Is Affirm outgrowing credit cards? Max Levchin explains

So reality is I co-founded this one, too. I've been a solo founder in the past. It doesn't work as well as people think it does. It's very challenging to be entirely in your own head all the time in the early days when you sort of think I got it. I have a plan. It's a vision and you don't have a counterpart to be like, hey, am I crazy or is this really a thing? And vice versa like are we really toast? Like are we done? Like you want to have that couch ideally fully reciprocated by your co-founders where they can sometimes come hang out on yours and complain to you. And I had that with Peter and I was very happy to have a handful of great co-founders early on at Affirm. The reason the co has gotten dropped at Affirm. It's been 15ish years actually. I didn't realize it. So yeah, it's been a long time and one by one my co-founders are like this is great but you seem to really enjoy running it and I got this other thing I want to do. So one by one they went off to do another startup, join a VC firm, other things. So, I think at some point our head of comms back in the day said, you know, you're a co-founder, but when people ask where the other ones are, you're like, oh, not here.

I am one of four. Okay. It turns out, but yeah, I've been the last remaining last standing. No, last remaining.

So the cool thing, so I'd love to claim extreme foresight, but I didn't know any better. I've been heard recently saying that this is by far the best time in history to run a company as an engineer where just the sheer abundance of engineering amazingness that's happening all around us with AI codegen tools and just all the things that you can do better than most people if you have a good understanding of how AI works, what you can do with it, what it cannot yet do, all the sort of technical details that we have and have taken for granted as computer scientists, as computer engineers is just it's an unbelievable advantage. Like a company that is not run by an engineer is not as good a stock bet in my opinion today as one run by somebody with a CS degree. And 15 years ago, that was not obviously a thing. But I did start a company with three other CS majors. And so we actually only knew professionally how to do more or less one thing, all four of us. But as a result, we sort of got into this habit of like, hey, you're gonna go do this thing and I'm going to do that other thing. Neither one of us really qualified, but both smart people. We have engineering mindsets. We know how to build stuff. We're just going to get it from first principles. That actually served me really well. The current executive team at Affirm. If you look at sort of we've been together for a long time and many people have had their roles fairly suddenly changed like you're doing this, ah, we're going to go do that other thing now. You take over, you go jump on that guy's work. And it works really well for us because huge percentage of people around me all have either the engineering mindset or engineering degrees. Not necessarily computer engineering. Incidentally, our chief operating officer is a mechanical engineer and a great one. And so we have the luxury of having engineering mindsets all around us who are, you know, served us very well in this moment in time.

I think it's going to be the best time ever. I think the rumors of we're all just going to whisper to our agents who are going to tell their agents what to build is like not even a little bit how it's going to work. I think the current moment for computer scientists is the baseline used to be you're an engineer and that other person over there they're 10x and like you will have to work a long time and have the gift and you know go to IOI and all the sort of special unattainable things and then you might get to like 5x but that person they're 10x and the new baseline is like you got a computer science degree cool you're 10x engineer come along we got a bunch of agents ready for you and I think there's still going to be people who are exceptional, but they will not be 10x, they'll be 10,000x. And so this idea of I have a thing over the weekend. I'm going to build this and like turns out to be a six-month long project. It will be a weekend long project. The productivity gains we're seeing are stunning. And so if you're graduating in computer science, like the world is your oyster. Like there are so many opportunities. It's all complicated. You know, some companies are laying off engineers and other companies are screaming, 'We need an engineer right now.' It's all going to be okay. Like your computer science degree is going to be an extremely valuable asset.

I'm just... Oh, I'm not sure I qualify for the other half, but and by I mean old. I mean original. I think the best advice I got is you made the right choice. Don't fret. It's going to be okay. There's going to be a lot of companies that are very interested in building their own software for the first time in a hundred years. Like they may have been great consumers of software and the long list of I wish my XYZ software package did this but the company that made it they just don't care and no matter how big and important we are and to replace that two years ago well I would have to have a 100 person engineering team and it's just not who we are. That's not what we do today. It's like I would need like a five person engineering team and a really clear sense for what I want. Like you brand new computer science major can be in that five person team like that. That's the opportunity ahead of us. And the open-mindedness that you have to bring to the table is you may not want or may not have the opportunity to go work at a frontier lab unless that is exactly what you're trained for and have the interest and the aptitude for. You might end up in like some company that manufactures combined harvesters. They too need software and right now they probably don't make their own.

