本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
我的意思是，每个行业都有人才争夺，但从未达到这种规模。
I mean, every industry has talent wars, but not at this magnitude.
对吧？
Right?
你几乎看不到有人被挖走时开出50亿美元的价码。
Very rarely can you see someone get poached for $5,000,000,000.
这很难竞争。
That's hard to compete with.
所以这几乎都成段子了。
So it's almost become a meme.
对吧？
Right?
意思是，如果你一年内不能从零增长到一百，你就没意思，这话说得简直太荒谬了。
Which is, like, if you're not basically growing from zero to a 100 in a year, you're not interesting, which is just the silliest thing to say.
当真正的能力突破出现时，需求就会存在，一旦启动，收入增长速度比我们以往见过的都要快得多。
When there's a real capability breakthrough, the demand is there, and so the revenue growth is much faster than we've ever seen once it's turned on.
可能存在一种系统性情况，即大模型公司能够筹集如此巨额的资金，以至于它们可以超越任何在其基础上构建的公司，这在以前我们从未见过，因为我们过去在工程方面一直受到严重限制。
There could be a systemic situation where the soda models can raise so much money that they can outpay anybody that builds on top of them, which would be something I don't think we've ever seen before, just because we were so bottlenecked in engineering.
在互联网建设时期，投资者把钱投进了没人使用的光纤网络。
During the Internet build out, investors put money into fiber that nobody used.
随后出现了四年的供应过剩。
Four years of supply overhang followed.
这一次，没有闲置的GPU。
This time, there are no dark GPUs.
每一笔投入计算资源的资金，另一边都有相应的需求。
Every dollar going into compute has demand on the other side.
但还有其他不同之处。
But something else is different.
一家模型公司可以筹集资金，在一年内用20人的团队发布一个模型，并立即产生市场需求。
A model company can raise capital, drop a model in a year with a team of 20, and produce something with immediate demand.
如果Frontier Labs筹集的资金是所有在其之上构建的公司总和的三倍，它们可能会吞噬整个应用层。
If Frontier Labs can raise three times more than the aggregate of every company built on top of them, they may consume the entire application layer.
或者市场会碎片化，价值会聚集在最接近最终用户的公司身上。
Or the market fragments and value accrues to the companies closest to the end user.
没有人知道哪条路径会胜出。
Nobody knows which path wins.
在这段之前在《Latent Space》播客中播出的对话中，十六资本的普通合伙人马丁·卡萨多和莎拉·王与阿莱斯蒂奥·费内利和肖恩·王讨论了资本飞轮、人才战争、为什么枯燥的软件被低估，以及是否每个任务都是AGI完备的。
In this conversation previously aired on the Latent Space podcast, Martin Casado and Sarah Wang, general partners at a sixteen z, speak with Alestio Finnelli and Sean Wang about the capital flywheel, talent wars, and why boring software is underinvested and Whether Every Task is AGI Complete.
大家好。
Hey, everyone.
欢迎收听由十六资本现场直播的《Latent Space》播客。
Welcome to the Laid Space podcast live from a sixteen z.
我是来自真实内核的阿莱斯蒂奥，今天和《Latent Space》的编辑斯威克斯一起。
This is Alessio from the real Kernel Lance, and I'm joined by Swiggs, editor of Laid in Space.
嗨。
Hey.
嗨。
Hey.
嘿。
Hey.
我们非常高兴能和你们一起参与这次对话。
And we're so glad to be on with you guys.
还有，这是顶级的AI播客，马丁·卡萨多和莎拉·王。
Also, a top AI podcast, Martin Casado and Sarah Wang.
欢迎。
Welcome.
非常高兴能来到这里，欢迎你们。
Very happy to be here, and welcome.
是的。
Yes.
我们很喜欢这个办公室。
We love this office.
你们把这里布置得真棒。
We love what you've done with the place.
新的标志现在无处不在。
The new logo is everywhere now.
是的。
Yeah.
是的。
Yeah.
是的。
Yeah.
它还是需要一些时间来适应，但它让我想起一种更雄心勃勃时代的回响，我觉得这挺有意思的。
It's it's still getting takes a while to get used to, but it reminds me of, like, sort of a callback to a more ambitious age, which I think is kind of
确实很有冲击力。
Definitely makes a statement.
描述一下。
Describe it.
是的。
Yeah.
是的。
Yeah.
不太确定这个声明具体是什么，但它确实发出了一个信号。
Not quite sure what that statement is, but it makes a statement.
马丁，我从Netlify时期就认识你了。
Martin, I go back with you to Netlify.
对。
Yep.
你知道的，你搞过软件定义网络之类的东西。
And, you know, you create a software defined networking and all all that stuff.
人们可以去了解你的背景。
People can read up on your background.
对。
Yep.
莎拉，我对你比较新。
Sarah, I'm newer to you.
你们差不多是从AI基础设施方面开始合作的。
You you sort of started working together on AI infrastructure stuff.
没错。
That's right.
是的。
Yeah.
现在已经七年了。
Seven seven years ago now.
整个行业里最出色的 Growth 投资者。
Best growth investor in the entire industry.
再多说一点。
Oh, say more.
毫无疑问。
Hands down.
对。
Yes.
确实有。
There is.
确实有。
There is.
我的意思是，谈到AI公司，我认为莎拉在AI模型方面提出了最激进的投资理念。
I mean, when it comes to AI companies, Sarah, I think, has done the most kind of aggressive investment thesis around AI models.
对吧？
Right?
所以她曾与诺阿·贾齐尔、米拉、伊利亚、李飞飞合作。
So she worked with Noam Jazir, Meera, Ilia, Fei Fei.
因此，就这些前沿的大型AI模型而言，我认为莎拉是最广泛的投资人。
And so just these frontier kind of like large AI models, I think, you know, Sarah's been the the broadest investor.
这样说公平吗？
Is that fair?
不。
No.
嗯，我本来想说，我觉得这其实是一个非常有趣的搭档组合，因为很多这样的种子轮投资，不仅融资金额很大。
I I well, I was gonna say, I think it's a really interesting tag tag team, actually, just because the a lot of these big seed deals, not only are they raising a lot of money.
但这仍然是对科技创始人的押注，这显然是早期阶段，但很多人需要的资源。
It's still a tech founder bet, which obviously is inherently early stage, but the resources many people.
这些资源一方面让他们增长非常快，另一方面，他们从第一天起就需要的是增长型资源。
The resource like, one, they just grow really quickly, but then two, the resources that they need day one are kind of growth scale.
所以我认为，我们这种混合搭档组合非常有效。
So I the hybrid tag team that we have is quite effective, I think.
如今什么是增长？
What is growth these days?
你知道，如果增长不到十亿，你都不会醒过来，实际上它真的非常，非常
You know, you don't wake up if it's less than a billion or, like It's actually actually it's actually very, like like
不。
no.
现在投资是一个非常有趣的时期，因为，比如说，围绕这个角色。
It's a very interesting time in investing because, like, you know, take, like, the character around.
对吧？
Right?
这些项目通常处于 monetization 之前，但资金规模足够大，需要更大的基金和更深入的分析，因为你拥有大量用户，而这类产品需求极高，需要更强的数据分析能力。
These tend to be, like, premonetization, but the dollars are large enough that you need to have a larger fund and the analysis, you know, because you've got lots of users because this stuff has such high demand requires, you know, more of a number sophistication.
因此，无论是我们还是其他机构对这些大型模型公司的投资，都是一种风险投资与成长型投资的混合模式。
And so most of these deals, whether it's us or other firms on these large model companies, are like this hybrid between venture and growth.
是的。
Yeah.
完全正确。
Total.
我认为，比如商务拓展（BD）这类工作，通常在种子阶段是不需要的，当时你只是在试图
And I think, you know, stuff like BD, for example, you wouldn't usually need BD when you were seed stage trying to
你在谈业务吗？
get Are talking about business?
商务拓展。
Biz dev.
没错。
Exactly.
但现在，比如
But, like, now
我不太清楚风险投资基金里的业务拓展是什么意思？
I'm not familiar what what what does BizDev mean for a venture fund?
因为我知道业务拓展对公司来说意味着什么。
Because I know what BizDev means for a company.
是的。
Yeah.
你懂的？
You know?
一个很好的例子是，我们常说购买计算资源，但这里面涉及大量的谈判，比如：你能否用计算资源换取股权？
So a a good example is I mean, we talk about buying compute, but there's a huge negotiation involved there in terms of, okay, do you get equity for the compute?
你打算找什么样的合作伙伴？
What what sort of partner are you looking at?
这有专门的市场推广团队吗？
Is there a go to market arm to that?
在这样的规模下，几亿美元的融资，可能在公司成立仅六个月时就会发生。
And these are just things on this scale, hundreds of millions, you know, maybe six months into the inception of a company.
以前你根本不需要谈判这些交易。
You just wouldn't have to negotiate these deals before.
是的。
Yeah.
如今这些大额融资变得非常复杂。
These large rounds are very complex now.
过去，如果你完成一轮A轮或B轮融资，比如几千万到六千万美元的注资，事情就结束了。
Like in the past, if you did a series a or series b, like whatever, you're writing a 20 to a $60,000,000 check and you call it a day.
现在通常会有财务投资者和战略投资者，而战略部分总是伴随着这类大型计算资源合同，这些合同可能需要数月才能谈成。
Now you normally have financial investors or strategic investors, and then the strategic portion always still goes with, like, these kind of large compute contracts, which can take months to do.
