20VC：模型性能如何遭遇瓶颈期 | 有效交易的两大关键法则 | 从Keith Rabois、Brian Halligan和Pat Grady身上学到的公司建设经验 | Harvey CEO Winston Weinberg谈为何企业AI应用仍需数年

你提到过一些存在性威胁。

You said about kind of existential threats.

是的。

Yeah.

如今最让你担忧的存在性威胁是什么？

What existential threat today concerns you most?

我认为是产品迭代速度是否足够快。

I think just moving fast enough on product.

这始终是所有应用层公司最大的存在性威胁。

Like that is always, I think, the biggest existential threat for all the application layer companies.

并不是说Anthropic或OpenAI明天就会把50%的资源投入到法律或税务等垂直领域。

And it's not necessarily that, you know, Anthropic or OpenAI are, you know, tomorrow gonna put 50% of their resources after the legal vertical or tax vertical or anything like that.

但他们只是在持续改进自己的产品和模型。

But they're just improving their product and models.

除非你的产品与企业级GPT授权所能提供的功能之间存在巨大差距，否则你的产品价值将会下降。

And the value of your product is going to go down unless there is a massive delta between what your product does and what you could get from an enterprise GPT license.

对吧？

Right?

所以，这始终是一个持续存在的威胁：你如何确保产品达到逃逸速度，建立起足够的产品护城河，以免被他们碾压。

And so it's just a constant existential threat of how do you make sure you get to escape velocity on product so you have enough of a product moat for them to not run you over.

我每天都会思考这个问题。

And I think about that daily.

当我考虑竞争对手时，我最常想到的是，我认为这些实验室比大多数人更值得看好。

When I'm thinking about competitors, the main thing I think about is I'm more bullish on these labs than most people, I think.

我的意思是，很多人都是这样想的。

I mean, a lot of people are.

我非常看好它们。

I'm very bullish.

它们拥有非凡的人才。

They have incredible talent.

我更多思考的是，我们的客户将来会面临哪些前沿问题，而这些问题它们最终会解决。

And I think more about what are the frontier problems that our customers have that they're gonna solve later.

四点五

Four point

五改变了Anthropic的格局，也改变了整个行业格局。

five changed the game for Anthropic and changed the game for the landscape.

你看到使用量完全转向Anthropic了吗？

Did you see usage shift entirely to Anthropic with that?

并没有完全转向，但我们基本上是根据使用场景进行路由的。

Not entirely, but there were definitely so we route basically based off of the use case.

我们会路由到最佳的模型组合。

We route to the best combination of models.

而且我们的流量确实大幅转向了Opus 4.5。

And definitely our traffic went up to Opus 4.5 significantly.

当OpenAI是你们公司的投资者，而你们却突然将大部分流量路由到Anthropic时，这会有冲突吗？

Is there a conflict with OpenAI when they're an investor in your company and you're routing the majority suddenly to Anthropic?

我们目前还没有路由大部分流量，但即使真的这样做了，也不会有冲突。

Well, we aren't routing the majority yet, but even if we were, there's no conflict.

基本上，他们希望我们赢，希望我们使用最好的模型。

Basically, like, they want us to win and they want, you know, us to use the best model.

对吧？

Right?

我们的协议或合作关系中，没有任何规定必须使用OpenAI的模型。

There's nothing in our agreement or our relationship or anything like that that you have to use OpenAI models.

而且，我认为对他们而言，来自应用层公司的反馈非常宝贵——告诉我们哪些地方你们的模型表现不佳，哪些地方表现极佳，以及哪些地方需要改进。

And if anything, I think that for them, getting the feedback from application layer companies on this is where your models aren't doing as well, this is where your models perform super well, and this is where you need to improve is super valuable to them.

你认为不同模型提供商的性能正在趋于平稳吗？

Do you think we're seeing a plateauing in performance across the different model providers?

我认为在消费级应用场景中，性能确实正在趋于平稳。

I think that we're seeing a plateau in performance for consumer use cases.

我认为这是一个误解，或者说是人们根本不该关注的事情，因为我们并不需要消费级应用场景的模型变得更好。

And the reason why I think this is like a misnomer or something that people actually shouldn't pay attention to is we don't need them to be better for consumer use cases.

我觉得很多人还没真正理解，对于很多消费级应用场景，4代就已经足够了。

Like, I feel like this is something that folks don't quite understand is like a lot of the consumer use cases four was like, we're done.

你不需要更强的推理能力来解决这些问题。

You don't need better reasoning to solve these problems.

你需要的是不同的上下文。

What you need is like different contexts.

你需要让它连接到你的日历。

You need it to connect to your calendar.

你需要连接到你使用的所有不同应用程序和其他类似的东西。

You need to connect to all of the different apps you use and things like that.

对他们来说，性能的提升就是指这个，对吧？

That's what an increase in performance is for them, right?

在面向消费者的部分，可能确实出现了性能瓶颈。

There might be a plateau in some of the consumer facing side of things.

在企业领域，我认为进展还会继续，尤其是cogen。

On the enterprise, I think things are gonna keep going and especially cogen.

我认为cogen不会出现性能瓶颈。

I think we are not going to see a plateau in cogen.

我认为这会变得好得多，而且会非常快。

I think that is gonna get much better really, really fast.

在未来十二个月里，你预计在cogen领域会看到什么变化？

What do you expect to see in cogen in the next twelve months?

我认为趋势只会继续上升。

I think the slope will only increase.

我认为它会变得越来越好，这将释放整个世界的大量生产力。

I think that it will get better and better and better, and I think that will unlock a lot of productivity just across the entire world.

在Harvey内部，是不是所有人都没有使用Claw Code或Cursor？

When you look within Harvey, is everyone not using claw code, not cursor?

这是一种混合情况。

It's a combination.

在英国，我们有一个叫‘Shag、Marry、Kill’的游戏。

In The UK, we have a game called Shag, Marry, Kill.

我不会在这里玩这个游戏，因为这太不合适了。

I'm not gonna do that here because it's wildly inappropriate.

但如果我们进行买卖操作，Anthropic估值35亿美元，OpenAI估值800亿美元，你会买哪个，卖哪个？

But if we were to do a buy and sell, and you had Anthropic at $3.50 and OpenAI at 800, Which one would you buy and which one would you sell?

是的，我知道。

Yeah, I know.

我们之前讨论过这个。

And we talked about this.

我会以两倍的价格同时买入它们。

I'd buy them both at double.

你会以160亿美元的价格买入OpenAI？

You'd buy OpenAI at 1.6.

也许还没到160亿。

Maybe not quite 1.6.

在考虑160亿之前，我需要看到他们的一些进展。

There's a couple of things I need to see from them before I would do 1.6.

