亚历山德·王，Scale公司创始人兼首席执行官

我是杰西卡·利文斯顿，和卡罗琳·莱维共同担任社交雷达的角色。在这档播客中，我们与硅谷一些最成功的创始人畅谈他们的创业历程。近二十年来，卡罗琳和我一直在Y Combinator携手帮助数千家初创企业。欢迎成为我们的

I'm Jessica Livingston, and Carolyn Levy and I are the social radars. In this podcast, we talk to some of the most successful founders in Silicon Valley about how they did it. Carolyn and I have been working together to help thousands of startups at Y Combinator for almost twenty years. Come be a fly on the wall as we talk to founders and learn their true stories. Carolyn, I'm so excited because today, we have Alexander Wang, the founder and CEO of Scale, which provides high quality data to AI companies.

YC在2016年夏季批次投资了他，如今公司估值已达138亿美元，客户涵盖各类机构，当然也包括所有大型模型公司如OpenAI和微软。欢迎你，亚历山大。

YC funded him in its summer twenty sixteen batch and now is valued at $13,800,000,000 and is used by, you know, many types of organizations, but certainly all the big model companies like OpenAI and Microsoft. Welcome, Alexander.

是的，非常感谢邀请，我特别兴奋。

Yeah. Thanks for having me. I'm super excited.

首先我想问你个问题。需要先向听众说明下，其实我对你了解不多，因为你加入YC那批时我正好搬去了英国。所以我想了解些你的成长经历。

I wanna first ask you. I gotta go back a little bit to your childhood. I need to preface this for our listeners. I actually don't know you very well because you started NYC at the same batch that I moved to England. So I I don't have a whole lot of history, so I'm gonna find out a little bit of your history.

好的。我在新墨西哥州洛斯阿拉莫斯长大，那里是原子弹的诞生地。现在别人问起时，我都会问对方有没有看过《奥本海默》，这通常就是我解释家乡的方式。

Yeah. So I grew up in Los Alamos, New Mexico, the birthplace of the atomic bomb. Yep. These days, I I I ask folks if if people have seen Oppenheimer, and that's usually how I'll I'll explain where I'm from.

我也是这么向卡罗琳介绍的，当时她说要告诉大家那里为什么出名。

That's how I described it to Carolyn when she said, let's tell them what's that famous for. Yep.

我父母都是物理学家，祖父母也是。可以说我出身于物理世家。最近和父母聊天时他们说，我成长过程中最让他们骄傲的是我学习物理的速度。洛斯阿拉莫斯的人均博士数量应该是全美最高的，这是个充满智慧的地方，聚集着众多专注科学的聪明人。

Both my parents are physicists, and and my dad's parents were also physicists. So I come from a long lineage of physicists. I recently talked to my parents, and they said that, you know, the thing that made them most proud when I was growing up was how quickly I picked up physics. Los Alamos, New Mexico has the highest number of PhDs per capita in The United States, I believe. So it's this very brainy place, lots of very smart people, very scientifically focused.

能在这么特别的地方长大我觉得很幸运。初中开始编程后，这几乎占据了我全部注意力。我小时候特别好胜，参加了许多数学、物理和计算机竞赛。白天在学校和朋友相处，晚上就和全美各地参加竞赛的网友交流。

And so it was I felt I feel very lucky to have grown up in such a special place. I started programming in middle school, and I I kind of, you know, that ended up sucking up all my attention. And I was I was very competitive as a kid, so I did a lot of math competition, physics competitions, computer science competitions. I was, you know, I would I would hang out with my friends at school during the day, and then I would go home and I would sort of chat with online friends all around The US who were sort of my competitive math and competitive computer science friends.

你在这些竞赛中表现很出色对吧？参加过很多次？

And you did very well in in these competitions, though. You did a bunch of them. Right?

参加过很多次。是的，我曾是全国排名靠前的数学竞赛选手。

I did a bunch of them. Yeah. I was I was a nationally ranked mathlete. And so

哇。它

Wow. It

那确实非常有趣。通过这些数学竞赛，许多科技公司愿意带我们参观他们的办公室。所以我高中时就参观了Dropbox的办公室，很早便接触到了科技行业的生态圈。于是我在高中和大学之间休学一年，去做了软件工程师。我在问答网站Quora工作过。

was it was it was a lot of fun. So then, through these math competitions, there were a bunch of tech companies that were willing to show us around their offices. So I toured Dropbox's office while I was still in high school and I I got sort of exposed to the tech industry ecosystem when I was really very early on. And and so I took a gap year between high school and college to work as a software engineer. So I worked at Quora, the question and answer site.

然后

And then

等一下，等等。你是高中毕业了对吧？

Wait a second. Hold on. Hold on. You were you graduated from high school. Is that right?

对。但你那时才17岁，是吗？

Yeah. Were you you were 17, though. Right?

没错，正是17岁。

Yep. Exactly. 17.

所以你休学了一年，你父母就同意你独自搬去硅谷了？

So you took this gap year, and your parents were okay with you moving to Silicon Valley all by yourself?

是的。我向他们解释这是间隔年，他们觉得‘好吧，反正你只是晚一年上大学’。

Yeah. I explained it as a gap year where they were like, oh, yeah. Okay. I'll just end up going to college, you know, a year later.

那时你已经拿到MIT的录取通知了吗？

Had been accepted at MIT at this point?

对，我被MIT录取后延迟了一年入学。

Yes. I'd been accepted at MIT and then enrollment for a year. Exactly.

不过你当初为什么独自搬到帕洛阿尔托？

What did you move to Palo Alto by yourself though?

是的。我最初搬到山景城，后来又搬到了旧金山。那时候我才17岁，独自面对山景城这个陌生世界。

I did. Yeah. I moved to Mountain View originally and then I moved up to San Francisco. But yeah, I was yeah, I was 17 in a in the in the wild world of Mountain View.

17岁是怎么生活的？我是说，那个年纪连租车都不行吧？

And how did you live as a 17 year old? I mean, can you even rent a car?

确实不能租车。幸好当时有优步，优步还能正常运营，还有就是经常坐加州火车。

No car rental. Yeah. No. Thankfully, there was there was Uber. Uber was a functional service and then a lot of Caltrain.

经常坐加州火车。哦，没错。

A lot of Caltrain. Oh, yeah. Okay.

但你最初是在乔·朗斯代尔创立的公司工作对吧？

But And you first worked for a company that Joe Lonsdale had started. Right?

对。我先在Atapar短期实习过，后来... 然后我基本上在Quora工作了整整一年，之后就去上大学了——在MIT读了一年。

Yes. So I interned briefly at Atapar, and then I Okay. Then I worked for the full year basically at Quora. And then and then I went off to college. So then I went to MIT for a year.

那一年对我来说很特别，因为我决定全身心投入AI和机器学习领域。那年发生了很多大事：DeepMind推出AlphaGo击败围棋世界冠军，谷歌发布TensorFlow框架，终于让普通开发者——甚至宿舍里的学生——都能开始接触神经网络、深度学习这些如今推动AI革命的核心技术。我完全沉浸在AI开发中，训练自己的神经网络，编写程序，很快意识到当时所有机器学习都依赖于数据。

That year was that was a very special year, I think, for me because I decided to just fully immerse myself into AI and machine learning. And this was a lot of big stuff happened this year. So this is when DeepMind came out with AlphaGo and beat the world champion at Go. This is when Google released TensorFlow, and it finally became possible for, you know, developers kind of in you know, developers in their dorm room to start playing with neural networks and deep learning and a lot of the original core technology that powers all of the AI revolution today. And and so I so I was sort of fully immersed into building with AI, training my own neural networks, doing my own programming, and I realized pretty quickly that that the that all of, you know, AI or machine learning at the time hinged on data.

数据才是智能的原材料——这个关键洞见最终促使我创立了Scale。

And data really was the raw material for intelligence, and that was kind of the key insight that led to ultimately led to starting scale.

稍微回溯下童年时光，虽然有点突然——能说说你名字的拼写由来吗？

Really quick to go back to your childhood. This is super random, but can you tell us about the spelling of your name?

是的。所以我的名字少了一个e，Alexander最后的e被省略了。这件事我问过我妈妈无数次了。好吧。

Yes. So my name is missing an e. The last e in Alexander is missing. That, you know, I I've always I've asked my mom a bajillion times about it. So, okay.

他们喜欢Alexander这个名字，但在中国文化里数字很重要，像 numerology（数字命理学），某些数字会带来好运。Alexander有八个字母，八在中国是非常吉利的数字。我名字之前的拼写方式正好是八个字母。

So they like the name Alexander, but then in in Chinese culture, numbers are very important, so numerology, like, there's a lot of of of sort of luck associated with certain numbers. And so Alexander has eight letters. Eight is a very lucky number in Chinese. The way it's spelled before my name has eight letters.

没错。

Right.

对。所以这是个重要因素。另外笔画数也很吉利。嗯。各方面都是个好数字。

Right. So that was a big factor. And then there was something about the number of strokes too being being a good number. And so Mhmm. It was it was all around a good a good number.

幸运的

A lucky

数字。可疑啊。好吧，既幸运又吉祥。

number. Suspicious. Yeah. Lucky and auspicious. Okay.

好的。我还有个快速问题：你在Quora时有没有想过'干脆放弃MIT吧，我太爱这份工作了'或者'我只想专注创造和建设，去他的上学'？

Okay. I do have one other quick question, which is when you were at Quora, were you ever like, oh, I'm gonna ditch this whole going to MIT thing because I love working or I like I just want to be creating or building to screw this whole school thing?