因此，情况已经大不相同了。
And so it's it's a very different ties.
听好了。
Listen.
我做这一行已经十年了。
I've been doing this for ten years.
我从未见过这样的情况。
This is the I've never seen anything like this.
是的。
Yeah.
你担心这些战略投资者的循环融资吗？
Do you have worries about the circular funding from some of these strategics?
不担心。
No.
听好了。
Listen.
只要需求存在，需求就存在。
As long as the demand is there, the demand is there.
问题是，互联网当初并没有需求。
Like, the problem with the Internet is the demand wasn't there.
没错。
Exactly.
好吧。
Alright.
这整个就是庞氏骗局的泡沫，只要你们按名义价值来市值计价，那就没问题。
This is this is, like, the the whole pyramid scheme bubble thing where, like, as long as you mark to market on, like, the notional value of, like, these deals, fine.
但一旦开始崩塌，那就真的不行了。
But, like, once it starts to chip away, it really Well, no.
这就像是
It's just like
只要还有需求，你知道的，这些说法虽然已经成了陈词滥调，但还是值得说一说。
if if as long as there's demand I mean, you know, this this is like, a lot of these sound bites have already become kind of cliches, but they're worth saying it.
对吧？
Right?
就像在互联网时代，我们筹集资金铺设了根本没有使用的光纤。
Like, during the Internet days, like, we were raising money to put fiber in the ground that wasn't used.
这确实是个问题。
And that's a problem.
对吧？
Right?
因为现在你实际上面临供应过剩。
Because now you actually have a supply overhang.
嗯。
Mhmm.
即使在互联网时代，带宽过剩的问题——尽管没那么严重——也只持续了大约四年。
And even in the the time of the the Internet, like, the supply and and bandwidth overhang, even as massive as it wasn't as as massive as the crash was only lasted about four years.
但我们并没有供应过剩。
But We don't have a supply overhang.
根本不存在闲置的GPU。
Like, there's no dark GPUs.
对吧？
Right?
我的意思是，不管是不是循环的，如果有人投资一家公司，他们确实会使用这些GPU，而另一端则是客户。
I mean and so, you know, circular or not, I mean, you know, if if someone invests in a company, you know, they'll actually use the GPUs, and on the other side of it is the is the customer.
所以我觉得这是个不同的时代。
So I I I think it's a different time.
我认为另一点是，也许可以补充一下，我要引用一下马蒂娜的话，但这也是一个独特的时候，因为第一次你可以真正地将资金追踪到成果。
I think the other piece, maybe just to add on to this, and I'm gonna quote Martina in front of him, but this is probably also a unique time in that for the first time, you can actually trace dollars to outcomes.
对吧？
Right?
前提是规模定律仍然成立，并且能力确实在向前发展。
Provided that scaling laws are are holding, and capabilities are actually moving forward.
因为如果你能把资金转化为能力、能力提升，就像马蒂娜说的，那里就有需求。
Because if you can put translate dollars into capabilities, capability improvement, there's demand there to Martine's point.
但如果这一点出了问题，显然，这是整个逻辑成立的重要前提。
But if that somehow breaks, you know, obviously, that's an important assumption in this whole thing to make it work.
但你知道，与其把钱投在销售和市场推广上，你现在是在投入研发，以提升能力，这某种程度上一直是需求的驱动力。
But, you know, instead of investing dollars into sales and marketing, you're you're investing into r and d to get to the capability, you know, increase, and that's sort of been the demand driver.
因为一旦实现突破，人们就愿意为此付费。
Because once there's an unlock there, people are willing to pay for it.
是的。
Yeah.
现在你们的某些成长型公司已经类似于早期初创公司的基础设施，你们在构建投资组合时是否因此有所不同？
Is there any difference in how you build the portfolio now that some of your growth companies are like the infrastructure of the early stage companies?
比如，OpenAI 现在的规模已经和早期的云服务商差不多了。
Like, you know, OpenAI is now the same size as some of the cloud providers were early on.
这种情况会是什么样子？
Like, what does that look like?
这两类公司之间能相互借鉴多少信息？
Like, how much information can you feed off each other between the the two?
现在有太多界限正在被打破或变得模糊。
There's so many lines that are being crossed right now or blurred.
对吧？
Right?
我们之前已经聊过风险投资和成长型投资。
So we already talked about venture and growth.
另一个正在变得模糊的领域是基础设施和应用之间。
Another one that's being blurred is between infrastructure and apps.
对吧？
Right?
那么，什么是模型公司？
So, like, what is a model company?
它显然是基础设施。
Like, it's clearly infrastructure.
对吧？
Right?
因为它是在做核心的研发工作。
Because it's like, you know, it's doing kind of core r and d.
它是一个横向平台，但也是一个应用程序，因为它直接接触用户。
It's a horizontal platform, but it's also an app because it touches the users directly.
当然，这些平台的增长速度非常高。
And then, of course, you know, the the the growth of these is just so high.
所以我真的认为，你正开始看到一种新的融资策略出现，我们也因此不得不做出调整。
And so I actually think you're just starting to see a a new financing strategy emerge, and, you know, we've had to adapt as a result of that.
所以已经发生了许多变化。
And so there's been a lot of changes.
你说得对，这些公司很快就会变成平台公司。
You're right that these companies become platform companies very quickly.
你正在构建一个生态系统。
You've got ecosystem build out.
所以这些都不是全新的，但它们发生的速度确实非常惊人。
So none of this is necessarily new, but the timescales in which it's happened is pretty phenomenal.
我们过去通常划分界限的方式现在变得有些模糊了。
And the way we'd normally cut lines before is blurred a little bit.
但话又说回来，很多内容其实也像是我们过去见过的东西，比如云计算的扩展和互联网的建设。
But but that that that said, I mean, a lot of it also just does feel like things that we've seen in the past, like cloud build out and the Internet build out as well.
是的。
Yeah.
对。
Yeah.
我觉得这很有趣。
I think it's interesting.
我不知道你们是否同意这一点，但我觉得新兴的策略是——这跟你们之前的问题有关。
I don't know if you guys would agree with this, but it feels like the emerging strategy is and this build off of your other question.
你们为计算能力融资。
You raise money for compute.
你们把资金投入计算能力。
You pour that or you you pour the money into compute.
然后取得某种突破。
You get some sort of breakthrough.
你将突破性成果投入到你垂直整合的应用中。
You funnel the breakthrough into your vertically integrated application.
这可能是ChatGPT。
That could be ChatGPT.
这可能是Cloud Code，不管是什么。
That could be Cloud Code, you know, whatever it is.
你会大幅增加市场份额并获取用户。
You massively gain share and get users.
也许在那时，根据你的策略，你甚至还在进行补贴。
Maybe you're even subsidizing at that point, depending on your strategy.
你在势头最盛时融资，然后重复、循环、再重复。
You raise money at the peak momentum, and then you repeat, rinse, and repeat.
而且，我认为即使是两年前，情况也不是这样的。
And so and that wasn't true even two years ago, I think.
是的。
Mhmm.
所以这其实与你的融资策略紧密相关。
And so it's sort of to your just tying into fundraising strategy.
对吧？
Right?
这还涉及招聘策略。
There's a and hiring strategy.
所有这些都相互关联。
All of these are tied.
我认为，如今这些界限变得更加模糊了，因为大家都在这么做，当然，这些公司都有API业务。
I think the lines are blurring even more today where everyone is and it but, of course, these companies all have API businesses.
因此，这些既是竞争对手又是合作伙伴的关系正在变得越来越模糊。
And so there are these these frenemy lines that are getting blurred in that.
很多公司，我的意思是，API收入高达数十亿美元。
A lot of I mean, have billions of dollars of API revenue.
对吧？
Right?
所以那里有客户，但他们是在应用层面上竞争。
And so there are customers there, but they're competing on the app layer.
是的。
Yeah.
所以这是一个非常非常重要的观点。
So this is a really, really important point.
所以我肯定地说，风投和成长阶段之间的界限已经变得模糊了。
So I I would say, for sure, venture and growth, that line is blurry.
应用层和基础设施之间的界限也变得模糊了。
App and infrastructure, that line is blurry.
但我认为这并没有太大改变我们的实践。
But I don't think that changes our practice so much.
但真正开放的问题是，这一层是否像传统计算那样发挥作用？
But, like, where the very open questions are, like, does this layer in the same way compute traditionally has?
就像在云计算时代一样，你知道的，不管怎样。
Like, during the cloud is, like, you know, like, whatever.
某个人在某一层赢了，但另一批公司则在另一层获胜。
Somebody wins one layer, but then another whole set of companies wins another layer.
但在这里，情况可能并非如此。
But that not might not be the case here.
很可能你实际上无法在令牌流上进行垂直整合。
It may be the case that you actually can't verticalize on the token stream.
比如，你无法构建一个，因为它必然向下发展，只是因为缺乏抽象层。
Like, you can't build an Like, it it it necessarily goes down just because there are no abstractions.
所以这些是我们提出的更重大的存在性问题。
So those are kind of the bigger existential questions we ask.
这次与计算机科学历史上的另一个巨大不同是，过去如果你筹集了资金，你就必须等待工程团队赶上来，而这众所周知是无法扩展的。
Another thing that is very different this time than in the history of computer science is is in the past, if you raised money, then you basically had to wait for engineering to catch up, which famously doesn't scale.
