What Founders Have To Unlearn To Become Great CEOs

到了一两年后，从理性角度来看，你可能真的应该放弃了。

There is a point that you get to a year, maybe two years in, where from a rational standpoint, you probably should quit.

但不知为何，那些成功的人并不会这么做。

But for whatever reason, those successful ones don't.

因此，这是最重要的筛选标准。

And so that is the number one filtering criteria.

我能给出的最佳建议是，清楚地知道自己想学什么，然后敞开心扉，接受来自任何地方的启发。

The best advice I can have is be clear in your own head about what you're trying to learn and then, you know, be open to where it comes from.

这就是为什么我认为很多人在这方面都搞砸了。

And that's why I think people fuck that up a lot.

他们没有真正清晰地弄清楚自己为什么想创办初创公司，或者要取得成功需要做什么。

They don't get really crystal clear on why they're trying to build a startup or what they need to do to be successful.

在未来几年，数据分析领域将迎来一次革新，而我们希望成为引领这一变革的人。

There is gonna be a reinvention of analytics on the next few years, and, you know, we we wanna be the ones to go lead it.

欢迎回到《光锥》的又一期节目。

Welcome back to another episode of the Lightcone.

今天，我们非常高兴邀请到Amplitude的首席执行官兼联合创始人斯宾塞·斯凯茨。

Today, we're really excited to be joined by Spenser Skates, CEO and cofounder of Amplitude.

Amplitude在2012年冬季参加了Y Combinator。

So Amplitude went through YC in winter twenty twelve.

Amplitude是全球领先的分析平台之一，被世界上一些最大的公司使用，比如斯宾塞、DoorDash和沃尔玛。

Amplitude is one of the world's leading analytics platforms, and they're used by some of the biggest companies in the world, like Spenser, DoorDash, and Walmart.

感谢你加入我们，斯宾塞。

Thanks for joining us, Spenser.

当然。

Absolutely.

很高兴在这里见到你，哈吉。

Good to see you here, Harge.

我非常期待你能来，因为最近我们俩都在互联网上惹怒了很多人，尤其是在X或Twitter上。

So I was really excited to have you here because both of us made people on the internet angry recently, specifically on X or Twitter.

我惹怒他们是因为我说，许多现有科技公司无法打造AI产品的原因，是工程师们情绪低落，不相信AI及其能力。

I made them angry because I said that a lot of the reason incumbent tech companies can't build AI products is that the engineers are kind of grumpy and don't believe in AI and its capabilities.

他们不想开发这些产品。

They don't want to build the products.

人们因此对你发火了。

People got mad at you

就因为这个？

for that?

是的，人们真的很生气。

Yeah, people were really mad.

结果发现，确实有很多对AI持怀疑态度的工程师。

Turns out that like there are actually a lot of grumpy engineers that don't believe in AI.

所以我很想知道，作为一家在AI浪潮之前就已成立、如今正试图转向并开发更多AI产品的公司，

So I was curious to hear for you as like a company that started well before the AI wave and is now trying to move in, is moving into AI and building more AI products.

这种转变对你来说怎么样？你们遇到了哪些挑战？

How has that change been for you and what have been some of the challenges you faced?

对于一家大公司来说，重新调整并重建公司以更好地使用AI是很困难的。

It is hard, as a larger company to reorient and rebuild your company to use AI well.

而这正是你所提到的，许多从零开始构建的早期公司的巨大优势。

And this is, I think, to your point, the huge advantage that a lot of earlier companies that can build from the ground up in this way.

我的意思是，我来跟你讲讲Amplitude的故事。

I mean, I'll I'll just tell you the the Amplitude story.

事实上，我们也曾对AI持怀疑态度一段时间。

So we we were frankly skeptics on AI for a while too.

AI在2022年、2023年开始变得相关。

So started to become relevant in 2022, 2023.

当时有过讨论，但我们并没有真正做太多。

There's discussion, but we didn't we didn't really do that much.

直到2024年底，我们才意识到：好吧。

And it wasn't until late in 2024 that we're like, okay.

我们需要认真对待了，因为我认为这有潜力重塑分析领域和我们的工作方式。

We need to get serious because I think this is has the potential to reshape analytics and what we're doing.

我想，为了替那些怀疑者说句话，我记得曾经参加过一次董事会会议。

I I think, you know, to just to defend the skeptics for a second, I think I I remember being in a board meeting once.

你有这么多董事会成员、投资者和财务人员，还有这么多销售人员都在说：嘿，你们不该看看这个AI吗？

You had, like, you know, all these, like, board investor finance people, and you had all these salespeople who were like, hey, Shouldn't you look at this AI thing?

这不是很火吗？

Like, isn't that getting hot?

你们难道不该做点什么吗？

Shouldn't you guys do it?

那你觉得怎么样呢？

And it's What's like

你们的AI战略是什么？

your AI strategy?

是的。

Yeah.

没错。

Exactly.

没错。

Exactly.

是的。

Yeah.

实际上，这真的是我们的一位高管问我一个问题。

Actually, that literally was a question, like, from one of our execs to me.

我们必须制定我们的AI战略。

It's like, we gotta get our AI strategy.

我们的AI战略是什么，斯宾塞？

What's our AI strategy, Spenser?

我觉得这种想法完全误解了我们正在做的事情。

And I'm this is the wrong way to think about what it is we're doing.

如果你觉得你们能搞清楚，那就去干吧。

It's like, if you guys think it is you guys think you can figure it out, go do it.

