Story of a startup that closed after ChatGPT

0:00You know, my vision for this channel is to help you build, not just software, but a company so that the stuff that you're building using AI will actually generate revenue for you, help you build a business that you can then use to hire people. It'll be good for them, good for you. Anyway, the reason I'm saying this is because one of the models for that was a company called Lex. It was an innovative business that allowed you to write on the screen and then have an AI

0:23assistant give you feedback. And I'm sad that the and upset and angry and scared that the company is closing down. And the reason for that is because I think it was basically all their features were taken over by ChatGPT and Claude and other AI tools. At least that's the theory that I have going into this interview with the founder of Lex, who was incredibly innovative, one of the first people to get into building with AI and building AI tools. And the reason I want to do this interview is so that we can learn what what he

0:57learn from him, from his experience. So that as we build we can build better, we can build more profitably, and he's willing to be open here. So, here's the founder of Lex on what happened to Lex. Presented by Zapier, the AI automation company. What's going to happen with Lex? So, it is winding down at the end of the

1:17year. How does that feel to you? This was one of the like early AI tools. This was showing the way of the future. How does it feel? Uh it's funny. It's a lot of mixed emotions. Obviously, there's a lot of sadness because I poured a lot into Lex and I think I'm incredibly and proud of the product and the team and all that kind of stuff. And at the same time like this is life. The whole point is to take big swings, try fun things, you know, build something that's cool. I think

1:49there's a ton to be proud of of that. And then to know when it's time to move on is like a very important life skill and to like not attach your whole identity to any one thing is also a very important life skill. And I got to say the vast majority of the time, I just spend my time thinking about all the

2:04exciting new things I'm working on now. And I'm very much a school of the Mark Andreessen like don't dwell on it. [laughter] You know what I mean? I I don't I don't take your point of view on this, but I think you're right. If I were in your shoes, I would. I am angry and scared. I'm angry because here was a tool that clearly was revolutionary. And we'll talk about what it was in the AI space and scared in the sense that it was so revolutionary. It was early, but it was also used. And if this doesn't work, then what chance does anyone else have of working? Tell you

2:34what, let's go back and tell the quick story of of what happened. What was the original idea for Lex? Yeah. Yeah, so the original idea was I just was annoyed at Google Docs and I wanted to make something that I would want to use. It was like one origin of the idea. And then the other origin of the idea was like, also that would be a good sandbox to play around with AI. Because this is

2:55before ChatGPT. This is the big If you thought of like AI writing at the time, the big thing was like Jasper or like Copy.ai, which were really designed around like generating a blog post, like a marketing blog post, or like generating ad copy or that kind of a thing. And it's sort of like at scale, turn it out. A little bit formulaic, but like it's an efficient thing for market for

3:15marketers, like an efficiency thing. And the kind of writing that I was doing was like newsletter writing that had to like attract attention and like thoughtful and substantive. And I wasn't interested in AI spitting that out, but I was interested in like having more of a, you know, this word thought partner gets thrown around a ton. But and then you want on the right to be able to say, "I have a question. I need you to go look at this instead of going into Google and then going down a wormhole that you never come back into

3:45the writing from." Yeah, exactly. Exactly. Um and so I mean, if you're the kind of person who wants that sort of experience, I think it was a great product. And also, one thing I learned is that saving people time is just a much bigger deal or helping people do things that they literally couldn't do before is a much bigger deal than helping someone spend a bit more

4:11time on something. Which is basically what Lex comes down to, right? It's like if you're writing a thing, you want it to be really good. You want to get a lot of detailed feedback. Um that's just really different economically from you literally couldn't write code and now you can. Or you could write code, but

4:27now you can write code 10 times faster. And you can produce 10 times more ambitious projects. That that's the promise of cloud code and why people went into coding tools like that. And you're saying with Lex, what I was doing was not helping you write 10 times faster. I was supporting the writing that you were doing. Helping

4:44you be more efficient and more accurate. And that's just a small incremental improvement. Am I understanding it right? Yeah. I think it's like it can be more than an incremental improvement for some people, but for a lot of people it's it's relatively speaking an incremental improvement, especially as compared to the difference that AI made in like coding use cases or lots of other use cases involving just like ambitious projects that you never would have been

