In this AI wave, we're in the middle of a new industrial revolution. We're seeing companies scale faster than anything I've seen in my lifetime. We've seen companies investing trillions of dollars in infrastructure. These new companies, many of them are doubling every quarter, and it's adding huge amount of value to the US economy. Productivity is starting to shoot up in a bunch of smaller areas that are going to grow into really big areas over the next few years. We're going to take you to the edge of the AI wave with some of my very favorite entrepreneurs. We're going to meet Scott and Russell at Cognition. Gonna learn about what's happening in the insurance world and finance with JD Ross and with coverage. And I love hearing from Joe Ben about what's happening in aerospace and with airplanes and the physical world with AI. And then we're going to break it down with Alice Cole and Jack, two of my partners at ABC to help you understand how to think about what's happening in the next 5 or 10 years, how are investors and builders approaching it, what's it going to look like as AI spreads throughout our society in terms of business growth and productivity.
Let's just fast forward for fun a few years in the future. When I talked to some of the people running the top labs, they're pretty convinced things are going to keep getting better at a pretty high pace for at least two or three years. It's like kind of like with Moore's law, you can't really see out what's going to get there five years, but but see it feels like things are going to change a lot. Like so so so tell us about 2028 2029 like is there certain things that just look very different? Are 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 be like doing massive projects 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 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 gains as a result. I mean obviously you know all of these big organizations or these these companies or governments or things like that are are seeing those same results and realizing 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 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 MER 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 interrupt it 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 write one line and it's like, 'All right, the next line is already wrong. Let's 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 was around 18 hours.
This is always very weird to me because it's implying that models work in human time though, right? Which is so weird. Why are they faster?
Yeah. So, so, so the answer they 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, okay? In in between each of the interruption points, right? And and the the AI might be doing that in in one hour or two hours or something like that, right? And and the thing that's really crazy about this 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, you know, which is insane, which means, you know, you wait two or three months and it's already doing twice as much work. Changing every two or 3 months in terms of what's possible.
Yeah. And this is this is what we've what we've kind of seen as well. And I think we're we're going to see even more of that. And to 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 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 autorespond to all of the you know, the upgrades or the potential vulnerabilities that are coming in with our reporting and just make sure all of that, right? So you're gonna see a lot more, I think, proactive work. You're gonna see a lot more basically like event driven kickoff work where it's it doesn't have to be human that's 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 a 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 chatbot 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 their own.
I actually I love this I actually have a thing on the side where I'm trying to help create tens of thousands of small business owners. So, so I'm totally in line with this. This is a really good theme right now for us for us to do. What one last thing I want to ask you guys about the business that I'm just so impressed by. So, I I hired a lot of the first 200 people of Palantir. I spent a lot of time on talent. Obviously, I even hired Scott at one point a long time ago with one of my companies with with God's help.
How did he do how did he do in his very very very impressive but 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, you know, he's like he's learning and growing as he goes and and definitely see him 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 are you hiring the best people in the world, you're hiring a ton of former founders like like why are you doing that? How are you doing that? Tell us a little bit about the talent stuff here.
Yeah. Yeah, for sure. No, I mean I I I think the I I I think the reality is we 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, Russell was 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 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 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 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 that's exciting for 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 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 at the company. 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 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 going to take initiative and you're working at a small fast growing company with with a you know with the problem space so big that you know your only constraint is like your own ambition. Well, it does seem like there's this renaissance or revolution going on in the world where the capabilities are doubling every two or three months and 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.
One one thing about the 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 we're and I think that has totally flipped honestly I think that was wrong right and and I think if 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 and and and so so so you know I think for 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, just build build your whole own product surface, right? And build 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 to kind of how familiar you are with these tools, what we actually want to test is is yeah, 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 and and so we found that that's that's helped us a lot.
Not a single one of our full-time engineers right now writes code anymore. They do everything through codeex, cloud code, or cursor.
It's all it's all agentic.
And they're a lot of them are like managing lots of agents at once, right? So they're working.