确实闪过这个念头。因为在Quora后期，我做出了一个产品原型，管理层非常看好。我当时考虑过要不要留在Quora继续开发。但很多高管都建议我去上大学——不是因为我表现不好，主要是...我记得Adam当时说的话起了关键作用，他说自己可能不会读完四年大学，但至少应该体验下大学生活，这很值得。

I the thought crossed my mind because at the end of my time at Quora, I, like, I had built out a prototype for a product and leadership was really excited about it. And so I was I was considering, oh, maybe I should stay at Quora and build this out. But then many of the executives at Quora told me, you should go to college. Not because I was doing a bad job, I don't think, but mostly because mostly because they said, you know, even Adam I remember Adam at the time told me this actually probably was was a big factor. He said, you know, I probably wouldn't go and do all four years of college, but I would do at least some college because it's really worth it.

所以

So

是的。为听众说明下，Adam DiAngelo是Quora联合创始人。你简直就是双料神童——智力超群又胆识过人。独自闯荡硅谷的举动非常勇敢，多数孩子做不到。

Yeah. Adam DiAngelo, for our listeners, one of the founders of Quora. So you were basically like a child prodigy, but in two different dimensions, in intellectually and in boldness. I mean, for you to leave and move all over yourself to Silicon Valley is pretty bold. A lot of kids wouldn't do that.

但这就是我困惑的原因。你之前说你在MIT时理解了数据的重要性。我想谈谈你的YC面试，因为你还记得吗？我当时在场，就在那场面试中。

But here's why I'm confused. You were saying you understood the importance of data at MIT. I'm gonna I wanna talk about your YC interview because do you remember? I was in the room. I was in that interview.

你们当时申请的是一个完全不同的想法，与规模扩展没太大关系。能详细说说吗？

You applied with a totally different idea that didn't really have that much to do with scale. Can you Yeah. Tell us about that?

完全理解。在大学时，我觉得和大多数申请YC的大学生一样，你会和联合创始人凑在一起，列出一个长长的谷歌文档，里面全是创业点子。嗯，很多时候这些点子五花八门——真的，有面向消费者的想法，

Totally. So in in college, I think like most people who apply to YC in college, you end up, you know, you get together with your your co founders and you just make a big old Google Doc of, like, startup ideas. And Mhmm. And a lot of times, they're all over the place. Like, Truly, there were, like, consumer ideas.

有B2B的想法，有纯API的想法，包罗万象。在最开始的头脑风暴阶段，Scale的雏形就在其中，所以我才说这个创意可以追溯到大学时期。不过...

There were B2B ideas. There were just API ideas. There was everything everything across the board. In that initial brainstorming, the idea for scale was in that initial brainstorming, which is why I say, hey, the idea dates back all the way to, like, you know, the very you know, really came from my experience in college. But then Okay.

我记得那时候大概是2016年，就在我们申请YC前夕，我们觉得'AI数据'这个概念可能不够宏大，所以决定换方向。最终我们提交了那个医生应用，因为想着'人人都需要医生'——这毕竟是全球最大行业之一。

But I remember at the time, this was like, I don't know, this was 2016. I remember the time, you know, thinking right before we applied to YC, it's like, okay, this data for AI thing, that's probably not a big enough idea, so we should figure out something else. And so that's why we ultimately applied with the doctor app because because we were like, oh, doctors? Everyone needs a doctor. That's, like, one of the biggest industries in the world.

所以就这样申请了YC。说实话我不完全记得您当时在场，但现在回忆那10-15分钟...虽然很...但挺有意思的。

And and so, yeah, that's when we applied NYC. I actually I didn't fully remember that you were in the room. So that's that's I but I'm as I'm, like, remembering that that ten, fifteen minutes, it's very it's very good. But

才十分钟而已。我...

It's only ten. Well, I

Alex我得向你坦白，我留着面试笔记。我当时投了赞成票——房间里三个面试官，我支持录取你们。但听听我的笔记怎么写的：

have to admit something to you, Alex. I have my interview notes. So I did vote a yes. There were three of us in the room, and I did vote yes to accept to accept you guys. But listen to what my notes said.

写着'可能傲慢或天才，倾向前者，但值得投资'。现在看真让人尴尬。

It said, maybe arrogant or brilliant. I think the former, but worth funding. I'm so embarrassed to this.

太狠了。真糟糕。

Savage. So bad.

天啊。

Oh god.

可能值得投资吧。我也不确定。

Just might be worth funding. I don't know.

我觉得我必须向你承认这一点。

I feel like I had to I had to admit that to you.

不，那个...所以我一直好奇，虽然从没问过YC合伙人，但我特别想知道他们在YC期间的笔记内容。因为我们在YC前半段一直在做这个医生预约应用，后来不得不面对现实。我记得特别清楚有一次和YC小组合伙人的办公时间，他们基本上是说：'这个医生项目可能对也可能不对，你们或许该花时间想想其他点子。'

No. I that's that's so I always wonder I haven't I've never asked the YC partners, but I'm so curious what the notes during YC look like as well, because we were working on this doctor app for the first half of YC, and then we had to come to Jesus. I remember one office hours in particular with our with our YC group partners. We were basically were like, they said, you know, this whole doctor thing may or may not be the right idea. So, you know, you should probably, like, take a moment to figure out what other other ideas you have.

当时我们反应是'哦糟糕'，然后我们重新规划，最终提出了Scale。但我真的很好奇，因为对YC合伙人来说，我们当时肯定显得完全迷失、非常混乱。对吧？

And and at the time, we were just like, oh gosh. And then we went back to the drawing board and and then ultimately came up with scale. But I'm really curious because I feel like for the YC partners, it must have just seemed, like, totally lost, very chaotic. Yes. You know?

等等，我刚找到一条笔记写着：'似乎对产品市场定位很迷茫'。你完全说中了，就是这样的。

Wait. I'm just I'm just gonna tell you. I just found a note, and it says seems lost on what their product market fit is. So you're you nailed it. That's the yeah.

精彩。不是傲慢。其实其他一些评论——我相信合伙人们不介意我说——有点像'这个团队波动性很高。团队看起来很棒，但点子不怎么样，不过值得投资'。你知道，年轻创始人经常遇到这种情况。

Brilliant. Not arrogant. Actually, it was some of the other comments, and I'm sure the partners won't mind me saying this, but it was kind of like high beta with this team. Team seems really good, but idea not so great, but worth funding. And, you know, that often happens with young founders.

你们当时是年轻创始团队，年轻人往往更难判断。带我回到你们意识到那个点子行不通的时刻——你们说过知道数据标注是个大问题，但是什么促使你们转向这个方向的？

You were a group of young founders, so it's often hard to tell, the younger people are. So take me back to when you sort of realized that the idea wasn't working and that data labeling, you you had said it you knew it was a big problem, but what made you change your mind to pivot to this?

有几个因素。首先是和YC小组合伙人谈话时，我们问：'做这个医生预约应用的话，演示日时大概需要多少预约量才能融资？'他们说：'可能每周需要一千个。'那一刻我们意识到：这几乎不可能完成。

Yeah. There were a few things happening. So the first thing that happened is we talked to our YC group partners and we asked them we're, you know, we're going on this on this doctor booking app, and we asked them, how many appointments do you think we'll need to to be able to raise at Demo Day? And they said, oh, you'll probably need a thousand a week. And I think we, like, realized at that moment, oh, that's gonna be nearly impossible.

我们看不到实现路径，因为让年轻人预约医生就像拔牙一样困难。我的意思是，你根本不想费力说服年轻人预约医生，这不是个有胜算的提案。所以...对，那就是我们面对现实的时刻。

We didn't really see, like, a clear path to that because it was it was like pulling teeth, getting young people to book doctor's appointments. I mean, it was you don't wanna be trying to convince young people to book doctor's appointments. That's not a that's not a winning proposition. So Yeah. So that was kind of the the sort of, like, moment of reality.

与此同时发生的另一件事是，那是在2016年，很多人可能不记得了，但那是第一波聊天机器人热潮的开端。现在回想起来很有趣，因为我们正处在这场持续多年的重大聊天机器人热潮中。但那时正是另一家YC公司Magic风头正劲的时候。Meta（当时还是Facebook）推出了他们的M机器人——Facebook M，那是他们正在研发的虚拟助手产品。还有一大批公司获得融资，致力于开发即时通讯和聊天机器人等相关领域。

And and then the other thing that was happening at the same time was this was 2016, and a lot of people won't remember this, but that was the first major chatbot craze. So which is so funny in retrospect now that we're, you know, in this, like, multiyear major chatbot craze. But that was right around when Magic, another IC company, was was really hot. Meta had create or Facebook at the time had their M bot, Facebook M, which was a their, like, virtual assistant product that they were working on. And there were a bunch of companies that were getting funded to work in messaging, chatbots, that whole area.

所以当我们思考这个问题时——这个市场是否足够大？感觉如果聊天机器人产业要发展壮大，它将极度依赖数据。有趣的是，现在回顾这段历程，看清整个发展轨迹后——当时我们以为聊天机器人时代会成为AI数据的关键驱动力。但最终情况是，这股聊天机器人热潮迅速消退，我们转而将重心放在了自动驾驶汽车上。

And so it was, you know, the when we when we thought about it, this question of, oh, is it gonna be a big enough market? It felt like it felt like, you know, the chap if the chatbot industry was going to be big, it was going to be super reliant on on data. The funny thing is, now that I look back on this and I, like, sort of, like, see the full arc, the you know, at the time, we thought it would be the chatbot era that would be the sort of the key for for data for AI. That was going to be the key driver. What ended up happening is that chatbot this, like, chatbot craze quickly died down, and then we ended up focusing on autonomous vehicles.

