创始人模式：杰克·海勒，Casetext创始人兼首席执行官

卡罗琳，今天我特别兴奋，因为我们今天有幸在Y Combinator创始人模式静修会上，与Casetext的创始人兼CEO杰克·海勒面对面交流。嗨，杰克。欢迎你，杰克。

Carolyn, today I'm so excited because we are here in person with Jake Heller, the founder and CEO of Casetext at the Y Combinator founder mode retreat. Hi, Jake. Welcome, Jake.

你好啊。

Hey there.

是啊，很高兴见到你。

Yeah. Nice to see you.

我也很高兴见到你。

Nice to see you too.

Casetext几年前被汤森路透收购了。

So Casetext got bought a couple years ago by Thomson Reuters.

没错。

That's right.

对。但我认识你们是在2013年夏季项目期间。

Yeah. But you guys I met you when you went through the summer thirteen, program

是的。

That's right.

在Y Combinator的时候。

At Y Combinator.

十二年前的事了。

Twelve years ago.

十二年前。十二年前。天啊，已经过去很久了。

Twelve years ago. Twelve years ago. Man, it's been a long time.

是的。是的。

Yes. Yes.

那么在进入创始人模式的话题之前，先跟我们说说公司当时的情况，简单介绍一下背景。

So tell us about and and and what the company was and a little bit about that before we get into founder mode things.

在创立Casetext之前，我是一名律师。执业期间我深感使用的技术不仅没有帮助，反而常常碍事。我想，我们必须能运用现代技术来实现信息检索、文件筛选等功能，让法律工作更轻松、快速、高效。比如当时我觉得，如果我想找一家营业中的好餐厅或电影场次

So before starting Casetext, I was a lawyer. And I felt when I was practicing that the technology I was using would often not even not help, it would get in the way. And I thought, you know, we gotta be able to apply modern technology to do things like search for information, sift through documents, etcetera, to make legal work easier, faster, more efficient. I felt the time, for example, that if I wanted to find a great restaurant that was open or movie times

嗯。

Mhmm.

瞬间就能找到。但若想查找能让无辜者免于牢狱之灾或保护当事人的判例法，可能得连续十天熬到凌晨五点才能找到那个判例。所以

I'd find it instantly. Yeah. But if I wanted to find precedent and case law that could make an innocent person, you know, not go to jail or, you know, protect my client, I might be up to 5AM, ten days in a row to find that one precedent. And so

卡罗琳正在点头表示赞同。

And Carolyn's nodding in agreement.

对。没错。你也是过来人。

Yes. Yeah. You've been there, done that.

是的。

Yes.

是的。所以我们创立公司的理念是用技术提升法律工作的效能。我们的使命就像是法律界的核磁共振仪——用技术赋能律师，让他们能为客户做得更多更好。最终希望实现的结果是让客户获得更公正的对待和更优质的法律代理。

Yeah. So we founded the company with the idea that we can use technology to make legal work effective and efficient. And the mission became if you know, almost like being an MRI machine before law, like technology that enables lawyers to do more and do better for their clients. And the hope for outcome would be justice for those clients, better justice for those clients, and better representation.

好的。那你参加YC时项目进展到什么阶段了？

Okay. And when how far along were you when you went through YC?

我们当时非常早期。我记得刚写完一个几乎算是演示版的网站。

We were very early. I think I just coded a, like, almost demo website.

你会编程吗？

Could you program?

我正想问，你是怎么学会编程的？

I was just gonna say, how'd learn to code?

我是在硅谷长大的。嗯。是吗？对，具体哪里？

So I grew up in Silicon Valley. Mhmm. You did? Yeah. Where?

森尼维尔。是的。

Sunnyvale. Yeah.

天啊。森尼维尔。摇篮之地。

Oh my god. Sunnyvale. The Cradle.

没错。就是摇篮之地。

Yes. The Cradle. Exactly.

当然了。

Of course.

对。霍姆斯特德高中。

Yeah. Homestead High.

没错。我们正在讨论这个。

That's right. We're talking about it.

是啊。昨晚我们就在聊这事。

Yeah. Were talking about this last night.

对。好的。

Yes. Okay.

其实我爸九十年代初就在我们家车库创办了互联网公司。那可是特拉华州第一家注册名称含'互联网'的公司，懂吗？

And my dad actually started Internet business in our garage in the early nineties. Like, the first ever company in Delaware registered the name Internet in You know?

天啊。你是认真的吗？

Oh my god. Are you serious?

是啊。大概是92或93年。而且——顺便说下——他是高中辍学生。那公司可不是谷歌级别的。

Yeah. That's '92 or '93. And, and he's he's a high school dropout, by the way. He's not like that company is not Google.

高中辍学？

High school dropout?

噢，没错。哇哦。辍学生。但他很早就意识到这项技术会大有可为。说清楚点，这家公司供我读完了大学，还有我妹妹也是。

Oh, yeah. Wow. School dropout. But he he figured out really early that this technology is going somewhere exciting. And to be clear, this this the company, it paid for my way through college and my sister through college.

这简直是...它让我们家实现了阶层跨越，从底层中产跃升到上层中产，彻底改变了生活。但它终究不是谷歌。懂我意思吗？虽然规模小，但对我人生影响巨大。别的孩子可能和父亲玩接球、修车什么的，而我是和我爸一起写网站代码。

It was it was immensely it took our our family to a new kind of, like, you know, from the saw lower middle class to the upper middle class as life changing, but it wasn't, Google. You know what I mean? Like, we're like but but it was it's so it was so impactful for my life. And and I think a lot of kids with their dads, maybe they, like, play catch or fix cars or what have you. I coded websites with my dad.

