know, the more I think about it, I feel like it is like a cost problem. Like the really magical things you have to pay a lot of money.

我们必须达到杰文斯悖论的那个点，对吧？就是你可以挥霍它。这在游戏中绝对是个问题。

We got to get to that Jevons paradox point, right? Where you can squander it. And it's definitely a problem in games.

是的，尤其是在大多数消费产品和游戏中，每一分钱都很重要，所以你怎么让它可行？你实际上无法进行那种计算和处理。这就是为什么，你知道，埃隆说游戏将完全由 AI 在你环顾四周时逐像素生成，我想那会非常昂贵。我是说，直到我们建造足够的 GPU。

Yeah, especially with a premium product in most consumer and games, every cent counts and so how do you make it work? You can't actually do that kind of compute and processing. That's why, you know, where Elon says games are just going to be created pixel by pixel by the AI while you're looking around, I'm like that's going to be very expensive. I mean, until we build enough GPUs.

我的朋友 Suddosi 刚刚提到，我们可能可以指望的一件事是推理能力提升一万倍，就像 Jeff Dean 说的那样，所以有一个关于这一切将如何发展的分解。

My friend Suddosi was just mentioning, you know, one thing that we could probably count on is like a 10,000x increase in inference, like from a Jeff Dean, so there's a breakdown of how all this stuff is going to go down.

你不能跟埃隆对着干，所以埃隆很可能是对的。只是时间点不同，可能比我们意识到的来得更快。他看到了一个点，你有那么多 GPU，任何一个周期或 token 的成本如此之低，以至于我们可以在游戏中到处挥霍。

You can't bet against Elon, so Elon is probably right. It's at a different point in time that maybe is coming faster than we realize. He sees a point that you have that many GPUs and the cost of any one cycle or token is so cheap that we can squander it everywhere in games.

这个时刻即将到来。对于那些看到最后的人，他们会想，消费领域的故事是什么？这可能就是那个故事。

This moment is coming. For people who are watching and they're watching to the end and they're like, well what is the story about consumer? That might be the story.

可能是，而那个时间机器时刻可能就是回到今天说，还记得我们不明白这就是智能的时候吗？我想你说过像水一样即开即用，对吧？我们还没到那一步。

Might be, and that might be the time machine moment is to come back to today and say, remember when we didn't understand that this was intelligence? I think you said on tap like water, right? And we're not there yet.

但如果你能进入那个思维空间，那可能是思考消费领域的正确方式。如果你想想当算力基本免费时，所有这些服务会变成什么样？如果我可以随时无限量地使用它，我会怎么用？你知道，我们会有我们的应用一直在监听……

But if you can get your head in that space, it's probably the right way to think about consumer. If you think how will all of these services be when that compute is basically free? How would I use it if I could just use unlimited amounts for everything all the time? You know, we'd have our app that would be listening and...

是的，因为你现在就可以有那个应用。只是每月要花 1000 美元。所以你负担得起，我负担得起，我们一群朋友也负担得起，但这不是消费级。这基本上是企业级。

Yeah, because you can have that app now. It'll just cost $1,000 a month. And so you could afford it and I could afford it and a bunch of our friends could afford it, but that's not consumer. That's basically enterprise.

对。对。

Right. Right.

所以这只是时间问题。实际上，之前我开玩笑说我们都不是时间旅行者，但实际上我们现在就是时间旅行者。大多数花不起那么多钱的人还无法得到它，这让我很痛苦。

And so it's just a matter of time. Actually, earlier I was making a joke about like neither of us are time travelers, but actually we are time travelers right now. Most of the people who can't spend that much money can't get it yet and that pains me honestly.

或者你可以受到启发，说每个人最终都会很快拥有这个。

Or you could be inspired to say everyone eventually will have this like soon.

所以我们现在必须搞清楚基本要素。你可以现在就开始构建，然后倒推。我的意思是，那是一个不可思议的先发优势。你知道，这是产生创业想法的经典方法之一：你拿任何成本曲线或能力曲线的对数图。你可以猜出 iPhone 可能出现的时刻，因为内存成本下降、显示成本下降、算力成本下降到人们能买得起 iPhone 的程度。

So we have to figure out the primitives right now. You could start building now and work backward. I mean that's an incredible head start. You know, that's like one of the classic ways to come up with startup ideas: you take any log graph of a cost curve or some capability curve. You could have guessed the moment that the iPhone could have been possible because memory came down and display cost came down and compute came down just to a point where someone could afford an iPhone.

是的。

Yeah.

所以你现在就可以用智能的成本来做这件事。我喜欢‘免费’的商业模式。如果你想说，你一直知道什么更好？免费，对吧？这是互联网的规则之一。任何可以免费的东西最终都会免费。

And so you can do this right now with the cost of intelligence. I love the business plan of free. If you want to say what do you always know is better? Free, right? And it's one of the rules of the internet. Like anything that can be free will be free.

我的第一家公司 Freeloader，我们有一个免费的互动屏保，与 Berkeley Systems 售价 35 美元的飞行烤面包机竞争，对吧？这是一个不错的胜利，对吧？它甚至不是新的，只是更好。你想要的。是的。

And my first company, Freeloader, we had a free interactive screen saver that competed with Berkeley Systems' $35 flying toasters, right? It's a good win, right? It's not even new. It's just better. You want. Yeah.

但新的是互联网，互动下载互联网，而更好的是免费的屏保。然后到了 Zynga 和社交游戏，游戏行业只有 230 亿美元，他们只有一个模式：去商店买一个盒子，60 美元。我们问，我们能做出质量不错的免费游戏吗？你知道，免费又奏效了。所以我喜欢思考，在 AI 和算力的时代，免费意味着什么？当它无限且免费时，我们将如何重新思考所有这些服务？

But the new was the internet, the interactive download the internet, but the better was just free screen saver. And then with Zynga and social games, again, like the game industry was only $23 billion and they had one model which was go to the store and buy a box and it was $60. And we said, can we make decent quality games free? And you know, the free worked again. And so I love thinking about well what does free mean in the age of AI and compute and like how will we rethink all of these services when it's unlimited and free.

我的意思是，有趣的是，黎明前总是最黑暗的，对吧？所以也许就是那个时刻。这就是为什么我们坐在这里感叹消费领域不可投资，但它会变得可投资。我们现在就知道了。是的，这是一个很好的论点，想象任何一个特许经营或消费服务，如果你在其中加入免费无限的 AI，它会有什么不同。

I mean what's interesting is like it's always darkest before dawn, right? And so maybe that's the moment. That's why we're sitting here lamenting that consumer is not investable, but it will be. And we know this now. Yeah, it's a good argument to imagine any one of these franchises or any of these consumer services and think how would it be different if you offered that with free unlimited AI inside of it.

那时我们将看到新的 Meta。那时我们将看到新的 Snap。那时我们将看到所有的未来。而且这甚至不是几年后的事。

That's when we're going to see the new meta. That's when we'll see the new Snap. That's when we're going to see all of the future. And it's not even a couple years away.

是的，Mark，这太棒了。非常感谢你抽出时间与我们交流。对于所有观众，请务必去买他的新书。它将在任何卖书的地方有售。感谢你的加入。

Yeah, Mark, this is so amazing. Thank you so much for spending time with us. And for everyone watching, make sure you go out and buy his new book. It's going to be available anywhere books are sold. Thanks for joining us.