Ep 147: Scott Wu & Russell Kaplan on the New Era of Software Abundance

you're really only limited to your ideas and to your imagination, where you can kind of just turn things into reality. - You were the three time gold medalist of the IOI top programming competition in the world now you're running one of the top AI companies here. - When we launched Devon in March of '24, it was the first autonomous agent. One hour of human time spent managing Devon was worth like six to 12 hours of that human time doing the work themselves. Elon had this phrase that he really drilled into us which is everyone is chief engineer. - Let's talk about this new era of software abundance. - For us at cognition, for example, our engineers don't type code anymore. - You really can just turn your ideas into reality. - The engineer and the designer and the product manager all look at each other and say, "I don't need you guys anymore." (upbeat music) - Scott Wu was a three time global gold medalist in programming. I worked with him in the past. He's now running one of the fastest growing AI companies in the world, helping to usher in this era of software abundance. He and Russell co-founders met up with us. We played some games, not gonna tell you who won. These are pretty smart guys, but it's always really interesting to hear from Scott and Russell about the cutting edge of AI, how the world's changing, and what we can create in this new era of abundance. Scott and Russell built Devon. It was a very first AI programming agent two years ago. They're already launching in all sorts of other areas. They're now transforming how governments work as well, excited to see where they're headed next. - Welcome to American Optimist. We have back Scott Wu, the CEO and founder of cognition and your co-founder, Russell. And through mine, people, Scott, you were the three time gold medalist of the IOI, a top programming competition in the world. I think you were the one time world champion at 17. And we worked together at Adapar after that. And now you're running one of the top AI companies here in the world. Russell, I think you started your career at Tesla as an ML engineer. You sold a company to scale. You guys are both in your 20s, still, right? - No, I'm a shout now. - You're 30. - Yeah, I turned 30 this year. - It's hard to say. - So you're turning 30, okay. Well, that's okay. You're getting old like me. You're still writing the heart of it here. What's the average age on the team? Actually, I'm curious. - I think it's probably, so on engineering, it's about 25. And then obviously on go-to-market, it's a little bit older. But we, yeah. - Well go-to-market is different. That's fair. You're running more go-to-market stuff. - Yeah, I think the engineering team, we have 17 year olds. We have 18 year olds. We have really young folks. And then, but we'll take anyone at any age as long as they're ready to grind and ready to have a big impact. - And you guys have like huge numbers of people with one gold medal globally in programming. This is a very advanced technical team here cutting edge of AI. - Yeah, I mean, some of our favorite people, I would say are people who are like 17 or 18 and like finishing a high school, but they had already played around with a ton of building agents themselves, like working with AI, training models and so on. And it's obviously, I mean, it's-- - I actually want to ask you about this really briefly because so you were gold medals in the world of 15, world champion of 17. It's obvious people can be really, really good at these things at a young age. Is there something about AI where like a young person's brain that kind of grows up in forms, using it can somehow like be more ahead of anyone who does it or something like that? How do you think about that? - It's a good question. Yeah, no, I mean, it's fun enough. So I went to what's called the USACO, the USACO, which is like the USA Computing Olympiad. And from there, that's like the training camps and all the selection camps that choose the national team that go represent at the international impact. And every year, there were about 20 kids, there was like the 12, 20 kids around the US. I was from Louisiana, most people were from like, California or like New York or like around like Massachusetts, like around MIT and Harvard and stuff like that. But in my year, actually, there were a ton of others who all kind of went into AI. And so obviously, Steven and Andrew, who started the company, who started cognition with us. But also a ton of others. And so Alexander Wang, who started scale, Demi Guil, who started Pika, the, let's see who else. Daniel Ziegler, who is one of the co-inventors of RLHF, Alex Wei, who is like now running a lot of the, the reasoning efforts at OpenAI, Johnny Ho, who started Proplexity. So we were all the same year, actually, out of that like group of 20 people. And it was kind of an interesting one. I mean, I think there are a few things there. I think for one, obviously, I think entrepreneurship is infectious. You know, and I think, I mean, Alexander was, I would say the first to really start a company and to see real success with the company. He left freshman year from college to start scale. And we see a lot of scale, obviously, for like 16 million to face, or whatever, some funny structure. But yeah, but then, probably inspired other people who would say, wait, I'm not smart too. I can do this too. Yeah, so I think that was a big motive. And then, you know, we all kind of like came up together and kind of got to go through some of these things together. And I think that was a big deal. I think the other thing I would say about AI particularly is I think in AI, what you see is that really excelling on the technical aspects just matters much more, I think, in AI than some of these other fields. And I think there's been lots and lots of businesses in the past that have been very, I'll say, like very intense logistics businesses or very tough kind of like marketplaces to get started. Or for example, businesses where a lot of your edges, just like how you figure out pure distribution or how you kind of like make the right little addicting loop and so on. And AI has a ton of these too, obviously. And then, I think all of those same skills are still necessary. However, I think in AI, in a lot of factors and a lot of these verticals, what you see is that, obviously, pure technical execution, it's like for every level that you push it, there's still like another level to go and hit. And a lot of the best companies that we see in the valley are the ones that are just able to roll out, like technology pushes or break throughs that others have not. - And to push you guys on this, there's a 17 year old today who's like the Scott Wu of today who's a world champion, who's maybe you're hiring. Do they have some special edge having grown up in this world where AI is already possible whether you're using it, like is it exciting things further? - Yeah, I mean, everyone starts with the same level of experience with AI for software engineering, right? Which is basically none. I mean, every three months, you have to throw out your previous experience and build new experience because the tools get so much better. - This is probably harder for someone who's my age, I'm 43 than someone who's like 18 is still learning or not. - Yeah, so because some people say, oh, you know, it's gonna be really hard for junior engineers now because, you know, the entry level tasks are being done automatically by AI. But I think a lot of what we see internally, it's kind of the opposite in some way where if you're coming in with no preconceptions about how things are supposed to be done or how things are supposed to work, then you can just go all in just really embracing this completely new way of working. But I think the AI technical depth pieces, it's actually not just in the sort of modern generative AI era, you know, when I was at autopilot, I was a machine learning scientist working on the sort of the vision neural network. And Elon had this phrase that he really drilled into us, which is, you know, everyone is chief engineer. You know, everyone on the autopilot team has to understand how the full stack worked. And this is actually extra important in AI because what happens is the abstraction boundaries between different teams start to break down. You know, the sort of classical way that the self-driving