在那之前，你需要看到他们哪些方面的表现？

What do you need to see from them before?

我认为OpenAI最主要的优势是它拥有强大的消费者品牌。

I think the main thing with OpenAI is they have so much consumer brand.

它的影响力非常强大。

It is so powerful.

尤其是在X平台之外，以及在我们这些科技圈之外的地方，它的影响力简直不可思议。

And especially outside of X and outside of kind of like our worlds, it is so unbelievably powerful.

我并不是来自科技行业。

And I don't come from the tech world.

因此，在进入这个行业之前，我所有的朋友和人际关系都来自科技圈之外。

And so all of my friends, all of my relationships before this were outside of tech.

这种品牌影响力极其强大。

That brand power is incredibly powerful.

我认为，更加专注于消费者市场，全力押注这一方向，是我最看好的地方。

And I think like more focus on consumer and just tripling down on that, that I think is where I'd be the most bullish.

我认为在企业端，将会出现多个赢家。

I think that on the enterprise side, there will be multiple winners.

企业界不会允许只有一个赢家，对吧？

Like enterprises don't allow there to be one winner, right?

所以不管怎样，OpenAI 会拿到一部分企业市场，Anthropic 也会拿到一部分企业市场。

So no matter what, there's gonna be OpenAI is gonna get some of the enterprise market, Anthropix is gonna get some of the enterprise market.

但消费市场，我认为 OpenAI 有机会拿下其中很大一部分。

But the consumer market, I do think OpenAI has an opportunity to take a lot of this.

而显然，那里的主要竞争对手将是谷歌。

And obviously the main competitor there is gonna be Google.

但我认为这两家公司都可能变得极其庞大。

But I think they both can be just astronomically huge companies.

另一件我觉得人们没有意识到的事情是，一切都有可能暂停。

The other thing that I feel like folks don't realize is everything could pause.

即使这两家公司现在就停止所有开发，AI 对经济的渗透程度仍会急剧上升。

Like both of those companies could stop developing things right now and the amount of saturation of AI that would just happen to the economy would still skyrocket.

我们离模型真正融入日常生活还有很远的距离。

Like, we're so far from just the ability of the models right now being integrated into daily life.

人们不知道如何使用这些系统。

People do not know how to use these systems.

别用。

Like, don't.

他们真的不会用。

They don't.

企业也完全不会用。

And businesses definitely don't.

而且这种能力过剩的情况非常严重。

And there's so much of, like, the capability overhang is so high.

我觉得这比任何人谈论的都要严重得多。

Like, I think it's higher than anyone is even talking about.

这简直是天文数字。

It's astronomical.

所以我认为，如果这两家公司真的停止推出新产品，它们的收入仍会爆炸式增长，因为会有无数公司基于它们的模型进行开发。

And so I think if both companies literally just stopped shipping things, their revenues would still explode because there are going to be so many companies building on top of their models.

会有这么多不同的经济领域采用这些技术，我认为我们还处于早期阶段。

There are gonna be so many different parts of economy that adopt these things that I think we're still in early days.

你觉得这个时间差是多少？

What do you think that time lag is?

我知道这是个很难回答的问题，但大概是两到三年？

I know it's a horrible question to ask, but is it like a two to three year?

是十年吗？

Is it like a ten year?

你和一些这些企业合作过。

You work with some of these enterprises.

他们不说X所使用的语言，而且通常也不这么做。

They don't speak the language that X and they generally do.

这个时间线实际上是什么样的？

What does that timeline actually look like?

我认为，直到我们看到企业中出现巨大且巨大的生产力提升，大概需要三到五年。

I think like three to five years until we see like massive, massive productivity gains in enterprise.

我认为这些能力已经具备了。

I think the capabilities are there already.

这些能力两年前就已经有了。

The capabilities were there two years ago.

很多情况下，如果你想想典型的企业的业务流程，他们为了完成一个流程，要从17个不同的系统中提取数据。

A lot of this is like, if you think about just like the average enterprise workflow, there's like 17 different systems they're pulling data from to get that workflow done.

literally，17可能都算少的了。

Like literally, 17 might be on the low end.

有时候甚至达到50个。

Sometimes it's like 50.

你得开着一百个标签页，在各个不同的应用程序之间切换，它们本应彼此连接。

And you have a 100 tabs open, you're opening in all these different apps and they kind of connect to each other.

但实际上，它们根本连不到一起。

They don't really connect to each other.

对吧？

Right?

因此，要让这些系统和代理真正从头到尾完成一项任务，其长尾难度非常大。

And so the long tail on actually getting these systems and agents to do a task from start to finish is so difficult.

而你最终会遇到的问题是，像我们和Sierra这样的垂直领域代理虽然存在，但许多垂直领域都需要与企业其他部分紧密连接。

And the problem that you're gonna end up having is you have these building vertical agents like us and Sierra, etcetera, but a lot of even the verticals connect to all of the other parts of the enterprise.

所以对我们来说，一件有趣的事情是，我们的收入正越来越多地来自全球2000强或财富500强公司。

So like one thing that's happening that's interesting for us is, you know, a lot of our revenue is starting to come from Global 2,000 or Fortune 500 companies.

而我们实际上并没有为税务合规和采购等职能开发很多功能。

And we actually haven't built many features for, like, tax compliance and procurement.

对吧？

Right?

正在发生的是，这些部门正在采用Harvey，尽管我们并没有为这些部门专门开发功能。

Starting to happen is those departments are adopting Harvey, even though we haven't built features specifically for those departments.

原因是法务部门实际上会与其他部门互动，而法律文件是企业运营的核心，因此它们与企业的各个部分都有交集。

The reason why is the legal department actually interacts and just like legal documents are such a core part of a business that they interact with all of these different parts of the business.

对吧？

Right?

所以我们发布了一个类似多人协作的功能。

And so we released basically a feature that's multiplayer.

它叫做共享空间。

It's called shared spaces.

最初，推动我们做这个功能的原因是，你希望像沃尔玛这样的大型企业能和他们的律师事务所在同一个平台上协作。

At first, like a lot of the impetus for doing it was you want a large corporate like a Walmart or whatever to work with their law firms in the same platform.

这种情况确实正在发生。

And that's happening.

但真正开始发生的是，这些企业让法务团队与合规部门、人力资源部门以及其他所有部门同时在Harvey中协同工作。

But actually what's starting to happen is these corporates are using the legal team, is working with the compliance department, is working with HR, is working with everything else, all in Harvey at the same time.

我从Legora那里听说，共享空间功能是从他们那里复制的。

I heard from Legora that the Shared Spaces was ripped from them.

这种说法公平吗？

Is that fair?

不。

No.

我们很早就开始做多人协作功能了。

We were working on multiplayer a long time ago.