对吧？这就是我们的亲子活动。每个暑假我都在互联网商业系统公司工作，就这样学会了编程。尽管后来我爱上法律与政策，上了法学院成为律师，还是忍不住继续编程。

Right? So that was like our thing. Right? I every summer, I was working at Internet business systems, and so that's how I learned how to code. And and even though I fell in love with law and policy, went to law school, became a lawyer, I couldn't stop myself.

我一直在编程。

I was always coding.

太棒了。我之前不知道你还有这段经历。

That's great. I did not know that about you.

从车库开始创业。你的故事充满了标志性元素。

Just starting a business in a garage. Everything about your story is very iconic.

没错。这很加州，对吧？就像在说‘是的’。

Yeah. It's very California. Right? It's like Yes.

这也是硅谷的特色。

It's also Silicon Valley.

是啊。

Yeah.

尤其是霍姆斯特德高中，乔布斯和沃兹尼亚克就是在那里创立了苹果公司。嗯。

Especially Homestead High is where Jobs and Woz started Mhmm. Apple.

对，没错。

Yeah. That's right.

那也是在车库里起步的。不是吗？我是说，是的。

That was also in a garage. No? I mean, yes.

是的。是的。是的。

Yes. Yes. Yes.

哇。好吧。所以你说这是法律界一个真正棘手的问题。

Wow. Okay. So you said this is a real broken problem here in in law.

作为律师，我每天都深有体会。是的。

I I was feeling it every day as a lawyer. Yes.

是啊。

Yeah.

所以你决定离开舒适的律所？你当时是在大律所工作吗？

And so did you decide I'm gonna leave my cushy law firm? Were you at a big law?

那时我确实在一家大型律所。对，瑞普格雷律师事务所。

I was at big a big law firm at the time. Yeah. Ropes and Gray.

哦，当然。那你有尝试编写些程序验证可行性吗？

Oh, yeah. Course. And so did you did you did you program something up and see if it could work?

其实我最初犯了个错误——我在律所私自编写程序后，才告诉他们这件事。哦，结果他们说...

So I made an initial mistake, actually, which is I started coding stuff for myself at the law firm, and then I told them about it. Oh. And they said

已经猜到结局了。

Already know where this is going.

版权归我们所有。没错，他们当然会这么说。所以我立刻停止了。

We own it. Yeah. Of course they said that. Yeah. And so I halted immediately.

你可是律师，早该料到这点的。确实。

You're a lawyer. You should have known that. Yeah.

我我当时

I I was

犯了个错误。

making a mistake.

律师，但我有点天真。你知道吗？我以为他们会很兴奋感兴趣。他们问我，杰克，如果你整天搞这些编程，显然你计费工时不够。

Lawyer, but I'm kinda naive. You know? I I thought they'd be excited and interested. They asked me, well, Jake, if you're doing all this programming, you're clearly not billing enough hours.

是啊，首先你没达到计费标准。

Well, yeah. First of you're not meeting your billable.

但问题是我远超计费要求。可能我就是个工作狂吧。但那一刻我意识到必须辞职，因为我无法兼顾创造事物的热情和律所工作。只能二选一。后来我离职时还自嘲，要么几年后公司大获成功，要么彻底失败回去当律师——我当时以为只有这两种结局。

But the thing is I was way over my billable. I'm just a workaholic, I guess. But, yeah, like, you know but it was, it that was the moment I knew I had to leave because I couldn't pursue that passion of building stuff and also do my law firm job. So either or. And so I left, and it's funny at the time, I left, and I told myself, oh, either I'll start a company that'll be super successful in a few years, or it will totally burn out, and I'll go back to being a lawyer.

当然现实是，我花了十年时间四次转型才让公司最终成功退出，这绝非一场速战速决的冒险。

And that was what I thought the two options were. And, of course, what really happened is I fought to build my company piece by piece over ten years in, like, four pivots to a finally successful exit, but it was not a a quick in and out adventure.

你的故事值得单独做期节目，但今天我想请教其中一个细节

And I think you'll have to come on the social radars for a full episode of your story, but there is one aspect of it that I wanna ask you about in our time today

嗯。

Mhmm.

你们应该是最早一批运用ChatGPT的公司对吧？能给卡罗琳讲讲

Which is I think you were one of the first companies to, like, leverage ChatGPT for. Yeah. Is that right? Can you tell Carolyn and

这绝对算我们的创始人高光时刻。

I And that that's definitely, I'd say, our founder mode moment as well. Okay.

哦，完美的过渡。完美的过渡。

Oh, perfect segue. Perfect segue.

是的。所以，你知道，我们最初认为应该运用现代技术来帮助解决法律问题。说实话，我们尝试了许多不同的版本，不同的方法。但真正在早期（大约2016、2017年）就开始奏效的，甚至在当时还不叫人工智能，而是称为自然语言处理和机器学习。

Yeah. So, you know, we initially thought we should apply modern technology to help solve legal problems. And we had, honestly, a bunch of different iterations of that, different approaches to that. But the one that that really started working for us way early, like 2016, 2017, was wasn't even called artificial intelligence at the time. It's called natural language processing and machine learning.