那确实成为了公司第一个重要的增长驱动力。但多年之后，聊天机器人又以另一种形式回归——如今的大语言模型和生成式AI（主要体现为聊天机器人）再次成为公司发展的重要篇章。想想这一切真是不可思议。

And that was really the sort of first major growth driver for the company. But then years and years later, chatbots ended up being, you know, right now, large language models and generative AI, which is mostly chatbots, has been Yeah. You know, a huge part of the company's story again. So it's kind of crazy to think about all in all.

你开始像个专注起名的人——我刚意识到公司最初叫Ava，后来在孵化期间改名为Scale。你们是怎么决定的？

You started to be like the person who just focuses on names, but I just realized you the company was called Ava at first and then you changed it to scale during the batch. How did you decide that?

最初公司是以聊天机器人命名的。科技行业的轮回很有趣——现在很多公司都采用拟人化名字，因为它们本质是聊天机器人，比如Claude。

It started out by being called a chatbot. It's funny because tech is so cyclical, like, now you have it's very common to have companies that are named that are sort of, like, personified names because they're chatbots, like Claude, for example.

嗯。

Mhmm.

对，这就是当时的想法。我们原计划做聊天机器人，但后来转向了AI数据领域。其实改名为Scale的过程没那么深思熟虑——记得有天晚上我在搜索域名，发现scaleapi.com这个域名可用，大家觉得听起来不错。

Yeah. So so that was the idea. At the idea, we thought we were gonna have a chatbot, but then but then we ended up focusing in on on data for AI. And the idea so then the move to scale, you know, it wasn't that thoughtful at the time, actually, because I remember one night, I was just sort of, like, searching for domains, and scaleapi.com was available as a domain. And so we're like, that sounds good.

就直接定下来了。后来AI行业的发展确实都围绕着'scale'（扩展）这个概念——不单指我们公司，而是指模型规模的扩大。所以这确实是个绝佳的名字。

Let's do it. That sounds great. And we just went with it. And then and then later on, the AI industry became all about scale, not us scale necessarily, but all about, like, growing the models. And so it was a it was a really good name.

多年后彼得·蒂尔甚至告诉我，他们投资的原因之一就是觉得这个名字起得特别好。

Actually, years later, Peter Thiel told me, he was like, one of the reasons we invested is, like, the name is just really good.

哇，真的吗？这可是很有力的认可。

Oh, nice. Really? That's some good validation.

真是意外。我敢打赌。

So surprised. I bet.

这这挺这这非常有趣，因为当时这并非一个深思熟虑的决定。就是就是，你知道的，运气好。当时只是在搜索域名。

It's it's pretty it's it's very funny because it was not a super thoughtful decision at the time. It was it was, you know Lucky. It was just searching for domains.

如果你回想一下，如果你们保留了Ava这个名字，Scale的发展轨迹会有什么不同？可能不会。

If you, like, think back, like, if you had kept the name Ava, how would the trajectory of your of Scale been different? Probably not.

我认为会有所不同，因为名为Scale的公司唤起的联想确实是基础设施，某种可以构建其上的东西。你知道，这非常吸引工程师、开发者和研究人员。所以我认为这非常重要。早期有些依赖我们的公司，其实本不该依赖我们这么小的公司。但我觉得这个名字在微妙地帮助吸引开发者加入方面起了作用。

I think it would have been different because because I think that the the thoughts that are evoked by a company called Scale is really, like, infrastructure, something you can build on top of. You know, it really appeals to engineers and developers and researchers. And and that was so I think that's that was really important. I mean, early on, there were companies that relied on us who really had no business relying on a company as small as as we were. But but I think the the name sort of, you know, helped in sort of a subtle way to get to get developers on board.

当时我们深受Stripe的启发，我觉得在公司名字上也有类似的感觉。

And we were very inspired by Stripe at the time, which I think has has sort of a similar sort of feeling in terms of the company name.

是的。我觉得你说得对。

Yeah. I think that's right.

然后和Cullison兄弟有类似的氛围。

And then the Similar sort of vibe with the Cullison brothers.

另外最终对我们有利的是，我们知道我们想从数据标注开始，为AI提供数据支持。但'scale'这个词的妙处在于它让我们未来能做更多事。作为早期创始人，你总有这些宏大愿景——先做数据，然后专注实施，接着做这个做那个，逐步覆盖所有领域。Scale就像一张空白画布，让我们能持续扩展以实现愿景。

And and the other thing that we ended up that ended up working out well was, you know, we we knew that, you know, we we wanted to start with data labeling and and and powering data for AI. But the the great thing about the word scale is that it it enables us over time to do so much more. And the vision always, you know, always as a as a as an early founder, you're you always have these these grand visions, which is, first we're gonna do data, then we're gonna, you know, focus on implementation, and then we're gonna do this, and we're gonna do testing, and, you know, you're gonna sort of over time do do everything. And scale has been a blank canvas for us to be able to to continue Yeah. Expanding to fulfill our vision.

最初你们是自己做数据标注吗？

Did you do the data labeling yourself at first?

是的。没错。我们做过。我记得我们的第一批客户都是YC的客户。

Yes. Yes. We did. Yeah. I remember so our first customers were all YC customers.

而且

And

他们是谁？你还记得吗？

Who were they? Do you remember?

哦，Teespring对吧？

Oh, Teespring. Right?

Teespring。Teespring是早期客户之一。我记得有天晚上不得不熬夜查看所有不同的T恤设计，并将它们精确分类到具体的设计类型中。顺便说一句，看到人们过去使用的各种T恤设计真是令人着迷。但他们确实是第一个真正开始使用我们服务的、具有相当规模的公司。

Teespring. Teespring was an early customer. And I remember I had to stay up one night looking at all the different t shirt designs and categorizing it into exactly what kind of design these t shirts were. It was fascinating, by the way, to see all the different t shirt designs that that that people used to be for. But they were they were one of they were probably the first, you know, company of any meaningful size that that really started using us.

那是在YC批次之前，他们是我们能在那轮融资中成功的重要助力。不过当时都是YC系公司。是的，我们最初是自己完成所有工作，然后逐步建立起运营体系。

And that was before the batch and and a huge part of us being able to raise during the batch. But but it was all YC companies and and, yeah, I we started out just by doing the work ourselves and then and then slowly built out our operation.

那时候你们主要是在标注图像吗？

And were you just labeling images at that point?

主要是图像。对，主要是图像。当时有图像和一些文本标注，不过文本标注那时规模还不大。YC结束几个月后，我们非常明确需要聚焦自动驾驶领域。

It was mostly images. Yeah. It was mostly images. It was there was images and some text, although text labeling wasn't as as big at that point. And then and then a few months after YC, it became very clear that we needed to focus on autonomous vehicles and self driving.

后来我们围绕自动驾驶汽车处理的复杂传感器数据，开发了完整的产品套件。包括激光雷达数据、雷达数据、GPS等等。所以业务复杂度很快就变得非常高。

And then we ended up building out this this entire product suite around all the complicated sensor data that self driving cars deal with. So LIDAR data and radar data and GPS and all of that stuff. And so it became it became very complicated very quickly.

你们在演示日的表现如何？

How did you guys do at Demo Day?

我们...挺不错的。有位YC系投资人——Excel的前创始人Dan Levine在演示日前就联系我们了。当时我们很紧张，因为知道数据指标只是中等水平。他提出在演示日前就投资，我们和团队合伙人商量后很纠结：该不该等演示日呢？

We it was it was cool. We ended up there was there was one y investor who actually was a former YC founder, Dan Levine from Excel, who reached out to us before Demo Day, and we were kind of nervous because we knew our metrics were were just okay. And so he offered to fund the company before Demo Day. And I remember we had a conversation with our group partners, and we were like, what do we do? Should we wait for Demo Day?

我们应该这样，他们说，这是个绝佳的机会。就这么做，然后，你知道的，减轻你们的负担。

Should we like, and they said, this is a great offer. Just do this and and and, you know, take a load off your backs.

不过你们还是做了展示，对吧？

You guys still presented, though. Right?

我们确实展示了。而且，是的。我们依然进行了展示，并专注于帮助客户摆脱困境，我们也确实做到了。我们成功争取到了不少客户。

We still presented. And and, yeah. We still presented and we focus on getting customers out of that, which we did. We did manage to get a good handful.

这会暴露我对AI有多无知。请多包涵。但当时，可以说这个想法——数据准备看似枯燥，现在却很激动人心吗？那时人们意识到它有多重要，以及它将成为一项庞大的独立业务吗？

So this is gonna reveal what a dummy I am about AI. So just forgive me. But, like, at the time, would would it be fair to say that this idea, you know, preparing the data is like kind of unsexy, but exciting now? Did people back then realize just how exciting it was and, like, that it's this, like, huge standalone business?

它一直被认为非常乏味。记得最初几年，投资者总觉得这是个，嗯，毫无吸引力的领域。我认为真正需要的是对AI的坚定信念。如果你相信人工智能、神经网络，相信技术会突飞猛进，那么作为基础设施提供商和奠基者的Scale，自然会成为极其令人兴奋的公司。

It was it was always perceived as very unsexy. I remember for the first many years, like, investors always just felt like it was it was such a, you know yeah. It was an unsexy, uninteresting space. And I think it what it really required, you know, I think to see what would have ultimately happened, you needed to be a big believer in AI. And I think if you believed in artificial intelligence and, you know, and neural networks and you believed in the technology just getting dramatically better, then then scale as an infrastructure provider and as, you know, as a company laying down the foundations would naturally become an incredibly exciting exciting company.

我们在早期投资者身上注意到，那些相信AI技术、认为它极其重要的人，投资决策对他们来说更自然——当然不是容易（早期投资总是棘手），但比怀疑者要顺畅得多。

And so this is what we really noticed in a lot of our early investors is there were there were people who believed in AI as a technology and believed that that was going to that it was going to be extremely extremely important. And for them, it was quite natural and easy for them to, to invest. I shouldn't say easy. We saw it to you know, investment is always tricky in the early stages. But, it was easier versus those who, were more skeptical.