我觉得我们公司有趣的一点是，我们一开始接触的就是最难搞定的客户，在内部团队方面我们也做了同样的事情。

I think one of the things that's interesting about our company is we started with, like, the hardest customers, and we did the same thing actually on the in house side too.

我们以前就有银行客户，为银行构建的安全和权限系统，其复杂程度和企业级准备远超其他客户。

So, like, we had bank customers a while ago, and the security and permissioning systems that you need to build for a bank are so much more in-depth and, like, the enterprise readiness than for a lot of the other folks.

而多人协作功能中最大的问题在于，我们允许内部团队或律所一方来启动协作。

And the biggest problem with multiplayer and the way that we're doing it is we're allowing the in house side to kick it off or the law firm side.

要做到这一点，双方所需的安全和权限体系都必须达到极高的标准。

And to do that, the security and permissioning that you needed in place for both is astronomically high.

所以我们花了非常非常长的时间来处理这件事。

So we were working on this like a very, very long time.

我们为此持续了大约六个月到接近一年的时间。

We were going on it for like six months to almost a year.

在做界面和其他上层功能之前，我们先彻底完成了所有的权限设置和相关工作。

We just did all the permissioning and all of that stuff first before you do kind of the UI and on top.

你为什么觉得他们总说你剽窃他们的产品创意？

Why do you think they continuously say that you rip their product ideas then?

我认为，如果你是市场上的第二名，其中一个能让你获得大量关注的方式，就是以任何方式与第一名挂钩。

I think that if you are number two in the market, one of the things that can get you a lot of attention is just attaching yourself to number one in any way, shape, or form.

对吧？

Right?

通过这种做法，你实际上能获得免费的媒体报道。

You kind of get free press from doing that type of thing.

这是一种很好的方式，可以搭上对方已有的分发渠道。

It's a good way to basically jump on to the distribution that the other front has.

你尊重他们吗？

Do you respect them?

因为你们之间某种程度上是互相憎恨的，不，真的，你们就是这样，而我特别喜欢这一点，因为我觉得我们太客气了。

Because you guys hate each other in a way that like, no, no, really you do, and I love it because it's like, I feel we got too kind in tact.

总是说，哎呀，我们都是朋友。

Like, oh, we're all friends.

我当时就想，不，我们来这里是为了赢。

I was like, no, we should be here to win.

你认识斯洛特曼吗？

You know if Slootman?

是的。

Yeah.

是莱斯。

It's Rice.

战争？

War?

对。

Yeah.

我们很喜欢他。

We love him.

是的。

Yeah.

而且你们真的能看到这种敌意和仇恨，这真是太棒了。

And you guys really it's wonderful to see the animosity and hatred.

但你们彼此之间还是相互尊重的。

But you guys respect each other.

我的意思是，我确实很尊重他们。

I mean, I I definitely respect them.

我认为，他们做得特别好的一件事是在欧洲的表现非常出色。

And I think, like, one of the things that they did really well is I think they did a great job in Europe.

这要追溯到2023年了。

And, you know, this is back in 2023.

我们并不是说他们比我们晚了六个月之类的。

We're we're not that I think they're, like, six months after us or something.

差距并没有那么大。

It's not that big of a gap.

我认为，如果当初能重新来过，我会在2023年就更多地投资欧洲。

And I think, like, one of the things that I would have done differently in the beginning is just invest like more in Europe in like 2023.

实际上，我们的许多早期客户都在欧洲。

And actually, we a lot of our first customers were in Europe.

但在这里有实地人员非常重要，要尊重不同的文化，以及如何将这些实现规模化生产等等。

But having folks on the ground here is just really, really important and respecting kind of like the different cultures and how to productionize that and all

所有这些方面。

of those things.

你为什么在考虑更早进入欧洲时点头表示兴趣？

Why did you nod out of interest in terms of coming to Europe earlier?

只是因为人力有限。

It was just bandwidth.

当我们签下第一个客户Eno Sherman时，我们只有四个人。

Like, when we signed Eno Sherman, which was our first customer, we had four people.

所以我们用四个人完成了四千人的企业级部署。

So we did a 4,000 person enterprise grade roll out with four people.

这是你当时在……我从帕特那里听说了这件事。

Was this when you were I got told this from Pat.

你们当时住的是Airbnb吗？

You were like in an Airbnb?

是的。

Yeah.

住在Airbnb里。

Were in Airbnb.

加入的那位工程师，也就是我们的第一位工程师加布，之前几乎独自编写了所有代码。

The engineer who had joined, was our first engineer, Gabe was basically coding everything before that.

他那时已经在那里待了一个月左右。

He had been there for, I think, a month.

然后我们为一家4000人的大型企业部署了安全系统。

And then we onboarded a 4,000 person, very large scale enterprise security.

那是一个极其重要的团队。

It was incredibly important team.

所以，只是人力不足。

And so it was just bandwidth.

当你快速扩展时，这会更困难。

Like, when you're scaling that quickly, it's harder.

你是不是担心平台会崩溃？

Are you shitting yourself that the platform's gonna fall over?

现在不会了。

Not anymore.

所以不会。

So No.

现在不会了。

Not now.

但也许在2023年那时候会。

But maybe then in 2023

就像一个人住在Airbnb里那样。

with, like, one person in an Airbnb.

但但但其实，我觉得这很重要。

But but but actually, think this is important.

这是一件很有趣的事情。

This is this is something that's interesting.

当我观察很多AI应用层公司时，如果你去看看它们在领英上招聘的工程师，会发现90%都是前端工程师，这让我觉得挺有意思。

When I look at a lot of AI application layer companies, if you go through their LinkedIns and you look at the engineers that they're hiring, it's like 90% front end engineers, which is interesting to me.

我认为主要原因在于，氛围编程在前端的效果远比在基础设施上好。

And a lot of that reason is I think like vibe coding works much better with front end than it does for infra.

而且我认为，AI公司正在经历或将要经历的事情——就像我们在2024年初所经历的一样——就是先做大量前端工作，做出非常漂亮的用户界面和出色的演示。

And a lot of, I think, what's happening to AI companies or is going to happen to them, and it happened to us in early twenty twenty four, is you do a bunch of front end and you make really pretty UIs and really nice demos.

对吧？

Right?

然后用这些成果去吸引所有客户。

And then you use that to land all the customers.

但现在你有了大量实际活跃的客户，却还没有为成千上万甚至上百万用户使用你的产品而投资架构和基础设施。

And now you have a lot of actual active customers, and you haven't invested in the architecture and the infrastructure of hundreds of thousands, millions of customers using your product.

我们在2023年就犯了这样的错误。

And we kind of made that mistake in 2023.

在2024年初，让我们交付速度放缓的一个原因是，我们在前一个季度末增加了数万用户。

And in the beginning of 2024, something that kind of slowed down our shipping velocity is we added just tens of thousands of users in like the Q4 before that.