没错。在现代人工智能浪潮兴起之前，我们就发现了各种令人兴奋的方式，可以优化信息搜索和总结。部分原因在于我们作为涉足AI领域的‘法律怪人’投入了大量精力，而当时从事AI相关工作的人还很少。

Yep. And we found all kinds of exciting ways that we could make searching for information better and all kinds of ways you can summarize information better even before the modern wave of artificial intelligence. And in part because we were so invested as like the weird legal guys doing AI stuff. And it's a pretty small community at the time of people doing anything AI stuff. Yeah.

所以我想我们引起了OpenAI的注意，当然他们也引起了我们的关注。我们互相交流经验，最终成为了他们早期测试计划的一部分。

And, you know, so I I think we got on the radar of OpenAI, and, you know, they're on our radar. We we shared notes, and we eventually became part of, like, an early tester program.

因为你们拥有大量数据。

Because you had a lot of data.

对，我们有数据。更重要的是，我认为对OpenAI而言，当时他们首先是一个研究实验室，最需要的是获得反馈。

Yeah. We had data. Yeah. And we had people I think I think importantly to OpenAI, was at the time, you know, as a research lab, first and foremost. And what they wanted is they wanted feedback.

比如我们开发的东西对医疗、法律、金融领域都有用。他们找到我们应该是看中了我们在法律领域的专长。而且我们整个公司都押注AI将变革法律行业，所以对他们做的事情非常兴奋。

Like, we're making stuff that's useful for medicine, for law, for finance. And I think they came to us as, like, the law guys. K. And we also you know, we we are bet the whole company that AI is gonna transform law, and so we were very excited about the stuff that they were doing. And yeah.

你刚才提到，你们从2013年就...嗯...Sam那时还在YC吗？

Well, was just saying, you were since you were twenty thirteen Mhmm. Was Sam at YC at the

他在。

He was.

所以你们从那时就认识，他肯定一直记着你。

Yeah. So you guys knew each other from that era, so he would have had you in his brain.

是啊。不过，我们几乎没有共事过。

Yeah. Although, we we barely overlapped.

哦，是的。

Oh, yeah.

但那时候确实如此。他在那里，我想那是在他成为YC总裁前不久，但他确实参与其中，绝对是合伙人，而且在YC过程中给了我们一些非常直接的反馈，比如'这不够快'。

But yeah. At the time. So but, yeah, he was there. This is right before I think he became president of YC, but he was definitely involved, definitely a partner, and, you know, gave us some pretty direct feedback during the YC process. Like, that's not fast enough.

就像这样，而且他是对的。明白吗？

That's like, and it was he's right. You know?

是啊。

Yeah.

所以快进到2020年、2021年。我们在GPT-3正式发布前看到了早期版本，还有GPT-3.5。每次我们都觉得这很有趣，也显示出潜力，但对法律工作没用。它当时只会生成听起来像法律术语的合理句子，但内容都是胡扯。

So, you know, fast forward to, like, 2020, 2021. We're seeing early versions of GPT three before it comes out and GPT 3.5. And each time, we were like, this is interesting and and shows promise, but it is not useful for legal work. It would do things would just make, you know, at the time, plausible sounding sentences that sounded legalese, but they were nonsense.

但它们确实不是。

But they weren't yes.

所以我们当时对OpenAI的产品实用性反馈相当严厉，几乎放弃了那条路。那时我们正在开发同类型的自有模型，但试图仅基于法律信息构建，我们认为那可能是更好的方法——结果证明并非如此。2022年某个时候，他们联系我们说终于有了可能适合我们的产品。

And so we we were actually pretty harsh with our feedback back to OpenAI about the usefulness of their products at the time, and it almost had written off that path. At the time, we were developing our own models in that same genre, but we're trying to do it just based on legal information, and we thought that had potentially a better approach. It didn't turns out. You know? But in 2022, I don't remember exactly when, they approached us and said, finally, we have something we think will actually work for you.

他们当时没这么称呼，但那是最早的GPT-4版本之一。历史背景是：这发生在ChatGPT发布前三四个月，GPT-4发布前七八个月，或许九个月。

And they didn't call this at the time, but that was the one of the first versions of GPT four. And to put this in historical context, this is three, four months before ChatGPT came out, and seven or eight months, maybe nine months before GPT four came out.

明白了。

Okay.

我们正在见证这件事。对我们最早的迹象之一是，我们用他们早期的模型去参加律师资格考试，只是想测试下它的表现。早期的模型得分在考生中排后10%。比某些律师强，但你知道，基本上就是那些醉醺醺来考试的人。

And we're seeing this thing. And one of the earliest indications for us is we ran their earlier models against the bar examination as just a test to see how it'd do. Yeah. Earlier model scores in the tenth percentile of test takers. So better than some lawyers, but, you know, they're the ones that show up drunk, basically.

后来我们接触到了这个新模型，它在考生中排前10%。比90%的人强

Like Yeah. And and then later on, we get access to this this new model, ninetieth percentile of test takers. Better than nine

而这中间只隔了四五个月。

And this was a span of, like, four or five months.

是的。没错。

Yes. Yes.