我记得我的一位联合创始人杰弗里特别沮丧，因为他看到太多人在AI领域搞噱头，说什么AI会取代所有工作，会做各种事情，而我们却什么也做不了，它会创造一个富足的世界，彻底改变我们构建产品、交付产品以及客户使用产品的方式。

And I remember one of my cofounders, Jeffrey, actually was the most frustrated because I think he saw a lot of grifting happening in AI being like, it's gonna replace all these jobs and like, you know, it's gonna do all these things and we're not gonna you know, it's gonna create this world of abundance and it's it's just gonna totally reshape how we're building products and shipping them and how our customers are using them.

但现实是，如果你现在使用这些模型，根本不清楚它们到底能做什么。

But the the reality was if if you use any of these models at this point, it's like it was not it was not clear at all.

我认为，如果你看看这些模型的能力，就会发现它们非常不均衡。

I think if you look at the capabilities of any of these models, it's like they're very, very jagged.

有些方面它们表现得极其出色，但有些方面却糟糕透顶。

So there's some things they're exceptional about, and there's some things that they're just absolutely terrible about.

令人沮丧的是，有人根本不清楚哪些功能属于哪一类，却还来告诉你该怎么做。

I think the frustrating part is to be told by someone who has no clue about, you know, what what stuff falls in what buckets.

是的。

Like, yeah.

你应该多做点这个AI相关的事情。

You should do more of this AI thing.

那真的让人很沮丧。

It was very that was frustrating.

因此，坦白说，我对此持怀疑态度。

And so I think there there's frankly skepticism on my part.

我的联合创始人以及Amplitude的其他一些团队成员，也对AI究竟能实现什么持怀疑态度。

There's there's skepticism on the part of of my co founders and and some of the broader team with an Amplitude about what was actually gonna be possible with it.

但有些事情发生了变化。

But something changed.

是的。

Yes.

嗯，我认为你看到了人工智能对软件工程产生的变革性影响。

Well, I think you you saw the transformative effect that AI had on software engineering.

毫无疑问。

Like, no question.

如果你使用过——我是说，从 Cursor 开始，然后还有这些其他出色的工具，比如 Claude Code、Codex，还有很多很多其他的。

If you were using I mean, started by Cursor, but then, you know, all these other amazing tools, Claude Code, Codex, like tons, you know, tons and tons of others.

很明显，使用这些工具会让你高效得多。

It was it was very clear that you would be a lot more productive using these things.

因此，这是我们最初开始说的：好吧。

And so that was kind of the first of us saying, okay.

那里确实有东西，我们得去抓住它。

There's something there there, and let's go after this.

所以我们真正开始这条道路是在2024年10月左右，当时发生了两件事。

So we really started on this path in earnest in around October 2024, and kind of two things happened then.

第一，我聘请了一位新的工程负责人，这位叫韦德·钱伯斯的人是硅谷的传奇人物。

One, I hired a new engineering leader, this guy Wade Chambers, who's a Silicon Valley legend.

第二，我们收购了一家公司，Command AI，也是一家YC孵化的公司。

The other is we acquired this company, Command AI, also a YC company.

在这两种情况下，他们都是Amplitude所需要的变革推动者。

In both their cases, they were kind of the change agents that Amplitude needed.

韦德之前在公司里一直在做AI相关的工作，认识很多走在模型应用前沿的人。

So Wade had been working on AI in his previous company and had known a bunch of people who were on the bleeding edge of leveraging the capabilities of models.

他们一直在开发一个产品，做了很多不同的事情，但核心是根据用户是否感到困惑，智能地触发引导提示。

They had been building a product where they were trying to they're doing a bunch of different things, but they were trying to smartly trigger guides to end users just based on if they were confused.

他们创建了一个聊天机器人，非常像Intercom Fin，会与用户互动，帮助他们解答支持问题之类的。

And they, you know, created this chatbot that very much like an intercom fin that would, yeah, interact with users and, you know, help them answer support questions and stuff like that.

他们一直在充分利用各种模型的能力。

And they had been, you know, on on leveraging a bunch of model capabilities.

那正是我们开始认真对待这件事的起点。

That was kind of the the first point that we started to to get serious about it.

从那以后，我们一直在稳步前进，我觉得我们正逐渐取得进展。

And then since then, it has been a very I'd like to think we're kind of coming up.

我的意思是，我们在过去几周推出了一系列AI产品。

I mean, we just launched a whole bunch of AI products in the last few weeks.

我们推出了AI反馈功能。

We did AI feedback.

我们推出了AI可视化功能。

We did AI visibility.

我们推出了我们的MCP服务器。

We did our MCP server.

接下来在十二月、一月和二月，我们将推出我们称之为‘分析版光标’的产品，我对此无比兴奋，相信这将彻底改变人们使用和利用分析的方式。

And then we're going to go much, much bigger in December, January and February, where we are going to be coming out with what we call the cursor for analytics, which I'm incredibly excited about, and I think will dramatically change how people use and leverage analytics.

但要实现这种转变，确实需要重塑Amplitude这家公司的整体架构。

But it's taken to transform, you know, Amplitude as an organization.

我觉得我们仍然算是个小公司。

I like to think we're still small.

我们大约有800人。

We're about 800 people.

产品、工程和设计团队大约有200人。

Product engineering and design, that organization is about 200 people.

所以，你知道，你几乎认识里面的所有人，能够很快地做出改变。

So, you know, it's like you can know most people in it and change pretty quickly.

但即便如此，我们还是花了一整年时间，才让团队完全融入、上手、相信并真正看到和构建起来。

It's still taken us a full year to get the team kind of fully on board and ramped and believing and seeing and building.

