TBPN Live - GameStop + eBay, Neural Computers | Nat Eliason, Michael York, Maddie Hall, Anjney Midha, Ben Lamm, Jake Stauch, Garth Sheldon-Coulson, Katie Haun, Nick Abouzeid

Starting point is 00:00:00 You're watching TVVN. Today is Monday, May 4th. Would you look at that? 2026. Would you look at that? Would you look at what, Jordy? Everything looks fine to me? We're back.

Starting point is 00:00:09 We're back. That's what I'm looking at. We're getting ready to last second. We may have installed Modern Warfare 2 on this large screen and gotten a little bit behind schedule with some of the interns playing. Yeah, Russ and Tyler were. I was cooking him. It was a knockout, dragout fight. You won.

Starting point is 00:00:30 a lot, right? Yeah. Well, yeah. It was, it was pretty embarrassing for Ben, considering that he is still in that, not,

Starting point is 00:00:38 not, not chief producer, but other Ben, considering he's still in like the main kind of like demo, video game demo. But at the same time, I mean, you were probably two years old when Modern Warfare 2 came out. So you got to sort of relearn the old,

Starting point is 00:00:54 the old tricks, the old tricks on Rust. Anyway, big show to be. weekend. Hope you had a great weekend. Big week. We'll be, we are going to a conference on Wednesday and Thursday. So we might be off both those days, but we have some great shows for you planned Monday, Tuesday, Friday. It's a conference. Oh, I thought you said concert. No. Although, I think there might be a concert. That would be extremely, that would be extremely out of character. I think there might be a concert or at least some sort of musical performance at this particular conference. But what do we have coming on the show today? We have a bunch of folks. Nat, Nat, Michael, your. from Kasa, Ajne from Amp. We got Garth from Panthalasa. A little more context on Nat joining.

Starting point is 00:01:38 Yeah, we talked about it. He joined Alpha School. Yeah, that's right. He's the head of founder development. They're launching a founder track at Alpha School, which I'm very curious to hear about considering that. A lot of the parents in my area are very Alpha School curious. Yeah.

Starting point is 00:01:58 So I'm excited to hear about that. Well, I mean, there's a bunch of news we're going to go through. But first, I was just sort of reflecting on, there was a great interview with Andre Carpathie at Sequoia AIS Sent. I think last week it went up on YouTube. And it was, he was sort of reflecting on how his workflow is changing around vibe coding. And I was sort of reflecting on how my knowledge workflows are changing, particularly around image generation now that image generation is really good at infographics.

Starting point is 00:02:29 and effectively designing slides or output. And so we're starting to see the rumblings of this idea of like the neural computer. There was this, people have been talking about this for years since the AI boom began. But the basic idea is like you would have a computer that basically has no software whatsoever on it. It would just have an LLM or just an AI model, just inference capability or potentially connection to cloud inference. That would generate whatever you want, whatever you need on demand on the fly.

Starting point is 00:02:59 And so I think Elon talked about this with macro hard a little bit. That was a piece of the vision. This has always been theorized, but it's becoming more and more real. And so Carpathie describes this idea of like a neural computer this way. And I think it's an interesting framing. Obviously, it'll have implications for SaaS products that might be used in a headless under the hood scene way or might be competed with against these neural computers. But Carpathie describes it as, he says, imagine a device that takes raw video or audio into basically what's a neural net and uses diffusion to render a UI that's unique for that moment. And so this sort of like on-the-fly instantiation of the exact UI that you need for that particular question or whatever problem you're trying to solve, whatever you're trying to do is an interesting paradigm shift that it feels like we're starting to see glimpses of.

Starting point is 00:03:55 So I most recently felt it when I was trying to understand Ryan Cohen's proposal for GameStop to take over eBay. This is a big story. We'll go through it today. But I haven't tracked either company closely. We've had Ryan on the show. And we've talked about eBay and GameStop intermittently, but I couldn't tell you off the top of my head, what's the revenue for each company? What's the profit like? What are the different multiples?

Starting point is 00:04:20 And so in a pre-chat GPT world, I would have gone to Google Finance. or Yahoo Finance and pulled some data, maybe had two tabs up, maybe used one of their comparison tools. If I wanted to be really advanced, I would have copy and pasted the stats into a spreadsheet. If you're really working on Wall Street, you might have like Cap IQ or Bloomberg plugging into a sheet,

Starting point is 00:04:41 an Excel sheet that then can build you a comparison table and do like comps. And then once we got into the chat GPT world, you might do a deep research report, pull all that data, put it into a table, which is effectively marked down. and sometimes the tables renders a little weird and like you can kind of bounce around. But now I, the whole process from start to finish

Starting point is 00:05:04 is just a single prompt. It outputs an image. So you can pull up the image that I generated. So this was one prompt. I said do a bunch of research on GameStop and eBay's valuation and key financial metrics, things like growth rate, top line, earnings, revenue, valuation, how the multiples fit together,

Starting point is 00:05:19 build a nicely designed side-by-side comparison of the two companies. and you wind up getting something that is very digestible. Like just looking at this, I mean, it's obviously a little zoomed out, but you can zoom in and see, okay, eBay has about three times the revenue, $50 billion versus $15 for GameStop, and revenue growth. eBay is growing while GameStop shrunk by 5%.