这真是

That's really

这是个未发布的模型，我们签了超级保密协议。我整天和这东西聊天，当时甚至还没ChatGPT。我就想，天啊，我是在和神明对话。

And it's this unreleased model that that we're under super NDA. Yeah. And, you know, I'm walking around chatting with this thing, and I'm like this is gonna be before we even chat GPT. And I'm like, oh my god. I'm talking to a a god.

这是个信息之神，而我不能告诉任何人。

Yeah. This is this informational god, and I can't tell anybody about it.

我的天啊。

Oh my gosh.

创始人模式是这样的：他们周五给我们演示了这个模型，当时我们有三位联合创始人，但他们只给了我和另一位创始人权限。我们被告知：能告诉第三位创始人吗？好吧。

Where founder mode comes in is so they show us this model, and I think that they show us a demo of it on a Friday, and they give just my just my cofounder and I, with three cofounders at the time, they give us just two cofounders access, and we're the ones that are it. Can you tell the third cofounder? Okay.

哇。太疯狂了。

Wow. Crazy.

正因为他们对此守口如瓶，而我尊重这一点。我们整个周末都无眠地测试这个模型。对吧？就我们两个人。我们遇到了所有这些问题。

And because they want to be so tight lipped about it, and I I respect that. We spent the entire weekend sleeplessly just testing out this model. Right? Just the two of us. We had all these problems.

客户上门就说，哦，你们搞的这个AI有点意思，但能实现这个吗？能搞定那个吗？而我们只能回答不行。那需要一整个公司来做，你知道，既要文件审阅又要法律研究还要合同起草——客户想要的就是全套服务。我们想要包揽一切。

Our customers come in saying, oh, you're doing with this AI stuff is kind of interesting, but can you do this? Can you this? And we're like, no. That's a whole company to do, you know, document review as well as legal research as well as contract drafting as well as that's what our customers want. We want we want everything.

但我们只能提供其中一小部分，因为公司只有100人左右的能力。但经过一个周末，我们突然意识到：等等，这个模型能全包。虽不完美，但人类也不完美啊。

And we can only give them a sliver of that because that's all we can do with, like, a 100 people at the company. And but over the course of, like, a weekend, we were like, oh, wait. This model can do all of it. Not perfectly, but you know what? Humans aren't perfect.

对吧？它处理所有这些事都出色得难以置信，速度也快得惊人。周末我们就构思出十几种应用场景。到周一时，我清楚必须让全公司100名原本专注其他业务的员工全部转向，彻底改变我们过去九年半的轨迹——全力押注这个方向。

Right? It's it's doing all this stuff so incredibly well and so incredibly quickly. We came up over the course of weekend with like a dozen different use cases for it. And by Monday, knew that I had to take all 100 people at the company who are focused on other stuff, and we had to pivot the business to focus on this instead, to change everything we were doing for the last nine and a half years Oh, to be on this.

是啊。这可是创始人级别的重大时刻。非常严肃的决定。

Yeah. That's a pretty serious founder man. That moment. Is serious.

说来有趣，对我来说这决定显而易见。本以为同期能接触这技术的公司不多，想着其他人肯定也在做——等GPT-4的API发布后全世界都会跟进。

It was I mean, it's funny because to me, it was obvious. Yeah. And, you know, when I tell the story to other people, I thought I thought, honestly, there are few other companies that have access to this technology at the same time as we did. I was like, obviously, everybody else would be doing this. And as soon as g b d four API comes out, you know, everybody in the world will be doing this.

我错了。大多数公司只把这当作现有技术栈的补充，搞个突击队应付下。而我认为最该做的就是抛弃此前所有业务，因为这才是未来——我们赌的就是这个。

I was wrong. Most companies are like, oh, this is kind of a neat add on to our current existing tech stack, and we'll, you know, have a tiger team attack this or whatever. To me, the most obvious thing to do was to drop everything we're doing before this moment and focus on this instead because this is the future, And that was the bet we made.

这就是创始人模式的硬核形态。没错。正是如此。

This is founder mode as grinder mode. Yes. Yes. That's what it is.

哇哦。所以你和联合创始人——唯二知情的两人——是同时顿悟的吗？

This is wow. So did you and your cofounder, the only two that were in the loop, did you both have this epiphany together? Oh, yeah. One

我们在Slack上，还有面对面时都疯狂讨论这事。凌晨两点、四点还在说：'我试了这个，完全可行！我们必须干！'具体日期记不清了，但绝对是个疯狂的时刻。

We were we were on Slack, you know, and also in person, just, like, talking ferociously about this. And we're like, you know, at 2AM, 4AM. It's like, I tried this thing, and it totally worked. We gotta we just gotta do this. And I don't remember exactly the the day we did, but it's something, like, crazy.

比如，可能在接下来的一周或两周后，我们有个高管闭门会议，讨论第三季度的情况，我们的销售表现如何？你知道的，就是那些常规事项。是的。到那时，我想很明显，我和Pablo这两个早期参与者已经，怎么说呢，彻夜难眠，大家都在谈论这件事。他们会说，嘿。

Like, maybe that next week or the week after, there was an executive off-site where we're talking, you know, quarter three, how did our sales go? You know, like, all the normal stuff. Yeah. And by that point, I think it was clear that we you know, Pablo and I, the two that early access, we're just, like, sleepless and, like, people talk about stuff. They're like, hey.

这真的不重要。相信我们。看，我们没法告诉你原因。这真的不重要。

It doesn't really matter. Trust us. See, like, we can't tell you why. It doesn't really matter.