AI怀疑论者从未投资过Scale。

You know, AI skeptics never invest in scale.

我从未问过这个问题，但现在很好奇：你还记得收到过最坚决的拒绝吗？就是对方完全看走眼的那种？

Do you I've never asked this question, but now I'm curious. Do you remember sort of the biggest, like, no you got and, like, just thinking you are not seeing this correctly?

记得很清楚。有位投资者在一次大型合伙人会议上——我们正向全体合伙人做演示——

Yes. I remember quite vividly. There was one investor who it was in one of these, like, big group partner meetings. We were presenting to a full partnership. And Oh.

某张早期幻灯片写着'AI将依赖日益增长的数据量'之类的内容。一位投资者当场纠正我说：'实际上专家认为可能不需要那么多数据。'我回答：'从整个行业来看，显然大家都极度渴求数据。'他们在这个观点上非常固执。

And in in one of our early slides, we said it it said something along the lines of, you know, AI will depend on growing amount of data or something along those lines. And there was a there was an investor who corrected me at that on that slide and said, well, actually, you know, we've talked to some experts and they think actually maybe you don't need that much data. And I said, I don't know what to say. I think that you can tell from the entire industry that everybody's extremely data hungry. And and that was I think that, you know, they they were very they dug their heels in on that on that point.

多年后，那位投资者向我道歉了。你知道，现在显然我们身处生成式AI的世界，所有人都极度渴望数据，各大实验室都在竭力获取尽可能多的数据。

And and years later, that investor apologized to me. You know, now that obviously we're in in generative AI world, it's clear that everybody is immensely data hungry and all the labs are trying to get as much data as possible.

那是哪一年的事？

What year was that?

应该是2018年。

This must have been 2018.

好吧。能收到道歉挺好的，这种事挺少见的。

Okay. Yeah. Nice to get an apology. Seems rare.

是啊，能收到道歉确实不错。

Yeah. Nice nice to get an apology. Yeah.

这确实罕见。我很高兴。嗯，他们大概是想参与你的下一轮融资吧。没错。抱歉。

That is rare. I'm I'm glad. Well, they probably wanna be, like, in your next round. Exactly. Sorry.

卡罗琳和我都是这么愤世嫉俗。对吧。你之前在为Cruise工作，但让我们回到2022年大语言模型爆发的时刻。你预见到会发生这种事吗？我是说，连Sam Altman可能都没料到它会变得如此重要。

Caroline and I are such cynics. Right. So you'd been doing stuff for Cruise, but then take me back to the time in 2022 when LLMs became huge. Did you know that that would happen? I mean, I don't know if even Sam Altman saw just how big it would become.

带我回顾下当时的感受和你内心的想法。

Take me back to what that felt like and what was going on in your mind.

好的。公司在2022年前主要有两条发展脉络，而2022年是最新重要阶段的起点。第一阶段是我们与所有自动驾驶公司合作，比如Cruise、Waymo，以及丰田、通用等大型汽车制造商，推动自动驾驶浪潮。看到Waymo车辆上路，大家如此兴奋，对我们来说真的很欣慰，因为自动驾驶行业经历了太多起伏，现在技术终于成熟，整个行业强势回归，这非常令人满足。

Yeah. So so there's sort of like there were there were two major arcs of the company before 2022, and 2022 was the start of, I think, the the most recent major arc. The first one is that we were working with all the self driving companies, you know, Cruise, Waymo, a lot of large automakers like Toyota, General Motors, etcetera, to fuel the autonomous vehicle wave. It's been really gratifying for us actually seeing, the Waymo vehicles on the road, and how excited everyone is about them because autonomous vehicles have gone through so many ups and downs as an industry, and it's really I think it's, like, very gratifying to see the industry sort of come back strong now, now that the technology works really well. Yeah.

然后在2020年，我们开始将大量精力投入与美国政府及国防部合作，研究AI在国家安全领域的应用。当时这其实相当不受欢迎。如今国防科技本身已成为巨大浪潮，我甚至看到YC投资国防科技公司，这很棒。但当时很明显，美国政府需要更好的AI合作伙伴，需要更多AI领域的盟友。

And then in 2020, we got started. We focused a lot of our efforts working with the US government and working with the Department of Defense and working on using AI for national security. And this was it was actually relatively unpopular at the time. Now defense tech is obviously this huge a huge wave in and of itself, and I'm I've been excited to see even YC invest into defense tech companies, which is awesome. But but at the time, it was it was very clear that, you know, the US government needed better partners on AI, needed more partners on AI.

大约在那个时候，谷歌已经公开声明不再参与各种国防应用项目。所以从2020年开始，这可以说是公司第二个重要发展阶段，真正聚焦于政府和国家安全领域的人工智能问题。嗯。在那第一年，也就是2020年，我们签下了第一份重大合同——一份价值9000万美元的国防部合约，随后几年持续拓展业务。所以...

This was around the time when Google was had sort of made public statements about not working on various defense applications. And so this was, you know, starting in 2020, this was a sort of the second major arc of the company was really focusing in on on the government and national security AI problems. Mhmm. In that first year, 2020, we closed our first major contract, a $90,000,000 contract with the DOD, and then continued building out the business over the next few years. And so Can

我能打断一下吗？你当时多大？和国防部签合同时，你才24岁左右吧？

I interrupt you? How old are you at this time? And you're closing a deal with the DOD. Are you like 24 at this point?

23岁。2020年时我23岁。是的。

23. 23. Twenty twenty, I was 23. Yeah.

当你走进会议室时他们肯定惊呆了，心想：哇，这完全是个孩子啊。

It must just blow their mind when you walk into the conference room and they're like, wow. This kid is this is a kid.

确实很疯狂。我去五角大楼时会穿非常考究的西装。但我们有真正重量级的支持者和倡导者——这显然是人生一切事情的关键。我们与他们合作完成了出色工作，对与国防部的所有合作成果都深感自豪。

That's insane. I wore a very dapper suit, I would say, when I was going to the Pentagon. But we had really big believers and big champions, which is obviously key to everything in life. And so we were and we've done great work we've done great work with them. We're very proud of all the work we've done with the DOD.

你们在乌克兰不是做过重要项目吗？

Haven't you done some important work in Ukraine?

没错。我们协助开发了能识别乌克兰境内损毁情况的图像识别模型。通过持续卫星图像分析，可以探测任何存在破坏或军事活动的区域。这对国防部协调行动至关重要，同时也帮助人道主义组织和救援力量精准调配资源。

Yeah. Exactly. So we we built we helped build models, image recognition models that could detect damage in Ukraine. And so on consistent satellite imagery, you can detect any place where there's, you know, damage or or any kind of, you know, basically military activity. And that's been that's been really critical for the DOD, a, to coordinate their actions, but also for humanitarian organizations and and and aid efforts to know where to divert resources.

这个系统是与国防部共同开发的，但我们也向多家非营利组织和人道主义机构提供了技术支持。

And so we built that with the DOD, but also, you know, provided it to, you know, various nonprofits and humanitarian organizations as well.

太棒了。

That's awesome.

到了2022年，我们一直在推进其他重要AI领域的工作。当ChatGPT问世时，虽然我们越来越关注生成式AI，但当时前景尚不明朗。随后的六个月内——即2023年上半年——我们将大量内部资源转向支持生成式AI，将业务重心调整为构建整个技术浪潮的数据基础。粗略估算，2022年我们约有10-20人从事生成式AI数据工作，而当时公司总人数约700人。

But yeah. And so then so then come 2022, you know, we we had been working on all these other major AI waves, so to speak. And and at the when ChatGPT came out, I mean, I think we we were focusing more and more on generative AI, but it really wasn't clear where it was all going to go. And over the next six months, so in the 2023, the first six months of ChatGPT, basically, we ended up shifting a huge amount of our internal resources to focus on supporting generative AI and and focusing our business on being the data foundation for this entire this entire technology wave. And so the numbers, roughly speaking, I think we were at the 2022, there was sort of maybe something like 10 to 20 people working on on generative AI data, and this was out of a whole company of about 700 people.

在接下来的大约六到九个月里，我们最终将超过一半的员工重新聚焦于生成式AI的数据工作。这最终成为公司的首要任务。我认为我们作为企业的许多成功，正是因为我们行动迅速。最近在YC重聚活动上与Jared Friedman讨论时，我提到过个人作为创始人的哲学之一就是：做得过多。这次情况感觉像是一场地震般的转变，这个跨越如此之大，以至于我们很快意识到——

And over the course of the next maybe six to nine months, we ended up refocusing more than half of all of our headcount onto onto data for generative AI. So it ended up becoming the top focus for the company. And and I think a lot of our success as a business is because of how quickly we moved. You know, one of the things that one of the I talked about this with Jared Friedman recently at the the YC reunion, but one of my personal philosophies is is as the founder is, like, do too much. And in this case, you know, it felt like this was such a seismic shift, and it was it was so it was such a big jump that we very quickly realized, hey.

如果我们没有反应过度，那就是反应不足。最终，我们几乎将公司的全部重心转向了生成式AI数据。

We need to if we're not overreacting, we're underreacting. And, and we just we ended up, shifting almost all of our focus on the company towards generative AI data.

你有个博客，在《做得过多》里写过这个。有句话我特别喜欢，几乎想裱起来挂在YC。你说：‘我从未见过普通的努力带来非凡的成果。’这句话简直太棒了。

You have, like, a blog, and this was you wrote about this in Do Too Much. There's a quote that I loved. I almost wanna, like, frame it and put it up at YC. You said, I've never seen ordinary effort lead to extraordinary results. I just I just love that.