system worked, which you had a, you know, a perception team, you had a planning team, you had a controls team and they had these like thin interface boundaries between them. But the nice thing about AI is you can optimize systems end to end. So if you want to actually optimize systems end to end, you have to have an accurate mental model about each of those pieces work. And so I think more sort of technical breadth than depth across the entire stack is becoming increasingly relevant. >> That is an interesting kind of way Elon does things, which I've seen a lot of really top people, not too many, but some people do things. In order to really be the best, you have to understand everything going on. So it's really breadth and depth in a way. And so you're saying, "Hey, I make that a lot easier to do that 'cause I can give you some of that breadth you wouldn't have otherwise." >> Yeah, I mean, like the way we onboard onto our own code base for new people, you just asked Devon all the questions of what's going on, why is this done this way, what's the historical context? >> Do you tell them they're also like equivalent of like the chief engineer where they have to learn everything? Or it seems like it's a take a while to learn, I mean, he's a big code base now. >> Yeah, I mean, I think in practice, it's so much of it is all obviously very connected. And so I mean, a simple example of this is like, we bought Windsor, if you know, seven, eight months ago at this point. But like, we don't have a distinction of like, oh, this person is an engineer working on Devon or this person is engineer working on Windsor. Like, there are a lot of the same people, should be people who are kind of, you know, working across both of these. >> So, catch us up since we last talked. Like, like, what's the state of cognition? Where are we now? You're probably not giving out revenue numbers, but you're drawing a lot. Like, what can you tell us? >> Yeah, no, it's, I mean, it's, we've had a ton of growth over the last, I guess it's just under a year since we last talked. And obviously, you know, back in July, we bought Windsor, but I think over the last several months, I think both Devon and Windsor have grown exponentially. One of the fun stats actually is today's March 9th. And we actually, at this point, have already done more Devon sessions in our customers in 2026 than we had in 2025. And so basically over the last like two months in change, we've already done more Devon usage in total than we had in all of 2025. >> So, it's more than six, six X or something. >> Yeah, yeah, and obviously, you know, we're working on making sure that that growth trend continues. And so we'll see how that goes. But, but no, I mean, I think the business has grown a lot. We've been working with a lot of the biggest companies in the world, you know, Citibank, Santander, and so on, on figuring out how we really, you know, transform their engineering efforts. >> Yeah, one of the interesting developments since last year is, you know, when we launched Devon in March of '24, it was the first autonomous agent, right? It was like very early for the form factor. And it was, I would describe it as kind of like just at the edge of possible, then in terms of doing, in terms of doing real useful work. >> It made a lot more mistakes back then, obviously. >> Yeah, yeah, as much less reliable, you know, our infrastructure around it, the connectivity with the rest of the code base was a lot less mature. I mean, it took us until summer of 2024 for Devon to become the number one contributor to its own code base, like which was like the first real milestone. And then towards the sort of the end of 2024, to really get deployed in production at large scale at meaningful companies. But one of the things we learned is that if you look at sort of where the technology was, in 2024, it was not reliable enough to be used as the primary source of software engineering for most tasks. You actually still need a tighter feedback loop in that year between AI and humans for most things. So one of the first niches we found where this was actually already really useful back then was in these very large code bases that just have tons of existing technical data complexity and you want to do a large-scale refactor across 10,000 services. If you're doing it the sort of normal way as a human engineer, you might have to define some new architecture and then manually go implement thousands and thousands of changes. And these changes would be complex enough that you couldn't just write a rejects, but not so complex that AI couldn't help. And so that's kind of one of the earliest places I think we got to a real strong product market fit is inside these very large complex code bases. And that's one of the reasons now if you look at the state of our business, we're deployed at a lot of the largest most complex organizations in the world, most of the top health insurers, retailers, banks, government agencies. And these places with large complex amounts of code have been actually surprisingly early adopters of the Genetics. Well, these guys just have massive amounts of code that's been doing the same thing for 30 years on a very old architecture and you can go in there and pretty easily accurately fix that, I guess. Totally. It's like to your point on is this a new skill for engineering? It's the mindset of it architect inside one of those organizations has become, you're almost like a CTO of an agent fleet. Now, you define the problem space of what you want to go solve and then you just spin up your army of devins to actually go do the implementation work and then you kind of review the results. It's a very different way of programming. There's all these memes in San Francisco where like nerdy guys are on dates, but they're too distracted watching their fleet of agents and to talk to the girl. It's like a problem. I have gotten in trouble with my wife for that. Let's talk about this new era software abundance as you call it. What does software abundance mean? What's cognitions role there? Yeah, I think the simple way to put it is, to Russell's point, all of these traditional industries, you know, it's in Silicon Valley the way we think about software engineers is obviously these tech startups or these tech companies that are building. The reality is every company in the world has software, right? I mean software is in many ways, I think the premier knowledge work job of this century. And so you're talking about like CBS or you're talking about Walmart or you're talking about UHG or you're talking about Goldman Sachs. All of these places have so many software engineers and they have so much software that they're building because obviously so much of what we do in the world now happens over the web or yeah, with computers, right? And I think what we kind of think of when we think of software abundance is just getting to a point where you really can just turn your ideas into reality and build what you want to build. And one way that I like to think about this is you can kind of think about software products on a scale of like let's say a log scale based on how much reach they have or how many users they have, right? You think about the best products in the world and the products that everybody uses all the time and this is like, you know, YouTube or Instagram or TikTok or something like that and it's, you know, these are incredibly good products. And it makes sense. I mean, they have billions of users and so they have, you know, 100,000 software engineers or tens of thousands of software engineers that are working on them, right? And you feel it in the product, you know, it's like the experience is perfect. Like there's never bugs. It doesn't go down. It's streaming you like gigabytes of data. The algorithm is super addicting, right? Like they've, they've really like perfected the software. I always use two to three you mentioned. They're too addictive. Yeah, they're too addictive, right? And then you go to the next tier of like, okay, instead of billions of users, let's talk about the products that have hundreds of millions of users, right? And you're thinking about like, you know, banks or you're thinking about like, you know, apps like Uber or DoorDash or you're thinking about like, you know, a lot of these various other kind of services and products that we all use, right? And similarly, it's like, you know, you feel the software and