It's a thing that harvests stuff from fields. You don't know what a combined harvester is?

Oh, it's all kinds of fun stuff. I built software. I've written more software with some help from my agents in the last six weeks than I've done in the last six years. I think...

I love this. Like if I would have my brothers, I'd just like live in a small room with the occasional visit to the outside world to get my vitamin D.

No, this is enjoyable. But I love making stuff. I love making products but making products in my mind is mostly making software that enables others to have great product experiences. So I still build software. I try to write code mostly because it's therapeutic and it's a lot of fun every day certainly every week. Some days are busier than others but every week is definitely the case. I have side projects, have a lot of side projects. They're easier now with agents. I built some home automation stuff recently which I really enjoyed through an agent.

So I designed actually so we went on this long exploration ourselves like I have all these devices I don't like some of them. I really hate the remotes they come with. What if I consolidated everything into a single product and initially I built a command line interface that with my agent friend we had to go get some access and some of these devices weren't really configured for user level access. So, we broke into them together and acquired some passwords that we weren't supposed to have and got through some other holes and we had a full complement of command line access to every device in my house and then I built a web page that I can control it and I showed it to my wife and she said if you're going to make a web page like I'm not going to go to it like you have to make an app like I've never built an iOS app before but this is a great time to start to learn. So I talked to my agent. We agreed that we should build an iOS app together. And then we did and now we have an app that is the consolidation of every remote in my household.

Yes.

It was like I mean I got a little bit rabbit hole into it. So that was probably I don't know like eight hours over the course of two days.

No. I don't... I think the value of human interaction, human leadership and human conversation is actually going up in value and so having a...

Maxbot, Levchinbot, that sounds funny.

I Maxbot actually. So I built a Maxbot one of my other side projects. So I got completely obsessed with the like, you know, Claude Code is a really cool harness as the fancy term of art. And like all these guys have built all these their own versions of like what their perfect bot is. And like these bots are all trained to do something or refined to do something. So some of them are for writing code and others for like doing menial tasks. Like I got to go get one of those for myself. So I built a Maxbot and it's great. It sits on many of my various communication system channels and answers my questions and like look stuff up for me.

It's not that hard. Like you don't need me.

But therein lies the answer to your question. So...

No. Well maybe. But I do think that the world is currently the headlines are dramatically underestimating how much easier it is to start building these things and to bring them to completion if you actually understand how the underlying stuff works.

We live underground. We never leave these sort of polyhedral rooms. Food drops in from the ceiling. We talk to our friends and peers through sort of a screen. Like this guy invented Zoom basically long before COVID. And like we just never visit each other unless it's like a true emergency. And a machine that is fully intelligent and is infinitely capable is there to do everything from bring us food to do medical exams when we don't feel so good and so on. And so it's just this very very dark and it ends badly. I don't want to spoil a surprise but it's not nice. And that story is like that's a future where we surrender more and more control to AI and say like we're not worthy. Feed us, clothe us. We just need to tell you what we want and like descend into this babylike experience where we just like get diapers and food from.

No, I don't think so. I think we're actually headed a much happier future where we are telling the machine, I understand what you can do and I am going to make you better and smarter and I'm going to guide you to the outcome that I want and I will be the judge whether you're successful or not and we're going to co-invent recipes to feed me and we're going to co-design clothes that you're going to make for me right in front of me just like the 3D print revolution never really took off because only very very very sophisticated sort of engineering minds could get into the CAD that you needed to get the designs or the stuff to print then you need to really understand how to operate these 3D printers and that was supposed to be the consumer make everything in their home with AI the interfaces are smarter the machines will understand us better but we will still be co-designers like the future I want the future we're working towards the future that does not look like The Machine Stops is one where we are getting 10x superpowers in everything including the things that we don't fully grasp the underlying realities of but have a good sense for theory of how these things work I think clothing designers are still going to make the best clothing with AI. Software engineers are still going to make the best software with AI. Everything is going to be done with the help and the amplification power of AI, but you'll still have specialization. Like you will make the best podcasts with AI giving you real time help, but you're still going to...

Exactly. But like you will use more and it's going to be even better.