Yeah. Cole who runs our engineering team has like six agents that are running at once at all times just doing different kind of projects. When we create a new ticket in linear, it automatically kicks off this thing that basically creates a dev environment which then tries to oneshot the solution and then creates a PR that we review. Like the whole way you build software is changing so quickly. And I I think right now like the cost of software is just asymptotically dropping to zero of building new software which means that the way you build a company completely changes. Like we can now have an enterprise customer come up who says, 'Oh yeah, this is so close. What we actually need it for is X. We can build X in one day, ship it live, and close that deal. We did this last week.
It's like a completely different way of even thinking.
No Fortune 500 company right now, I don't think, works that way. They they always have these like three month and six month plans and we'll get back to you and we can't sell you cuz you don't have it yet. And like in the old days, you were told I was told a lot cuz I'm too expensive. Like, Joe, stop trying to Yeah, you can't do that. Stop trying to like make promises that are going to get in the way of our road map in the next 12 months. But and they still probably don't want to do some promises because they probably still are pretty hard. But certain promises you just go around and just do.
You can just make it happen. I think even the the job historically the job of product engineering and design the sort of three-legged stool is that product was responsible for like the why and the prioritization. Design was responsible for the what like what does this actually do and how does it work and engineering owned the how like how do we actually implement it? What's possible? And today like that's still kind of true, but the loop doesn't go from let's come up with a hypothesis, let's build it, let's learn. It becomes more like we're a group of wizards who are creating spells that the entire organization can use to make things better. And that is we're at the very early stages of this transition where now most companies including us are just like accelerating that loop of hypothesis learn deliver. But it's quickly becoming, hey everyone in the company, tell us what you need and like and we're going to ingest that into a self-learning loop of AI and that's going to tell us what we should build and it's going to try and prototype it and we're going to use that to determine what to ship.
It's interesting. It's almost like there should be like some kind of AI intelligence guided by your top product mind just to listen to everything going on and help you organize it, right? Because that was always a hard part of the product job for me. Like I love the product job. is like the vision of what I think everyone wants, but then there's like this 14-hour a day job of prioritizing everything you're hearing from everyone in the company. And I I feel like AI could be much better at that. Is that is that is a little bit what you're saying? Like just everything you hear, you should be pri using that to prioritize than just constantly trying to build stuff.
Yeah. If the job like in the past was writing, now it's editing. And shifting from being a writer to an editor is a very I think for some people they're thriving in it and some people are really shocked by it.
I mean, what's the output difference, right? because it's not it's not some people are saying Joe it's 100x it doesn't really seem to me like it's a 100x yet like like no as an example like Adapar is doing a lot with AI now and there's a lot of people there each managing tons of agents and there's you know several hundred engineers around the world. I mentioned this because we built this together a long time ago and you know they've probably invested almost $2 billion now into the product right because it's like 16 years old and this is what happens you and and it's definitely the case that like it wouldn't take that much money to build it now and like but the question is is it like 2x is it 10x I don't think it's 100x I don't think it's like $20 million to build adapar but but it might only be like two or 30 hundred million I I don't know like how should and and I think out of part because it's using AI to build more now it's like harder to catch right but it's like how should we think about this like all these SAS companies the lowend ones got crushed last week right and like a lot of the the firms you probably saw that invest in lowend SAS companies got crushed because in anthropics new stuff came out and everyone's scared they can rebuild it like like how are you thinking about how this is evolving there.
There's so much in that. I think one like Anthropic, OpenAI, Google, they're all going to in the same way that you probably don't want to build an email company while Google's on its rise because they're going to come and eat that. Like every one of these companies is looking at where are our tokens being used? Where's a lot of tokens being used? Great. Let's go after and like attack that surface.
You think you think so? You think those companies are themselves are going to try to go build a bunch of the very most important assets? So the most important areas 100% in the same way that Google would never let you know messyloma become like its own they're going to capture as much of that value as possible because it's the highest value keyword.
That's actually different than what I've heard because I'm worried about the low-end SAS companies, but maybe maybe I should be worried about both like the bottom half and like the top 1% as well or something. I think it's a barbell thing like the the large AI companies are going to eat the top end of value and everyone in the middle is going to just commoditize or people who aren't even in the software business like in the next two or three years this isn't true right but if you take the 5-year view kids who are in college today who come out and start working in a construction company or something are going to look at every problem they're paying for for software and say this doesn't do exactly what I need and I can oneshot a much better version of this right now. All of those SAS companies are worthless.