对吧？

Right?

而我们当时并没有准备好相应的基础设施来支持这些用户。

And we didn't quite have the infrastructure, like, to support that.

现在如果你看我们的团队，整个EPD部门中将近40%是来自Databricks之类的顶级公司的资深基础设施工程师。

And now if you look at our team, it's almost like 40% of our entire EPD org is very senior from like a Databricks or something like that, infrastructure engineers.

这是一个长期投入，因为当你拥有这些能够处理数千万级任务的智能系统时——对我们来说，去年我们几乎处理了五十亿份文档——你就必须有能支撑这一切的基础设施。

And it's a long term bet that as you get these agentic systems that are processing tens of millions, for us, I think last year we almost did like half a billion documents or something like that, that you need the infrastructure to actually support that.

所以，这不仅仅是如何赢得演示、如何拿下订单，更是如何为你的产品构建一个真正可扩展的企业级基础设施。

So it's not just about how do you win the demo and how do you win the deal, but how do you actually create an enterprise, very scalable infrastructure on your product.

我看到很多AI应用层公司都没有做这件事。

And I see a lot of AI application layer companies not doing this.

所以，如果你今天要给这些创始人建议，你会不会说：‘嘿，一定要更早地优先招聘基础设施人才，这样才能兑现你对客户的承诺？’

So if you were advising those founders today, would you say, Hey, really focus on prioritizing infra hiring earlier so you're able to fulfill what you say you will do to end?

我实际上会告诉他们，你的GRR很重要。

What I would tell them actually is your GRR matters.

我认为AI领域很多投资者忽视的一点就是GRR。

And I think like one thing that a lot of investors in the AI space have been not paying attention to is GRR.

他们只关注新增ARR，觉得流失率无所谓，因为增长太快了，也许可以转型或者还有其他客户。

They've been basically just looking at net new ARR and kind of being like, Ah, churn is fine because they're growing so fast that maybe they'll pivot or they have some customers.

我觉得这是个巨大的错误。

I think that's a huge mistake.

你会看到很多公司在一个垂直领域迅速签下大量客户，因为可能这个领域原本只有一个玩家，现在出现了第二个。

You're gonna see a lot of companies in a lot of verticals that go really, really fast to signing a bunch of customers because maybe there's only one in the vertical and now there's a second player, etc.

但随后他们必须实际支持所有这些客户。

But then they have to actually support all of those customers.

如果你没有相应的基础设施，又过早做出大量承诺，一旦系统崩溃，客户就会迅速流失。

And if you don't have the infrastructure in place and you make a bunch of promises upfront and then all of that falls down, you'll start losing customers really, really fast.

我看到很多AI公司都没有关注这一点。

And I see a lot of AI companies not focusing on this.

我认为，当他们的年经常性收入超过一亿美元时，这将是一次巨大的清算。

And I think that's gonna be like a huge reckoning for folks once they get past 100,000,000 ARR.

不过，当我看到GRR数据，想到像Sierra这样的公司时，我担心的是，Brett的总收入确实惊人，堪称顶尖中的顶尖。

This is what I worry about though when I look at kind of, when you look at GRR figures and you think about like a Sierra, it's and like the gross Brett's amazing and like the best of the best.

天啊，Brett，把我的钱也拿去吧。

Jesus, take my money too, Brett.

我在质疑你，但我正在

I'm questioning you, but I'm

天啊，你拥有的东西

like, Gosh, what you have

要从每个客户的实施服务提供方角度，将客户数量从100增加到400，这任务非常繁重。

to now fulfill from a per customer implementation service provider aspect to go from 100 to 400.

这太多了。

It's a lot.

太多了。

A lot.

这可不是即插即用的。

It's not like a plug and play.

通过产品驱动增长（PLG）模式，在消费者端从100增长到400要容易得多。

It's much easier to go on the consumer side from 100 to 400 with a PLG motion.

同意。

Agree.

我认为这需要回到我们之前说的：你怎么实现产品市场契合，然后又怎么实现公司市场契合？

And I I think that what that requires then is going back to what we were saying, which is how you go product market fit and then how do you get company market fit?

这实际上关乎你如何构建公司。

And that's actually like how you structure your company.

我认为这部分和过去有所不同，过去如果你有很长的实施周期，最终会发生的是，这些垂直领域都会落在一家大型财富100强或200强企业身上。

And I think part of that is different than how it used to be in the past, where if you have long implementation cycles or one thing that is gonna end up happening, I think, is a lot of these verticals are gonna land at a big Fortune one or Fortune two or whatever.

对吧？

Right?

而他们的产品将会大幅扩展。

And their product is gonna expand massively.

微软的这一点非常有趣。

This is a really interesting thing about Microsoft.

我不知道你对微软的销售团队了解多少，但最初他们的大部分销售人员都是售前人员。

I don't know how much you know about Microsoft Sales force, but they started in the beginning and the vast majority of their sellers, it was presales.

对吧？

Right?

就像渔夫一样。

So spear fishermen.

比如，有大量的这种人。

Like, have tons of that.

对吧？

Right?

老派的渔夫。

Old school spear fishermen.

而他们最终逐渐转向了大量售后服务。

And they actually eventually migrated to a lot of what they have is post sales.

他们的大量投资实际上是在售后阶段。

Like a lot of their investment is actually in post sales.

原因在于，他们的客户净留存率（NDR）一直在持续上升。

And the reason why is because their customers just their NDR goes up, up, up, up.

他们不断购买更多产品，购买更多计算资源等等。

They keep buying more things, they buy more compute, etcetera.

我认为许多这些企业公司也应该以这种方式思考自己的业务。

And I think a lot of these enterprise companies should start thinking about their company that way.

意思是，虽然现在确实存在一些抢占市场的行为，但真正重要的是，如果你对人工智能持乐观态度，你就应该对你产品的价值持乐观态度。

In the sense of like, sure, there's some of a land grab right now, but really what's gonna matter is if you are bullish on AI, you should be bullish on your product, the value of your product.

我们还处于产品开发的第一天。

Like we are in day one of product development.

它将会发生翻天覆地的变化。

Like it's gonna change astronomically.

因此，比起获取新客户和实现极高的年度经常性收入（ARR），更重要的是能否留住这些客户，因为今天支付你一百万美元的客户，未来完全有可能支付你一亿美元。

And so what's more important than landing new customers and getting really high ARR is can you retain those customers because that customer that pays you 1,000,000 today, there's a real world in which they pay you a 100,000,000 at some point.

我认为Databricks在这一点上做得非常出色。

I think Databricks is a company that's done an incredible job with this.