ChatGPT发布后，AI开始兴起。

ChatGPT launches, AI starts taking off.

听起来，你的投资者、董事会成员，甚至一些不参与日常运营的人，都会问：你们在AI这件事上到底在做什么？

And it sounds like your investors and board members and maybe people who aren't in the day to day up sort of, hey, what are doing with this AI thing?

他们关于

What about the They rent about

在TechCrunch之类的平台上。

it in TechCrunch or something.

是的，没错。

Yeah, exactly.

对，没错。

Yes, exactly.

没错。

Exactly.

那内部呢？

What about internally?

比如在Amplitude团队内部，氛围如何？

Like what was the vibe internally at Amplitude on the team?

有没有人内部问过，我们是不是应该开发AI产品或做点与AI相关的事情？

Like was there anyone internally asking, hey, like should we be building AI products or doing something with AI?

有几个人在尝试一些想法，但我认为我们当时太专注于日常的事务了。

There were a few people that were kinda testing out ideas, but I I think we're so focused on our regular motion.

我们当时正在推出许多其他非分析类产品。

Like, we were coming out with a lot of other products outside of analytics.

我们推出了实验功能，你知道的。

We were doing know, you we launched experimentation.

我们内部正在开发会话回放功能。

We were building session replay internally.

我们正在开发一个名为激活的功能，根据用户行为来定位他们。

We're building this thing activation, which targets your users based on their behavior.

因此，眼前就有大量明确要做的事情，比如，好吧。

And so there there was a lot to do that was just kind of right in front of us that was clear, like, okay.

嘿。

Hey.

如果我们能更具有竞争力，你知道的？

We can be much more competitive with you know?

而且，如果我们要开发这些功能，就有现成的收入在那里等着我们。

And and there's this revenue that's just sitting here if if we go build these things.

你知道，有几个人在摆弄它，还有一些人用过Cursor和其他工具。

You know, there are a few folks that were messing around with it, and there there were some people that had used Cursor and other things.

但整个组织并没有意识到，或者说并没有清醒地认识到即将发生的这一巨大变革。

But the the the organization as a whole did not have it's not like it was conscious and aware of this change, this this massive change that was about to happen.

我要给韦德和指挥团队很多赞誉，他们是率先展示可能性并推动整个组织接受变革的先锋。

I give a lot of credit to Wade, and I give a lot of credit to the command team of being the tip of the spear in terms of showing people what is possible and then getting the organization to to embrace.

到2025年初，我已经确信了，我觉得我们必须在这件事上采取非常积极的行动。

By early twenty twenty five, I was convinced, and I was like, all right, we have to be very aggressive on this.

所以第一步是让整个组织了解AI的能力，特别是工程方面的能力。

So the first order is to train the organization of what the capabilities of AI are, specifically engineering

对你来说，那个顿悟的时刻是什么时候？

What was, like, the light bulb moment for you?

你是什么时候突然意识到，这必须成为我的优先事项？

When did you flick into, and this needs to be my priority?

很难说有一个具体的瞬间。

It's hard to say that there was there was one moment.

我们一直有这样一个愿景，即Amplitude的产品能够自我进化，能够动态响应用户的反馈。

We've always had this vision at Amplitude of a self improving product, where you have a product that dynamically responds to your user feedback.

它知道你喜欢哪些功能。

So it knows what features you like.

它知道你何时感到沮丧或卡住。

It knows when you're getting frustrated and stuck.

它知道如何根据你的用户输入来调整产品。

It knows how to change things based on your input as a user.

我曾经认为这个愿景至少需要十年才能实现，甚至可能都没那么快。

And I had always thought this vision was like, you know, ten years out, if if, you know, if even that close.

但如今越来越清楚的是，这个目标其实已经近在咫尺。

And one of the things that is becoming clear, it was actually a lot closer.

由于我们在编码端看到AI带来的变化，这个转折点实际上比我们意识到的要近得多。

That that moment was actually probably a lot closer than we had realized because of what we saw happening on the coding side with AI.

我说：好吧，不管自我进化的产品最终能否实现，至少为了朝这个方向迈出一步，我们必须去行动。

I said, Okay, look, we you know, whether or not a self improving product is going to be possible, it's clear at least a step towards it, we're going to have to go do it.

所以，要做到这一点，我们必须让整个组织接受这方面的培训。

So and the way to do this is we're gonna have to train the the organization on this.

我开始与Command的创始人兼CEO James以及我们的工程负责人Wade合作，探讨如何让组织接受AI培训。

I started working with James, the founder CEO of Command, as well as Wade, our engineering leader, to figure out how do we train the organization on AI.

然后我们提出了一个‘AI周’的计划，但由于种种原因，我们直到六月才真正实施。

And then we we came up with an AI week, and unfortunately, for a bunch of reasons, we weren't able to actually do that until June.

但那确实是一个关键的转折点。

But that was a that was a key pivot point.

我们召集了组织内的一批现有领导者。

What we did was we got a bunch of the existing leaders in the organization.

比如我们的产品副总裁、工程经理，让他们亲自使用这项技术，看看它能实现什么可能性。

So, you know, our VPs of product, our engineering managers to use this technology and to see what was possible about it.

在那场AI周期间，我们会对团队进行培训。

And then during that AI week, what we do is we train the team.

我们让一位产品负责人当着整个组织的面，用Vibe Code为Amplitude快速实现了一个深色模式。

We had our like, one of our product leaders, you know, vibe code like a dark mode for Amplitude in front of the entire organization.