it's obviously built quite well. But it's already, you know, I think you notice a different level of like how much execution there is. And then there's next level and next level and next level and it goes all the way down to it. You know, you're like, you know, your, your, your kids website, a school website, which is from like 2001 or something. And it's like an elementary school and it has like a picture and it's like super outdated and has no other information about that. Right. And so like maybe one way to put this is, you know, I think software abundance means making it much easier for everyone with every idea or every product or use case that they want to serve to be able to climb that ladder and build products as well as the best products in the world are built right now. And for someone for a lot of listeners or not as in the air world, they might be CEOs of a bank or something or running other businesses like, like what is actually changing about how engineers work, right? Like what does a great engineer's workflow look like now? Yeah. It's a great question. And the simple way that I put this is that for us at cognition, for example, our engineers don't type code anymore. Like that's just reality. And this is the hours of the last several months. And this is within the last, yeah, three to six months, honestly, when the shifts has really happened. And there are other steps to be clear, obviously, but I think at this point, you know, you used to maybe one way to put it is like, I mean, you used to work with punch cards, for example, and you used to work. And now in many ways, the medium has shifted away from code and a lot more of it has become basically English, right? And so, so, you know, we obviously use a ton of Devon internally. We use, you know, Windsorf and then the agents inside Windsorf internally as well. But at this point, either way, whatever tools you're using, you know, it's not really you typing out the lines of code yourself. It's you looking, understanding what it is that you want to do, thinking about, okay, how do I want to handle this case or this behavior? And you just tell the agent what you want it to do in English, right? And it goes and goes. Yeah. And in terms of then, like what the impact of that is, especially, you know, if it's like a CEO level objective, what we're seeing across the board, across the customer base, is people are just getting more ambitious. You know, like it used to be the case that you have this entire software development life cycle. And every step of that cycle is oriented around not wasting the super precious time of engineers writing code. And now you have suddenly this overflowing abundance of the ability to generate code from ideas from prompts. So you can iterate faster in ideas to use tri stuff more. You can try stuff a lot faster. You know, you don't necessarily need to spend three weeks on design before you then handed off to engineering because engineering is so fast that the turn, you know, the cycle time is much tighter. And so, you know, increasingly, especially at the executive level, it's like, oh, do we have to choose between A or B? Let's just try both. I mean, could the product people themselves create some things now or how does that work? 100%. I mean, we see it every, there's like a joke where it's sort of, you know, the, the, the engineer and the designer and the product manager all look at each other and say, I don't need you guys anymore. I mean, is there all they're all just doing it all themselves, right? It's like every person is empowered to do the other aspects of the product development life cycle. And I think it's really rewarding people who are, you know, actually personally highly agentech and thinking about, okay, what's the impact I can go have and be sort of like self relying that way. And I know, you know, with Devon, a lot of product managers, one of the very first things they started using Devon for was actually to not bother the engineers with questions with silly questions, you know, how often have, you know, you're a new employee to company, you don't understand what's going on somewhere and you're a little nervous to say, hey, can you explain this to me? You know, Devon is very non-judgmental. You know, you'll ask a question. You get the answer. I actually ask it a lot. Then question is, yeah, it's because I'm the boss in Switzerland. It's something like crap. What is that acronym you mean again? But you know, in judges, you do it right away. Totally. Yeah. And you'll cite the, it will sort of actually cite the source code alongside it too. And, you know, give it kind of puts everyone actually more on the same playing field in a way it wherever it has the same content. And by the way, agency something I'm thinking a lot about for society recently, it seems like highly agentech people. This is like a big advantage for everyone, right? But along those lines, like what are the habits or instincts of someone who's going to be particularly good at leveraging Devon? Like, like, other certain things you're seeing? Yeah, yeah, for sure. I mean, one way to put it is, you know, I think software engineering for people who've grown up doing programming and so on have been through all these previous errors. You know, usually what software engineering has looked like is like, why do people who love coding love coding, right? And I think the answer is like 10% of that job is basically this really fun part of just pure problem solving, thinking about what you want to build, being creative, like understanding the different solutions, deciding, okay, what architecture makes sense here. How exactly am I going to achieve my goals here? What do I want to build? Right? And then 90% of the job is once you've figured out all of those parts, just doing all the dirty work of the implementation to go make that happen, right? And there's like a million bugs that your customers have reported for you and you have to deal with this messy migration or this upgrade to make your stuff still work or like, you have to go implement all the little cases and all the little details and like write all the, you know, the front-end code that serves this thing that you just built, right? And I think what we're seeing is that the best engineers are just doing 10 times more of the first part because you don't have to do that 90% right? You have an agent, you have Devon that's going to go and do that for you. And I think to your point, that obviously means that, you know, having high agency is really, really important. We think about this ourselves internally. It's one of the jokes is like, you know, we're a reasoning lab building agents and so the things that we really value are agency and reason. And I think it's to your point, like a lot of the skills that really matter, you know, are you the kind of person that's going to think about, okay, this should be this way instead of that way or like, what is the right way to solve this problem or what do we want to do here? And also, and also just like internally embracing the abundance mindset to where, you know, a lot of times you might be in your head, oh, should we do it this way or that way? I think a lot of our best engineers, they just rip it all of the way at the same time. And then, you know, you get a bunch of Devons come back and then you can sort of analyze the results and try and parallel. It reminds me again, it's kind of like the mindset of actually being a good machine learning researcher is now relevant for every type of software engineering. You know, if you look back a few years ago, what were machine learning researchers doing? So, you know, at Tesla, we had one mantra, which was, you know, never go to sleep while the GPUs are idling. You know, if you let your cluster idle overnight, that's just a huge waste of resources. Just kick off some experiments, you know, before you go to bed, so you wake up, you get some more data. And it's the same for all of software. Now, like, why would you go to sleep while the devons are idling, right? You could just be ripping a ton, you could be ripping a ton of these. working on your sleeping. Exactly. Exactly. And so I think there's actually a lot of parallel in my head where it's not just about that, it's also some comfort with non-determinism. A lot of engineering, it's in incredibly precise craft and it's naturally uncomfortable and natural to not have full control over everything