比如，神话般的人月需要很长的时间，但这里的情况并非如此。
Like, the mythical man must take a very long time, but, like, that's not the case here.
比如，一家模型公司可以筹集资金，并在一年内推出一个更优秀的模型。
Like, a model company can raise money and drop a model in a in a year, and it's better.
对吧？
Right?
而且它只用二十个人或者十个人的团队就能做到。
And and it does it with a team of 20 people or 10 people.
所以这种资金进入公司后立即产生需求和增长，并用它来筹集更多资金的方式，是一种我们前所未见的全新资本飞轮。
So this type of, like, money entering a company and then producing something that has demand and growth right away and using that to raise more money is a very different capital flywheel than we've ever seen before.
我认为每个人都试图理解这些变化的后果。
And I think everybody's trying to understand what the consequences are.
所以我认为这与其说是关于大公司和增长，不如说是关于我们实际上还没有答案的更系统性的问题。
So I think it's less about big companies and growth and this, and more about these more systemic questions that we actually don't have answers to.
是的。
Yeah.
比如在Kernel Labs，我们的一个想法是，如果你有无限的资金可以高效地用于将令牌转化为产品，整个早期阶段的市场就会完全不同。
Like, at Kernel Labs, one of our ideas is like, if you had unlimited money to spend productively to turn tokens into products, like the whole early stage market is very different.
因为今天你投入一定量的资金来赢得一个项目，是基于价格结构等因素，你实际上是在对某种策略进行长期承诺。
Because today you're investing x amount of capital to win a deal because of price structure and whatnot, and you're kind of pot committing to a certain strategy for a certain amount of time.
但如果你能迭代地孵化公司和产品，比如今天我就想花一百万美元的推理算力，明天就推出一个产品。
But if you could, like, iteratively spin out companies and products and just throw I I wanna spend a million dollar of inference today and get a product out tomorrow.
是的。
Yeah.
我们应该达到这样一个阶段：从代币到产品的转化摩擦如此之低，以至于你可以轻松做到这一点。
Like, we should get to the point where, like, the friction of, like, token to product is so low that you can do this.
然后你可以将早期风险投资模式转变为更加迭代的形式。
And then you can change the the early stage venture model to be much more iterative.
这样一来，每一轮融资可能只是百万美元的推理算力，或者来自ASIC的上亿美元资金。
And then every round, it's like either 100 k of inference or like 100,000,000 from ASICs in
在城市里。
the city.
不再有八百万美元的C轮融资了。
There's there's not like $8,000,000 C round anymore.
但确实有一个我们尚未解答的行业结构性问题，涉及前沿模型，比如Anthropic。
But but but there's a there's a the an industry structural question that we don't know the answer to, which involves the frontier models, which is, let's take Anthropic.
假设Anthropic拥有一个最先进的模型，并占据了相当大的市场份额。
Let's say Anthropic has a state of the art model that has some large percentage of market share.
假设有一家公司正在构建更小的模型，这些模型在后台使用更大的模型，比如Open 4.5，但在此基础上增加了价值。
And let's say that, you know, a company is building smaller models that, use the bigger model in the background, they use Open 4.5, but they add value on top of that.
如果Anthropic每次融资都能获得三倍于前一轮的资金，它可能筹集到的资金将超过所有基于它构建的应用生态系统的总和。
Now if Anthropic can raise three times more every subsequent round, they probably can raise more money than the entire app ecosystem that's built on top of it.
如果是这样，它就能超越所有基于它构建的产品。
And if that's the case, they can expand beyond everything built on top of it.
想象一颗正在不断扩张的恒星。
Imagine a star that's just kind of expanding.
因此，可能会出现一种系统性情况：前沿模型能够筹集如此多的资金，以至于它们可以支付得起所有基于它们构建的公司，这在以前是从未见过的，因为我们过去一直受制于工程瓶颈。
So there could a systemic situation where the Soda models can raise so much money that they can outpay anybody that builds on top of them, which would be something I don't think we've ever seen before just because we were so bottlenecked in engineering.
这是一个非常开放的问题。
And it's a very open question.
是的。
Yeah.
这简直就像给创业行业上了一堂苦涩的课。
It's almost like bitter lesson applied to the startup industry.
百分之百。
A 100%.
是的。
Yeah.
它实际上变成了一个问题：筹集资本，直接将其转化为增长，然后用它来筹集三倍更多的资金。
It literally becomes an issue of, like, raise capital, turn that directly into growth, use that to raise three times more.
如果你能持续这样做，你就能彻底超越任何构建在你之上的公司。
And if you can keep doing that, you literally can outspend any company that's built not any company.
你可以超越所有构建在你之上的公司的总和，因此你必然会抢占它们的市场份额，这太疯狂了。
You can outspend the aggregate of companies on top of you, and therefore, you'll necessarily take their share, which is crazy.
你会说这种情况在Character身上发生过吗？
Would you say that kind of happened to character?
这是否就是对发生之事的事后复盘？
Is that the the sort of postmortem on what happened?
不。
No.
对。
Yeah.
因为我觉得
Because I think
所以我的实际复盘是他
so I actual postmortem is he
想回到谷歌。
wanted to go back to Google.
没错。
Exactly.
是的。
Yeah.
好的。
Okay.
但那又是另一个不同的事情了
But, like that's another different that's
我们之前说过的。
what we said.
我们应该谈谈
We should talk
我们其实应该谈谈这件事。
we should actually talk about this.
是的。
Yeah.
对。
Yeah.
没关系。
That's alright.
去吧。
Go for it.
拿去，拿去
Take it take it
你想要的，拿上去。
up, you want.
我想说，角色这个事情实际上引发了一个不同的问题，这也是Frontier Labs将面临的挑战。
Gonna say, I think the the the character thing raises actually a different issue, which actually the Frontier Labs will face as well.
所以我们看看他们如何处理。
So we'll see how they handle it.
但我们是在2023年1月投资了角色这个项目，感觉像是上个世纪的事了。
But, so we invest in character in January 2023, which feels like eons ago.
我的意思是，三年前了。
I mean, three years ago.
感觉像是隔了几个世纪。
Feels like lifetimes ago.
但后来他们在2020年8月，也就是‘24年，和谷歌达成了IP授权协议。
But, and then they did the IP licensing deal with Google in August 2020, '4.
所以，当时，诺姆，你知道的，他公开谈论过这件事。
And so, you know, at the time, Noam, you know, he's talked publicly about this.
对吧？
Right?
他希望谷歌不让他把产品推向市场。
He wanted to Google wouldn't let him put out products in the world.
这当然已经发生了巨大变化，但他还是去做了。
That's obviously changed drastically, but, he went to go do that.
但他当时有一个具体的产品。
But he had a product attached.
目标是，哦，说到底，这是诺姆·沙齐尔。
The goal was oh, I mean, it's Noam Shazir.
他想实现通用人工智能。
He wanted to get to AGI.
这一直是他个人的目标。
That was always his personal goal.
但你知道，我认为通过收集数据，尤其是在这种最初和现在仍然是人类使用场景的字符产品中，这是一种实现这一目标的途径。
But, you know, I think through collecting data, right, in this sort of very human use case that the character product originally was and still is, was one of the vehicles to do that.
我认为，真正让任何公司在最终做出抉择前感到压力的原因，是AGI与产品之间的权衡。
I think the real reason that, you know, if you think about the the stress that any company feels before you ultimately go in one way or the other is sort of this AGI versus product.
我认为，很多大型公司，比如OpenAI，现在正感受到这种压力。
And I think a lot of the big I think, you know, OpenAI is feeling that.
Anthropic如果还没开始感受到的话，鉴于他们产品的成功，很快也可能开始感受到这种压力。
Anthropic, if they haven't started you know, felt it, certainly given the success of their products, they may start to feel that soon.
而且确实存在真正的取舍。
And they're real I think there's real trade offs.
对吧？
Right?
当你想到GPU时，这是一种有限的资源。
It's like how many when you think about GPUs, that's a limited resource.
你该把GPU分配到哪里？
Where do you allocate the GPUs?
是偏向产品吗？
Is it toward the product?
是偏向新研究吗？
Is it toward new research?
对吧？
Right?
是偏向长期研究吗？
Is it or long term research?
是偏向中短期研究吗？
Is it toward, you know, near to midterm research?
所以在资源受限的情况下，当然你可以玩融资这个游戏。
And so in a case where you're resource constrained, of course, there's this fundraising game you can play.
对吧？
Right?
但市场在2023年时可是大不相同的。
But the fund the market was very different back in 2023 too.
我认为世界上最好的研究人员都面临这样一个困境：好吧。
I think the best researchers in the world have this dilemma of, okay.
我想全力投入通用人工智能，但正是产品使用带来的收入飞轮，才让公司有足够的收入来购买GPU，从而实现通用人工智能。
I wanna go all in on AGI, but it's the product usage revenue flywheel that keeps the revenue in the house to power all the GPUs to get to AGI.
因此，这确实让人意识到，对于任何在融资上难以达到这一水平的初创公司来说，都构成了一个有趣的困境。
And so it does make you know, I think it sets up an interesting dilemma for any startup that has trouble raising up until that level.
对吧？
Right?
而且，如果你没有取得这样的进展，就无法继续这种融资飞轮。
And, certainly, if you don't have that progress, you can't continue this fly you know, fundraising flywheel.
我会说，因为我们一直在追踪
I would say that because because we're keeping track of
所有那些
all of the things that
不同的事情。
are different.