这其实挺吓人的，但他们碰巧遇到了一个bug，不过还是解决了。

It was actually very scary, but actually, it it happened you know, they ran into a bug, but they'd happen to sort it out.

这其实是个很酷的时刻，因为整个工程、产品和设计团队都看到了，哇，原来如此。

It was actually kind of a cool moment because the entire engineering product and design organization saw what was it's like, wow.

好吧。

Okay.

你知道，所有Amplitude的领导者都在说这是关键，而且他们正在展示自己是怎么做的。

You know, all of the leaders with an Amplitude are saying this is the thing, and they're showing how they're doing it.

我也得赶紧跟上这波潮流了。

And, you know, I better get with this as well.

接下来的一周，我们大概培训了两天，剩下的时间就像一场黑客马拉松，让大家继续做自己手头的工作，只是用Cursor和其他一些工具来更快更好地完成。

And then the rest of the week, you know so we did training for about two days, and then the rest of the week was just, you know, like a hackathon, work on stuff you're already doing, except, you know, do it faster and better by, you know, getting set up with Cursor, as well as a bunch of these other pieces of tooling.

所以第一步并不是直接说：‘我们得用AI在Amplitude开发什么新功能。’

So step one was actually not like, here's what we need to build at Amplitude with AI.

第一步只是让现有团队先用起这些工具

Step one is just get the existing team using the tools and

是的。

Yes.

并且相信它们。

And forth in and believing in them.

这是在SaaS中构建产品与使用AI构建产品之间的核心区别。

This is the core difference between building product in SaaS and building product with AI.

在SaaS中，我的意思是，这是有史以来最好的商业模式和最佳的产品交付体系。

In SaaS, SaaS, I mean, it's the best business model and the best product delivery system of all time.

你去问客户他们想要什么、愿意为哪些功能付费，然后对这个列表进行优先级排序，开始开发，交付给客户，再一遍又一遍地重复整个过程。

You go to your customers, you ask them what they want and what they're gonna pay for, you prioritize that list, and you start building it, you get them to those customers, and then you just do the whole thing over and over again.

在过去十年里，Amplitude已经精通了这种交付循环。

That's the delivery loop that Amplitude has mastered over the last decade.

而你知道，这就是我们的核心竞争优势，作为一家分析公司，我们在这一点上比任何人都更出色。

And, you know, that's that's our core competitive advantage, you know, as as an analytics company is we're better at that than anyone else.

关键在于，你可以去问你的客户。

The key is you can go to your customers.

他们能告诉你他们想要什么。

They can tell you what they want.

由于AI的能力如此参差不齐，你需要从技术角度出发，理解什么是可能的。

With the capabilities of AI, because they're so jagged, you it's a technology first understanding of what is possible.

所以，如果你去问客户他们想要什么，他们甚至无法描述出什么是可能的。

And so if you go to your customers and tell them and ask them what they want, like, they're not even gonna be able to describe what's possible.

给我一匹更快的马。

Give me a faster horse.

是的。

Yeah.

给我一匹更快的马。

Give give me a faster horse.

你知道的。

You know?

或者他们会要求一些根本不可能实现、或者以错误方式实现的东西，比如，嘿。

Or or it'll be asking for something that's not quite possible or possible in the wrong way or like, hey.

你知道，我想要一个能向我推送洞察的东西，并且以这种方式实现，你会看到各种不同的版本。

You know, I want something that ships insights to me and, you know, does it in this particular way, and you'll get all sorts of different visions of it.

更重要的是，你必须熟悉这些模型的能力，然后理解它们如何映射到你的产品功能上。

And what's much more important is you you have to be familiar with the capabilities of the models, and then how those can map back into what your product does.

我想要的是

The thing I

关于这一点，我感到特别惊讶的是，通常新技术的采用模式是工程师作为早期使用者，他们会不断向公司施压。

just find really surprising about this is usually the way it works with new technologies is like the engineers are the early adopters and they are usually like bugging their company.

我真的很想用这个工具，而公司却总是抵制，说它还没经过测试，不安全。

I really wanna use this tool and it's like the companies are resisting saying well it's not tested yet, it's not safe.

但AI似乎正呈现出相反的趋势。

But it seems to be a more common pattern with AI that it's going the other way.

是的，这是自上而下的。

Yeah, it's tops down.

是的，你认为为什么会这样？

Yeah, why do you think that's happening?

我的意思是，这听起来可能过于简化了，但我认为萨姆·阿尔特曼是这一代最出色的推销者，没有之一。

I mean, it's gonna sound extraordinarily reductive, but I think Sam Altman is the best salesperson of this generation by no bar none.

我认为他出色地阐述了一个极具雄心的愿景，成功凝聚了OpenAI内部的大量人才，展示了这项技术的潜力，并说服了全世界，让他们认识到这项技术将产生的深远影响。

Like, I think he has done an exceptional job stating a very ambitious vision, getting a lot of people rallied behind it at OpenAI, showing what's possible and convincing, you know, the entire world of the impact that this technology is going to have.

因此，你看到投资者已经信服，企业高管也信服了，甚至世界领导人也都信服了——作为社会整体，掌握权力的人们已经表示：这项技术很重要。

And so you have these, you know, investors already bought in, executives are bought in, know, world leaders are bought in, like, you know, as a society, you know, the people in power have said, Hey, this technology matters.

但现实是，技术能力仍在追赶这些宏伟目标，目前还不清楚它们是否能真正实现这些抱负。

And the reality is the capabilities are still trying to catch up, and it's not clear if they'll catch up to those aspirations.