5:07able to undertake before. You know, there's a ton Also, to be clear, there's a ton of people that are just like using AI to spit out books and like publishing them to Amazon. Not the market I was going after, right? No. You wanted to be the support for the person who's actually writing. Did you have users early on, right? I remember it was you on like a screen share loom on the homepage saying here's how I use it. This is what it is. I remember using it and getting a similar experience at

5:32work. Yeah. You got my My theory was that you were getting users early on and then ChatGPT came out and people could get a full article. And then ChatGPT added artifacts and people can kind of talk back and forth with the article. And that that's what took it away. Am I right? Yeah, I think there's a huge part of it that's that. Um I also think that um there are so like

5:58ChatGPT didn't take coding away, right? Why did Cursor grow so fast? Why did Cloud Code grow so fast? Like I think that there's probably it's not just that ChatGPT is like the end-all-be-all for any AI product. I think for this specific type of AI product, it makes a lot more sense why ChatGPT was, you know, incredibly hard to compete with, obviously, versus, you know, as as like a kind of standalone thing. I think another big thing is um

6:26being vertical and specific helps a ton. And Lex was like kind of vertical and specific, but just in a vertical that was like not a very good vertical, to be honest. What was the vertical? Um uh people trying to write essay-ish like blog post-y type things or even like books. Um the most people who mostly do that, like programmers, are incredibly enthusiastic

6:49about using AI. You know, writers I think that there's so much fear and uncertainty and doubt and the idea of slop and the idea of like AI just like writes the thing for you and it's obviously bad. Like writing is thinking. I basically agree with all those things. I just also think AI is super helpful as like a sparring partner and an idea generator and like a way to make me think more about what I'm writing. And so, you know, you're like it's like a muddled message at that point. It's not near as clear as just like programmers being like AI good. I want to use more

7:19AI. Amazing. It's like selling into a much tougher market. And also just economically code is extremely valuable, right? Um like essays, they're they can be incredibly valuable, but they're sort of only valuable to the extent that they really like, you know, whatever, take off. Uh and uh that's just a really hard thing to do. It's

7:40It's a lot of people's job. And so, if you had to redo this, go back in history, what are the changes that you would make? And the reason I'm asking is because a lot of us are creating AI-based software. This is guidance for what we should be doing. Yeah. It's It's so hard to know. I think that um obviously knowing what I know now, I would have done some things differently, but it's hard to I mean, the the the future has to unfold one day at a time, right? Like, it's hard to There's a lot of things that would have been impossible for me to

8:07know then. Yeah, yeah. Let me be a little bit of a of a sparring partner for you. It might be a little bit It seems like one of the things I'm saying I'm hearing from you is go after a more valuable market. Somebody who is writing ads, which is what Jasper was doing, is more likely to stick with their software and pay for it than

8:25somebody who's writing essays, right? So, that's that's one takeaway that I've got. Number one. Number two, if you're selling AI, go after AI enthusiasts. So, artists are less enthused about AI because they feel it's entrenching on their creativity than coders who just want to move or even business people who are just looking for the fast answer. That's a second thing. What else would you take away from this that someone

8:49who's building AI software should learn? [snorts] Like I agree with both of your takes. Also, I think it's really important. Yeah, yeah, yeah. I think it's really Well, I think it's a little bit of an extrapolation of what I of what I of what I said. Uh what I actually think is at any given time, if there's something you're excited about, you should do it, and you shouldn't count on it being a business until it's sort of proven otherwise, but just do it because you want to do it, right? Because I think people who overly fixate on uh what the consensus feels like is a good idea for like a business,

9:27um it's just going to like most things that seem like a bad idea are a bad idea, but the best ideas also seem like bad ideas, and it's just good like Have you ever There's this Paul Graham essay called like it's like about stamp collecting or something like that. And the whole idea is like a lot of the best founders are kind of like stamp collectors in the sense that the reason why they get obsessed with something has nothing to do with it's going to be this huge brilliant business. They just got obsessed with the thing and it's hard to explain why. They're just kind of

9:53fixated on it. Mhm. And I think basically my advice to anyone would be do the thing that you're fixated on and like don't worry about it too much. Like don't think about if it's going to be a big thing or not. But also don't count on it becoming a big thing. In general, it's hard to like set out to like count on something becoming a big thing. I think it's much better just to tinker with things that seem exciting to you and to develop your taste over time with the lessons you learn rather than it is to try and triangulate like let me like average all the business advice and see

10:20if I can follow all the rules. Like it just doesn't You know what I mean? But would you would you have done it differently then? Would you have then maybe had fewer people on the team, kept it more as a side project, had a full-time gig while you did this if this was your stamp collecting?