对吧？
Right?
比如，你知道的，风险增长和应用基础设施，其中一个关键因素绝对是创始人的个性。
Like, you know, venture growth and app infra, and one of the ones is definitely the personalities of the founders.
这次真的非常不一样。
It's just very different this time.
我的意思是，我已经做了十年这行，创业也二十年了。
Mean, I've been I've been doing this for a decade, and I've been doing startups for twenty years.
所以，很多人开始做这个就是为了实现通用人工智能。
And so, I mean, a lot of people start this to do AGI.
而我们以前从未有过像现在这样明确统一的北极星目标。
And we've never had, like, a unified north star that I recall in the same way.
过去，人们创办公司只是为了创办公司。
Like, people built companies to start companies in the past.
那就是当时的情况。
Like, that was what it was.
我会创建一家互联网公司。
Like, I would create an Internet company.
我会创建一家基础设施公司。
I would create an infrastructure company.
这更像是工程型建设者，而这是一种不同的思维方式。
Like, it's kind of more engineering builders, and this is kind of a different, you know, mentality.
有些公司在这方面做得非常好，因为它们的方向明显朝着人们所认为的AGI前进，但其他公司则没有。
And some companies harness that incredibly well because their direction is so obviously on the path to what somebody would consider AGI, but others have not.
因此，人员方面总是存在这种张力。
And so, like, there is always this tension with personnel.
所以我认为，我们正看到比以往更多的创始人流动，是的。
And so I think we're seeing more kind of founder movement Yeah.
作为创始人比例，这比以往任何时候都高。
You know, as a fraction of founders than we've ever seen.
也许自肖克利和多塞特的那批人以来，也就是行业初期以来，就从未有过这样的情况。
And maybe since, like, I don't know, the time of, like, Shockley and the trade at Dorset or something like that way back to the beginning of the industry.
我的意思是，现在的人才状况非常非常特殊。
I mean, it's a very, very unusual time of personnel.
完全正确。
Totally.
而且我认为，这种情况因人才竞争而加剧了——每个行业都有人才竞争，但从未达到这种规模。
And it I think it's exacerbated by the fact that talent wars I mean, every industry has talent wars, but not at this magnitude.
对吧？
Right?
几乎不可能看到有人被挖走时开出五十亿美元的价码。
Very rarely can you see someone get poached for $5,000,000,000.
这很难竞争。
That's hard to compete with.
其次，如果你是AI领域的创始人，你只要打个嗝，都可能成为当今信息时代的头版新闻。
And then secondly, if you're a founder in AI, you could fart, and it would be on the front page of, you know, the information these days.
因此，这种‘鱼缸效应’我认为加剧了这些AI创始人所感受到的深层焦虑。
And so there's sort of this fishbowl effect that I think adds to the deep anxiety that that these AI founders are feeling.
是的。
Yes.
我只是简要评论一下创始人和人才竞争这个问题。
I mean, just on briefly comment on the founder the sort of talent wars thing.
我觉得2025年只是个短暂的插曲。
I feel like 2025 was just like a blip.
我不知道我们是否还会再看到这种情况，因为Meta已经组建了团队。
Like, I I don't know if we'll see that again because Meta built the team.
我觉得他们可能已经差不多完成了，还有谁会出比Meta更高的价格呢？
Like, I don't know if I think I think they're kinda done, like, who's gonna pay more than Meta?
我不确定。
I I don't know.
我同意。
I I agree.
这感觉确实也让我有同样的感受。
It's so it feels it's so it feel it feels this way to me too.
这就像扎克伯格一开始强势出击，现在又回到专注建设了。
It's like it's like basically Zuckerberg kinda came out swinging, and then now he's kinda back to building.
是的。
Yeah.
对。
Yeah.
你知道，你得付高薪才能把团队凑齐，赶在截止前完成工作，等等。
You know, you gotta, like, pay up to, like, assemble the team to rush the job, whatever.
但现在你已经做出了选择，接下来就得把产品推出去了。
But then now now you, like, you you made your choices, and now they gotta ship.
对吧？
Right?
我的意思是，另一方面，我们现在其实正处于招聘市场中。
Like I mean, the The Us other side of that is, you know, like, we're we're actually in the job hiring market.
我们这里已经有600人了。
We've got 600 people here.
展开剩余字幕（还有 480 条）
我一直在招聘。
I hire all the time.
如果有人在听这个并且感兴趣，我这里有三个职位推荐。
I've got three open recs if anybody's interested that's listening to this.
或者是投资者？
Or investor?
是的。
Yeah.
在团队里。
To on the team.
比如在投资团队这边。
Like, on the investing side of the team.
我们接触的很多人手头都有年薪一千万左右的现成offer。
Like and a lot of the people we talk to have acting, you know, active offers for 10,000,000 a year or something like that.
而且，你知道，我们给的薪酬真的非常高。
And, like, you know, and we pay really, really well.
只是看看市场上有什么，真的非常惊人。
And just to see what's out on the market is really is really remarkable.
所以我想说，你其实说得对。
And so I would just say it's actually so you're right.
比如，那些特别炫目的公司可能会花十亿美元雇人，但泡沫正在消退。
Like, the really flashy one, like, will get someone for, you know, a billion dollars, but, like, the inflated Chickles down.
是的。
Yeah.
今天市场依然非常活跃。
It is is still very active today.
我的意思是
I mean
对。
Yeah.
你即使是个L5，也能拿到数千万美元的报价。
You could be an l five and get an offer in the tens of millions.
是的。
Yeah.
很容易。
Easily.
确实是。
It's Yeah.
所以我认为你是对的，这感觉像是一个短暂现象。
So I think you're right that it felt like a blip.
希望你是对的。
Hope I hope you're right.
但我认为这已经是常态了。
But I think it's been it the steady state
现在已经回升了。
has now pulled up.
是的。
Yeah.
是的。
Yeah.
没错。
Exactly.
是的。
Yeah.
确实上升了。
Pulled up for sure.
是的。
Yeah.
是的。
Yeah.
我认为这打破了早期创始人的数学模型。
And I think that's breaking the early stage founder math too.
我认为以前很多人会想，也许我应该直接去当创始人，而不是像这样拿工资。
I think before a lot of people were like, well, maybe I should just go be a founder instead of, like, getting paid Yeah.
在谷歌，80万到100万美元。
800 k, 1,000,000 at Google.
但如果我拿到5600万美元，那就另当别论了。
But if I'm getting paid $56,000,000, that's different.
但另一方面，如今的战略资金比历史上任何时候都要多。
But on but on the other hand, there's more strategic money than we've ever seen historically.
对吧？
Right?
所以没错。
And so Yep.
从经济角度来看，其计算方式在很多方面都截然不同，这给市场带来了大量变化和混乱。
The economics the the the calculus on the economics is very different in a number of ways, and it's it's caused a ton of change and confusion in the market.
有些是积极的，有些是消极的。
Some very positive, some negative.
比如，联合创始人被收购的另一面。
Like, so for example, the other side of the the cofounder, like, acquisition.
你知道，马克·扎克伯格花大价钱挖人，实际上我们正目睹史上最大规模的并购活动，基本上都是为了收购人才。
You know, Mark Zuckerberg poaching someone for a lot of money is, like, there's we're actually seeing historic amount of m and a for basically acqui hires.
对吧？
Right?
从风投的角度来看，你确实能看到一些非常成功的案例，这些本质上都是为了收购人才而进行的并购。
That you you, like, you know, really good outcomes from a venture perspective that are effective acqui hires.
对吧？
Right?
所以我认为，从投资角度来看，这总体上是积极的，尽管从新闻标题上看，它似乎以一种负面的方式造成了巨大冲击。
So I would say it's probably net positive from the investment standpoint, even though it seems from the headlines to be very disruptive in a negative way.
是的。
Yep.
我们来聊聊那些没有被投资的领域吧，比如一些你可能会看到更多人去打造的有趣想法。
Let's talk maybe about what's not being invested in, like, some interesting ideas that you will see more people build.
或者说，某种程度上，随着YC越来越受欢迎，X也变得越来越热门。
Or it it seems in a way, you know, as YC is getting more popular, it's like X has gotten more popular.
很多创始人走的是创业学校路径，他们知道风投圈里什么热门，也知道什么能拿到融资。
There's a start school path that a lot of founders take, and they know what's hot in the VC circles, and they know what gets funded.
而对于那些超出这个范围的事情，风险承受意愿可能没那么高。
And there's maybe not as much risk appetite for things outside of that.
我想知道你是否觉得这是真的。
I'm curious if you feel like that's true.
你认为哪些领域是被忽视的呢？
And what are maybe some of the areas that you think are under discussed?
我的意思是，实际上我觉得我们在很多传统的软件公司上已经忽略了重点。
I mean, I actually think that we've taken our eye off the ball in a lot of, like, just traditional, you know, software companies.
比如，我觉得现在几乎形成了一个双峰格局。
So you like I I think right now there's almost a barbell.
一方面，最热门的是下一代技术，比如深度科技。
Like, you're like the hot thing in the next, you're deep tech.
对吧？
Right?
但我觉得，其实有一长串很好的公司，它们所处的市场非常大，而且会长期存在。
But I you know, I feel like there's just kind of a long, you know, list of, like, good good companies that'll be around for a long time in very large markets.
比如说，你在开发一个数据库。
Say you're building a database.
你知道的。
You know?