哦，那两个人会神神秘秘的

Oh, those two would be cloak and

对，有点神秘兮兮的。是的，就一小会儿。然后我们出现在这个高管闭门会议上，我个人已经构建了几个我们应该开发的演示产品，然后说，哦，好吧，各位。

dagger. Yeah. Little bit for a tiny bit. Yeah. And then we show up at this executive off-site, and I had personally built a number of different demo products that we should be building, and said, oh, okay, folks.

我知道我们为这次闭门会议准备了完整的议程。但我需要把这个放到一边。这才是我们要专注的事情。而且

I know we had this whole agenda for this off-site. I need to push this to the side. This is what we're gonna be focusing on. And

你听到大家的骚动了吗？是的。然后你给他们看了演示，

Did you hear people stirring? Yeah. Then did you show them demos,

他们反应像是，天哪。

and they were like, oh my god.

你知道，这真的很有趣，因为一开始会议室里的人意见不一。哦。会议室前半部分的人反应是，哦，是的。这显然太酷了。没错。

You know, it was really interesting because the room at first is divided. Oh. The first half of the room front part of the room was like, oh, yeah. Obviously, this is fucking cool. Yeah.

抱歉说了脏话。

Sorry for swearing.

没关系。好吧。在这个播客里说脏话没问题。

That's okay. Okay. Swear on this podcast.

这节目偶尔会爆粗口。好吧。我们能接受。

Occasional f bomb on this show. Okay. We're okay with that.

好的。不错。因为我本来就这样。所以他们当时觉得，就是它了。而且...我觉得房间里有些人会想，我简直不敢相信我们正在看这么神奇的外星科技。

Okay. Good. Because that's how I am anyways. So they were like, this is it. And and and I think for some of the people in the room, they're like, I I cannot believe we're looking at this amazing alien technology.

而另一半人则在想，天啊杰克，我们两年前才转型，刚做完那个新项目，现在又来个全新的...你知道...他们需要时间消化。这些人聪明勤奋又优秀，但他们确实需要时间理解我展示的全新事物。我得花点时间让他们慢慢接受。我当时想...这周高管闭门会议的目标就是——因为OpenAI那时允许我们将保密协议扩展到高管层了。

And for the other half of the people in the room, they were like, oh my, Jake, we just pivoted like two years ago, and we just did this other new thing, and now it's a whole new like, know, and and they they had to take some time to get wrapped their head. These are smart, hardworking, good people, but they had to take their time to wrap their heads around, you know, this whole new thing that I was showing them. And it took a little bit of time to kinda warm them up. My my I was like, my goal for this executive off-site for this week is going to be because they allowed us to OpenAI extend the NDAs to our executive team at this point. Okay.

只是当时还没覆盖全公司。我说我们必须在这周达成共识，因为下周一的全员大会上，我们要宣布这就是公司的未来方向。结果...哎呀...我们最终都接受了。天啊。

Just but not the rest of the company at the time. I was like, we need to get all together on the same page this week because next Monday, at the all hands, we're gonna announce this as what we're doing with company. And Oh, boy. And we all got there. And Oh.

这就像法律界的某些原则——虽然我很久没接触了可能说错，你可以纠正——比如‘事情不言自明’（res ipsa loquitur）。

Again, it's one of those things in law. It's been a long time. It may mess us up. You may be able to correct me. Think res res ipsa loquitur, the thing speaks for itself.

这就是我对这项技术的感受。它强大到根本不需要我多费唇舌推销，大家只要稍微试用就能明白它对用户的价值。有意思的是周一宣布时，反应还是两极分化。

And that's how I felt about this technology. But, you know, it it is not it is so evident and so obvious that this stuff is so powerful that I didn't even have to do that much selling at the end of the day. They just had to, you know, work with it a little bit and see exactly how powerful this would be for our users. Yeah. And what was really interesting is that when we announced it on Monday, same kind of reaction split.

有些人特别兴奋，有些人嘛...深入说说那些持怀疑态度的人吧。经营企业十年间，我见过很多一飞冲天的公司，像Facebook和Stripe对吧？有些企业几年就能登上《福布斯》封面融资无数——我们从来不是那种。

Some people very excited. Some people you know, and to maybe go a level deeper into the people who were, you know, a little skeptical. After running a business for a decade and there are many businesses that that take off immediately, and that's amazing to watch, like Facebook and Stripe. Right? At least some businesses that that a few years in, and they're already, you know, on the cover of Forbes and making you know, raising we were never that.

我们总是艰难前行，不断试错。十年间我们确实发展得不错，活下来了，奋斗到了这一刻。嗯。

We had always struggled, and we'd always like, okay. We'll try this thing. It didn't quite work, but it worked a little bit. You know, over the course of ten years, we had grown well, and we had survived, and we had fought to be at that moment. Mhmm.

但说实话，那时我们绝对称不上是爆发式增长。

But we were, we're, by no stretch of the imagination at that point, you know, a super takeoff.

不是火箭式上升。

Not a rocket ship.

是啊，还不是火箭飞船。而且，你知道，作为CEO，你的可信度是有限的。是的，我

Yeah. Not a rocket ship yet. And, you know, you only have so much credibility as a CEO, Yeah. I

那很棒。

That's great.

因为你说了那会成功，结果只成功了一点点。然后你说，

Because you said that's gonna work, and then it only worked a little bit. And you say,

每次都要拼命推销。

well selling it every time really hard.