我最近邀请了Alex Rampell做客节目，他当然是你们的投资者安德森资本的人，而且

I had Alex Rampell, who's obviously at Andreessen, one of your investors on the show recently, and

他说了一句话，我特别喜欢，而且

he said something that I loved, and

听起来有点糟糕，但我真的很喜欢。

it sounds a bit awful, but I loved it.

他说，我希望看到的是拥有‘人质’而不是‘客户’的公司。

He said, I want companies who have hostages, not customers.

明白吗？

Okay?

尽管这话听起来有点糟糕，但我确实很喜欢。

And again, as kind of bad as that sounds, I did like it.

在这个领域，他们是‘人质’还是‘客户’？

In this space, are they hostages or customers?

他们迁移起来有多容易？

How easy is it for them to move?

是的。

Yeah.

我的意思是，现在还出现了第三种情况，那就是这些AI产品如此强大，我认为随着时间推移，投资回报率会变得非常高，高到不再像人质关系，而更接近Palantir的思维方式。

Mean, there's a third one that now develops, which is these AI products are so powerful and I think over time the ROI is so high that your ROI can become so massive that it's less of a hostage and it's maybe closer to how Palantir thinks about things.

换句话说，你为客户创造的价值越多，你赚得就越多。

In other words, the more value that you create for the customer, the higher you get paid.

我认为越来越多的公司会开始向这种模式靠拢。

And I think that more and more companies are gonna start aligning to that.

原因在于，让我想想。

And the reason why is let me look.

我来分别从律所和企业内部两个角度说一下。

I'll do the law firm side and I'll do the in house side.

对于律所来说，他们是按小时计费的。

For law firms, they bill by the hour.

所以很多人会问，你怎么可能卖给他们呢？

And so a lot of people are like, how would could you ever sell to them?

这根本不可能成功。

There's no way that this is gonna work.

对吧？

Right?

可能会发生两件事。

Two things might happen.

第一，他们可能会转为固定收费。

One, they might switch to fixed fees.

那样的话，我们就没问题了，效率也会非常好。

Well, now we're good to go and efficiency is really good.

我不认为会发生第一种情况，而实际上正在发生的是第二种情况：我们有众多律师事务所客户通过在Harvey中构建定制化解决方案，声称‘我们将用我们在Harvey中开发的这个定制方案来处理这笔并购交易’，从而从其他律所手中赢得了业务。

I don't think it's gonna happen that The second thing that actually is happening, and we have so many law firm customers that have gained new business by building something custom in Harvey and saying, we'll do this M and A with this custom solution that we did in Harvey, and they win that deal over another law firm.

这可不是什么人质。

That's not a hostage.

这款我每年花一百万美元购买的产品，竟然帮我赢得了一笔两千万美元的交易。

That's this product that I'm paying maybe like a million dollars for a year just earned me a deal that's 20,000,000.

那这个投资回报率是多少？

What is the ROI on that?

太惊人了。

Incredible.

而在企业内部，这一点更加明显。

And on the in house side, it's even clearer.

只要你节省了时间，就是在节省大量金钱。

It's just if you save time, you're saving tons of money.

所以我认为，所谓的‘人质’概念实际上可以转变为像Palantir所理解的那样——我认为B2B SaaS的价值即将变得极其巨大。

So I think that hostage thing can actually change to more like how Palantir thinks about it, which is I think the value of B2B SaaS is about to become astronomical.

如果你能弄清楚如何让产品与这种投资回报率对齐，它就不再是人质了。

And if you can figure out how to align your product to that ROI, it's not a hostage.

你只是完全与客户的目标保持一致了。

You're just completely aligned with your customer.

当你的客户群体不愿意采用按使用量付费的模式，而只想支付他们熟悉且可靠的按席位付费模式时，你如何将产品与投资回报率对齐？

How do you align product to ROI when your customer base doesn't wanna pay for a consumption model and they just wanna pay for a seat model that they know and can rely on?

是的，我认为至少在我们的领域里，并不是这样。

Yeah, I think that that's not, at least in our vertical, what I've seen.

我认为，在很多领域，我们即将转向基于使用量的定价模式。

I think, like, there are a lot of areas where we are gonna start moving to consumption based pricing.

对我们来说，我们的客户群体完全能接受这种模式。

For at least for us, for our customer base, that would be completely fine with them.

罗里·奥德里斯科尔，那位爱尔兰人，我经常引用他的话，他太出色了，简直该直接取代我了。

Rory O'Driscoll, the Irish guy, who I quote so often, is brilliant, should basically just replace me at this point.

他比我聪明多了。

He's much smarter than me.

但他总是说，人工智能将

But he always says like, AI will

如果我们能把支出从人力预算转移到技术预算，那对我们所有人来说都将是非常美好的。

be magnificent for us all if we see spend shift from human labor budgets to technology budgets.

正确。

Correct.

我们在这里会看到这种转变吗？

Will we see that shift here?

已经看到这种情况发生了。

Already seeing it happen.

这如何发生？

How does that happen?

是的。

Yeah.

有一些公司已经表示，Harvey的预算来自他们的专业服务支出，而不是他们的技术预算。

So there are a couple companies that have basically said that the Harvey budget comes out of their spend on professional services, not out of their tech budget.

专业服务的预算每年高达数十亿美元，而G&A部门的技术预算则要小得多。

The budget for professional services is in the billions a year versus the tech budget for that G and A group is astronomically smaller.

抱歉，专业服务预算并不是指他们组织中的初级人才。

Sorry, the professional services budget is not the junior talent that they have in their organizations.

不，

No,

我认为我们业务中最有趣的一点是，我们为大型企业所做的工作，与我们律师事务所客户所做的工作大不相同。

and I think that's what's really interesting about our business is a lot of the work that we're doing for like a corporate is not the work that our law firm customers are doing.

这就像替代性法律服务提供商。

It's like alternative legal service providers.

这是那种低端的工作。

It's this like lower end work.

律师事务所和外部客户的收入各占多大比例？

What percent of revenue is law firm versus external?

现在吗？你是说企业客户和律师事务所客户之间的对比吗？

Right now oh, so you mean corporate versus law firm?

是的。

Yeah.

我认为大约40%的收入来自企业内部，60%来自律师事务所。

I think it's around 40% of our revenue is in house corporate, and 60% is law firm.

差不多是这样。

Something like that.

这就是你原本以为的吗

Is that what you thought it

吗？

would be?

差不多是这样。

Something like that.

我的意思是，如果你看看世界上律师的分布情况，看看有多少律师在公司内部工作，有多少在律所，比例基本是一样的。

I mean, if you just look at the breakdown of how many lawyers exist on earth and how many of them are at companies versus in house, that's pretty much the same.

五年后，这个比例会是多少？

In five years' time, what will that be?

我觉得会差不多。

I think it'll be similar.