比如说，你在做监控、日志记录、工具开发之类的。
Say you're building kind of monitoring or logging or tooling or whatever.
现在确实有一些不错的公司，但它们很难获得投资人的关注。
There's some good companies out there right now, but they have a really hard time getting the attention of investors.
这几乎都成段子了，对吧？如果你不是一年内从零增长到一亿，你就没意思，这话说得简直太荒谬了。
It's almost become a meme, right, which is if you're not basically growing from zero to a 100 in a year, you're not interesting, which is the silliest thing to say.
想想你自己，作为一个普通人。
I mean, think of yourself as like an individual person.
比如你个人的钱。
Like like, your personal money.
对吧？
Right?
所以你的个人资金，是放在股票市场里赚7%，还是投给这家在巨大市场中五年增长的公司？
So your personal money, will you put it in the stock market at 7%, or you put it in this company growing five x in a very large part?
当然，你可以投给五年增长五倍的公司。
Of course, you can put it in the company five x.
所以我们才说这些蠢话，比如你必须从零做到一百，但那些公司的利润率到底有多少，谁知道呢？
So it's just like the we say these stupid things, like, if you're not going from zero to a 100, but, like, those like, who knows what the margins of those are?
我的意思是，这些显然是对任何人来说都很好的投资。
I mean, clearly, these are good investments for anybody.
对吧？
Right?
比如我们的有限合伙人希望，在基金的整个生命周期内获得三倍的净回报。
Like, our LPs want whatever, three x net over, you know, the life cycle of a fund.
对吧？
Right?
所以，在一个大市场中增长五倍的公司是一个很好的投资。
So a company in a big market growing five x is a great investment.
我们都对这些回报感到满意，但我们对这种强劲增长产生了一种狂热。
We'd everybody would be happy with these returns, but we've got this kind of mania on these these strong growths.
因此，我认为这可能是目前最受忽视的领域。
And so I would say that that's probably the most underinvested sector right now.
枯燥的软件。
Boring software.
枯燥的企业软件。
Boring enterprise software.
没有传统的。
There's no traditional.
比如，一家非常好的公司。
Like, really good company.
这里涉及人工智能。
AI here.
不。
No.
嗯，AI呢，
Like, well, well, the AI,
当然，AI正在推动它们的应用场景，但那并不是它们的本质。
of course, is pulling them into use cases, but that's not what they are.
它们并不在代币路径上。
They're not on the token path.
对吧？
Right?
我们就这么说吧。
Let's just say that.
它们是软件，但并不在代币路径上。
Like, they're soft, but they're not on the token path.
这些公司本质上是很好的投资，除了那些在Twitter或X上随意发帖的风投说它们增长不够快之外。
Like, these are like, they are great investments from any definition except for, like, random VC on Twitter saying VC on x saying, like, it's not growing fast enough.
你怎么看？
What do you think?
也许我会回答一个稍微不同但相关的问题，那就是我们目前可能没有投资、但无论是否最终出手都在花大量时间关注的领域。
Maybe I'll answer a slightly different question, but adjacent to what you asked, which is maybe an area that we're not, investing right now that I think is a question and we're spending a lot of time in regardless of whether we pull the trigger or not.
实际上，这可能是在硬件方面。
And it would probably be on the hardware side, actually.
机器人。
Robotic.
对吧？
Right?
还有机器人领域。
And the robotics side.
对吧？
Right?
这方面的投资情况是，我不敢说它没获得资金，因为事实上，现在几乎不投资机器人反而成了非共识观点。
Which is it's I don't wanna say that it's not getting funding because it's clearly, it's it's sort of nonconsensus to almost not invest in robotics at this point.
但我们在这个领域投入了大量时间。
But, we spent a lot of time in that space.
我认为对我们来说，硬件方面还没有出现类似ChatGPT的突破时刻。
And I think for us, we just haven't seen the ChatGPT moment happen on the hardware side.
是的。
And Yeah.
投入到这个领域的资金似乎已经默认了这一点。
The funding going into it feels like it's already taking that for granted.
对。
Yeah.
对。
Yeah.
但我们还经历过无人机领域。
But we also went through the drone.
你知道的？
You know?
那里有个
There's a
高空滑索就在外面。
zipline right right out there.
那是那个
Was that The the
高空滑索。
zipline.
是的。
Yeah.
无人机和AI时代是什么样的？
What are the drone what's the AV era?
其中一个关键点是，谈到硬件时，大多数公司最终都会走向垂直整合。
And, like, one of the takeaways is when it comes to hardware, most companies will end up verticalizing.
比如，如果你投资一家农业机器人公司，你实际上是在投资一家农业公司，因为这才是真正的竞争对手，这很令人意外，也涉及供应链。
Like, if you're if you're investing in a robot company for an for agriculture, you're investing in an ag company because that's the competition, and that's surprising, and that's supply chain.
如果你是为采矿业做投资，那就是采矿业。
And if you're doing it for mining, that's mining.
因此，ADT 做了很多这类工作，因为它们本身就具备对这类业务进行尽职调查的条件。
And so the ADT does a lot of that type of stuff because they're actually set up to diligence that type of work.
但对于横向技术投资来说，机器人领域的投资机会非常少，因为机器人太过于专为特定用途设计了。
But for, like, horizontal technology investing, there's very little when it comes to robots just because it's it's so fit for for purpose.
所以我们更倾向于关注软件解决方案或横向解决方案，比如来自自动驾驶浪潮的 Applied Intuition，以及同样来自自动驾驶浪潮的 DeepMap。
And so we kinda like to look at software solutions or horizontal solutions, like Applied Intuition clearly from the AV Wave, DeepMap clearly from the AV Wave.
我认为 ScaleAI 实际上是一个面向机器人早期阶段的横向平台。
I would say Scaleai was actually a horizontal one for
这很公平。
That's fair.
你知道的，在机器人发展的早期阶段。
You know, for robotics early on.
所以这类东西，我们非常感兴趣。
So that sort of thing, we're very, very interested.
但机器人与现实世界交互的实际部分，可能更适合另一个团队来处理。
But the actual, like, robot interacting with the world is probably better for a different team.
是的。
Yeah.
嗯。
Mhmm.
是的。
Yeah.
我很好奇，这些 supposedly 投资机器人公司的团队究竟是哪些。
I'm curious who these teams are supposed to be that invest in them.
我觉得每个人都说，机器人很重要，人们应该投资它。
I feel like everybody's like, yeah, robotics, it's important, and, like, people should invest in it.
但当你看数据时，早期的资金需求和真正确认它能成功之间的差距。
But then when you look at, like, the numbers, like, the capital requirements early on versus, like, the moment of, okay, this is actually gonna work.
让我们继续投资。
Let's keep investing.
这似乎很难以一种不那么确定的方式预测。
That seems really hard to predict in a way that is not.
我的意思是，CO2、COSLA、GC。
I mean, CO2, COSLA, GC.
我的意思是，这些都投资于硬件公司。
I mean, these are all invested in in hardware companies.
你就知道吧？
You just you know?
然后听好了。
And then listen.
我的意思是，这次肯定能成功。
I mean, it could work this time for sure.
对吧？
Right?
我的意思是，如果马斯克在做这件事，那肯定是对的。
I mean, if Elon's doing it, he's like Right.
仅仅因为埃隆在做这件事，就意味着会有很多资本和长期的大量尝试。
Just the fact that Elon's doing it means that there's gonna be a lot of capital and a lot of attempts for a long period of time.
所以单凭这一点，也许就足以说明我们应当投资机器人领域，因为有一个明确的北极星——埃隆在做人形机器人，这将直接催生一个产业。
So that alone maybe suggests that we should just be investing in robotics just because you have this North Star who's Elon with a humanoid, and that's gonna, like, basically will into being an industry.
但我们历史上一直坚信，这件事一定会发生。
But we've just historically found like, we're a huge believer that this is gonna happen.
只是我们觉得自己并不具备足够的能力去深入调研这些项目，因为机器人公司通常都是垂直整合的。
We just don't feel like we're in a good position to diligence these things because, again, robotics companies tend to be vertical.
你必须真正理解它们所面向的市场。
You really have to understand the market they're being sold into.
真正重要的是，机器人与人类之间的竞争均衡关系。
Like, that's like, that competitive equilibrium with a human being is what's important.
而不是核心技术本身，而我们更偏向于做横向的核心技术型投资。
It's not like the core tech, and, like, we're kind of more horizontal core tech type investors.
这是莎拉
And this is Sarah
还有我。
and I.
是的。
Yeah.
AD团队，是的。
The AD team Yeah.
是的。
Yeah.
他们实际上可以做到
They can actually do
这类事情。
these types of things.
为了澄清一下，AD代表
Just to clarify, AD stands for
美国活力。
American dynamism.
好的。
Alright.
对。
Yeah.
对。
Yeah.
对。
Yeah.
是的。
Yeah.
是的。
Yeah.
我实际上确实有一个相关的问题，首先我想确认一下，关于芯片方面。
I actually I do have a related question that first of I wanna acknowledge also just on the on the chip side.
是的。
Yeah.
我记得你参加过一个播客，我想应该是ACCC的播客。
I I recall a podcast that where you were on I I I think it was the ACCC podcast.
大概两三年前，你突然说了一句话，让我印象深刻，说的是在某个规模下，制造定制ASIC是有意义的，是的。
About two or three years ago where you where you suddenly said something which really stuck in my head about how at some point at some point kind of scale, it makes sense to build a custom ASIC Yes.