And then they're going to go after the top too. I mean, why don't why don't they go after you? Like, how do you think about that?
I I think it's really important to understand what business you're in. And our business is we are in many ways in the business of answering unknown unknowns for people who know it's really important, right? We're selling to CFOs, GC's, people who know that insurance is there for when something really bad happens, it covers them. They know that risk management is really important, right? They know that when someone, you know, falls in a factory or cuts their hand that they need to deal with that and they want it to not happen in the first place, but that's not their full job. That's not their expertise.
So, you're so you're going to be the trusted expert partner in the area of understanding their insurance and what works for them and how to get a better deal than they are now.
Exactly. So, we help them manage risk and we help them save money.
And so, when you think maybe stepping back, maybe is it more for me than for you because I'm the venture capitalist and I'm investing in lots of things, but I know I know you angel invest. Think about it too like what are the categories of things that the model companies are going to eat versus not as it relates to this like is like obviously the model companies are pretty good at being experts in a lot of things like how how do you think about the modes here?
I mean the first one obviously is anything that's adjacent to software development they're like they've decided that they're going to war to win that and I think it's a really hard place to you have to be you got to be the very best in the world.
The very best in the world. I I think I'm pretty bullish on cognition because I think Scott and like these like 20 other like gold medal like Olympiad winners like probably could do something that's very very valuable despite everyone going after it. But if you don't have 20 other global gold medal physics and math leaders around, you probably shouldn't be bothering to compete in that area.
You have to know that you're competing. You have to like at least be aware that we are directly going to be at war with these companies for a long time and we're going to win.
Or or or maybe Yeah. And there's probably different territories they can each capture or something like this, but but they're going to try to win as much as they can.
Yeah. And then there, you know, there's companies like Harvey who are going in saying, 'We're going to be the ones who win on professional service, you know, AI for legal or for accounting and things like that.'
And that's a big big big area for to conquer. Yeah.
Yeah. And I I think over time I would be surprised if more of that doesn't get attacked by the large AI foundational companies.
It is very interesting because some of these foundational companies are really good at doing the foundational research but not at enterprise stuff, right? But I guess the question is kind of get good at enterprise stuff which you know.
I I think the the crazy thing is how quickly this is all changing right and so in many ways we're all forecasting but if you look at cloud co-work which came out you know two weeks ago and the types of spreadsheets and models that they're creating and you're like oh this is actually like valuable good work like we we use this internally for simple things like logging into carrier portals to download documents and ingest them and do a whole thing like that didn't exist three weeks ago. Yeah, it's crazy.
It's an entire entire area of our business that now is automated that zero people work on.
It feels like most of the world is just not aware of like how fast everything is going going to change like like what what does this mean for for business a year from now?
I I think right now the ability to absorb and adopt the tools is going to be the limiting factor for years to come. And I think that the capabilities of these models is already so far ahead of our ability in most enterprises to just absorb and adopt change that that is like the wave behind us. Right now we have a system that let's say we're doing like email drafting for example, right? All of our all the emails that we send are drafted by AI before we send them and a person reviews them and sends them out. If you switch that model from let's say you know sonnet to opus you know anthropics middle to their best model did the evaluation set get better if it does don't do anything stop working on it just let the mod let the hundreds of billions of dollars of investment behind you push you forward if it doesn't get better that's actually a good sign that you can invest more in getting everything ready because right now I think like the models are already so good that the constraining factor is actually your ability to provide them the context information to win.
Let's talk a little bit about about AI and and autonomy. So, you're saying like at scale if you have thousands or tens of thousands of these at some point enough pilots would become a blocker and you're going to want to like legalize autonomy. That's probably not a focus. I assume the next couple years, but but how are you thinking about AI in general in your operations and and you know, how how does how does it scale?
Yeah. So let let's take the autonomous aircraft and and and the autonomy systems that is a absolute game changer whether that's on the safety side whether that's on the productivity and efficiency side. Really really excited and again the the work that's happening on airspace from the administration from the DOT and from the FAA is absolutely phenomenal. That's laying the foundation for a really really exciting new future with the EIP. You're going to see a lot of that start getting tested.
So, they're going to start testing at Tommy in addition to the pilots flying it.