我觉得会一样。

Think it'll be the same.

我们将会看到律师事务所发生怎样的变化？

How will we see law firms change?

我们会看到初级律师被取代吗？

Will we have a cannibalization of juniors?

我不这么认为。

I don't think so.

我觉得我们会接到更多的工作。

I think we'll just get more work.

很有趣，我最近和Are聊过一次

So interesting, I had a conversation with Are

你打算让我女朋友失业吗？

you gonna make my girlfriend unemployed anymore?

不会。

No.

我最近和一家规模很大的私募股权公司聊过，他们说他们今年将会非常出色。

I had a conversation with a pretty large private equity shop recently, and they were talking about their year is gonna be incredible.

他们觉得今年将会是大规模并购的一年。

They think it's gonna be, like, a big M and A year.

这将会很棒。

It's gonna be great.

他们还谈到了他们如何看待法律费用。

And they were talking about, like, how they think about legal fees.

他们对法律费用的看法是，事实上，今年会是大年，每当我们遇到大年，就会支付更多的法律费用。

And the way that they thought about legal fees is the reality is, like, it's gonna be a big year, and whenever we have a big year, we pay more in legal fees.

这本来就是这么回事。

That's just like how it happens.

对吧？

Right?

但有一些事情我不再想付费了。

But there are certain things I don't wanna pay for anymore.

比如交易中的某些部分，等等，我只是在对保密协议之类的东西加价。

Like, there are certain parts of the deal, etcetera, that I just marking up NDAs, whatever it is.

我不再想为这个付费了。

I don't wanna pay for that anymore.

但我现在还要为律师事务所支付许多新项目。

But there's all these new things that I'm paying law firms for.

AI风险，比如你该不该收购这家公司？

AI risk, like should you buy this company?

在X Y Z国家有没有什么法案之类的问题，可能会改变现状？

Is there a, you know, a problem in x y z country with, you know, an act or something like that that's gonna change it?

专业服务领域出现了这么多新工作，但我的直觉是，这并不会真的发生。

There's so many new pieces of work for professional services that my gut is that's not what's gonna happen.

事实上，我认为真正会发生的是，专业服务市场会和GDP保持同样的增长速度。

In fact, think what's gonna happen is the professional services market is going to actually keep growing at the same as GDP.

看待这个问题的一种方式是，大多数专业服务都是周期性的。

One way to think about this is most professional services is cyclical.

所以，如果你有一年特别好，专业服务行业也会有一年特别好。

So So if you have a really good year, professional services have a really good year.

这几乎总是如此。

That's almost always how it works.

除了破产和诉讼在某种程度上是逆周期的，这取决于具体领域。

Other than bankruptcy and litigation is somewhat countercyclical, it depends on the area.

所以我认为人们想到这一点时会说：哇，真的吗？

And so I think people think about this and they're like, Oh, wow.

人工智能将影响法律行业，并彻底摧毁这些工作。

AI is gonna impact legal and it's gonna just destroy all these jobs.

他们没考虑到的是，他们的所有客户都在使用人工智能来创造更多产品。

The thing they aren't thinking about is all of their customers are using AI to create more products.

当你创造更多产品时会发生什么？

What happens when you create more products?

你需要更多的产品法律建议。

You need more product to legal advice.

当你更快地拓展到其他国家时会发生什么？

What happens when you're expanding into other countries faster?

你需要什么？

What do you need?

监管建议，对吧？

Regulatory advice, right?

所以我认为人们在所有这些行业中看待人工智能时，就像它是个真空一样。

And so I think people are thinking about AI in all of these industries as like a vacuum.

但现实是，你应该把人工智能看作整个经济将会发生的变化。

And the reality is you should think about AI as like the entire economy what's gonna happen.

很可能发生的情况是，经济将爆炸式增长，这些公司会对自身能实现的目标抱有疯狂的期望，而专业服务提供商必须对此做出回应。

And probably what's gonna happen is the economy is gonna explode, these companies are gonna have crazy expectations for what they can do, and the professional service providers are gonna have to respond to that.

你觉得经济会继续爆炸式增长吗？我都不敢相信我会问这个问题，因为这感觉像是面试官最基础的问题，但也许我现阶段就是一个糟糕的面试官，说实话。

Do you think the economy is gonna I can't believe I'm asking this question because it feels like the most base question that shit interviewers ask, but maybe I'm just a shit interviewer at this stage, to be honest.

你觉得经济会继续爆炸式增长吗？

Do you think the economy is gonna continue to explode?

我们对外部环境有这么多担忧，比如外面的人说，那些循环交易简直疯了。

We have so much external concern outside of the x sphere, which says, hey, the circular deals are fucking nuts.

美国的借款从未如此之高。

US borrowing's never been higher.

欧洲就像一个完全缺乏生产力的博物馆。

Europe is a fucking museum that is completely unproductive.

真美。

Beautiful.

我们将经历一次严重且实质性的放缓。

We are gonna have a serious and material slowdown.

你觉得这不对吗？

Do you think that's wrong?

我不认为今年就会发生。

I don't think it'll be this year.

我觉得会有一些波动。

I think there will be bumps.

比如，我确信我们会经历更多像深度搜索那样的时刻，那时所有人都会惊慌失措，对吧？

Like, I definitely think we will have more moments like the deep seek moment where everyone freaks out, right?

我认为我们已经非常接近临界点了，如果足够多的人说即将迎来衰退，那么只要一件事发生，大家就会恐慌，衰退就会真的到来，这完全是一个自我实现的预言。

And I think we're close enough to an edge of If enough people say that there's gonna be a bust, it's pretty easy for one thing to happen and for everyone to freak out and there to be a bust and it's a self fulfilling prophecy.

这些通常都持续得很短。

Those are usually pretty short.

我认为我们会经历很多短暂的波动，但从长远来看，人工智能将彻底重塑经济的每一个部分。

I think that we will have a bunch of short ones, but I think long term AI is going to completely reshape every part the economy.

我非常坚定地相信这一点。

Like, I very strongly believe that.

我确实要问一下，你提到了欧洲，我们之前也简单聊过欧洲的生产率问题。

I do have to ask you mentioned Europe, and we spoke a little bit about kind of Europe's productivity there.

你说你希望当初能更主动地介入欧洲事务，但你能做的毕竟有限，诸如此类的话。

You said you wish you'd been more proactive earlier on Europe, but there's only so much you can do, blah blah blah.

关于组建团队，你现在知道了哪些当初起步时应该知道的事情？

What do you know now about building teams in that you wish you'd known when you started?

我觉得在欧洲，情况和很多地方类似，那就是你不能一进入某个国家、某个领域，就一副自己什么都懂的样子。

Oh, I think that it's similar in Europe to where it is in a lot of places, which is you don't wanna go into a country or a domain or anything like that and act like you know how to do something.