用于每次运行。
For per run.
是的。
Yes.
这太疯狂了。
It's crazy.
是的。
Yeah.
我们正在广泛传播。
We're airing way.
五百亿，或者类似的数量。
500,000,000,000 or something.
不。
No.
不。
No.
不。
No.
一次十亿美元的训练运行。
A bill a billion dollar training run.
一次十亿美元的训练运行，如果你能及时完成，那么定制ASIC是有意义的。
A $1,000,000,000 training run, it makes sense to actually do a custom ASIC if you can do it in time.
现在的问题是时间表，是的。
The question now is timeline Yeah.
不是钱。
Not money.
因为只是粗略的计算。
Because just just just rough math.
如果训练成本是十亿美元，那么该模型的推理成本必须超过十亿美元。
If it's a billion dollar training run, then the inference for that model has to be over 1,000,000,000.
否则，它将无法盈利。
Otherwise, it won't be solvent.
所以，假设你能节省20%，而使用ASIC实际能节省更多。
So let's assume it's if you could save 20%, which you save much more than that with an ASIC.
20%就是两亿美元。
20%, that's $200,000,000.
你可以用两亿美元流片一块芯片。
You can tape out a chip for $200,000,000.
对吧？
Right?
所以现在，你真的可以从经济角度来证明它的合理性。
So now you can literally, like, justify economically.
而不是从时间预警的角度。
Not time warning wise.
这是另一个问题，每个芯片都要单独设计。
That's a different issue, an ASIC per Yeah.
模型。
Model.
仅仅因为每次我们使用通用的英伟达芯片时，都会白白浪费这么多钱。
Just because that that's how much we leave on the table every single time we we we do it, like, generic NVIDIA.
没错。
Exactly.
没错。
Exactly.
不。
No.
实际上还要多得多。
It's it's actually much more than that.
你很可能能实现两倍的效率，那就是五亿美元。
You could probably get, you know, a factor of two, which be $500,000,000.
典型的MFU大概是50。
Typical MFU would be, like, 50.
是的。
Yeah.
是的。
Yeah.
这很好。
And that's good.
没错。
Exactly.
是的。
Yeah.
100。
100.
所以，是的。
So so yeah.
我的意思是，我想承认一下，我们现在身处2025年，正在打开眼界，确认像博通以及其他公司的情况，没错。
I mean, and and I just wanna acknowledge, like, here we are in in 2025 and opening eyes confirming, like, Broadcom and all the other Like, right.
定制芯片交易，这太惊人了。
Custom silicon deals, which is incredible.
我觉得，谈到AD，其实有一个非常有趣的关联，显然你们也受到了影响，那就是所谓的‘美国优先’运动，或者说是这里的再工业化，没错。
I I think that, you know, speaking about AD, there's there's a really, like, interesting tie in that, obviously, you guys are hit on, which is, like, these are the sort of, like, America first movement or, like, sort of re industrialized here and Yeah.
如果可能的话，把台积电迁到这里。
Move TSMC here, if that's possible.
AD和这个有多少重叠？
How much overlap is there from AD?
是的。
Yeah.
至于增长和投资那些计算能力受限明显的美国AI公司，情况又如何呢？
To, I guess, growth and investing in particularly, like, you know, US AI companies that are strongly bounded by their compute?
是的。
Yeah.
是的。
Yeah.
所以我的意思是，我会
So I mean, I I would
我认为AD更应被视为一种市场细分，而不是一种使命。
view I would view AD as more as a market segmentation than, like, a mission.
对吧？
Right?
市场细分在于它涉及监管合规问题，或与政府的销售或硬件交易有关。
So the market segmentation is it has kind of regulatory compliance issues or government, you know, sale or deals with, like, hardware.
我的意思是，他们就是专门用来对这类公司进行尽职调查的。
I mean, they're just set up to to to to to diligence those types of companies.
所以这更像是一种市场细分。
So it's more of a market segmentation thing.
我会说，整个公司自成立以来，一直存在地域偏向。
I would say the entire firm, you know, which has been since it's been intercepted, you know, has geographical biases.
对吧？
Right?
我的意思是，很长一段时间以来，湾区都会是资金主要流向的地方。
I mean, for the longest time, like, know, Bay Area is gonna be, like Great.
大部分资金都会流向那里。
Where the majority of the dollars go.
是的。
Yeah.
而且听我说。
And and listen.
地理位置偏见实际上会带来很多累积效应。
There there's actually a lot of compounding effects for having a geographic bias.
对吧？
Right?
你知道的？
You know?
所有人都在同一个地方。
Everybody's in the same place.
你拥有一个生态系统。
You've got an ecosystem.
你就在那里。
You're there.
你有存在感。
You've got presence.
你有一个网络。
You've got a network.
而且，我觉得湾区真的已经回来了。
And, I mean, I would say the Bay Area is very much back.
你知道的？
You know?
比如，我记得在疫情前，加密货币几乎迅速拉走了大量初创公司。
Like, I I remember during pre COVID, like, it was, like, almost crypto had kind of pulled startups so quickly.
纽约之所以如此，是因为它离金融圈非常近。
New York was, you know, because it's so close to finance.
洛杉矶曾经因为靠近消费领域而兴起过一阵子，但现在这里又重新崛起了。
Came out the like, Los Angeles had a moment because it so close to consumer, but now it's kind of come back here.
所以我会说，我们历史上一直非常聚焦于湾区，尽管当然，我们投资遍布全球。
And so I would say, you know, we tend to be very Bay Area focused historically even though, of course, we invest all over the world.
再往外一圈，毫无疑问是美国，因为我们对这里非常了解。
And then I would say, like, if you take the ring out, you know, one more, it's gonna be The US, of course, because we know very well.
再往外一圈，就是美国及其盟友。
And then one ring more is gonna be kinda US and its allies.
嗯。
Mhmm.
是的。
Yeah.
然后就从那里继续延伸。
And it goes from there.
是的
Yeah.
抱歉。
Sorry.
不。
No.
不。
No.
我同意。
I agree.
我认为从内部来看，那大概是公司总部所在地。
I think from a but I think from the intern that that's sort of, like, where the companies are headquartered.
也许你的问题在于供应链和客户群。
Maybe your question's on supply chain and customer base.
我会说，从这个角度来看，我们的客户或公司相当国际化。
I I would say our customers are or our companies are fairly international from that perspective.
他们是在全球范围内销售。
Like, they're selling globally.
对吧？
Right?
在某些情况下，他们拥有全球供应链。
They have global supply chains in some cases.
我认为，风险投资和成长型投资之间的客户粘性也非常不同，是的。
I would say also the stickiness is very different Yeah.
历史上，两者之间差异很大。
Historically between venture and growth.
在风险投资阶段，需要做大量的公司建设工作。
Like, there's so much company building and venture.
非常多。
So much.
比如招聘下一个产品经理、引入客户，所有这些事情。
So, like, hiring the next PM, introducing the customer, like, all of that stuff.
当然，我们只会更专注于自己有网络优势的地方，我在这里做生意已经二十年了，呃，我在湾区已经待了二十五年。
Like, of course, we're just gonna be stronger where we have our network, and we've been doing business for twenty I mean, I've been in the Bay Area for twenty five years.
所以，显然，我在这里比在其他地方更有效率。
So, clearly, I'm just more effective here than I would be somewhere else.
但我认为，对于一些后期的融资轮次，这些公司其实不需要太多帮助。
But I think I think for some of the later stage rounds, the companies don't need that much help.
它们在历史上已经相当成熟了。
They're already kind of pretty mature historically.
所以，它们基本上可以遍布各地。
So, like, they can kinda be everywhere.
因此，这种黏性就弱多了。
So there's kind of less of that stickiness.
在人工智能时代，情况就不同了。
This is different in the AI time.
我的意思是，萨拉现在几乎是湾区一半人工智能公司的首席助理。
I mean, Sarah is now the chief of staff of, like, half the AI companies in the Bay Area right now.
她就像是运营忍者、业务拓展和业务运营的高手。
She's like ops ninja, biz dev, biz ops.
你有没有在工作中发现很多人工智能自动化应用？
Are are you are you finding much AI automation in your work?
你的技术栈是什么？
Like, what what is your stack?
哦，你是指我的个人技术栈吗？
Oh, my in my personal stack?
我的意思是，之所以有这个话题，是因为它确实很有触发性。
I mean, it's because, like by the way, it's the the the reason for this is is triggering.
是的。
Yeah.
我们正在招聘运营人员。
We are like, I'm hiring ops ops people.
我认识的很多合伙人也在招聘运营人员，而我知道你也在帮很多公司处理运营事务，所以这无疑是个机会。
A lot of partners I know are also hiring ops people, and I'm just you know, it's opportunity since you're you're also, like, basically helping out with ops with a lot of companies.
现在人们都在做些什么？
What are people doing these days?
因为据我所知，这仍然非常依赖人工。
Because it's still very manual as far as I can tell.
是的。
Yeah.
我觉得我们帮忙做的事情主要基于人脉，就是说，嘿。
Think the things that we help with are pretty network based in that it's sort of like, hey.
我该怎么简化这个流程？
How do I shortcut this process?
不如我帮你联系对的人。
Well, let's connect you to the right person.
所以目前还没有针对这个的AI工作流。
So there's not quite an AI workflow for that.
作为一位增长型投资者，我觉得Claude Cowork挺有意思的。
I will say as a growth investor, Claude Cowork is pretty interesting.