没错。是的。所以我不得不再次向公司强调，这次会成功的。有些人已经跟随我九年了。

Exactly. Yeah. And so I had to come to the company again. This is going to work. And some people have been with me for nine years.

我会说，比如，杰克，你知道，那次、那次和那次我都信任你，结果还算可以。但我觉得我必须格外努力确保每个人都认同这一点。我主动承担起责任，明确表示这就是定义，这就是愿景，这就是我们要构建的东西。

I would say I'd say like, well, Jake, you know, I trusted you that time and that time and that time, and it kind of worked. But, you know, I felt like I had to go out of my way to really make sure that everybody was on board with this. And I I took it upon myself to be like, this is, you know, defined. This is the vision. This is what we're building.

这就是未来的样子，这是它的运作方式。顺便说一句，现在每个人都在做某件事，你们得立刻停下，马上投入这项工作。在这个技术剧变的特殊时刻，你不能指望每个员工都有远见，也不能指望他们都知道所有部分如何拼合。

This is what's going to look like. Here's how it's going to work. And also, by the way, everybody's doing something right now, you just stop doing that and work on this right now. That, you know, I felt like in this especially a moment of of extreme technological change, you can't expect every one of your employees to be visionary. You can't expect every one of them to know exactly how all the pieces are going fit together.

他们不像你那样考虑全局，也没有像你那样与客户进行所有对话。在那个时刻，他们需要明确的指引。我认为其他时候你可以信任他们自主承担更多，但此刻，回到创始人模式的讨论，部分表现就是需要非常清晰地说明要完成什么、何时完成、成功的标准是什么，甚至在某些方面需要微观管理。我当时在亲自编写代码。

They are not thinking about the big picture like you are. They they aren't having all the same conversations with customers like you are, and they need that direction, you know, in that moment. I I think there are other times where they can you can trust them to to take on more on their own. But in this moment, and to go back to kind of the founder mode conversation, I think part of it looks like is is being moments where you need to be very, very clear about what needs to get done, what timing needs to get done on, what good looks like, and to be, in some ways, of micromanaging. I was coding stuff.

我对代码库做出了重大贡献——这是多年未曾有过的情况。是的，在GPT-4 API还未问世前，我们就在探索提示工程的创新方法。天哪。

I was I was a big contributor to the code base, that that was something that was not true for many years before that. Yeah. I was we were inventing ways of prompt engineering before that was, you know, even talked about because the the GPT four API was not available. Right? Oh my gosh.

我们真的、真的深入其中。我个人、我的联合创始人都非常投入。我认为如果没有这样做，就不会取得后来的成功。

We were, you know, getting really, really, really deep into it. And personally, me, my co founders, very deep into it, and I think had we not done that, it would not have been as successful as it was.

你们的投资者也存在类似的分歧吗？

Were your investors similarly divided?

是的。是的。我们的董事会，你知道，有些人立刻就明白了。其中一位投资者甚至专程飞去找他们，坐下来花了一整天时间...对，试图说服他们，没错，如果我们走老路也能达到2000万的目标，但这次...这次是我们等待了十年的机遇。

Yes. Yes. Our board, you know, some of them got it immediately. For one of my investors to fly out to see them where they were, and sit down with them and spend the day Yeah. Trying to convince them, yes, we're going to hit our target of 20,000,000, whatever, if we had to go the old way, but this is the this is the opportunity that we've been waiting a decade for.

这就是...这就是我们关于AI在法律领域应用的核心论点。这就是它将如何改变世界。而他们...他们起初持怀疑态度...对。但最终被说服了。是的。

This is this is our thesis around AI applied to law. Here's how it's gonna change the world. And and they they were skeptical at first Yeah. But came around. Yeah.

我们能否快速聊下这个。你后来，大概几年后，站在公司全员面前宣布：猜怎么着？我们要出售了。如果我没记错的话，那是个很大的数字。

If we can talk about this really quick. You then, probably a couple years later, got up in front of the company and said, guess what? We're selling. And it was a big number, if I remember.

对。

Yeah.

那么，带着所有这些历史包袱，十年的转型和折腾，当时是什么感受？

So, like, how did that feel with all of this history, ten years of, you know, pivoting and doing this stuff?

不是几年后，是几个月后。

It wasn't a couple years later, it was a few months later.

几个月后。哇。是的。我...我可能把时间线搞混了，因为...

Months later. Yeah. Wow. Yes. I I guess I screwed up the timeline there, because Yeah.

因为

Because

因为我们...它存在的时间还没那么长。

because we we it hasn't been around that long.

是的。对。

Yeah. Yes.

没错。

That's right.

我们又一次，我们奋斗了十年。是的。为了达到当时的位置，最终我们开始腾飞。我们就像火箭发射一样迅猛发展。

We we again, we had been fighting for ten years Yeah. To be where we were, and, finally, we were taking off. We were doing the rocket ship thing.

是啊。对啊。

Yeah. Yeah.

对吧？这很典型。就像，是的。立刻。我们有很多有利条件，但其中一个优势是，因为ChatGPT在2022年11月问世

Right? And this is classic Like, Yeah. Immediately. We had so many things in our favor, but one of the things in our favor was that because ChatGPT came out, in November 2022

确实如此。

That's right.