所以，正如你提到的，哈吉，这里出现了一种相反的脱节：人们非常渴望这件事发生，但技术能力是否已经到位，仍然不明确。

And so you have, to your point, Harge, the opposite disconnect happening where, you know, yeah, you have a lot of desire for this thing to happen, but, you know, it's not clear if the capabilities are are there yet.

因此，这就是为什么你看到了大量的挫败感。

And so this is why, you know, you had a lot of frustration.

以Amplitude为例，工程师们感到非常沮丧，他们认为AI领域充斥着大量空谈者，真正实干的人却很少。

You know, in Amplitude as a case study, had lot of frustration from engineers being like, man, I see just tremendous what they feel is grifting in AI, where it's a lot of talkers, not many doers.

直到去年，大家才开始意识到：哦，技术能力终于跟上了，足以改变我们所处的分析行业。

You know, it's it's not till the last year that, you know, okay, hey, think the capabilities are here for it to transform the the business that that we're in in analytics.

我想到的一点是，公司以创始人模式运作的方式更加明显了。

One thing that comes to mind is this mode of running the company more like founder mode is even more present.

是的。

Yes.

关于案例标签的转型，有一个故事。

And there's a story of the transformation with case tags.

我们完全是由创始人主导完成这一转型的。

We're just founder led to do that transformation.

听起来，Amplitude 二号的这种变化正是源自你们。

Sounds like this is what's going on with Amplitude two is coming from from you.

绝对如此。

Absolutely.

在过去十年里，我不得不学习如何从创始人转变为大型企业的高管。

So I've had to learn over the last ten years how to be go from being a founder to being a large company executive.

除非亲自使用这项技术并站在最前沿，否则无法真正理解什么是可能的。

There is no way to understand what is possible than being using the technology and being on the front lines of it.

所以，这就是为什么你会说，要培训整个组织，关于什么是可能的，这是一种自下而上的方式。

And so, you know, that's why it's like, yeah, you train the entire organization, and it's a much more bottoms up thing on what is possible.

因此，正是在那场AI周之后，我们目前所有在做的AI相关工作都诞生了。

And so out of that AI week came basically all the AI stuff that we're working on now.

我们的MCP服务器，其实我们根本没计划过，这其实是我们的工程师Brian Giuri提出的，他对此非常热情。

So our MCP server, which we didn't even plan for, like that was actually one of our engineers, Brian Giuri, who's incredibly excited.

我们有一位工程师，叫刘章，他开发了AI可视化功能。

You had one of our engineers, this guy named Liu Zhang, who built AI visibility.

他原本打算离开Amplitude去创业。

And he just wanted to He was actually going to leave Amplitude to start a company.

我们就跟他说：别走，留下来，边拿工资边学怎么做这件事，我们会教你、指导你。

And we're like, look, just come here, stay here, learn how to do this while getting paid, and we'll teach and coach you.

等你将来真要去做点什么的时候，我们会提供资金支持。

And then, you know, we'll fund whenever you do go something.

我本来不希望他做AI可视化，但他却说：我觉得这里有个巨大的机会，可以免费开发并开源它。

And I didn't want him to do AI visibility, but he was like, oh, I think that there's a huge opportunity here to, like, build something for free and give it away.

而且，你知道，现在这已经像一场爆发了。

And, you know, now that's now, like, you know, it's been this explosion.

我们的产品发布让Amplitude的新注册用户翻了一倍，这太惊人了，你知道吗？

Like, we have that product launch doubled new sign ups to Amplitude, which is wild, You know?

每周，我们免费计划的新注册用户数量都是AI可视化发布前的两倍，一直到我们一月即将推出的项目——我们将推出一个叫‘Ask AI’的全局聊天界面，非常类似于Cursor，你可以和AI聊天，让它为你提取特定图表、进行分析，或者帮你弄清楚你数据中某些现象发生的原因，并把这些结果反馈给你、我们的代理，以及众多其他功能，这一切都源于自下而上的创新，而对我来说和Wade来说，关键在于如何在组织中为这些创新设定成功路径。

Like, every week, it's like we're getting twice as many sign ups to our free plan as we did from before our AI visibility launch, all the way to what we're going to be launching in January, which is we're going to launch a what we're going to call Ask AI, which is a global chat interface, very much like Cursor, where you will be able to chat with yeah, chat with AI and ask it to pull certain charts for you or do analysis or figure out why something has happened to you for within within, you know, your data, and pull that back out to you, to our agents, to like so many different things have come out of this from but it's all been very bottoms up, and then it's about, you know, for me and for Wade, to sculpt, Okay, how do we set these up in the organization for success?

像你们这样规模的公司，很少有能完成这种转型的。

So not many large companies like yourself have been able to do this transformation.

很多公司都

Lot of them are

仍然很困难。

still hard.

陷入僵局，克服这一切可能非常痛苦。

Stuck, and it's been probably very painful to overcome.

为了达到你现在的位置，你们不得不放弃哪些东西？

What are the things that you had to give up to get to where you are right now?

我的意思是，自从今年年初以来，我们已经在工程、产品和设计团队中进行了两次重组。

I mean, we've done two reorganizations in the the engineering product and design organization since the since the start of this year.

所以，你知道，当时有一些领导者和高管，他们非常适应SaaS模式，但并没有紧跟AI的前沿，不幸的是，他们并不完全符合我们未来要打造的产品方向，因此我们不得不让他们离开公司。

And so, you know, there were leaders and executives and and different people who were very much kind of in the SaaS modality, but were not on the bleeding edge of AI that just unfortunately were not quite the right fit for what we were trying to build in the future, and so ended up having to move folks out of the business.