比如，第一次你可以真正一次性正确地完成数据分析，你知道的，如果你要处理客户数据库，分析用户留存率。
Like, for the first time, you can actually get one shot data analysis right, which, you know, if you're gonna do a customer database, analyze a cohort retention.
对吧？
Right?
这些以前都得手动做。
That's just stuff that you had to do by hand before.
我们团队有一次，半夜三点，我们三个人还在玩Claude Cowork。
And our team, the other it was, like, midnight, and the three of us were playing with Claude Cowork.
我们给了它一个原始文件。
We gave it a raw file.
搞定。
Boom.
完全准确。
Perfectly accurate.
我们核对了数据。
We checked the numbers.
这太棒了。
It was amazing.
那就是我的高光时刻。
That was my, like, moment.
听起来很无聊，但你知道吗？增长型投资者以前常常熬夜苦干的事情，现在几秒钟就能搞定。
That sounds so boring, but, you know, that's that's the kind of thing that a growth investor is, like, you know, slaving away on late at night, done in a few seconds.
是的。
Yeah.
你不得不想，整个Anthropic Labs，也就是他们新成立的产品工作室，值多少钱呢？
You gotta wonder what the whole, like, Anthropic Labs, which is, like, their new sort of products studio Yeah.
如果作为一个独立的初创公司，它值多少钱？
What would that be worth as an independent startup?
你知道的。
You know?
比如，
Like,
很多。
a lot.
是的。
Yeah.
没错。
True.
你得承认他们做得很好。
You you gotta hand it to them.
他们一直表现得极其出色。
They've been executing incredibly well.
是的。
Yeah.
我的意思是，像Anthropic这样基于Cloud Code进行开发，我觉得很合理。
I I I mean, to me, like, you know, Anthropic, like, building on Cloud Code, I think it makes sense to me.
真正全力以赴的时候，就是当他们开始直接与OpenAI争夺消费者市场时，那对OpenAI来说就是红色警报。
The the real pedal to the metal, whatever the the the phrase is, is when they start coming after consumer with against OpenAI, and, like, that is, like, red alert at OpenAI.
我认为他们一直很明确地专注于企业市场。
I think they've been pretty clear they're enterprise focused.
他们确实如此。
They have been.
但问题是，他们公开打破了这一界限。
But, like, here's They break their publicly.
他们是专注于企业的。
It's enterprise focused.
他们是做代码的。
It's coding.
对吧？
Right?
然后还有Cloud Cowork。
And then and but here's Cloud Cloud Cowork.
而且，据我所知，他们正在为CloudAI在Instagram上投放广告，面向像你这样的用户。
And and here's, like, well, we they're apparently they're running Instagram ads for CloudAI on you know, for for people I just
一直都有它们。
have them all the time.
对吧？
Right?
所以，就像这个颠覆性的事情，你知道，OpenAI 一直在做消费端，一直在追求各种模态下的通用智能。
And so, like, it it's kinda like this the disruption thing of, you know, OpenAI has been doing consumer, been doing the just pursuing general intelligence in every mod modality.
而这里是 Ethopa。
And here is Ethopa.
他们只专注于这一件事，但现在却在以一种颠覆性的方式，对其他所有领域进行打压。
They only focus on this thing, but now they're sort undercutting and doing the whole Innovator's Dilemma thing on, like, everything else.
是的。
Yeah.
这非常有趣。
It's very interesting.
对。
Yeah.
我的意思是，这还是一个非常开放的问题。
I mean, there's there's a very open question.
所以对我来说，你知道那个梗吗？就是有个人走在路上，一条路往这边，另一条路往那边。
So so for me, there's like do you know that meme where there's like the guy in the path and there's like a path this way, there's
一条路往这边，像这样，哪条路，西方人？
a path this way, like, Which way, Western man?
是的。
Yeah.
是的。
Yeah.
是的。
Yeah.
是的。
Yeah.
对我来说，整个行业实际上都取决于两种可能的未来。
And for me, like like, all the entire industry kind of, like, hinges on, like, two potential futures.
所以在一种可能的未来中，市场是无限庞大的。
So in in one potential future, the market is infinitely large.
存在一种反常的规模经济，因为一旦你发布了一个模型，它就会逐渐升华，其他所有模型都会跟上。
There's perverse economies of scale because as soon as you put a model out there, like, it kinda sublimates and all the other models catch up.
就像软件被不断重写和碎片化，到处都是机会，市场持续增长。
And, like, it's just like software's being rewritten and fractured all over the place, and there's tons of upside, and it just grows.
而另一种路径是，也许这些模型实际上已经具备很好的泛化能力，你只需要多花三倍的钱来训练它们。
And then there's another path, which is like, well, maybe these models actually generalize really well, and all you have to do is train them with three times more money.
你只需要做这一件事，它就会吞噬掉所有其他东西。
That's all you have to do, and it'll just consume everything beyond it.
如果真是这样，最终你会得到一个几乎垄断一切的寡头格局。
And if that's the case, like, you end up with basically an oligopoly for everything.
因为它们是完全通用的，所以这条路径就是AGI路径：这些模型是完全通用的。
Like, you know, because they're perfectly general and, like so this would be like the the AGI path would be like, these are perfectly general.
它们能做任何事，而另一种情况则更像是普通的软件。
They could do everything, and this one is like, this is actually normal software.
宇宙是复杂的。
The universe is complicated.
你有，但没人知道答案。
You've got and nobody knows the answer.
我的观点是，如果你仔细看看这些公司的数据，一般来说，如果你看看这些公司的收入和它们训练上一个模型所花费的金额，它们的毛利是正的。
My belief is if you actually look at the numbers of these companies so generally, if you look at the numbers of these companies, if you look at, like, the amount they're making and how much they they spent training the last model, they're gross margin positive.
你会想，哦，这确实有效。
You're like, oh, that's really working.
但如果你看看它们目前正在为下一个模型进行的训练，它们的毛利是负的。
But if you look at, like, the current training that they're doing for the next model, they're gross margin negative.
所以有一部分我认为，很多公司实际上是在向未来借贷，而这迟早会放缓。
So part of me thinks that a lot of them are kind of borrowing against the future, and that's gonna have to slow down.
总有一天这会反噬他们，但我们真的不知道。
That's gonna catch up to them at some point in time, but we don't really know.
是的。
Yeah.
这说得通吗？
Does that make sense?
比如，也许这一切之所以能运作，只是因为他们能融到下一轮资金，然后就能训练下一个模型，因为这些模型的生命周期太短了。
Like, and it could be it could be the case that the only reason this is working is because they can raise that next round, and then they can train that next model because these models have such a short life.
所以某一天，他们再也融不到下一个模型所需的资金，那时情况就会再次发生变化。
And so at some point in time, like, you know, they won't be able to raise that next round for the next model, and then things will kinda convert your fragment again.
但目前还不是这样。
But right now, it's not.
完全同意。
Totally.
另外，顺便说一句，有个宏观的观察。
I think the other by the way, just a meta point.
我认为过去三年的另一个教训是，我们经常谈论这个，因为我们身处推特x泡沫之中。
I think the other lesson from the last three years is and we talk about this all the time because we're on this Twitter x bubble.
但真的很棒。
But Very cool.
你知道吗，如果回溯到2024年3月左右，那时候感觉开源模型，尤其是像F这样的、在基准测试中领先的模型，几乎每天都在发布。
You know, if you go back to, let's say, March 2024, that period, it felt like a I think an open source model with an f like a, you know, benchmark leading capability was sort of launching on a daily basis at that point.
所以，那是一段时期。
And and so that you know, that's one period.
突然间，开源似乎席卷了整个世界。
Suddenly, it's sort of like open source takes over the world.
将会涌现出大量模型。
There's gonna be a plethora.
这并不是一个寡头垄断的局面。
It's not an oligopoly.
你知道，如果你再往前推，早在那之前，GPT-4曾连续九到十个月位居榜首。
You know, if you fast you know, if you if you were wine time even before that, GPT four was number one for nine months, ten months.
那可是很长一段时间了。
It's a long time.
对吧？
Right?
当然，现在我们正处于一个感觉像是寡头垄断的时代，可能还有一些非常稳定的结构性变化。
And, of course, now we're in this era where it feels like an oligopoly, maybe some very steady state shifts.
而且，你知道，未来也可能看起来像这样，但真的很难预测。
And and, you know, it could look like this in the future too, but it just it's so hard to call.
我认为让我们夜不能寐（既有好的一面也有坏的一面）的事情是，能力的进步实际上并没有放缓。
And I think the thing that keeps, you know, us up at night, in a good way and bad way, is that the capability progress is actually not slowing down.
所以在那之前，你根本不知道未来会是什么样子。
And so until that happens, right, like, you don't know what's gonna look like.
但我敢肯定，这远未达到收敛状态。
But I I would I would say for sure it's not converged.
毫无疑问，系统性的资本流动还没有收敛。
Like, for sure, like, the systemic capital flows have not converged.
意思是，目前我们仍在借未来的钱来补贴当前的增长，这可以持续一段时间，但最终市场会对此做出理性判断，只是没人知道那会是什么样子。
Meaning, right now, it's still borrowing against the future to subsidize growth currently, which you can do that for a period of time, but, you know, at the end at some point, the market will rationalize that, and just nobody knows what that will look like.
是的。
Yeah.