这真是

That's a

至理名言。

good one. To me.

太精彩了。

That's great.

但你经常写作，是为了理清思路吗？

But you do a lot of writing. Do you write to help you figure things out?

没错。作为领导者，我一直在探索如何最好地带领这个近千人的组织。写作对我来说是最有效的方式之一，能帮助我把想法落到纸上，再传达给整个团队。

Yeah. Totally. I mean, I think as a as a leader, I one of the things that I've, you know, struggled with or worked on as a leader is, like, what's the best way for me to lead this organization? Like, we have, you know, we have almost a thousand people now. It's like a lot of people.

最初写作是种领导工具，用于向全公司传递背景信息。后来我开始对投资者这样做，他们反馈说：‘这太棒了’。

And I found that writing was was one of the best ways for me to do that, to actually sort of get my thoughts on paper and then and then help explain that to the whole team. And so writing, it started out as a, I think, really a leadership tool, as a way for me to help give context to the entire company. And then I started doing it more with investors. And then my investor said, wow. This is great.

其实是投资者和一位新入职员工共同建议的：‘你应该公开发表这些文章，让外界了解我们的文化和价值观。’于是我们开始在更广的网络平台上发布。

You should you should it was actually a combination of investors, and there was someone who joined the company who was like, you should publish these so that we can help, you know, help everybody understand what our culture is and what we stand for. And so we started publishing them on the on the broader Internet.

这是个很好的过渡，Alex。MEI。我必须问问这个，因为我太着迷了。Carolyn，你还记得这个吗？

This is a good segue, Alex. MEI. I gotta ask about this because I'm so fascinated. Do you remember this, Carolyn?

我记得。是的。是的。是的。

I do. Yeah. Yeah. Yeah.

是的。所以对听众来说，你发布了关于MEI的内容，它代表优点、卓越和智慧。我要引用一小段。有一种错误的观念认为精英制度在某种程度上与多样性相冲突。我强烈反对这种观点。

Yeah. So for listeners, you posted about MEI, which stands for merit, excellence, and intelligence. I'm gonna do a little quote. There's a mistaken belief that meritocracy somehow conflicts with diversity. I strongly disagree.

没有一个群体能垄断卓越。我非常喜欢这一点。哦，这真是个大胆的举动，Alex。告诉我们你为什么这么做，人们的反应是什么？

No group has a monopoly on excellence. And I loved that. Oh, that's This was a bold move, Alex. Like, tell us about why you did that, and what did people what was the response?

是的。当时有几件事在发生。我认为第一件事是，我们当时非常专注于公司的快速发展，既要引进新人才，也要继续奖励现有的人才。我开始注意到，对于一个我认为我们总体上尽量保持非政治化的公司来说，即使如此，仍有一些小迹象表明，可能我们在多样性、招聘什么样的人以及如何在边缘情况下做决定方面的政策不够清晰。在这种不清晰的情况下，我认为公司里的很多人都感到沮丧。

Yeah. So so there were a few things happening at the time. I think the I think the first one was really, you know, we were focused on focusing on growing the company a lot and both bringing in new talent as well as continuing to reward our existing talent. And I started to notice, you know, for a company that that I think we, in general, try to stay very apolitical, even then, there were still, you know, small things that indicated that, you know, maybe we the policies weren't very clear for how we think about diversity, how we think about the kinds of people we hire, how we think about making these decisions on the margins. And in that lack of clarity, I think a lot of people in the company, you know, kind of felt frustrated.

经常会有这样的对话，当我与招聘经理或招聘人员交谈时，他们会因为不太清楚公司关注的重点而感到沮丧。对我来说，这感觉有点疯狂，因为我认为，就像任何公司一样，我们只需要专注于招聘最优秀的人。所以MEI实际上是我和我们对于没有觉醒主义的多样性的看法。我们关心拥有多样化的员工队伍，但我们也不想混淆这一点，我们总是想为每个职位招聘最优秀的人，我们想招聘卓越的人，并确保我们专注于招聘卓越的人。如果我们这样做，多样性自然会随之而来，因为事实证明，各种各样的人在某些方面都非常出色。

And there were there were oftentimes conversations when I would talk to people who are sort of hiring managers or recruiters or whatnot, where there would be there would be some frustration around not really knowing what the company was was was focused on. To me, this felt kind of crazy because I think, you know, just like any company, we just need to be focused on hiring the best possible people. And so MEI really was was my take on, and our take on diversity without the wokeness. So how do we, you know, we care about we care about having a diverse workforce, but we also don't want to confuse that with, you know, we always want to hire the best person for every possible job, and we want to hire excellent people, and we want to make sure we're we're focused on hiring excellent people. If we do that, then diversity will come in tow because it turns out, like, there's all sorts of people who are just incredible at things.

YC就是一个很好的例子。YC的创始人群体非常多样化。我们认为，如果我们专注于精英制度，真正招聘那些聪明、卓越、智慧的最优秀的人，那么多样性就会自然而然地实现。这就是我们的重新定义，即不是把多样性作为最高目标，而是把精英制度作为最高目标，然后承认如果你正确地做到这一点，多样性就会作为其结果而出现。这是基本理念。

You know, YC, I think, a testament to this. You know, the group of YC founders is incredibly diverse. And and we thought that if we just focus on hiring if we focus on meritocracy, really, hiring the best possible people who are who are brilliant, excellent, intelligent, then diversity will fall out as an outcome of that. And so that was kind of the the reframe, which is instead of being an organization that is focused on diversity as the the top level goal, being focused on meritocracy as the top level goal, and then and then acknowledging that if you do that correctly, then diversity will come as a as a sort of result of that. That was the basic idea.

我认为我们发布这个并希望专注于它的原因是，我们也认为科技界有很多人希望加入在这个问题上有明确立场的公司。无论是工程师、产品人员还是科技生态系统中的其他人，我认为在这些问题上，清晰度实际上对建立正确的团队有很大的好处。我们在实践中发现了这一点。在我们发布之后，当然，网上有很多人说这很疯狂。

I think that the the reason that we published it and the reason that we wanted to be focused on it was that we also thought that there were a lot of there were a lot of people in the tech world who wanted to join companies with a clear stance on this issue. I think that Mhmm. As whether it was engineers or or product people or whomever in the in the tech ecosystem, I I think that in a lot of these issues, clarity actually is a huge benefit to to building the right team. And we found this in practice. After we put it out there, you know, sure, there were a lot of people on the Internet in general who said, this is crazy.

这很糟糕。这不好。但也有很多非常优秀的人说，这正是我想加入的公司。这表明这家公司拥有我会欣赏的文化，我认为我也会欣赏。这实际上是我加入Scale的一个重要吸引点。

This is awful. This is bad. But there were a lot of very incredible people who said, this is the kind of company I wanna join. This demonstrates that this company has the the culture that I would that would appreciate me, that I think that I would appreciate. And and this is actually a huge attractor for me to join Scale.

所以这真的是我们的目标。最终它变得非常流行，就像有些人说的那样。它真的传播开了。我认为在某些地方，它被过度政治化，或者在不同的论坛上被误解了。

And so that was really the the goal. It ended up becoming I mean, frankly, it, like, got really, you know, quite quite, you know, quite viral, as some people call it. Like, it really spread around. And and I think in certain places, got over politicized or sort of it got misconstrued in various forums on both

也许在推特上或者对。在那个

Perhaps in Twitter or Yeah. In the

如果当初我知道现在所知道的，我可能会以不同的方式思考这个问题。但这个核心想法确实被以各种有些疯狂的方式曲解了。不过我认为核心理念极其正确——我们至今仍秉持这一点——即专注于雇佣最优秀的人才，这是我们的北极星。我们的北极星是任人唯贤、追求卓越与智慧。只要做好这一点，自然会收获许多美好的成果。

If I knew then what I know now, like, I would have I would have been able to, you know, we probably would have, you know, we would have thought about it a bit differently, but but it did get sort of picked up and misconstrued in in a variety of of somewhat crazy ways. But but the core idea, I think, is is incredibly sound, which is and we still live by it, which is we wanna focus on hiring the best possible talent, and that's our North Star. Our North Star is meritocracy, excellence, intelligence. And if we do that right, we have a lot of great things that come out of it.

亚历克斯，我记得2014年你刚来城里。但我特别记得——卡罗琳你应该也记得——GitHub撤掉了他们那块写着'GitHub精英合众国'的椭圆形办公室风格地毯。还记得吗？天哪，我完全不记得了。

Well, I remember you were just coming into town in 2014, Alex. But I remember specifically, and Carolyn, you probably do too, when GitHub removed their rug that was like the Oval Office rug that said, like, the United Meritocracy of GitHub. Do you remember? I don't remember that. Oh my gosh.

不。这事很重要，因为从此不能再把初创公司称为精英体制了。我记得当时在想：天啊，作为这个行业的女性，我确实因此遭遇过困境，但总体而言，初创企业就是精英体制。用户才不管创始人是谁，他们只会为好产品买单。

No. Big deal because you were not no longer allowed to refer to startups, you know, as a meritocracy. I remember thinking, gosh, you know, I'm a woman in this business. And, yeah, I've faced some hardships as a result of that, but overall, you know, it's very like, startups are a meritocracy. Your users are gonna pay you whatever.

如果你做出人们想要的东西，自然会成功。所以我清楚记得那个时期。亚历克斯，看到你的MEI帖子时我很欣慰，觉得这个观点很明智。

They don't care who the founder is. They're gonna pay you for a product that's good. And if you make something people want, it should do well. So I I definitely remember that time. So I was really pleased, Alex, to see your MEI post, and I thought, it seems sensible to me.