我的意思是，一年内进行两次如此大规模的重组，是非常具有破坏性的。

I mean, doing that level of reorganization, you know, twice in a year is very disruptive.

我们还收购了其他公司。

We've also acquired other companies.

所以我们吸纳了Craftful团队，还有Yana，她表现得非常出色。

So we brought in the Craftful team and Yana, who's been phenomenal.

我们还引入了来自Inari的Eric和Frank。

And we brought in Eric and Frank from Inari.

我们还引进了一家YC孵化的公司。

We brought in also a YC company.

我们还带来了来自June的Enzo和Frutio，我想他们也是YC孵化的公司。

We brought out Enzo and Frutio from June, also I think YC company.

是的

Yeah.

所以我们引进了这些优秀的YC创始人，并将他们与许多长期的Amplitude员工融合在一起。

So we brought in all these great YC founders, and then kind of melded them with a lot of longtime Amplitudeers.

这种组合非常特别。

And that combination has been very, very special.

具体有什么不同？

What are the specific differences?

对于那些在前AI时代的SaaS世界中非常成功的人，尤其是工程师，他们成长于那个环境，那么相比之下，他们缺乏的是什么？我不确定‘缺乏’这个词是否准确，但他们在哪些方面不如那些在后AI时代成长起来、真正深信AI的AI原生工程师？

For people who are really successful, think especially engineers, who are successful in sort of pre AI SaaS world, kind of grew up in that environment, and then you compare, what are sort of the things that they're, I don't if lacking is the right word, but what are the things that they can't see or they're not as naturally adept at as people who came up in sort of post AI world and are like AI native engineers and believe in it deeply.

说得更直白一点，是年龄差距吗？

To say the more uncomfortable thing, like, it an age gap?

是年龄差距吗？

Is it an age gap?

这并不是纯粹的年龄问题。

It's not an age thing per se.

这是一种心态问题。

It's a mentality thing.

比如，在SaaS世界里，你可以做我之前提到的那个循环。

Like, in the SaaS world, you can take you can do that loop I talked about earlier.

与客户交谈，询问他们的需求，对这份需求列表进行优先级排序，然后开发、交付，再重复一遍。

Talk to customers, ask them what they want, prioritize that list, build it, deliver it, do it again.

就像，我不知道怎么说。

Like, kind of I don't know.

说起来可能听起来很老生常谈，但你希望以你所处领域的前沿技术为基础，然后开始思考：如果我要用AI原生的方式重新做一遍，为什么？

It feels very cliche to say, but you want to take kind of the state of the art in whatever field you're building for, and then start to say, Okay, if I were to redo this in an AI native way, why?

另一方面，我认为这些AI原生团队真正缺失的是，他们还没有学会理解产品、问题，以及为什么事物会以当前的方式被解决。

On the flip side, I think what a lot of these AI native teams are really missing is they haven't learned the product and the problem and why things are solved the way they're solved.

因此，他们试图从零开始创建全新的界面，却没有借鉴过去十年中在解决该问题时积累的既有经验。

And so, like, they try to create these new interfaces from scratch without drawing on the previous expertise that's happened over the last decade on whatever problem that they're solving.

所以，如果你试图从头构建一个分析界面，比如只从问题出发，而看不到你的数据，这实际上会带来自己的一系列挑战。

And so, you know, if you try to build an analytics interface from the ground up where you're just, say, starting with questions and you're not seeing your data, that actually has its own set of challenges.

所以我觉得，说到Amplitude团队，那里确实有一些非常出色的工程师。

And so I think, you know, what I'll say with the Amplitude team is like there's some incredible engineers there.

我认为适应得最好的那些人，一直都清楚代码本身并不是最终目的。

I think the ones that have adapted the best have always been very in tune with, okay, the code is not an end in itself.

交付产品只是为了解决客户的问题，代码只是这个过程的副产品。

It just like shipping it, that's just a side effect of solving whatever problem for the customer.

然后，嘿，我现在需要快速掌握这项新技术。

And then, hey, now I need to get up to speed with this new technology.

如果我能把这两者结合起来，我就能创造出令人惊叹的东西。

And if I can marry those two things, then I'm gonna create something amazing.

我想我注意到自己行为中一个奇怪的地方是，我们在YC内部有一个代理基础设施，它可以访问我们所有的数据库和其他资源。

I guess one of the weird things that I've noticed in my own behavior is we have an internal agent infra at YC, and it has access to all our, you know, databases and everything like that.

最近Jared告诉我，我算是这个系统的超级用户。

And then Jared recently told me I'm, like, sort of a super user for it.

然后我意识到，实际上它大部分时候还是失败的。

And then what I realized is I don't basically, it actually fails most of the time still.

但我只是试着换种方式问它。

But I just like try and ask it in a different way.

我试过，如果Gemini 3失败了，我就换Claude试试。

I tried, you know, if it failed with Gemini three, then I'll try it with, you know, Claude.

我会试试GPT 5.1。

I'll try it with GP 5.1.

我会调整推理层级。

Like, I'll change the reasoning levels.

然后最终，我找到了方法，它就成功了。

Like and then eventually, I figure it out, and it's like and it works.

而在B2B SaaS或者构建标准的Web 2.0软件时，如果哪怕出过一次错，这东西就是垃圾，我再也不会用它了。

So whereas, like, in b to b SaaS and or building, you know, sort of standard web two point o software, it's like, man, if it broke even once, like, this thing is a piece of crap, and I'll never use it again.