全世界都在讨论AI。是的。而我们，在那之前就在研发GPT-4，直到2023年3月GPT-4也上线后，我们才被允许推出产品。

The whole world was talking about AI. Yeah. And we, from, you know, from before that working on GPT four, we're only allowed to launch our product after GPT four also went live in March 2023.

明白了。

Okay.

所以市场需求非常旺盛

So the market was so hungry

是啊。

Yeah.

人们迫切渴望将AI应用于金融、会计和法律领域，对吧？法律可能是最顶尖的应用场景之一，至今仍是大型语言模型最能大展拳脚的地方。毕竟，处理文本就是它们的本职工作。

So hungry to apply AI to finance and accounting and legal. Right? Legal is, like, maybe one of the top ones, still is the top ones of where okay. Large language models. Like, text is, like, the job.

嗯。

Mhmm.

市场反响极其热烈——律师事务所、司法部、无罪计划组织、公设辩护人，还有各大公司都争相询问：这项技术能否改善他们的工作，最终惠及客户？我们就像火箭般一飞冲天，前九年赚的钱在几个月内就翻了一番，明白吗？

And and so the market was really, really excited about. Like, the law firms and DOJ and the Innocence Project and public defenders and companies all wanted to talk to us about can can they use this technology to make their lives better and to help their clients ultimately? And so we we just came out like a like a just just rocketed straight up. And the amount of money we made for the first nine years, we basically doubled that in a few months. Right?

天啊，这数据太疯狂了。

My god. Crazy statistic.

哇哦。于是我们开始融资。说实话，过去九年每次融资都像在碰运气——你四处奔走，谁都愿意见你。嗯，懂吧？

Woah. And so we we started to fundraise. And I'll tell you, every single fundraise for nine years was one of those fundraises where you go out and you just anybody will talk to you. Mhmm. You know?

你总会接受那些会面，因为作为创始人，你总对自己说：只要有一个投资人点头就够了。

You just take that meeting because, you know, you you say to yourself as a founder, all it takes is one.

没错，没错。

Yep. Yep.

所以我们见了所有人，其中就包括汤森路透风投。最难忘的是——我们当时差点拒绝会面，因为已经收到几份投资意向书了，这简直难以置信。我当时就想：哇...

So we talked to everybody. One of those people we talked to is Thomson Reuters Ventures. And, you know, I'll never forget. Almost didn't take the meeting because we already had some term sheets, and this is incredible. I was like, wow.

我们居然已经拿到投资意向书了，太不可思议了。

We have term sheets already. That's crazy.

主动找上门来的。

Coming to us.

是啊。人们正打电话来，他们互相抬价。这真是前所未见。

Yeah. They're coming people are calling us, and they're, like, bidding up each other. Like, this this is new.

太棒了。

That's so pretty.

没错。这确实非常新鲜。

Yeah. Exactly. This is very new.

感觉如何？

What it feels like.

对，完全正确。这就是应有的感觉。

Yeah. Exactly. Exactly. Exactly. This what it's supposed to feel like.

哇，这太不可思议了。我们差点就没参加那次会议。我去开会时，发现他们在某种程度上存在竞争关系，我当时就想，

Oh, wow. This is amazing. So And we almost didn't take the meeting. We I went to the meeting, and and they're quasi competitive in a sense, and I was like,

哦，我正想说他们绝对...

oh, gonna say that they definitely We

我们几乎要错过那个会议了。在那短短三十分钟的会面中——我觉得时间也很晚了——首席产品官和首席战略官都出现了。我意识到这和我预想的会议完全不同。当晚他们就打来电话说：杰克，我们知道你手上有其他投资意向书，也知道你承诺明天回复他们。

we were almost to take the meeting, and, like, in that thirty minute meeting, which I think is, like, late too. Like, the chief product officer shows up. The chief strategy officer, like, okay, this is a different meeting than I thought it was. And that that night, they called and they said, Jake, we know you have other term sheets. We know you told them you get back to them by tomorrow.

我们清楚估值情况（他们做了功课）。但我们不打算投资，而是想提出收购邀约。请给我们宽限到周一准备正式报价。

We know what the valuation is. They did their research. Yeah. We're not gonna offer you an investment, but we would like to make an offer for you to buy the business. And just please give us until Monday to put together that offer.

相信你会满意。最终如你所说，以六亿五千万美元成交了。

We think you'll be pleased. Okay. And and and like you said, ultimately, sold for $650,000,000.

你去参加全员大会。你告诉大家这就是我们要做的，大家肯定都反应，比如，现在我明白这个愿景了。对。这真是个了不起的成果。

You go to the all hands. You tell everybody this is what we're doing, and everyone must have been like, Like, now I see the vision. Like Yeah. This is an amazing result.

我是说，而且你知道，即使在董事会和公司内部这也是有争议的。因为

I mean and and, you know, even that within the board and the company is controversial. Because

有些人希望你能乘势而上更久一些。

Some people want you to ride the wave longer.

是啊。这很合理。他们当然会这么想。

Yeah. That makes sense. Of course, they did.

没错。而且说实话，如果我们再坚持一下，可能还能卖得更高。当时我们决定出售时，感觉像是，也许几个月后这个价值能翻三倍。对。也许五年后我们会后悔没接受那个报价。

Yeah. And and honestly, we probably could have sold for even more had we held on. At the time when we made the decision to sell, it was like, maybe we can triple this in a few months. Yeah. Maybe we're talking about five years from now how we wish we took that offer.