我在想，你从19岁起就担任这家公司的CEO了吧？

You know, I was just thinking, have you you've been CEO of this company since you were 19 years old?

是的。

Yeah.

是这样吗？对。而且你提到公司有上千名员工。这个过程是缓慢稳定的上升（这个词合适吗）？还是像你说的那些AI浪潮一样突然扩招？

Is that right? Yeah. So that and and so and I was also you know, you have you said you mentioned over a thousand employees. And has that been one of those things where it's been a slow, steady up ramp? That's a word.

更宏观的问题是：这对年轻CEO来说压力很大。你肯定经历了很多吧？

Or is it more like, you know, you're mentioning these waves of different kinds of, you know, AI things. Did you, like, do tons of hiring when those happened? Like, how and and and I guess sort of my larger question is, like, that's a lot for a young CEO. Like, you must you've probably been through a lot.

确实。大部分时间里我们发展得很平稳。早期发展平稳主要是因为当时确实很难招到人。

Yes. Yes. For sure. For the most part, we grew quite steadily. Initially, we grew steadily because it was actually just hard to recruit people.

说服人们加入初创公司确实很难。

It was hard to convince it's hard to convince people to join a startup.

是啊。

Yeah.

当时我们其实很想加快招聘速度，但你知道，要招到优秀人才真的很难。后来和许多公司一样，在2020到2021年疫情期间，我们大规模扩招，团队规模可能从150人激增到了700人。

And then, we would have loved to have hired more quickly, but, you know, it's just it was it was hard to to get great people. And then, like many companies, during the pandemic, from 2020 to '20 you know, 2020, 2021, we hired a ton. And we maybe went from, like, a 150 to 700 people in that time frame.

哇哦，这增幅可真不小。

Oh, wow. That is a that's a big jump.

确实巨大。这个阶段让我深刻认识到：当人员规模如此快速增长时，很难维持企业文化，也很难保持那些让公司良性运转的无形要素。这实际上深刻影响了我们近年的发展策略——从2022年底至今，我们可能只从700人增长到约1000人，人力增幅非常有限。

Big jump. Big jump. And and my I learned a lesson at that point, which is it's very, hard to maintain culture and and maintain all of the sort of unsaid things that make a company work if you grow headcount that quickly. And that's actually really informed, you know, the past few years for us at scale, which is, you know, from late twenty two to now, we've maybe grown from, you know, we were 700 people, now we're about a thousand. Like, we haven't grown headcount very much.

但同期业务规模却翻了四倍多。我认为这是最重要的经验之一。Brian Chesky可以说是这个核心理念的鼻祖——比起盲目扩招，更重要的是以适当节奏稳步发展，引进顶尖人才并建立完善的入职流程，维系公司成功的组织架构。

But the business has has, you know, more than quadrupled in that time frame. I think the that's been one of the key thing key lessons. And I I think Brian Chesky, I think, you know, is kind of the OG, I think, this core insight. But it's more important that you, like, slowly grow the company at the right rate and bring on people who are amazing and have the right processes to onboard them, and and you sort of, like, maintain all of the structure within the company that makes it successful than to just flood the system with, you know, tons of people.

你们当时允许远程办公吗？招聘过远程员工吗？

Do you do you did you guys hire remote and allow remote workers?

我们确实招聘过远程员工，但后来调整了这个政策。最初开放了大量远程岗位，之后又回归到以办公枢纽为主的模式。现在绝大多数员工都在主要办公点工作。

We did hire remote. And then that was another policy that we changed. So we moved to remote we allowed a lot of remote work, and then Yeah. We moved back. And so we have a few hubs, and and the vast majority of people are hired into those hubs and come into the office.

说到年轻创始人面临的挑战...其实我常和朋友开玩笑，感觉自己现在像是三四十岁的人。

Yeah. And then in terms of, I think, the, like, challenges as a as a young founder with a growing organization, I mean, I think that I often I often tell my friends now that I kind of feel like I'm in my mid to late thirties.

你这是活出了别人几辈子的经历啊。

You've lived a couple lifetimes. Is that

你是什么意思？

what you mean?

是的。我想说的是，我17岁就搬到硅谷工作了。大多数人是在22岁左右那个阶段才这么做的。然后我...我觉得我确实学到了很多。

Yeah. And I think I mean, I I I moved to Silicon Valley to work when I was 17. Most people do that when they're, you know, 22, around that time frame. And then I've yeah. I I think I've I've, you know, been able to learn a lot.

我非常感激能在规模建设中学习到这么多。

I'm very grateful to how much I've been able to learn in building scale.

我有个相关的问题。我曾读到你说，年轻时在硅谷工作感到有些孤立，因为无法真正与已婚有孩子的同事建立联系，所以你和斯坦福学生相处。但又因未上大学而难以融入他们。我想知道，作为最年轻的白手起家亿万富翁，你现在是否还会感到孤立？

I have a question that's related to this. I've read somewhere that you said when you were working at such a young age in Silicon Valley, you sort of felt maybe a little isolated because you couldn't exactly connect with your coworkers who had were married and had kids. So you spent time with Stanford students. You couldn't exactly connect with them because you weren't in college. I'm wondering, like, if now do you ever feel isolated that you're now the youngest self made billionaire?

嗯，我在新闻上看到了。你现在是否感到孤立，还是觉得'嘿，我有我的同龄人和伙伴，现在与人相处很融洽'？

Mhmm. I saw that in the news. Do you feel at all isolated now, or do you feel like, hey. I have my peers and my my people, and I feel connected with people now?

是的。我认为组建团队和创建公司最大的乐趣之一就是能与所有队友和共建者共事。我们...我们共同奋斗，作为一个团队完成艰巨任务。所以我一直很幸运。

Yeah. I think I think one of the great joys of building a team and building a company is are all the, like, all all of your teammates and all the people you you get to build with. And and so I I spend, you know, you know, we we work we work really hard together. You know, we we do we do hard things altogether as a team. And so I've been I've been very lucky.

Scale也是个特别的地方，许多团队成员离职后又回来了。这就像个紧密的社区，苹果公司也常有这种情况。部分原因是我们有独特的文化，做事方式新颖有趣。

You know, Scale's also a place that where a lot of our team members have have been in the company. They left, and then a lot of them come back. And so it's it's one of these, I think, like tight communities. I think Apple's another company where it often happens, where people leave and then come back. Partially, I think, because we have a a unique culture and we do things kind of in a cool in an interesting way.

因此我对所有共事过的人都心怀感激。现在我不像17岁时那样感到孤立了——虽然那个年纪本身就容易感到孤独。

And so I think all the people I've worked with, I'm very grateful towards. And then, yeah, I think I feel I feel I don't feel quite as isolated now as I did, you know, also when you're 17. I think in general, you just sort of have a predisposition to feel isolated.

说实话，今天下午我和保罗聊过，我说要采访Alex该问些什么。他确实引导我问些技术问题...不过我得提醒你，不知你是否记得——

Yeah. Now I have to admit, so I was talking to Paul this afternoon, and I said, you know, I'm talking to Alex, what what might be an interesting question? So he did sort of lead me with a few technical questions that I'm gonna ask. I just have to remind you, though. I don't know if you remember this.

你第一次见保罗时（他在你申请YC时已退休），他说你们初次谈话就超过一小时，可能就在我们家。他觉得你太有趣了，所以又约时间继续聊。还记得吗？

When you first met Paul, because he had retired by the time you applied to YC. Paul said that when he first talked to you, it was for over an hour. It might have been at our house, actually. And he was found you so fascinating that he like, you guys booked more time to talk because it was such a fascinating conversation. Do you remember that?

是的。那次对话让我受益匪浅。我试图回忆具体时间，应该是2022年。那是在ChatGPT出现之前，我们深入探讨了AI发展的各种可能性，并借鉴了科技史上诸多精妙的类比。

I do. Yeah. It was I I got I got a lot out of that conversation. I mean, it was it was it was I'm trying to remember the exact time frame, but I wanna say it was it was '22. It was before ChatGPT, but it was in, like we were, like, talking about all of the nuances of how AI could play out and and taking all of these incredible analogies throughout throughout throughout the history of technology.

那真是一次令人惊叹的对话。

It was it was, it was an incredible conversation. Yeah.

嗯。这是他提出的一个问题，只有在你愿意回答的情况下：有哪些类型的数据是软件特别难以标注的？同时，最难生成的数据会是什么？

Mhmm. So he here's one of his questions, and this is only if you're you can answer it. Are there kinds of data that are really hard for software to label? And and also, what would be the hardest data to generate?

这需要从两个维度回答。首先是数据演进方向——整个数据生态正加速向智能体（agent）转型。就像AI产业正从聊天机器人转向行动型智能体，数据领域也在同步演进。我们最重要的趋势之一就是为智能体获取海量行为数据，这类似于人类执行任务时的认知流程：先形成思路，再收集信息，经过反复思考后最终采取行动。

So there's kind of two different answers. So one is, where is where is data going? And and, you know, the data the data ecosystem is moving more and more towards agents. So just like the entire AI industry is moving from chatbots to agents, from talking to doing, the same thing is happening for data. So, we one of the major trends, the biggest trends is really getting lots and lots of data for agents, which ends up looking a lot like, you know, how should you know, in humans, when you go about and you're doing tasks and you're doing things, you know, you have a thought process and then maybe you go collect some information and then you think a bit more and then eventually, you know, you take the action.