对吧？

Right?

你需要与某个行业或某个地区建立合作伙伴关系。

Like, you really need to partner with an industry or you need to partner with a geography.

对吧？

Right?

当我提到我们应该在那方面投入更多时，我的意思是我们在2023年和2024年并没有投入太多。

And when I say, you know, we should have invested more in that, it's more like we didn't invest in it that much in 2023 and 2024.

去年我们投入了大量资金，今年投入得更多，差距非常巨大。

We invested tons last year and we're investing even more this year, and the difference is pretty massive.

去年我们的团队质量、合作伙伴关系以及产品在各地区的本地化程度，与现在相比有着天壤之别。

The difference in kind of the quality of our team last year and the partnerships and things like that and how our product is localized for each geo is just a huge difference.

你不可能只坐在旧金山，凭空思考怎么做这件事。

But you can't do this from sitting in San Francisco and, like, kinda thinking about how to do it.

你必须亲自去旅行。

You gotta travel.

美国和欧洲在人才方面最大的区别是什么？

What's the biggest difference in talent between The US and Europe?

不是人才上的区别。

It's not a difference in talent.

只是招聘人员需要很长时间。

It just takes a long time to hire people.

所以你必须用长得多的时间视角来考虑这个问题。

And so you have to just think about it with a way, way longer time horizon.

因为Gardening Leaf的原因。

Because of Gardening Leaf.

是的。

Yeah.

招聘人员真的很难。

It's just really hard to hire people.

这对我来说有点有趣，或者说我以前没经历过，而在美国，你有时甚至可以第二天就让新人入职。

And so that was just kind of interesting to me or something that I wasn't used to, whereas in The States, you can hire someone and they start quite literally the next day sometimes.

确实如此。

Literally.

或者如果他们需要提前两周通知，那就正好两周后开始。

Or if they have to give two weeks notice, they start exactly two weeks later.

这让你能稍微更灵活一些，我需要立刻招聘这个人来补救。

That allows you to be a little bit more just, I quickly need to hire this retroactively.

我可以解决一个问题。

I can fix a problem.

在欧洲，你必须提前更久地规划。

In Europe, you have to plan out more.

所以我们已经进行了很多大型办公室的开设。

And so we've done a lot of really big office openings.

比如，我们刚刚宣布了巴黎、都柏林和其他几个地方。

Like, we just announced Paris and Dublin and a bunch of other ones.

但你必须以更长远的眼光来考虑这些事情。

But you have to just think about this stuff at, like, a longer time horizon.

你不可能瞬间完成。

You can't do it instantaneously.

美国人认为欧洲人工作不努力，这种刻板印象公平吗？

Is The US trope of Europeans not working as hard fair?

这并不是我的发现。

That's not what I found.

但我要说，我主要接触的是律师。

But I will say we interact I interact mostly with lawyers.

而且，律师有可计费小时数的目标。

And, like, lawyers have billable hour targets.

而且归根结底，他们要么在国际律所工作，要么要与国际律所竞争。

And at the end of the day too, they are either at international firms or they're competing against international firms.

所以我根本没发现这种情况。

So I have not found that at all.

对吧？

Right?

我的意思是，有太多极其勤奋的律师了

I mean, there are so many incredible hardworking lawyers

英国律师工作相当努力。

UK lawyers work pretty hard.

英国律师工作极其拼命。

UK lawyers work insanely hard.

是的。

Yeah.

所以也许我并没有太注意到这一点，因为我们所处的领域，全球的律师工作方式都差不多。

And so it might be that I don't notice it as much because the domain that we're in, they work the same across the globe.

律师们只是极其自律、勤奋的人。

Lawyers are just incredibly disciplined, hardworking people.

我完全理解你的观点，并且同意。

I totally get you and agree there.

关于对人的评估，帕特告诉我，你在理解人方面是世界级的。

In terms of the people assessment, Pat told me that you're world class at understanding people.

如果我要问你，在加入Harvey的人中，除了你通常看重的基本诚信或抱负之外，还会寻找什么不太明显的特质，我会说：我寻找的是那种痴迷的偏执狂。

If I were to ask you for a trait that you look for in someone joining Harvey that is less obvious than the foundational integrity or ambition that you normally get, Like, I look for obsessed psychopaths.

这说法不错。

That's a good one.

你确实必须非常痴迷。

You definitely need to be obsessed.

是的。

Yeah.

那你认为最重要的特质是什么？

What would yours be?

所以痴迷确实非常重要，但我现在特别看重的是责任感。

So obsession is definitely very important, but the one that I look for right now a lot is ownership.

评估这一点有多种方法。

And there's a bunch of different ways that you can assess this.

但随着时间推移，你渐渐就能看出一个人是否真正具备承担责任的能力。

But you do, over time, you start to be able to just read if someone actually can take ownership over something or not.

这之所以变得非常重要，是因为随着公司规模扩大，要找出问题的根源会变得非常困难。

And the reason this becomes really important is as you scale as a company, it becomes really hard to figure out where a problem is stemming from.

变得很难了。

Becomes It hard.

事情往往是这样的：以前我了解公司里每一件事的进展，可以直接说，嘿，问题就出在这里。

This is how this ends up going, is it used to be I knew every single thing that was going on at a company, and I can just be like, hey, that's where the problem is.

我去帮他们扫清障碍。

I'm gonna unblock that.

对吧？

Right?

现在我们到了这样的阶段：我了解公司里大部分情况，但有时某些问题太底层了，我根本不知道问题出在哪，如果我问五个人，他们都会这么做。

Now we're getting to the point where I know most of what's going on at the company, but sometimes something is, like, so low down that I don't know what the problem And if I ask five people, they'll all do this.

就像蜘蛛侠一样。

It's like the Spider Man.

这就像风投公司里的糟糕交易。

That was like bad deals in venture firms.

谁做的这个交易？

Who did the deal?

是的。

Yeah.

这就像是到处都乱七八糟的。

It's just like It's like all over the place.

是的。

Yeah.

没错。

Exactly.

对吧？

Right?

是约翰尼他

It was Johnny who

是约翰尼，但我也不确定。

It was Johnny, and I don't know.

我发现科技行业中有许多人非常擅长向上管理。

And I have found that there are a lot of people in tech that have done a really good job managing up.

他们借助团队的成功而受益，但自己并没有真正取得成功。

They have ridden a wave of their team's success without being successful themselves.

所以我经常关注的是：人们是否愿意承认自己的错误？

So the thing that I look for a lot is can people admit their mistakes?

当一个人真正承认错误时，和他们说‘自己最大的错误’其实是某种非常出色的表现时，这两者之间的区别非常明显。

It is so obvious when someone is actually admitting mistakes versus they're saying that one of the biggest things they've done wrong is actually like something that's really great.