that's going on. But if you get a little bit more willing to embrace the non-determinism and exactly, how should this be implemented as long as we can validate system performance end to end and I can actually understand the results. I think machine learning went through that lesson years ago and now we're going through it for the rest of the software. I mean, everyone's in all long while someone would probably check the machine learning to machine, sorry, so much like the machine assembly code actually on something. If they're really trying to optimize something, do people, there's still people checking the actual regular code on things? If they really have to these days, how do you think about that? Yeah, for sure. And for what it's worth, I think we are still in the midst of a lot of this change. And so I think while people are producing code with just English, often you do come back and you're still reviewing that code, you're making sure that it looks right or you're looking at the code as it is to understand what's going on. And I think to your point, it is kind of like in the right cases when you want to peel back the layer of abstraction. And I think what we'll see over the next 12, 18 months is that we will continue to get further and further to this point where we do have to use English for almost everything. You don't have to do any more. And what does this mean? So obviously a lot's happening a lot faster. We just covered this. And this is money movement, this is military, this is healthcare, it's flights, it's maintenance, scheduling, building new things, permitting. Everything in the world, a lot more people realize is software. And we're going now, what, five times faster, 10 times faster, soon on software. There's almost like we're getting multiple years done in one year, right? So what does that mean for society? Yeah, I think about this a lot. Have you ever seen the graph of inflation by different sectors? And you can see the sort of highly regulated constraints sectors at the top, like tuition, healthcare spending. The cost of these sectors are getting everything else. Right. And then the plasma screen TVs are just way down. Jeff, please taught us what to do. Exactly. Yeah, I kind of think that that chart is going to happen for like all of society. Basically, and in particular, if you think about where is software, we're going to enter this like hyper-deflationary cycle of software where it's so easy to build, it's so abundant, that doesn't mean that people are going to stop producing software. They're actually going to produce thousands of times more software, right? And so you're going to have this total software abundance, but basically any problem that's kind of solvable in the digital realm should just get solved in the digital realm. And so I think what you're left behind with is, now how do we direct a lot of this new energy to going improving the physical world, going improving everything up? You actually have to do things with real operations, real people in the real world. Now I guess robots could change that too. But for now anyway, with the robot society, the returns to actually do things in the real world should go up by comparison because it's harder to make those cheaper, right? Yeah, we think about that a lot at cognition of how do we not just improve software for software's sake, but how do those improvements basically translate into real world benefits for everyone? Ultimately, these things are tools and service of our own lives, our companies, our businesses and our livelihood. The real world's messy and you have to interrupt them. It's super messy and I think a lot of engineers fall into the trap of, oh, I want to solve this very pure problem, you know? There's nothing wrong. It's actually really fun to solve. A lot of our team, they spend their entire programming careers just optimizing their ability to solve the purest algorithmic programming problems in the world. But it's an interesting contrast to one of the things that differentiates cognition in the market right now. What we keep hearing from customers is we have this forward to plan engineering team that will go partner really deeply with large organizations and say, hey, it's not just about the tools we're delivering, but how do we get into the weeds together to actually go drive big structural changes? Tell me a bit more about that. You guys have obviously deployed with a lot of the biggest companies in the world. What are some of the more impressive results you guys have seen in the wild? Yeah, so the earliest results that we, where we saw, we were like, oh, there's something here where they started with basically like modernization programs. So people that had large legacy existing things, they needed to transform. And if you sort of did the math, scope out how long it would take, maybe it would be like a two year project. And, you know, relatively quickly by like late 2024, we were measuring, you know, somewhere between a six to 12 X productivity gain for those types of projects, meaning that, you know, one hour of human time spent managing Devon was worth like six to 12 hours of that human time doing the work themselves. And so that was a big, that was a big early result. We started doing lots of, lots of engagements where our customers would use, it would use Devon to just refactor, migrate modernization, large systems. Now the interesting trend that we're seeing is a shift from really reactive to proactive engineering work. So you know, if you think of the early days of the internet, most of the packets that were sent on the internet, it was like a human clicking a button or visiting a link or initiating some requests. And then at some point it totally flipped and now most of the packets are initiated by machines, talking to other machines. And I think we're now seeing the sort of the flippening there for software itself, which is the process of deciding what code to write. It used to be entirely a human scope thing. Now we have people, you know, we have people wiring up Devon to all sorts of events and alerts inside their organizations to say, hey, let's just start the engineering work right away when something happens. I'll give you one example. One of the largest regulated firms in the world, they do very thorough sort of security vulnerability scanning on their code. And they make sure to understand, oh, is this potentially insecure here or there? And there's great existing tools for scanning that, you know, so in our cube, Vera code, Sneak, there's this whole, there's this whole market of tools. And what they did was they hooked up all of those alerts they were getting from those tools. And they just started piping them to Devon saying, Devon, can you take the first pass triage? Because if you're a human engineer at a company drowning in these alerts, it's not, it's honestly not a really fun part of your job. And they're remediating 70% of these automatically now in production with Devon. And so it's really flipping the, you know, kind of flipping that. I've been thinking a little bit about the pyramids with regard to this lately, whereas like, it's all as impossible to see how people with like no tools at all built the pyramids. And it does seem amazing if you look at the software that we have built up until like a couple of years ago that all this exists, all by people. When you, and then, yeah, I feel like that would be like something you'd look at 20 years from now. And it just uses it. Yeah. I don't know if you've ever built all of that. I don't know how to imagine that. It seems like totally impossible. Yeah. No, I think to your point too, I mean, I think there's sometimes a question about, you know, we're talking about, okay, this is going to get way easier to build. It's going to be way cheaper to build these things. And so then what happens? And I think the reality is like, we have so much more to do. I mean, I just think about this even like today, you know, it's like you wake up, you're like, you know, you're like, let's say you're like logging in to check your medical records. And it's like, not a very good example. You're like logging into your bank. It's like not very good. You're like, so much to fix. I just think so much to fix. So many of these things. And to your point, I think the reality is there's so much more that we can do with software. I mean, I think some of the things that people started talking more