以订机票为例：你会先查询选项，评估时间等限制条件，查阅日历确认可行性，综合各种约束条件后完成订购。这种完整的思维链和行动链数据目前完全空白——从简单的订机票到复杂的合同审查、软件开发或重大产品决策，人类执行任务时的完整思维过程和行动轨迹都未被系统记录。

So, for example, if you're gonna book a flight, you're gonna first check out what the options are and then you're going to, like, figure out what your constraints are and, you know, figure out, oh, do I can I leave at that time? You'll check your calendar or and they also figure out, hey, are there like, do I you know, what are all the constraints? You keep gathering information, then you ultimately book the flight. That kind of chain of thought and that kind of chain of activity is doesn't exist anywhere right now. Nobody no none of this data is really captured of the entire thought process and actions that people take over the course of doing these these tasks, even something as simple as booking a flight, but also, obviously, more complex things like reviewing a contract or or, you know, building a software feature or, you know, making an important product decision.

这类数据目前完全缺失。因此我们的核心课题就是构建双重机制：既要自然捕获人类行为数据，又要开发能自主生成这类数据的软件系统。数据演进的方向就是记录人类执行任务时的行为轨迹与思维过程——我们称之为智能体数据（agent data）。

You know, none of this this data is out there. And so a huge a huge trend for us or a huge thing that we're focused on is how do we actually go out and build the the mechanisms to generate a lot of this data, which is a combination of how do you capture a lot of this data from just the things that people are naturally doing, as well as how do you use how do you build software systems that are able to generate this kind of data? So that's where the that's where data is going, is towards more and more kind of data about people doing things and what their thought processes are when doing things. We kind of refer to this all as agent data. Okay.

关于机器自主处理的难点，我认为人机协同将是长期主题。AI总会犯某些特定错误，需要人类持续监督其长期表现。本质上就是人类协助机器突破困境——当模型偏离轨道、产生幻觉、遇到障碍或需要现实干预时，就需要人类介入。

And then the I think the answer in terms of, like, what is what is going to be hard for machines to do on their own in general, I think that, you know, I definitely foresee a very long path of of human AI symbiosis. And there's going AI are gonna are going to always make sort of some kinds of mistakes or some kinds of weird mistakes, and we're always gonna need to keep an eye on how AI functions long term. And so it's really like humans aiding the machines and sort of, like, getting unstuck. And, like, you know, whenever the models are are sort of, like, going down the wrong path or they're hallucinating or there's there's something that they're they're getting stuck on or they have to, like, they have to make some, change in the real world or something, that's where that's where humans will have to get involved.

他还提到：我打赌像Alex这样的人在数据标注方面肯定有些绝妙创意。有什么能分享的吗？

He also said, I bet someone like Alex has found some amazingly clever things to do in terms of labeling data. Does he have anything he can talk about?

当然。我们最重要的成果正是这种人机协同模式。在人类端，我们专注于汇聚各领域顶尖专家——最优秀的程序员、律师、医生、物理学家、数学家等，请他们为AI系统贡献专业数据；同时最大限度自动化流程，因为这些专家的时间极其宝贵，必须确保其工作效率最大化。

Yes. Well, I think that the biggest thing that we've we've worked on is is exactly this, like, human AI symbiosis. So on the on the human side, what we're really focused on is how do we get the most brilliant experts in every field? So how do we get the, you know, like, truly the best programmers and lawyers and doctors and physicists and mathematicians and, you know, computer scientists, you know, across every field to help contribute data for these AI systems? And then how do you automate as much of that process?

虽然谈不上特定技巧，但我们看到整个AI生态的关键在于：如何调动全球顶尖人才的智慧，让他们有机会塑造这些卓越的AI模型——这正是我们重点关注的领域。

Because these people are busy. They're you know, there's a lot of their time is very valuable. How do you make sure that everything they're doing is as efficient as possible? And so I would say that the you know, there's nothing particularly clever per se, but I do think the thing that we see in the overall AI ecosystem is, how do we leverage sort of the this community of the best and brightest in the world and then give them the opportunity to help influence and and leverage and sort of really shape these incredible AI models. And that's that's one of the things that we're really focused on.

有趣的是我们提到了科里森兄弟，因为你确实散发着科里森的气质。我坐在这里思考，虽然我觉得自己已经知道答案了，但规模是你毕生的事业吗？或者换个说法，你是否考虑过规模之后的生活？

It's funny that we brought up the Collison's because you definitely give a Collison vibe. And I was sitting here wondering, is I think I already know the answer to this, but like, is scale your life's work? Like or maybe another way to phrase that is, do you think about life after scale?

是的。其实...这是我和保罗初次见面时讨论的话题之一，他说长期来看，要想胜过所有人，最好的方式就是把公司当作毕生的事业。确实如此。我感到无比幸运甚至有些侥幸的是，当初选择了这个问题。

Yeah. I don't. Actually, this is one of the things that Paul and I talked about the first time we met where he would he said, you know, the best way that you're gonna outcompete everyone over time is if is if your company is your life's work. And and, yeah, no, I I think it's true. I mean, I think that one of the one of the things that I feel incredibly fortunate, and frankly, just kind of lucky, is that, you know, I picked this problem.

我们专注于AI数据领域，始终如一。通过OpenAI、Meta等合作伙伴，我们助力推动了人类历史上最激动人心的技术征程之一。我们看到了在奠定基础和促进行业繁荣方面的持续作用。最初我们甚至放弃过这个想法，觉得它成不了大气候，如今却成为我和整个Scale团队引以为豪的事业——它已成为AI产业未来几年的核心支柱。

We focused on data for AI, and and we've been focused on it for this whole time. And then we've helped really enable through our partners like OpenAI or Meta or other companies enable one of the most exciting technical journeys, frankly, of like human history. And so And so and we see a continued role in helping lay the foundations and continuing to sort of ensure that this industry can flourish. And so we've I feel very, very lucky that, you know, Scale, what started out as an idea that we even threw out we threw out the original idea because we thought it wouldn't be a consequential enough company, has ended up being something that I, you know, myself and the entire scale team are very proud of as as being quite central to the AI industry and and everything that's going to happen over the next few years.

这确实是AI产业的核心。虽然与直接研发AI不同，但至关重要。没有你们他们做不到。确实如此。

It's hugely central to the AI industry. You know, it's a different business than making AI, but it's hugely important to that. They couldn't do it without you. Yeah.

我们内部常说，Scale和英伟达就像是行业的幕后支柱，支撑着整个生态，让OpenAI、Meta、微软等合作伙伴能够大放异彩，成为耀眼的明星。

We usually talk internally that like, you know, scale and Nvidia are kind of are the behind the scenes folks that help, you know, hold everything up, in the industry and enable, you know, our partners like OpenAI or or Meta or or Microsoft or others to really, you know, shine and succeed and be the be sort of like these superstars.

是啊。

Yeah. That's

我有两个随机问题。知道你赶时间，但能否再聊聊ChatGPT早期？当时对你个人而言那种兴奋感——当一切发生时你究竟在想什么？

true. I just have two random questions for you. I know you probably have to go soon. But can we just please once again go back to the early days of ChatGPT? I want you to describe to me I feel like we didn't capture, like, how exciting and interesting that was to you personal like, what were you thinking back then when all of that happened?

因为那就像一场爆炸。

Because it was this explosion.

记得2020年GPT-3刚问世时，我给朋友演示。对某些人来说，GPT-3标志着新时代的开始——'天啊这太神奇了，这才是真正的成熟技术'；而另一些人则认为'这不过是个玩具，还不成气候'。

Yeah. So I remember in I think it was in 2020 when GPT three first came out. And and I was showing a friend of mine you know, g p t three was kind of it was an interesting moment because I think for some people, g p t three was was was like that was really the start of this whole thing, was, oh my god, g b three is so incredible. This is like, this is, you know, this is this is really prime time. And then for and then for others, was like, okay, this is still a toy.

当时朋友还不擅长与AI对话，但尝试交流时被AI惹恼了——因为AI没认真听他说话，反而有点戏弄他。那一刻我记忆犹新：'哇，这太有意思了'。

This is, you know, this is nothing serious yet. And I remember I showed one of my friends, and he was he was sort of chatting with it. And he wasn't very good at chatting at the time, but he was sort of chatting with it. And he got really angry at the AI because it was it was not really listening to him, and it was and it was sort of teasing him a little bit. And there was, like, a very clear moment in my memory where it's like, oh, wow.

这项技术很快就会成为一件大事。因为仅仅通过与这个人交谈就能引发如此强烈的情感，这确实是一个重要信号。但想想看，那是在2020年。要知道，那是在ChatGPT问世前两年。

This technology is is gonna be a really big deal pretty soon. Because the fact that it could elicit just in talking to this person, it could elicit strong that strong emotion was a real indicator. But that was think about it. That was 2020. It wasn't you know, it was two years before ChatGPT.

所以我一直觉得，嘿，这显然意味着我们发现了某些即将产生重大影响的突破，但我没想到它会来得这么快。当ChatGPT出现时，从行业内部看是个有趣的现象——它确实比之前的同类产品更优秀，不过之前也有过不错的聊天机器人。但OpenAI团队做得太出色了，他们让这项技术突然之间引起了全球所有人的兴趣，而不仅仅是AI圈内人。

And so I I had always felt like, hey. This is this is clearly this is, like, we're that we're on to something that's gonna, like, result in stuff, but I didn't know how quickly it was gonna come. And then when ChatGBT came out, you know, was kind of an interesting thing from a within the industry, ChatGBT felt a little bit better than what had come before it, but there were there were good chap there were, like, reasonable chatbots that had come before that. But I think the OpenAI team, like, really, they did such a good job with that because they got it to the point where all of a sudden, it was interesting to everybody in the world. It wasn't just interesting to people in the AI world.

没错。我开始注意到生活中各种人都在试用ChatGPT并讨论它。那时我就意识到：哇，AI已经突破了小众技术的范畴，即将成为主流。记得2022年初我在某个会议上发言时...