I am. I wouldn't be there if I didn't love it. I'm definitely enjoying it tremendously. I think probably the most interesting stuff we're going to start seeing, we're seeing a little bit of it already, so maybe it's not the most interesting stuff. The most interesting stuff is something that no one's seen yet. But the things that I'm quite blown away by is getting multiple flavors of models helping each other get better. So AI as judge of other AI output is like very much a thing. It's happening everywhere. It's really really cool. The thing that I'm quite excited about that we're certainly working on at Affirm and in my copious spare time building agents that are uniquely taught to do a thing better. The general purpose LLMs that sort of teach you one of everything and kind of have greater depth than many topics that you and I can have just through sheer limitations of memory is still like that much deeper. The thing that's going to be amazing is when you have embodied AI. So you have robots that actually have really good grasp of the world around us and are taught to be like the best manufacturing robots we could ever possibly imagine. You're still going to need to help them decide what to build. Like they're not going to know the next thing they're going to build on their own. But you'll be able to take advantage of both software and hardware agents. They're like so much deeper than any human than any generic LLM or generic technology out there right now. And which is also why I think like this whole like oh you know the jobs are just going to go slowly dwindle down to zero is total insanity. Like there's going to be so many humans needed to train them and to guide them and to co-design with them. Like your job is safe.

I think on my memory a similar moment was when the web emerged. So I had the extreme good luck of matriculating at University of Illinois in the fall of '93. And that summer I was still chatting on IRC and direct messaging on ICQ and emailing from Pine, like all the green screen tools that you may be too young to remember, but as young as you think, but yes, I was there and it was great. I thought the internet was the most amazing thing ever. I could talk to people from across the world and it were like, you know, 33 milliseconds, you could be coast to coast. And then the browser came out on my campus thanks to Marc Andreessen. And it was like, oh my god, it's not always going to be just like the green screen of my mom's generation. She was a software engineer. She's now retired, but she spent her entire career writing code on green screens that were like mostly Fortran. And you know, I mean, and that was very exciting to me when I was much younger. But I was like there's you know, only one color, green or not green, $2. And the idea that you could have these amazing very visual experiences with people around the world with tools that are built in real time in these pages that like you click and it becomes the next thing. You don't have to wait like everything suddenly came together and it's like the world's going to change and it's going to be like TV meets mainframes. That's what I felt like that's how I explained to myself when I was starting college majoring in computer science at the place where this was invented at the time where it was invented was like the best thing ever happened to me obviously but also just like an unbelievable moment of like the world is changing in front of me and I can have my fingers in it if I want it. And then right after I graduated, I moved to Palo Alto where it was being commercialized. Like that moment in time like I would go to Tresidder Union or outside at Stanford where kids would eat lunch and people who were kids like 10 minutes before maybe have graduated Stanford or dropped out were pitching VCs that drove down from Sand Hill Road to talk to them and like literally explaining, oh well, let me tell you about what an IP address means. And so that moment was just like the whirlwind moment where everything changed and it was so exciting and you would run into someone and be like, 'Oh, you're not working on a startup. What's wrong with you? Everyone's working on a startup. It's so cool. Get going.' And this is what it feels like right now in AI but just AI adjacent world where like every one of us and again that's why it's so cool to understand how these things work.

So you know the famous line, future's already here, it's just not evenly distributed. It does feel that way. It's definitely the case when you talk to people who are not in the soup right now. It feels a lot like 1993-94 where I would tell someone like have you seen the latest version of Mosaic the very first graphical browser and like what's that is that like Windows I can have it there too like so it's not quite here for everyone yet but that's where the opportunities are. The reason there's so much venture capital being raised, the reason there's so much funding events going on is because it's very clear it's going to benefit everyone, but we have to do the work of bringing it to everyone who could benefit from it. And so I wouldn't expect to wait too long. I think you'll see it quite quickly in a lot of different places, but physical world is complicated. You know, the great thing about the internet is it's not that hard to bend to your will. So long as you're willing to write code or you have an agent to write code for you. To get the food cart downstairs to fully benefit from the AI revolution, there's going to be some atom movement that's going to be a little bit more complicated.

Yeah, sometimes it's got pretty cool. It is fun. When she retired fairly recently, actually, she coded basically non-stop for 30 but like more like 40 years.