about, which is I think what we're going to get to over the next little bit are things like even like generative UIs or like single U software. Right. If you get to a point where, you know, so much of the work that you want to do is it can be done in code. It just doesn't make sense to like write code for a single, you know, for something that you're only going to do one time or even something that you're only going to do like 10 or 20 times. I love the AI can actually write code for this instance coming up and give you the right UI for what's going on. Exactly the right thing. It builds it. Yeah. And so this is I mean, people talk about Jevons Paradox. And I think I mean, in software, it is perhaps, I think there's nowhere that it is more true than in software, right? As we have produced more and more and as it's gotten easier, the reality is that actually demand has only gone up. That's actually a really fun idea. Like right now luxury is like, it's like Vakunya. It's really nice. It's stitching is right. What if luxury was like, it's a perfect UI for you. It's for this instance in your life where you happen to need it, right? Yeah, either. No, either that or actually luxury might flip the other side because artisanal handcrafted software is going to be so rare. So terrible. It's like, you know, the in the early days of the industrial revolution was like, Oh great, mechanized, mechanized labor for goods. Like that was the higher status good because of course the machines were more precise. The bags were more, you know, were more perfectly made. And now it's totally flipped if something's handmade that's obviously much higher status because it's so rare. How could you expend the resources for that? I think site was handmade. No, you can tell because of the little bugs. And I think we're actually going to see that in the software. Yeah, it's like handcrafted artisanal code. It's like the homage except that there's a soft. Yeah, yeah. My hands. It's like a 2020 society frozen in time. People making you handmade because that's very silly. So speaking of speaking of like things that are broken and that there's like infinite needs effects like government to me is like one of those areas where I think if you like photograph everything in terms of getting better or worse and efficient or less efficient is this like probably unfortunately comes off as like really, really messed up. And I've always started to hear some pretty interesting things in government. My friend Jared Kushner who's obviously you know been involved in these administrations. He worked with Qatar with Elad and I and they built something where the permits only take 120 minutes there. So if you want a permit to build something like a back to you in two hours, it's pretty cool. That's like there's always ways government could do things better with software. You guys just launched cognition for government like what's the goal with that? What's going on? Yeah, I mean, I think in a high level, I mean we talk about places where there are so much more software to build so many things to fix. And I mean, I government is is an obvious example of that in all these departments. Similarly, you know, things that we used to do by hand, not even that long ago, honestly, I mean, 20, 30 years ago, like lots of these things were done by hand at the treasury. Now so much of that is software and yet so much of that software obviously still has such a such a long way to go. And I think from our perspective, you know, when we think about how do we make sure that the US stays in pace with what it needs to be? How do we make sure that, you know, the break throughs that are coming through an AI in the private sector or also coming through in the public sector? It's something that's a really important problem for us to work with. I also think there's something poetic about it because I think people don't appreciate this, but the government is really responsible for a lot of modern technical innovation. You know, going back decades and centuries, Scott drew the analogy earlier to, you know, punch cards. And I think I think people don't appreciate this, but really the first wide scale production use of punch cards was the 1890 census. In the 1880s, they did the census by hand. They tallied it. It took like seven years, and they were doing the math, and they realized that 1890, if they did the census the same way, it was going to be longer than 10 years. And the census every 10 years. So they were screwed. And so the government basically put out this call for technology help. Can we solve this problem with technology? There was a guy by the name of Hollerith, who he invented, but later became the Hollerith machine to use punch cards for actually running the 1890 census. And it was on time. It was under budget. And it kind of kickstarted-- actually, a lot of modern summer, where I studied at Stanford, a lot of that-- the Silicon Valley ecosystem was actually really invested in by the government back decades ago, partially for defense. And now we're in this sort of this flipping point, even by the self-driving. I mean, the DARPA doesn't get enough credit, I think, for really kickstarting the self-driving revolution, with the DARPA grand challenge. And now we're kind of in this state of the world where the government spends $100 billion a year on just IT modernization. We have huge amounts of governments still written in cobalt with people who have left or no longer understand how the code works. And it's really holding us back. And I mean, you feel it as a citizen day to day. Think about your interaction with the DMV or trying to pay your taxes. Or waiting for a permit to come for-- Or waiting for a permit? Yeah, I mean, these things have real world consequences. And that's one of the things I'm personally really excited for is, you can bring technology like this across the private and public sector. It's like a great equalizer for entering work. Can you deal with the cobalt stuff? Is that something, David? Cobalt is actually a massive use case for us. Yeah, one of the early best and investments we made was in what we call code-based intelligence. So if you think about a modern language model, there's a limited context window, right? But a lot of the largest organizations in the world, they have very, very complex code bases that do not fit inside a single context window. We've done a lot of work, both on the model training on our L side, but also on the sort of harness engineering around that and the indexing around that to figure out how do you work with really messy custom languages. And so cobalt, it's actually not even the hardest one for us. One of the examples I like is Goldman Sachs has invented some of their own programming languages. People don't know this. But Goldman has a pretty insane internal engineering team. And they've literally written their own programming languages. And they've been able to customize and tune Deb into work on those internal languages too. So cobalt is actually easier in some sense than that. So in the government, I mean, they spend 100 billion a year in IT. It's ironic because you're right. Sometimes government in the past, especially when innovation was really expensive, they pushed some new things that otherwise wouldn't have happened. Today, most $100 billion is spent on special interest. They don't seem to be using the money well. So it's a giant mess. What are the types of projects you're working on? Totally. I mean, the incentives are obviously super screwed up for a bunch of reasons that your listeners are probably familiar to. One of the less known ones that I think we actually might be able to just sidestep is the government is a really unique buyer of software for a bunch of reasons. But one of them is that a lot of times they want to own the IP of the software they're using. And this is a really big implications for most SaaS businesses. If you make scheduling software and your business is a SaaS business, you don't want the government to own your IP. You want them to have a license to it to use it for scheduling. And actually, that desire is literally incompatible with how a lot of government contracting has worked and happened historically. So you end up in this situation where the government says, oh, this scheduling provider is a real example. This scheduling provider that has great SaaS that can do scheduling, I can't use it because I wouldn't own the