Right. And I started seeing this, like like, all these random people in my life were trying ChatGPT and were talking about it. And that's when I knew it was like, oh, wow. AI's crossed over to being a like, instead of just being this, like, really interesting, more niche technology, it's it'll be the it'll be the the major thing. You know, I remember there was I was I spoke at, like, a conference early twenty twenty two, I wanna say.

当时我还对主办方说：'不确定大家是否还想听AI话题，可能大家都听腻了，我们应该关注其他主题'。现在回想起来很可笑，因为过去两年人人都在谈论AI。确实如此。

And and I remember I told the organizer. I was like I I literally said, oh, I don't know if everybody wants to, like, hear about AI yet again. I feel like maybe, like, you know, maybe everybody's heard enough about AI and we we should focus on other topics. And it's so funny in retrospect because over the past two years, AI is all anyone's talked about. That's true.

但当ChatGPT成为史上增长最快的产品，当企业CEO们开始询问'我们该如何应对AI'，当政府召开各种紧急会议讨论AI政策，当它成为全球瞩目的现象时，我最大的感受是：我们必须系好安全带，要为这个行业奠定基础做大量准备工作，必须全力以赴。

But but, yeah, no, it was and then I think in the moment when when all of a sudden when it became, you know, Chachapi became the fastest GPT became the fastest growing product ever and and everybody was focused on it. And and enterprise CEOs were asking, like, what are gonna do about AI? The government held all sorts of crazy meetings about what what we were gonna do with AI, and it sort of became this the this very notable global phenomenon. I think the main feeling I had was, like, we we gotta buckle up. We gotta, like we have so much work to do to to prepare the ground for the industry, and we just have to, like, we have to, like, you know, put our heads down and and and really go after it.

即便现在，我们也很少有机会退一步看清发生的一切。但数字确实惊人——2024年竟有超过2000亿美元投资用于构建先进AI系统，这相当于美国国防预算的三分之一，太疯狂了。

And so even now, I think it's, like, it's rare for us, many people to even, like, take a step back and realize everything that's happened. But it's I mean, the numbers are staggering. I mean, like, I would no world would I have predicted that in 2024, something like 200 plus billion dollars of investment are going into building advanced powerful AI systems. I think that's, like, a third of The US defense budget. It's it's just Crazy.

它如何成为世界焦点之一绝对令人震惊。我常对团队说：在我有生之年，恐怕再难见证这样的机遇——我们长期耕耘的技术突然成为全球最重要的事物，这种情况绝无仅有。

Absolutely staggering how how much this has become, like, the one of the central focuses of the world. And I don't think I mean, I I I tell everybody on the team this. Like, I frankly don't think in my lifetime I'm gonna witness another, like, have this have this amazing opportunity where we were we were working on a technology for so long. We were we were helping power for so long, and then all of a sudden, it became the, like, most important thing in the world. That's that just doesn't happen.

确实。我觉得你将来应该把这些经历写成书。

I know. I feel like you should write a book someday about this and your experiences.

是的。最近尤其有趣的是，AI从业者都非常狂热且派系林立。AI就像罗夏墨迹测验——有人视其为大国竞争的核心要素（那些关注地缘政治的人），有人认为是下一颗原子弹...

Yeah. It's been it's been especially interesting as of late because I think the AI industry people in the AI industry are very, are very passionate, and there's there's so many different camps. You know, AI is kind of a Rorschach test in the sense that different people see such different things out of it. I mean, some people see it as a as the major element of, you know, US competition versus other countries, like people who are very geopolitically focused. Other people see it as as, you know, the next atomic bomb.

还有人看作下一代互联网或计算机，将推动经济增长。因此很难与大众进行理性讨论，因为每个人对技术的认知都不同。而技术发展如此之快，它确实可能演变成任何形态。能参与其中实在令人着迷。

Other people see it as the next Internet or the next sort of the next computer that'll, like, enable all of this growth and all this economic activity. And so there's it's it's almost impossible to have a clear, reasonable conversation about AI with with large groups of people because everybody just sees something so different out of the technology. And the technology is moving so quickly that it honestly could become any of the above. It's been extremely fascinating to have the opportunity to be a part of it.

太棒了。是的。嗯，这简直不可思议。非常感谢你今天能来分享所有的见解。我觉得和你交谈让我学到了很多。

Awesome. Yeah. Well, it's just incredible. Thank you so much for coming on today and sharing all of your insights. I feel like I've learned a lot just from talking to you.

所以我非常感激。希望今年夏天我们回到湾区时能见到你。

So I appreciate that. And I hope I get to see you maybe this summer when we're back in the Bay Area

是啊。

Yeah.

面对面。

In person.

是的，我很乐意。

Yeah. I would love to.

好的。那么，亚历克斯，再次感谢。这期节目会很精彩，我很高兴我们终于做了一家正经的AI公司，因为我觉得我们的观众肯定会对这些内容非常感兴趣。再次感谢。

Alright. Well, Alex, thanks again. This is gonna be a great episode, and I'm so glad we've finally done like a proper AI company because I think our audience is is definitely gonna be very interested in this stuff. So thanks again.

是的，谢谢邀请我。

Yeah. Thanks for having me.

很高兴能和你交谈

It was great to talk to

，亚历克斯。再见。再见。再见。卡罗琳，和亚历克斯的对话真是太有趣了。

you, Alex. Bye. Bye. Bye. Carolyn, that was such an interesting conversation with Alex.

是啊，我知道。我觉得我们本可以问上千个超级聪明的问题，关于Scale在做什么、他们怎么做以及他们下一步的计划，但我们真的是问得最多的吗？不，我们不是。

Yeah. I know. I feel like there's probably a thousand super smart questions we could have asked about, what scale does and how they do it and what they're gonna do next, but Are we really the ones asking the most? No. We're

不，我并非要贬低我们，但这确实涉及数据标注、清洗及生成这类工作。这对规模化AI公司来说是个巨大组成部分。任何希望在公司中应用AI的企业都必须能获取真正优质的高质量数据。

not. I don't mean to rag on us, but it is this sort of labeling and cleaning data and then generating data. It's such a big component for Yeah. For scaling AI companies. And anyone that wants to have AI in their company has to have access to really good high quality data.

我喜欢他把Scale和NVIDIA相提并论，因为几乎所有人都清楚NVIDIA在这场AI热潮中的角色。所以当他说'没错，就是我们'时——我们虽然知名度稍逊，但同样重要、同样举足轻重——我觉得他这个评论非常精彩。

I love that he put scale and NVIDIA together because everybody pretty much everybody knows and understands the the role of NVIDIA in this whole AI boom. And so for him to say, like, yeah, it's us. It's we're, like, even less you know, not quite as well known, but just as important, just as huge. That to me was a that was a great comment he made about that.

我感觉Scale AI将会成为主流大众从未听说过的最大的Y Combinator系公司

I feel like Scale AI is gonna be the biggest y combinator company that the mainstream has never

听说过

heard of.

从未听说过

Never heard of.

对。我现在就要把这个预测抛出来。

Right. Yeah. I really think I'm just gonna throw it down there now.

我认为这个预测可能是对的。它确实低调潜行，但表现绝对惊艳。

I think I think that might be right. And, yeah, sort of the under the radar, but absolutely killing it.

惊艳。而且他作为如此年轻的创始人，看起来真是个杰出的领导者。

Killing it. And he's just seems like such a great leader. I mean, for such a young founder.

是啊。最让我震撼的是，他正在壮大这个巨型企业，但通常——不一定是这个播客里，而是我在Well I See的见闻中——我们总听说领导者们的各种'脏衣服'，呃不该这么说，应该说是创始人领导者们面临的挣扎。而他似乎...

Yeah. I guess that's what I thought was so striking is, like, he's growing this huge company, and so often well, not even not on this podcast necessarily, but just in my life at Well I See. Like, we know all of this dirty laundry about people and all this well, I should it's not dirty laundry. It's more like struggles that leader founder leaders have. And he just seems like yeah.

或许他不谈论这些，但他要么完全掌控了这些挑战，要么因其管理风格和营造的文化而根本无需面对。这太了不起了。

Maybe he doesn't talk about them, but he's he's either just totally managed them, or he doesn't have them because of his style and the culture he's created. That's amazing.

这太不可思议了。我知道。我只是被我们的对话震撼到了。我想我得回去一趟，得在2016年对他的采访笔记里加个附录，当时我写了，可能是傲慢或天才。

That is amazing. I know. I'm just struck by our conversation. And I think I have to go back. I have to add an addendum to, like, my interview notes with him from 2016 where I said, maybe arrogant or brilliant.

我决定就写他是天才。

I'm just gonna put down he's brilliant.

是啊是啊。我完全不觉得他有傲慢的气场。我是说，格兰特，这都是很久很久以前的事了。对吧。

Yeah. Yeah. I I don't think there's any arrogant vibe. I mean, Grant, this is this is a very, very long time ago. So Right.

没错。辩护一下。

Exactly. Defense.

容我辩解一句。而且那只是基于十分钟的谈话得出的结论。

In my defense. And it was only based on a ten minute conversation.

十分钟的谈话。对，对。就说他是天才。纯粹的天才。

Ten minute conversation. Right. Right. And say, he's brilliant. Just brilliant.

是啊。而且非常接地气，非常随和。

Yeah. And, very down to earth and very chill.

等不及想看看Scale的下一步动作，我觉得这期节目会很精彩。我都等不及了。

Can't wait to see what is next with scale, and I think this is gonna be a great episode. I can't wait. Yeah.

能和他交谈真是太好了。

It's great to talk to him.

好吧。卡罗琳，我们下期再见。

Alright. Well, Carolyn, I will catch you on the next one.

好的，待会儿见。再见。再见。

Okay. See you later. Bye. Bye.