我们回到你之前说的一件事。

We'll go back to something you said earlier.

就像你说的一些事情是你要改掉的坏习惯时，

Like when you said some of the things are bad habits that you're trying to

我的意思是，是的，是的。

Well, I mean, yeah, yeah.

你说过，比如，

And you said, like,

检查，对。

check Right.

太松懈了，对。

Being slack too Right.

而且，我觉得这是一个很好的例子。

And and, like, that I think is a good example of

我在想，比如碱性，欺凌。

I was thinking, like, alkalism, bullying.

哦，确实如此。

Oh, well, yeah.

对。

Yeah.

而且，如果我面试自己并看到这样的回答，我会深入追问：你为什么这么做？

And and, like, I would actually if I had interviewed myself and I saw that as a like, the way that I answered, the way that I would push on this is I would say, why do you do that?

你为什么这么做？

Like, why do you do that?

我对此的真实回答是我有信任问题。

And my genuine answer to be to that would be I have trust issues.

我有信任问题。

Like, I have trust issues.

很难相信别人会处理好这个问题。

It is hard for me to trust that somebody else is going to handle that problem.

而现在，这突然变成了真正的责任。

And now all of a sudden, it actually is ownership.

这正是领导者面临的一个实际问题。

That is an actual problem of being a leader.

如果你一直有信任问题，无法信任其他领导层，就不可能将一家优秀公司扩展到数十亿美元的营收规模。

You cannot scale a really good company and get to tens of billions of revenue if you have constant trust issues and you can't trust other leadership.

你觉得这源于什么？

Where do you think that comes from?

我有信任问题，因为我发现，当关系破裂时，总是因为他们以金钱要挟我。

I have trust issues because I've found that generally when relationships break down, it always comes down to them extorting me for money.

很好。

Great.

确实如此。

That's true.

是的。

Yeah.

我觉得我的信任问题，我的意思是，这部分可能很难说清楚是天生还是后天形成的，对吧？

I think my trust issues, I mean, I think part stuff is, it's hard to tell whether it's like nature or nurture, right?

但我觉得我年轻时确实对权威有些抵触。

But I think that I definitely had some problems with authority when I was younger.

我没有典型的硅谷背景和成长环境。

And I had not the normal Silicon Valley kind of background and upbringing.

因此，我觉得我从小就不得不独自闯荡，非常独立。

And because of that, I think that I had to I really went out on my own at like a pretty young age and was pretty independent.

当你领导一家公司时，你必须学会的一件事是：你既是领导者，也是团队的合作伙伴。

One thing that you have to learn when you're leading a company is you are a leader and you are a partner to the rest of your team.

这不仅仅是你一个人的问题。

It is not just you.

我不是哈维。

Like, I am not Harvey.

哈维不是我。

Like, Harvey is not me.

这是一个团队在共同建设这家公司。

Like, it is a group of people that are building this company.

我认为，有时候创始人会开始说，我想成为第一，我把这看作是一个运动团队。

And I think that sometimes what founders can end up doing is they can start basically saying, I want to be the number one I think of this as like a sports team.

我认识一些人，他们并不在乎赢得冠军。

There are people I know that they don't care about winning the championship.

他们只想成为得分最多的人。

They want to be the person who scored the most points.

只要他们是得分最多的人，即使输掉冠军他们也无所谓。

And they're okay with losing the championship as long as they're the one that scores the most points.

那些正是我不想共事的人。

Those are the exact type of people that I do not want to work with.

我想和那些在意自己得了多少分的人共事，但他们之所以在意，是因为他们帮助球队赢了比赛。

I want to work with people that do care about how many points they've scored, but they care about that because they helped win the game.

对吧？

Right?

我认为这在科技行业是个大问题：太多人只想着‘我、我、我、我、我’，而不是‘公司、公司、公司’。

And I think that's a huge problem in tech, is where you have too many people that it's me, me, me, me, me, not company, company, company.

天啊，我觉得美国到处都是追逐品牌的人。

God, I think The US is just full of logo chasers.

这种现象太多了。

There's a lot of that.

你们就是喜欢去热门公司工作。

You guys just love to, like, work at a hot company.

我称之为‘热门公司门卫’，就是他们一待就是两年、两年、两年，像是在用公司股权做风险投资组合。

It's the hot company bouncer, I call it, where it's like they just go two year, two year, two year, and it's like they're doing venture portfolios with, like, company equity.

是的。

Yeah.

而且简直让人无语。

And and it's just like, god.

我正想说风险投资公司呢。

I was about to say VC's

同样的情况，我参加一些运营者小组，他们就说：‘哦，听说Clay现在超火。’

the same And I sit in these operator groups and they're like, oh, I hear Clay's really hot.

或者：‘哦，听说Notion现在超火。’

Or, oh, I hear Notion's really hot.

然后他们就立马跳过去。

And they just jump.

是的。

Yeah.

我的意思是，我觉得

I mean, I think The

美国运营者的随意性

promiscuity of American operators

令人难以置信。

is incredible.

同样的事情。

The same thing.

但是但是，你看啊

But but but, like, look at

英国人太消极了。

Brits are too negative.

我们觉得这太烂了。

We're like, that's crap.

这很好。

That's good.

这个我也忘了是谁发的推。

This also I I forgot who tweeted this.

我不记得是谁了，但这是2022年的事，就在ChatGPT发布之后。

I don't I don't remember who it was, but this was, like, 2022, and it was, like, right after the ChatGPT launch.

有个人，我不记得是哪位风投做的了，抱歉没提到他们的名字。

And someone I forgot which VC did this, so sorry for not giving credit to them.

但他们基本上发了条推文，预测会发生的事是：很多风投因为不理解AI生态，会重新回到看简历的方式，因为他们根本不懂。

But they basically tweeted, my prediction is what's gonna happen is a lot of VCs, because they don't understand the AI ecosystem, they're gonna revert back to looking at resumes because they don't understand.

这就是人们会做的事。

And this is what people do.

当出现混乱局面、人们不清楚状况时，最安全的做法就是去看其他社会信号来做决定，而不是依靠自己的直觉。

When there are situations of chaos and folks don't know what's going on, the safest thing is to go look at other social signals to make decisions instead of using your own gut.

所以我认为这在AI领域正大量发生，希望随着市场成熟，这种情况能停止。

And so I think that that's happening a lot in AI and hopefully, you know, as the markets mature, this stops happening.

但有很多事，我真的不太理解。

But there's a lot of, I don't really understand this.

所以我打算去看看简历、看看Logo，或者看那个，因为看起来好像

And so what I'm gonna do is look at the resume or look at a logo or look at that because that seems like