IP. So I have to go work with the systems integrator and completely custom build my own, right? Insane. Now, what we could-- could we lobby and go try to convince the government to change their policies? Yes. But actually, I think easier for us to just sidestep the problem and say, look, Devon can just write this thing for you. You know, this bill is really fast. Just bill for you. We're actually because of AI agents. You're in the regime where actually now everyone can own their own IP more easily than before. So I think that's one of the ways we're sort of trying to sidestep some of this. Can you be your own government contractor? Are you going to have your own government contract doing this thing? Are you just going to power others? How are you thinking about it? Right now, we have dozens of kind of fed ramps deployments of cognition, both with agencies and with primes. You know, we work with US Army, US Navy, the treasury we work with, folks like Palantir, with Andrew, and other primes. And so we just want to build sort of the most capable, most useful, agentex software and engineering platform. And then work like heck to kind of get in the hands of people to make it useful. And in Cartier's government, this is a fast growing business for you then right now. I would say government is one of those things that it really takes some time to kind of build and be compliant and work in the way that people want to work. And then once you're there, once you're there, it's much easier to be helpful. And so we've over the past year, we've really done a lot of the light work to, you know, how do you get your fed ramp certification? How do you understand the needs of these agencies, which are actually pretty different in a lot of ways. Again, just in the civilian sector, these, they're operating in a completely different trade office, right? A lot of the software they use is actually by statute not allowed for them to write themselves. Talk about regulatory capture and intrad, you know, there are literally laws saying you government agency are not allowed to maintain your own website. You have to bid this out to a contract. That's true. And so it kind of sounds insane, but you know, that's the way the system works. And I guess my experience working on AI with technology is that a lot of times it's actually easier to solve like a frontier science or engineering problem than it is to sort of change the molasses of the existing world, right? When we were working on self-driving, a Tesla, a lot of folks would ask us, hey, like, why don't you make the cars talk to each other? Wouldn't that be way easier for self-driving if all the cars could talk to themselves? And the answer is, yeah, it totally would. If you got every car talking to itself, but until then you're going to have to deal with human driven cars. And so then you have to actually solve the much harder science problem of predicting the motion of all those human cars. I think it's the same for our government. So yeah, I mean, right now, we've dozens of these deployments, you know, I think folks are giving us really good feedback on how much it's accelerating their work. And some of the missions are really exciting, you know, think of an organization like NASA JPL. Who doesn't want to help us get back to space faster? - I love it. Yeah, we just actually interviewed Gerard Isaac van who's running NASA. So it's a great guy to partner with. It is true in government. A lot of times people say, well, this solution would work. And we just put this thing in the middle and made everything talk to it. And I'm like, sure, let's put everyone always try to see the government. But these guys are never going to all work together. They're all going to debate. And so you actually have to design it, knowing it's distributed. It's very interesting. The real world's messy and you deal with it as it is. - Yeah, it's like the XKCD comic. Like we have a dozen standards. Like this is a mess. We need one more. Now you know, you have to get one more. - Yeah, exactly. So, so what I, it's very honorable to go work in government. America needs that you're fixing it. Obviously you're growing even faster in the enterprise in general. So these are both big businesses. This is obviously a very competitive time right now. There's a lot of the smartest people in the world. You have a lot of them here. There's some of them in other places as well. I think very famously, some of the big labs like in Thropic have been, I think they've doubled into the tens of billions in the last few months. It's obviously the highest growth thing right now in the general area. You guys are obviously, you know, don't say you're exactly revenue, but you're likely to get into the billions soon if you're not already there. What's the competition? Like do people use that with you? Is it help you when they do well? Like how do you think about these things? - Yeah, yeah, for sure. So no, I mean, it's an exciting one obviously. And software engineering and code is just so big that I think there's a ton to do. I mean, we've seen a ton of growth as well. I mean, our usage, for example, of Devon and our customers has, I think, roughly tripled already since the start of this year. So just the cost of, inside the customers alone, if the customer is alone, that's right. But what I would say, I think a couple thoughts here. One, again, there's so much different work to do and code. And so you'll see a lot of others who are, for example, building products that will help you make a quick little website or something like that or build something like fun, which is, you know, you can use Devon for that, but I think that's not necessarily what we specialize in. On the other hand, a lot of what we really, really focus on to Russell's point is working with enterprises, governments, regulated industries, working with massive, massive code bases, and trying to make sense of that and work in all of those systems. And so a lot of the problems that you have to go deal with and work on are, how do you absorb all of the messy context and the knowledge of this code base? How do you work with a massive, you know, something that has hundreds of thousands of different files and work across that? How do you test and iterate against your existing unit testing framework? Or how do you, you know, click around and use these products to yourself and make sure that the edits that you made were good? And that's a lot of what we've always focused on with cognition. And so, you know, something, I mean, basically all the guys, that you mentioned here in terms of the foundation labs and so on, we actually partner with them and we work very closely with them. I think what we tend to find is that, you know, for obviously they're kind of like base research and the work that they do with models, there's a ton of interesting work for us to do together. Whereas I think for a lot of this work with really enterprise transformation, you know, we want to be like on the ground working very deeply with folks and figuring out with them how we use software to help them achieve their goals. - So, Palantir originally we created this thing called for-to-played engineers, which didn't really make sense to people 10 or 15 years ago. And it turns out there is like certain types of workflows where you could build a product to do a lot, but then the for-to-played engineer would actually have to understand the business value and kind of connect the dots. And then and then take things that they created that oftentimes go back to the core. So the core gets better, so the core could do it next time. Like do you have something similar to this? Do you have a lot of these? Is that it's a big part of the value you're providing? - Yeah, so it's interesting. There's some similarities and some differences for us with the sort of the Palantir model. So one of the interesting differences is that a lot of times, you know, our customers, they just use our product on their own, even without us talking to them or without us discovering them. You know, people bring in our tools and they just start using them immediately. But once, once Devon is inside an organization, you know, the ceiling of what you can accomplish, if you're a world class agent manager, you know, versus like a, you know, a new engineer who's just sort of learned the tools, it's an enormous delta. And so a lot of the, there's kind of some parallels to our