所以这感觉有点奇怪，是的。

So it's like a weird yeah.

你得重新调整自己的思维，就像在和一个孩子打交道一样。

You have to, like, rewire your brain to be it's like working with a child right now.

这就是为什么我认为AI会取代SaaS的很多炒作都被夸大了，因为对于许多业务流程来说，高性能的高可靠性是根本性的要求。

This is why I think a lot of the hype in that AI killing SaaS is way overblown because particularly for a lot of business workflows, like, very high guarantees on performance are fundamental to it.

你知道，如果你把一条记录录入到你的CRM系统里，这些东西必须得在那里。

You know, if you put a record into your CRM, that stuff better be there.

你知道，你并不希望它出错。

You know, you don't want it.

哦，它只是以80%的概率把这条信息放了进去。

Oh, well, it put this in with 80% probability.

你希望它确实存在。

It's like you want it to be there.

而SaaS在这方面做得非常好。

And SaaS does a great job of that.

它实际上是把原本在本地或通过纸质处理的一系列现有流程，直接迁移到了云端。

It actually took a set of existing workflows that were handled on prem or by paper and just moved them to the cloud.

所以这根本不是什么流程上的变革。

And so it wasn't really like it wasn't any workflow transformation.

我认为安德烈·卡帕西的观点完全正确，许多企业试图过度超越目标，直接说：好吧。

And this is where I think Andre Karpathy's point is totally right, where a lot of these businesses are trying to overshoot the mark and just say, okay.

嘿。

Well, hey.

我们干脆让这个代理完全接管整个工作流程。

We'll just have this agent handle this workflow end to end.

但实际上，正如你提到的，加里，编辑和重做这部分工作至关重要。

Actually, the editing and redoing it, like you like you talk about, Gary, is actually incredibly important.

因此，如何打造一个能让你很好地完成这些操作的产品，我认为是许多AI B2B公司正在解决的关键挑战。

And so how do you create a product that allows you to do that really well, I think, is the key challenge that a lot of AI b to b companies are figuring out.

说到为了做这件事你必须放弃的东西，Amplitude原有的产品路线图怎么样了？

Speaking of things that you have to give up in order to do this, what happened to the existing Amplitude road roadmap?

你之前不是还有其他很多在做的项目吗？在公司陷入AI狂热之前。

Like, you had all these other things that you were working on before the company got AI fervor.

你们把整个路线图都扔掉了吗？

Did you throw out the whole roadmap?

只有一半左右？

Only, like, half of it?

你们是怎么考虑在这些新的AI项目和现有的盈利业务之间分配资源的？

How are you thinking about, like, allocating resources between these new AI bets and the, like, existing stuff that makes the money?

并不是说有完全独立的产品。

It's not like there's there's totally separate products.

比如，我们称之为Ask AI的Amplitude聊天界面，目标是让现有产品更容易使用。

Like, the goal of, for example, what we're calling Ask AI, which is the chat interface in Amplitude, is to make it easier to use the existing product.

因此，当我们展望明年时，我们有四个主要优先事项。

And so as we've looked at next year, there are four big priorities we have.

第一个是将Amplitude重建为原生AI产品。

One is rebuild Amplitude to be AI native.

第二个是让产品更容易使用。

The second is make it much easier to use.

第三个是确保我们分析之外的其他产品与竞争对手持平。

The third is make sure our other products outside of analytics have parity with the competition.

第四个目标是为营销人员提供卓越的服务，以便我们能够取代大量传统的营销技术公司。

And the fourth is serve marketers really well so that we can take on a bunch of the legacy Martech guys.

我们仍在我们的会话回放产品、实验产品以及指南和调查产品上做一些基础性或根本性的工作。

There's still like basics that we're doing or foundational things we're doing on our session replay product, on our experimentation products, on our guides and surveys product.

比如，我们将在未来几个月内推出的会话回放功能之一是区域划分，它能让你查看网页时直接看到叠加在上面的分析数据，这是一项非常酷的功能。

Like, you know, for example, one big thing that we're gonna be coming out with in session replay in the next few months is zoning, which is this way where you can look at a web page and then see the analytics overlaid on top of it, which is super cool functionality.

所以我们仍然以同样的方式在推进这些工作。

And so that we're still doing in the same way.

我认为，让团队从底层真正拥抱AI，使我们能够做到的一点是：他们变得更快、更高效，交付的内容也多得多，这简直太棒了。

I think what getting the team to embrace AI from the the ground up has happened has allowed us to do is, one, you know, they're a lot faster and more productive, so they're shipping a lot more, which is absolutely amazing.

其次，他们开始用这种全新的视角来看待这些问题。

And then two, they are looking at these problems with this lens.

因此，我们在组织上做出的重大调整是：过去这些AI项目只是分散在各处的辅助性工作。

And so the big change we did organizationally is we had a lot of these AI projects as kind of side projects for a bunch of people.

于是我们决定：好吧，成立一个专门的团队来集中攻克这些项目。

And we said, okay, let's have a dedicated team that kind of goes after them.

但现有团队仍然会继续努力打造优秀的Amplitude产品，这项工作同样重要。

But the existing team is still gonna do you know, work on making the Amplitude product great, and that work is is is just as important.

因为我认为商业世界变革的时间周期要长得多、长得多、长得多。

Because I I I think the timeline for change in the business world is the cycles are just much, much, much longer.

你们如何在负责新AI项目和传统SaaS项目的团队之间分配人员？

And how do you allocate people between the teams working on the new AI stuff and the more traditional SaaS stuff?