或者，计算成本的下降会拯救他们。
Or or, like, the drop in price of compute will will will save them.
谁知道呢？
Who knows?
是的。
Yeah.
是的。
Yeah.
我认为模型需要针对特定任务趋于饱和。
I think the models need to asymptote to specific tasks.
你知道的？
You know?
好吧。
It's like, okay.
现在OPUS 4.5可能在某个特定任务上达到了通用人工智能，这样你就可以更长时间地摊销模型成本。
Now OPUS 4.5 might be AGI at some specific task, and now you can, like, depreciate the model over a longer time.
我觉得现在根本没有旧模型。
I think now right now, there's, like, no old model.
没有。
No.
但让我换个思路。
But let but let me just change that mental.
那曾经是我的思维模式。
That's that used to be my mental model.
让我稍微调整一下。
Let me just change it a little bit.
如果你能筹集到三倍的资金，或者比所有使用你模型的人的总和还多，那根本无关紧要。
If you can raise three times if you can raise more than the aggregate of anybody that uses your models, that doesn't even matter.
那根本无关紧要。
It doesn't even matter.
你明白我的意思吗？
See what I'm saying?
就像
Like,
是的。
so Yeah.
所以我有一个API业务。
So I have an API business.
我的API业务的利润率是60%、70%或80%。
My API business is 60% margin or 70% margin or 80% margin.
这是一个高利润率的业务。
It's a high margin business.
所以我清楚每个人都在用什么。
So I know what everybody is using.
如果我能筹集到比所有使用它的人加起来还多的资金，无论我是否是AGI，我都会吞并他们。
If I can raise more money than the aggregate of everybody that's using it, I will consume them whether I'm AGI or not.
我会知道他们在使用，因为他们确实在使用。
And I will know that they're using it because they're using it.
而且，不像过去那样，工程上的限制会阻碍我这么做，嗯。
And, like, unlike in the past where engineering stops me from doing that Mhmm.
这非常直接明了。
This is very straightforward.
你只需要进行训练。
You just train.
所以我以前也觉得，你可能得问一些关于AGI的通用问题，但我也认为，资本市场有可能直接给他们提供弹药，让他们去全面压制所有竞争对手。
So I also thought it was kind of like, you must ask some code AGI general general general, but I think there's also just a possibility that the the capital markets will just give them the the the ammunition to just go after everybody on top of them.
我确实好奇，按你的说法，是否存在某种任务，即使略微提升也没带来多大改善。
I do wonder, to your point, if there's a certain task that getting marginally better isn't actually that much better.
比如，我们已经逼近极限了，不管称它为AGI还是别的什么。
Like, we've asymptote it to you we know, can call it AGI or whatever.
你知道，实际上，阿里·戈德赛谈过这个问题。
You know, actually, Ali Godsey talks about this.
比如，在企业环境中，我们其实已经达到了AGI的水平。
Like, we're already at AGI for a lot of functions in the enterprise.
不过对于这些任务，你很可能可以建立一些非常专业的公司，专注于从这些任务中榨取尽可能多的价值，而这些价值并非来自模型本身。
That's probably though for those tasks, you probably could build very specific companies that focus on just getting as much value out of that task that isn't coming from the model itself.
那里很可能存在一个丰厚的企业级商业机会。
There's probably a rich enterprise business to be built there.
我的看法可能不对，但确实有很多有趣的例子。
I mean, could be wrong on that, but there's a lot of interesting examples.
比如，如果你关注法律行业之类的领域，也许这并不是个好例子，因为模型在这一领域也在不断进步。
So for if you're looking about the legal profession or or whatnot, and maybe that's not a great one because the models are getting better on that front too.
但如果是那些已经趋于饱和的领域，价值就会来自服务。
But just something where it's a bit saturated, then the value comes from services.
价值来自实施过程。
It comes from implementation.
对吧？
Right?
价值来自所有这些真正让最终客户觉得有用的东西。
It comes from all these things that actually make it useful to the end customer.
抱歉。
Then sorry.
我再补充一点，我认为在所有这些讨论中，有一个被忽视的问题是：每个任务在多大程度上是通用人工智能完备的？
One more thing I think is is under discussed in all of this is, like, to what extent every task is AGI complete?
嗯。
Mhmm.
对吧？
Right?
我每天都写代码。
I code every day.
这太有趣了。
It's so fun.
这是核心问题。
That's core question.
是的。
Yeah.
而且，当我跟这些模型交流时，不只是写代码。
And, like, when I'm talking to these models, it's not just code.
我的意思是，什么都涉及。
I mean, it's everything.
对吧？
Right?
比如，你知道的，它就是
Like, I you know, like, it's
是医疗保健。
it's health care.
是法律领域。
It's legal.
但每个领域都是如此。
But it's every it's exactly that.
我的意思是，比如客服。
Like, I mean, support.
是的。
Yeah.
那就是一切。
That's everything.
我是让这些模型去理解合规性。
Like, I'm asking these models to, yeah, to understand compliance.
我让这些模型去网上搜索。
I'm asking these models to go search the web.
我让这些模型谈论我知道的历史内容。
I'm asking these models to talk about things I know in the history.
这就像在我进行工程设计时，它们能和我进行完整的对话。
Like, that's having a full conversation with me while I I engineer.
所以，也许最接近AGI完备的——我不是AGI专家。
And so it could be the case that, like, the most, you know, AGI complete like, I'm not an AGI guy.
我觉得，不管什么任务，最接近AGI完备的模型总会胜出。
Like, I think that's, you know but, like, the most AGI complete model will always win independent of the task.
我们也不知道这个问题的答案。
And we don't know the answer to that one either.
是的。
Yeah.
但在我看来，听好了。
But it seems to me that, like listen.
根据我的经验，Codecs 在编码方面肯定比 Opus 4.5 更好。
Codecs, in my experience, is for sure better than Opus 4.5 for coding.
它能找出我工作中最难缠的 bug。
Like, it finds the hardest bugs that I work in with.
你知道，我认识的最聪明的开发者都在做这个。
Like, it's it's you know, the smartest developers I know work on it.
它很棒。
It's great.
但我认为 Opus 4.5 其实非常棒，它特别有亲和力，当你在构建非常复杂的东西时，这一点真的很重要，因为它真的能成为你的伙伴和头脑风暴的搭档。
But I think OPUS four point five is actually very it's got a great bedside manner, And it really it it really matters if you're building something very complex because, like, it really you know, like, you're you're you're a partner and a brainstorming partner for somebody.
而且我觉得我们讨论得不够多，每个任务都某种程度上
And I I think we don't discuss enough how every task kind
都具有这种特质。
of has that quality.
嗯。
Mhmm.
这对我们所说的资本投入、前沿模型和子模型意味着什么？
And what does that mean to, like, capital investment and, like, frontier models and submodels?
是的。
Yeah.
那些专门的编码模型都去哪儿了？
Like, what happened to all the special coding models?
它们都没成功。
Like, none of them worked.
对吧？
Right?
所以它们中有一些
So did some of them
甚至没有发布。
Didn't even get released.
Magic dot dev
Magic dot dev
或者不是。
or No.
有一整套，有一大堆。
There was a whole there's a whole host.
我们看到了很多，而且，有一个完整的理论认为，可能，我认为其中一个结论是，根本不存在所谓的编程模型。
We saw a bunch of them, and, like, there's this whole theory that, like, there could be a and I think one of the conclusions is is, like, there's no such thing as a coding model.
你知道吗？
You know?
像这种东西根本不存在。
Like, that's not a thing.
你的意思是，你在和另一个人交流，它虽然擅长编程，但必须在各方面都出色。
Like, you're talking to another human being, and it's it's good at coding, but, like, it's gotta be good at everything.
我略有不同意见，因为我非常有信心，OpenAI 一定会发布 GPT-5 和 GPT-5 Codex。
Minor disagree only because I I'm pretty like, have pretty high confidence that, basically, OpenAI will always release a GPT five and a GPT five codex.
因为那是核心产品。
Like, the that that's the core one.
是的。
Yeah.
对。
Yeah.
对。
Yeah.
对。
Yeah.
我把它称为一个给 Riz，一个给 Tiz。
The way I call it is one for Riz and one for Tiz.
然后，有个在开放空间的人说，是的。
And and then, like, someone in turn on open air was like, yeah.
这是成名的好方法。
That's good way to fame.
这太有趣了。
That's so funny.
但也许它最终会简化为Riz和Tiz，就这么简单。
But maybe they maybe it collapses down to Riz and Tiz, and that's it.
它不是一百个维度。
It's not like a 100 dimensions.
它只有两个维度。
It's two dimensions.
是的。
Yeah.
对。
Yeah.
是的。
Yeah.
是的。
Yeah.
大部分互动与编码无关，但结果却占据了很大一部分。
Much of the interaction has nothing to do with coding, and it just turns out to be a large portion of it.
所以，我认为，无论是什么任务，最好的模型都会依然非常重要。
And so, like, you're I think, like, like, the best model, you know, is gonna remain very important no matter what the task is.
是的。
Yeah.
说到编码，我斗胆问一下，你到底在写什么代码？
Speaking of coding, I I'm gonna be cheeky and ask, like, what actually are you coding?
因为显然，你可以写任何代码，而你显然是一位忙碌的投资者，还管理着一个庞大的团队。
Because, obviously, you you could code anything, and you're obviously a busy investor and a manager of, like, a giant team.
你到底在写什么代码？
What are you coding?
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。