business that are more like, you know, kind of a data breaks or snowflake type thing where you just kind of get in and then people start using your products more and consuming more. On the other hand, if you actually want to go drive major outcomes, like real business change that results in structural, oh, I can launch this entire product line that I wouldn't have had the capacity to otherwise. That type of outcomes, R4 deployed engineers are among the best in the world at managing agents at really high scale, because that's exactly what they sort of focus on and do every day. So at Palantir, we had these frameworks of building an ontology of all the data, making it talk to each other on top of the processes. We had all sorts of different frameworks over time that are conceptual frameworks we'd use when you go in. Do you guys have your own conceptual frameworks for business value and do you have something called ontologies? Not to get the secret sauce, but other things like this. What we really look at it from the perspective of the software development life cycle. So we go inside an organization. They have a way of doing things, of developing software, starting from planning and deciding what they even want to write, understanding all of their existing code and process to then maybe scoping it out and maybe writing some code, testing it, fixing it when it's wrong, iterating on it, deploying it in production, monitoring it. There's a pretty standardized software development life cycle this point. And what's happening is agents are just eating more and more of the cycle. And it kind of started with the writing of the code. And now we're well past that. And so in fact, one of the more recent products we shipped is called Devon Review. It's because we observed that there's this totally new bottleneck in the software development life cycle that wasn't the case previously, which is there's this abundance of code being written by AI now. How can humans even keep up with it all to understand what's going on? Again, we work with a lot of regulated, large, complex organizations that are running mission critical systems. And you can't just sort of vibe code, you know, your way and like, yolo-mers. The translation be vibe code. No, it's actually really important. And eventually, you know, are we going to have English as a source of truth, and people are just going to be collaborating on specs? Yes. I think, you know, March, 2026. No, you still need to understand the code that you're merging. And so we, you know, a big thing we care about cognition is we're building tools that are like future-looking, but still meet people where they are, right? We still want to meet people where they are. And so Devon Review, it's not just having like an AI auto comment on every PR and say, oh, this was good or not good. It's actually tools for humans to really deeply understand huge quantities of AI generated code. Is there some sense that at some point, one of the AI models, if it gets a little too tricky, could like sneak something into the things that are being built to like do something we didn't want it to do? Yeah, so it's a great question. Obviously, I think with a lot of these things, that the simple answer and software is you want to be working with the same review process and the same QA processes that we all have, right? And so any big enough engineering organization, frankly, already has to think about this with their humans, you know, and it doesn't have to be on purpose. Obviously, maybe you accidentally introduced a security risk and so on, right? And that's why you have review and that's why you have QA and that's why you have user testing and that's why you have release cycles and all these other things. And I think, you know, I do think this is one of the, one of the things that often comes up, which is like, how do you make sure that that your agent behaves correctly, which of course is like a very important problem. The reality is that I think in software, we have actually a lot easier or maybe at least a lot more grounded of a path to do that because we have the same thing already with all of our humans, right? And so when you work with Devon, for example, Devon is making commits, is submitting, you know, code diffs and so on. Devon is not allowed to go and deploy your code to production by itself or anything like that, right? It works in all those same systems and has the same guard rails. And so let's just fast forward for fun a few years in the future. I talk to yeah, some of the people running the top labs are pretty convinced things are going to keep getting better at a pretty high pace for at least two or three years is like, I like with more is like, you can't really see out what's going to get there five years, but it feels like things are going to change a lot. Like so, so, so to tell us about 2028, 2029, like as there's certain things that just look very different or certain things that are that are like, we have to do now, we don't have to do it all there. Like what are the unsolved problems for Devon to just be like doing a massive project on itself in three years? Yeah, I think a couple shifts. I mean, one of the obvious ones, which I'll just call out is just much more widespread usage of all of this and just good knowledge on how to use these things. I think right now, you know, you have this core group of we'll call it like agent forward engineers, right? Or agent forward companies that understand how to use this and they are seeing these, you know, 5X 10X productivity games as a result. I mean, obviously, you know, all of these big organizations or these companies or governments or things like that are seeing the same results and re-eising, wait, we can't just sit here and be five times slower and, you know, we have to go learn how to do this right now. And so that's really happening, I mean, even this year, I would say, in terms of the the continued capabilities gains that we're going to see. Yeah, I mean, I think there's, I think that I think the models are going to get better and better. You know, one of the stats that people talk about a lot is this METR report, which basically says for each different model that comes out roughly how much human work can it do in an automated fashion before you have to go enter up and say, oh, that was wrong. Let's go do this, right? And so it's like, you know, just two or three years ago, the answer was like 10 seconds or something. You know, you would have it right one line. And then it's like, all right, the next line is already wrong. You look stop here, right? And at this point, it's already gotten to the point where it's in the scale of, you know, 10, 20 hours is what the latest one has been. I think Opus 4.6, for example, I think it was around 18 hours. This is always very weird to me because it's applying that models work in human time, though, right? Which is so weird. So how did I see that? Yeah, so the answer, they do typically do the tasks in less time than a human would. And then the 18 hours is basically, this is how long it would take a human to do that amount of work in between each of the interruption points, right? And the AI might be doing that in one hour or two hours or something like that, right? And the thing that's really crazy about the stat is you just see it very consistently double. And I think for the last few years, it's doubled about four or five times every year, which is insane, which means you wait two or three months and it's already doing twice as much work. So the whole world is changing over two or three months in terms of what's possible. Yeah. And this is what we've kind of seen as well. And I think we're going to see even more of that. And to some of Russell's previous points, the thing that's kind of interesting for us is that the form factor of what you want to deliver or how you want to work with the AI changes a lot as you're going through that, right? And so when we're saying, okay, you do 10 seconds of work, obviously the answer is you as a human need to be staring at your file of code and like shepherding it and handholding it with every single step, right? If you're talking about it's doing days of work or weeks of work, now you're actually giving it whole you know, output level tasks of like, hey, you know, we really need to go make this app much faster. Can you go and run this whole thing and then you know, do a smoke test and make sure all