他们是自行选择加入不同团队的吗？

Did they self select into different groups?

是否存在一种具有挑战性的文化隔阂，需要你们去应对？

Is like is there like a like a like a challenging cultural divide there that you have to navigate?

没有。

No.

你知道，通过举办AI周并明确我们的整体隐喻，我们做到了这一点。

You know, I think by doing the AI week and make it clear we had this whole metaphor.

我们谈到了焚船断后。

We talked about burning the boats.

我们所有人都全力投入这件事。

Like, we're all in on this.

有些人甚至更加投入，他们自然而然地选择了参与这些工作。

And there were some people that got even more into it, and they very naturally self selected into a bunch of a bunch of this work.

我们真的疯了。

Like, we and it's it's crazy.

这不仅仅是对工程师们说的。

It's like it's not just, hey, engineers.

设计师们也是。

It's like designers too.

我们最好的设计师，叫威尔·牛顿，他之前被分散在太多项目上。

Our very best designer at Amplitude, this guy named Will Newton, we actually had him spread among way too many projects.

他说：不行，我得专注起来。

And he was like, no, I got to focus in.

让我们暂时拒绝一些项目，这样我可以深入钻研这个聊天界面，确保明年年初发布时，它会非常出色。

Let's say no to some of these for a little bit so I can go really deep on this chat interface at you know, and and make sure that when we launch this thing in early next year, it's it's gonna be awesome.

所以，是的，人们会自然地选择加入或退出。

So, yeah, people will naturally kinda self select in or out.

这是一家200人的公司，因此你会看到这一现象的完整谱系。

And it's a 200 person org, so you get a whole spectrum of this.

我们能聊聊你在网上引发争议的那件事吗？

Can we talk about the thing that you said that made people angry on the internet?

对。

Yeah.

是的，是哪一件？

Yeah, which one?

我觉得，是的。

I think, yeah.

这是个问题。

That's a problem.

我想是那个关于AI可见性的言论吧，总的来说，你当时想表达的是，很多初创公司其实只是功能堆砌，而不是真正的公司。

Think it was the AI visibility treat, I think like generally, I think you were making the point that, the, a lot of these startups are sort of like features, not companies.

显然，在YC，我们会站在另一方。

And obviously like at YC we're gonna take the other side of it.

天啊

Hell

是的

yeah.

我会说——

And I would argue that-

你们有一个每次孵化营都出现的可见性公司。

You guys have a have a visibility company that fills like every batch.

对。

Yeah.

这是一个观点。

It's it's a point.

这已经成为一个流行的观点。

It's become a a popular idea.

这是真的。

That is true.

我们寻找的是创始人，而不是点子。

We find founders not ideas.

我喜欢这个观点。

I love it.

但抛开AI的可见性不谈，我想你已经走过了整个历程。

But AI visibility aside, I guess you you've gone through the whole journey.

你曾经参加过孵化项目。

Like, you were in the batch.

你创办了一家公司。

You started a company.

现在你正在运营一家上市公司。

Now you're running a public company.

我认为我们会主张，初创公司普遍的优势在于，它们没有现成的客户群，没有收入，当产品尚未成熟时，客户也会更加宽容。

I think we would argue the side of just that the advantage that startups have in general is just they don't have an existing customer base, they don't have revenue, their customers are gonna be more forgiving if the thing doesn't work yet.

这些优势似乎对那些用AI产品挑战现有公司的新兴初创企业非常有利。

And these have seemed like advantages for like new startups attacking incumbent companies with like an AI product.

但从你的角度来看，现有公司相对于初创企业有哪些优势呢？

But from your perspective, are, what are like, what are the advantages that like incumbent companies have over the business?

你怎么看待这个问题？

How do you think about it?

我认为真正的商业价值必须在AI可见性之后才能实现。

I think the real business has to be downstream of AI visibility.

这里面确实有价值。

It's a valuable there's value in it.

就像在SEO领域一样，也是有价值的。

Like, you know, just like there's value in the SEO world.

但实现起来太容易了。

But it's so easy to do.

你知道的，我们做了，正如我提到的，大概只花了几个星期，但我们很快就完成了，并且免费提供出去。

You know, we did it, you know, as I mentioned, probably more like a few months and a few weeks, but we did it very quickly and gave it away for free.

这为我们带来了惊人的潜在客户引流。

And it's been this incredible lead generation for us.

所以商品化将会非常、非常快地发生。

And so the commoditization is going to happen real, real fast.

我的意思是，你听人们说，嘿，SaaS 会因为 AI 而消失。

I mean, you talk about like, you know, people talk about, hey, SaaS going away because of AI.

我觉得这是个很好的例子。

I think that's a great example of it.

相比之下，如果你看看那些在这里做得非常成功的企业。

And so in contrast, if you look at, I think, businesses that have done very well here.

比如有一家业务，我认为非常可行，那就是 Air Ops 所做的——他们确实有一些可见性方面的功能，但他们还有一个完整的内容生成业务，帮你创作博客文章和其他内容，这才是他们的核心业务。

So like there is a business, I think, that is very viable here, which is what Air Ops is doing, where they're yes, they have some visibility aspects, but they have a whole content generation business to help you create blog posts and other material on that, you know, and that's their real business.

所以我认为，所有这些可见性类业务，其实都可以从我们这里免费获得。

And so I think all of these visibility businesses, like, you can get it for free from us.

你可以从很多其他地方免费获得这些服务，因此你必须构建一个真正下游的业务。

You can get it, you know, if you're gonna be able to get it free from a lot of other places, and you're gonna have to construct a real business kind of downstream.