For certain of these use cases. That's why having all those 12 states is absolutely fantastic. So, we're going to see this is fast forwarding this really fantastic future all across the country. So that's that's on on on autonomy and and in fact over the last two days I was at NVIDIA's GTC conference. Nvidia is an incredible partner. They've put invested vast sums of both people hours into building this really safe autonomy stack. They're using that for autonomous driving, using that for robotics, and we're able to layer on top of that for all the work we're doing for autonomous aircraft. So, this is really fantastic, incredible partner, very grateful for all the work. And I mean, and then they've been working on this for for more than a decade.
So, and you think they actually end up there for being NVIDIA chips potentially in the in in the Joby Aviation?
Oh, absolutely. We've been we've been partnered with them for many years and we're using the latest generation of of their autonomous chipsets.
Very cool. So each each of these things will have these chipsets and I I guess you want that for safety anyway. If your pilot somehow has a problem, you're going to want the autonomy backup either even if you did have a pilot. I assume.
The perfect initial use case.
Yeah. Then eventually it's like maybe you don't actually need the pilot. Right now by law you do. We got we got figure that out. I guess you're going to have to train a bunch of pilots to start though. And I guess maybe there's some things you're going to want to hopefully have a bunch of pilots for. So, so, so how how's that work? And is is this awkward? Like, we're going to have autonomy, but also please come learn to be a pilot for us, for thousands of pilots. How how are you thinking about that?
There there is an insatiable demand for
Pilots across the aviation ecosystem today, just everywhere. And so we have built something called the Joby Academy where we are training pilots. We've taken delivery of our first simulator which we developed in partnership with CAE. CAE is the leader in flight simulators. This is an incredible experience where you get in and it feels like you're flying. You're in a cockpit that has exactly the same avionics.
Oh wow. So you actually feel like you're sitting in a Joby.
Absolutely. And you've got the displays and it looks just like a flight simulator that you'd learn to fly a Boeing or Airbus aircraft in. So these are the same pedigree of...
You practice crashing it and see what happens.
Not allowed the... not allowed flight controls.
This is what I always did as a kid on these simulators. Is that weird?
No. One of the really cool things about our aircraft is because of fly-by-wire, you can take your hands off the controls and the aircraft maintains the attitude. You can get into the simulator and you can fly the aircraft flawlessly the first flight. Hundreds of people or thousands of people have done it now.
So, it's pretty easy to do.
It's super easy to fly. So, if people are interested in learning to fly, they should go to jobacademy.com. We would love to have them. This is going to be a fantastic opportunity to get to be part of this next stage of aviation.
And these things can go like... almost how fast are you letting them fly right now when they go up to 200 miles an hour?
Really? It's got to be up to 200 miles an hour. I love it. That's pretty fast.
It is. And it's point to point, right?
Yeah. So, no. Of course, you're not going over the road system. You're going directly as the crow flies, as they say. It's being able to take a trip that takes you half an hour or an hour on the ground and turn it into a five or 10 minute trip. You know, sometimes you're replacing a two-hour trip because you've got traffic.
Of course you can fly over, you see the traffic jam, you get a little shot in Freud right and just go right by it. It's pretty awesome. So I mean if you were to train yourself like could you actually then just like fly yourself I guess places or how would that work? I mean you know because...
That's what I'm getting excited about. I came over from Santa Cruz and if I could have been here in 10 minutes...
The what is it the 237 is it there what is it 680 what what's that road there I forget.
The 17, that's right. It's been a while. I'm in Texas now. So yeah, the 17 is probably a lot faster to hop over the hill. It's a pretty hill.
There's probably plenty of room by my pool next time we have to do one of these episodes where we get you landing somehow. I don't know, the Woodside might not allow it. They're like the last places to allow things is like these little communist like strongholds of California. I don't know. We'll see. We'll see. We can hope.
My bias was always that even though it's really quiet, you could still sort of hear it if you're a neighbor. So, if you'd have like a tall building, you couldn't hear it on top of a tall building. Shouldn't we just build one tall building per nice neighborhood? Landed there. Wouldn't that make sense?
Uh landing on rooftops is a fantastic opportunity.
Especially if it's high. You can have like one building where you have like housing for all the staff in the neighborhood for a nice neighborhood and then they can land on the top of it.
Tall buildings got nice views.
So some people pay premium prices for penthouses.