the changes look right. And it's going to go off and do that entire project, right? Versus or or you know, even even bigger initiatives of like, yeah, can you auto response to all the, you know, the upgrades or the potential vulnerabilities that are coming in with our reporting and just make sure all that, right? So you're going to see a lot more, I think proactive work. You're going to see a lot more basically like event driven kickoff work where it doesn't have to be human. That's moderating it every step of the way. And then and then in terms of like, what does that mean for society or where does it go? I actually think one thing I'll go on the record on is I think there's going to be an explosion in small businesses. I think AI is actually like an extremely small business enabling technology in particular where you think about what's hard about about building or starting small business. It's, you know, you don't have the resources of a large company, the specialization of labor in each part of the, in each part of the process. And I think AI, it's extraordinarily enabling, right? Think about, you know, the quality of quick legal gut check you can get from, you know, from a chap off from a frontier lab, the quality of, you know, analysis of your financials, the quality of software that you build, right? It's like, it's all coming together to empower each individual person again if you exercise that agency to just do so much more on the run. I actually, I love this. I actually have a small thing on the side where I'm trying to help create 10,000 to small business owners. So I'm totally aligned with this. This is a really good theme right now for us for us to do. One last thing I want to ask you guys about the business that I'm just so impressed by. So I hired a lot of first 300 people at Palantir. It's been a lot of time on talent. Obviously, even higher. It's got at one point a long time ago with one of my companies with, with Bob's help. How did he do it? How did he do it? He's very, very, very impressive. I did not, I did not see him at the time as someone who was like a CEO person. So he really grew a lot, which is good. I mean, he's like, he's learning and growing as he goes. And it definitely seemed as a CEO and founder now. But one thing you guys have done is like a significant percent of the cognition hires are actually former founders. So not only you hiring the best people in the world, you're hiring a ton of former founders. Like, like, like, why are you doing that? How are you doing that? Tell us a little bit about the talent stuff. Yeah. Yeah. For sure. No, I mean, I think the, I think the reality is we just have such a massive problem that we're going after. We're just, you know, solving all of code. And I think the way we even started this company is, you know, a wrestler is a founder before this. I was a founder before this. All of us were, you know, I think of our, of our kind of like initial crew. And, and the idea for us, I think was, let's make this one the big one. You know, we're going to go for it all. We're going to go for the most ambitious, I mean, even, I mean, the, the, the play of solving software engineering feels like a big enough one, you know, that, that we can all do that together. And, and, and I think that's a lot of what it comes down to, honestly, is just like, yeah, are we working on something and doing something that's, that's exciting for folks who, to your point, are, are, you know, I think a ton of the folks at cognition are, could, could very, very easily go off and start their own companies and get funded and build their teams and so on. And, and the question I think for us always has been about how, how do we make this the place that, that, that makes more sense for them to do that. And it's, it's honestly easier to do that now than ever before. Also, as a, as a company, like, I'm thinking, you know, we have one team of the company, they're, they're called special projects engineers. And basically every person on that team is a former founder, like every single one. And, and, they, they do, they do, you know, a really interesting mix of engineering work of product work, of talking to customers, of driving commercial outcomes, you know, the problem space to, to your point is so big that actually if you're, again, a high agency person, you're going to take initiative and you're working at a small, fast growing company with, with, you know, with the the problem space so big that you're only constrained is like your own ambition. Well, it does seem like there's just a renaissance, a revolution going on in a world where the capabilities are doubling every two or three months. I'm like, this is a place you can come and be around some of the other smartest people in the world who are part of growing something with that, which I guess is pretty fun for people. We have a good time. One thing about the interview process or the selection process to your point, which I think is kind of interesting to call out is I think, you know, one or two years ago, I think a lot of people had this mentality in terms of how you interview of like, okay, there's all these AI tools. How do we interview people in a way that, you know, how do we make sure that people aren't using AI while we're interviewing them, because they're worried. And I think that has totally flipped. Honestly, I think that was wrong, right? And I think if you're asking the question of like, how do we evaluate people on exactly the thing that AI can already do that's kind of the wrong? Yeah. And so, so, you know, I think for us, and this has always been the case for us, you know, our interview process has always been, you can use as much AI as you want to use, you know, it's just like, we're going to give you a few hours, build your whole own product surface, right? And build a, build your own, a lot of these are projects that frankly is like, if you were trying to do this by hand, you would not be able to get done in a few hours. So you kind of have to use AI for this, right? But, but in reality, what we actually want to test is in addition to it, to kind of how familiar you are with these tools, what we actually want to test is, is yeah, what, what, what do you think is the right thing to build or how do you make these product decisions? How do you make these trade-offs? How do you decide or collect information about what you should be doing? And so we found that that's helped us a lot. I love it. What we started the American Optimist podcast to try to push back on cynicism and pessimism in our country. What's the best case for an optimistic AI future? Wouldn't one inspire you about what you're seeing? Yeah, it's a great quiet, you know, so funnily enough, I think recently we had this whole Satrini piece come out, which I thought was frankly ridiculous. I mean, I think it's, I think it gets a lot of the basic economics wrong is maybe a simple way to put it. And it's, I think it goes off of some of the same things that we are calling out, which is it's going to be easier to build things, it's going to be cheaper and so on. And then somehow concludes that the, you know, the outcome is going to be much worse for all of us. And I think maybe that, you know, from, from, from an economics perspective, there's a distinction between what is real versus nominal, you know, deflation, which I think is maybe one thing to call out here, which is of course prices are going to get cheaper. But why should that mean that we're all worse off, right? But, but, but I think the simple thing that I'd call out about AI is I think we right now are at a point where so many of the things that we want to build and so many of the things that we want to do are just hard bottlenecks by pure execution, right? And I think we're pretty quickly getting to a point where that's not the case. And my, my favorite line on this is, you know, our co-founder Walden says this, which is for so long we've all been living in Minecraft survival mode and now we're going to be in creative mode. And I think that's really, I think that's what we're going to see over the next, honestly, the next five or 10 years is getting to a point where you're really only limited to your ideas and to your imagination, where you can kind of just turn things into reality. And I think that's going to be a great future. Awesome. Well, it's inspiring now to leave it on Thanksgiving. Cool. Thanks for having us. Yeah.