你知道吗？因为可能GPT-5就要发布了。我们...因为我们完全没预料到。在2023年的时候，可能GPT-5一个月后就问世了。是啊。然后让一切都过时了。

Did you know? Because maybe the the g b d five will come out. We we because we had no idea. At the the '20 at the 2023, maybe g b d five will come out in a month Yeah. And and make everything obsolete.

我们当时完全不知道。

We had no idea.

你有没有想过，比如，也许我们会成为一家上市公司，而我将成为上市公司的CEO？

Do ever you think, like, maybe we'll be a public company, and I'm gonna be a public company CEO?

是的。那...那对我们来说是个可能性。对。第一次感觉那么真实

Yeah. That that was, you know, a possibility for us. Yeah. For the first time, felt real

是啊。

Yeah.

你知道，经过九年的努力，在一个大约1500万到2000万美元年收入的规模上，艰难地建立了一家相对较小的私营公司。

You know, after nine years of struggling to build a relatively small private company in, like, a 15 to $20,000,000 kind of ARR

范围。

range.

是的。现在感觉就像是，哇，我们真的可以起飞了。说实话，我认为这个领域的其他公司很可能也会上市。

Yeah. Like, now it's it's it's like, wow. We this really could take off. And I and to be honest, I think there are other companies in this space who will probably become public companies.

是的，没错。

Yeah. And That's right.

事实上，我一秒钟都没有后悔过。

The truth is, I don't regret it for a second.

是的，那很好。

Yeah. That's good.

你知道吗？我是说，

You know? I mean,

这是银行里的钱，宝贝。

it's Money in the bank, baby.

不仅仅是那样，虽然那很棒。是的。我的意思是，部分原因是我看了我的员工名单，而这个决定最困难的部分是你必须在没有咨询他们的情况下做出。通常是这样。

It's not just that, although that's great. Yeah. I mean, a part of it was I looked at the list of my employees, and the hard part about this this decision is you have to make it without consulting them. You you know, usually.

对，没错。

Right. That's right.

对我们而言，我们不想分散公司其他员工的注意力。我们需要双线并行——我来处理这次谈判，其他人则继续全力推进业务。因为如果收购不成，所有人都需要为未来做准备。我们必须持续增长，必须领先于竞争对手等等。

And for us, you know, we didn't want to distract the rest of the company. We didn't we needed to do dual process. I handle this conversation. Everybody else build, build, build, because if we do not do this acquisition, then everybody else, you know, needs to we need to prepare for our future. We need to keep on growing, we need to be ahead of the competitors, etcetera.

但我打印了一份Excel表格，列出每位员工的潜在收益。当我看到那些数字时——这位跟随四年的工程师将获得改变人生的金额，那位效力七年的办公室经理也将得到足以改变命运的报酬。在我的价值观里，没人能拒绝这样的机会。

But I I printed out a list of Excel spreadsheet of every employee, how much money they'd be making. And I saw people like, that that life changing amount of money for this engineer has been with us for four years. This is life changing amount of money for this person, you know, the office manager who was here for us for seven years. Like, they each can't say no to that in my village. Yeah.

这对员工来说是巨大的责任。你改变了他们的人生轨迹。

It has to It's such a responsibility for your employees. You've changed their lives.

是的。希望我做了正确决定。当时我们坚信，要将事业推向巅峰，汤森路透是最佳选择。多数人只知道他们的路透社新闻服务。

Yeah. And I hope I made the right call. I mean, one thing that we also felt strongly about the time is there is no better place, we felt, to take what we were doing and make it really big than Thomson Reuters. Most people only know them for Reuters, which is the news service. Yeah.

但他们还拥有最权威的法律产品Westlaw和Practical Law，在会计、金融和风险管理领域也举足轻重。最让我震撼的是我们的CoCounsel产品——在汤森路透官网上，这个巨型企业用小字标注着它。

But they also have the biggest used legal products, Westlaw and practical law, but also in accounting and finance and risk, like, the board. And one thing that is, like, mind blowing to me, that product that we have is called CoCounsel. If you go to, like, thomsonreuters.com, it's, like, mega company. It's like Thomson Reuters small print co counsel. Aw.

明白吗？这种感觉...太疯狂了。能为全球数万乃至数十万专业人士带来影响，既令人振奋又深感谦卑。

You know? It is That feels good. Wild. Yeah. Is wild and humbling to to, you know, have that kind of impact now for tens of thousands of professionals, hundreds of thousands of professionals across

全世界。真喜欢你刚才说的毫无悔意。这种无悔的感觉太棒了。

the world. Love that you just said, like, no regrets. I have no regrets about this. That's a that's a great feeling.

是啊，我懂。

Yeah. I know.

嗯。

Yeah.

太精彩了。幸亏你保持着创始人思维，在那个与巴勃罗共度的周末听从了直觉。

I love it. And thank god you were in founder mode. Yeah. And you listened to your gut Yeah. Over that weekend with Pablo.

是的。是的。

Yeah. Yeah.

太棒了。干得好。听到这个故事我超级兴奋。这太有趣了。感谢你的到来。

That's amazing. Well done. Well am so psyched to hear the story. This was so fascinating. So thanks for coming on

我们今天的节目。谢谢你，杰克。

our show today. Thank you, Jake.

太棒了。再见，杰克。再见。

It's great. Bye, Jake. Bye.

再见。

Bye.