10:32Yeah. You would? Uh That's a That's a good question. I think that given the information that I had at the time, it seemed like a great idea. And so it's hard to like I just basically don't think it could have unfolded any differently really. It's It's like my honest take and I had a great time. So I have no regrets. So basically you're saying the way that I handled it was fine. I saw an idea that I got obsessed with. I like writing. I like reading essays. I like helping people do this better. I don't think I should have taken this more seriously as a business and looked

11:06for advertisers cuz I don't give a rip about people who are looking to buy better write better Google ads. Yeah. Yeah. Kind of. Kind of. Like I just think that basically you have to do the thing that you want to do and you just have to keep learning and keep sort of adjusting and we did that. And like I

11:24think yeah. Anyone who's listening the idea is that I'm only cutting you off because I I know we only have very limited time. You've got four more minutes before you've got to leave the room here at Notion, you're saying, "Hey, folks, you have an idea? Don't worry about whether ChatGPT's about to steal your business. They might, they may not. Keep obsessing on the thing that you love. Make it stamp collecting so good that you can't stop it. If the world doesn't want to come around to it, or there's not enough money, that's fine." And I would add one

11:47other thing to it. Before you started working at at Notion, I saw a tweet from one of the guys at Notion, I forget who it was, who said, "This guy, Nathan, is one of the big influencers on me." That I had followed his work, I'd seen what he'd done, and fair to say Look at the smile of pride. Who was it that that tweeted

12:04that out? Uh Ivan and Akshay, they both They both tweeted nice things about me. Yeah, yeah, it's The two freaking founders of Notion, they were influenced, and they said, "This is the person that we want to work with." So, maybe the other takeaway is, listen, if the thing doesn't end up working out great, look at the people who are going to respect you for the work that you've done because your passion is in it. So, vibe code your stuff, or AI code, let ChatGPT and

12:25Claude come at you if they want to. Build something you love, and the outcome, even if it's not the ideal outcome of of a billion-dollar business, is still a great outcome. Totally. And also, the thing I would add, like to to sort of drive the point home, the reason why it's so important to just do the thing you're excited to do, is because that's the only way you're actually going to learn. So, like the things I'm excited to do now are a product of the things I was excited to do in the past, plus what I learned from

12:52the experience. And if you have a thing you're excited about, you're always going to be able to think of reasons why it's not a good idea. And then you just won't ever learn anything. So, you just have to touch hot plates, and you have to get hurt, and you have to lose. But like you have to do the thing that you're actually excited to do. You have to bet on your own instinct. And to be clear, instinct is going to be dead wrong so much, but the only way you end up with good judgment is by continually betting on your instinct, and learning whatever lesson you need to learn from

13:20it. Because whether you're right or wrong, you need to learn from it. Sometimes you'll learn you're right, or aspects of what you got You got aspects right and aspects wrong. You just have to keep betting on your instinct basically. And that's the only way that I actually know to learn because if you're sort of like, well, I really want to do this and I think it's a good idea, but other things the butt is like you will you at some fundamental level emotionally disagree with those butts. So like you have to just go complete the emotional arcs each

13:50time. Okay. What's the high of the business? How high like when you look at it this is when it got really good. What's that image in your head? Of like what Lex? Yeah. When it got really good. Oh, I mean like there's there's moments that are like oh, it seemed like people were loving the product or like the metrics were looking good for a while or

14:10whatever. There's ultimately all that really matters is the building with the team. Right? Just like going through the stuff, making decisions, like hanging out setting ambitious goals, shipping the things that we're proud to ship. Um and then you know, stories from people about how it helped them. That's all that really matters. How how were you able to afford the

14:34team? Did you raise money? We raised money. Yeah, yeah. Oh, I how much did you raise? 2.8 seed round. Wow. Okay. we were generating revenue too and so that 2.8 like went a lot longer than just the 2.8, right? But like yeah, in the end it doesn't you know um make sense to continue as like a independent business, but um like

14:54Okay. the the journey of it was uh yeah, it was like a real thing for a while. Hell yeah. Hell yeah. It was the one of the first companies, which is why I want to learn from you. Thanks for doing this and um if you like this interview, we've got another one for you right there.