Starting point is 00:05:42 eBay grew 8%. Operating income. eBay has 10 times the operating income at $2.28 billion versus $2.332%. million for GameStop. And so you just get this like very easy, okay, what's the operating margin? eBay is up at 20% game stops down at 6.4%. And so you can start to see on a price to sales ratio, Game stops at 4x, eBay's at 4.5x, but on a market cap to net income, GameStop's higher, has a higher value, 34x versus net income versus eBay is 25x. So you can just sort of see this table and this is something that usually would have been like three or four steps to get.

Starting point is 00:06:22 here and instead it's just this single prompt. And so I think that there's like this is not a perfect result. Like in that image you can see that like it chose red as the color for all of GameStop's financials, which is not what you'd normally do because red is usually for negative numbers, but those revenue figures are positive. Like it could be better. I could probably go further and prompt it a couple more times to get exactly what I wanted. But it solved my problem of having like here's the summary of the question that you were

Starting point is 00:06:50 actually asking, which is like, how? How do these companies stack up to each other? What's the relative size of the business? What are the strengths and weaknesses of each of them? And then boom, you have a square image that you can easily text to someone and it's ultimately shareable. And more importantly, I don't care what it used under the hood. It could have puppeteered a spreadsheet and put it all in comma separated values and make

Starting point is 00:07:14 a CSV and it could have transformed it with Excel or Google Sheets under the hood. It could have written Python. It could have used Pandas or Scikit Learn. could have done anything it wanted to, but it's all abstracted to me and I don't even think about it. And this is different from the previous era of like, okay, well, if I wanted to do some sort of stock comparison tool, I could vibe code a stock comparison tool with API integrations, make sure you have the data connections, but it's just kind of less necessary as the models get fatter and they sort of eat more and more of the process.

Starting point is 00:07:43 And so Carpathie describes this concept as Software 3.0, and we should pull up his example because it's very similar. Of course, it happened like months ago because he's ahead of me on everything, obviously. But he gave a good example of shifting from like, you have a problem that there's no solution for, so you're going to vibe code an app to just a few weeks or months later,

Starting point is 00:08:04 like the AI tools can just do it. And you don't need any code. You don't need any system to build. Even though it's fun to build a system and it's interesting and it allows for more, maybe more speed and more reliability. Like more and more things are like one shotable by the model. So let's pull up,

Starting point is 00:08:19 Andre Carpathie talking at Sequoia AIScent about his experience with Software 3.0. I think one more maybe example that comes to mind that is even more extreme than that is when I was building menu gen. So menu gen is this idea where you come to a restaurant, they give you a menu, there's no pictures usually. So I don't know what any of these things are. Usually I like 30% of the things I have no idea what they are, 50%. So I wanted to take a photo of the restaurant menu and to get pictures of what those things might. look like in a generic sense. And so I built, I vicoated this app that basically lets you upload a photo and it does all this stuff and it runs on Varsel and it

Starting point is 00:08:58 basically re-renders the menu and it gives you like all the items and it gives you a picture that it uses an image, you know, generator a 4 to basically OCR all the different titles, use the image generator to get pictures of them and then shows it to you. And then I saw the software 3.0 version of this which is which blew my mind which is literally just take your photo, give you it to Gemini and say use Nanobanana to overlay the things onto the menu and Nanobanana basically returned an image that is exactly the picture of the menu that I took but it actually put into the pixels it rendered the different things in the menu and this

Starting point is 00:09:35 blew my mind because actually all of my menu gen is Perius it's working in the old paradigm that app shouldn't exist and yeah the software 3.3.3 paradigm is a lot more kind of raw it just your neural network is doing more and more of the work. And your prompt or context is just the image and the output is an image. And there's no need to have any of the app in between. So I think that people have to kind of like reframe. Yeah. I mean, it's real.

Starting point is 00:10:06 And it's, and it's, and there were, I had some takeaways from this. Like, what are the implications for this? And I think there's a few things. The first thing that it was on my mind was that although we had, have gone through this crazy vibe coding boom where everyone is vibe coding apps, it feels like a very temporary aberration. And also, I know that, like, even though there are millions and millions of people that have used Codex and Claude Code and OpenClaude, like the numbers are big, but it's not at 20% of the U.S. population. Like, it's just not at that level of adoption,

Starting point is 00:10:43 as opposed to chat apps, which are at like 70, 80% penetration, right? Yeah, the other thing that's been interesting is non, like people outside of tech that have gotten into vibe coding that have been pitching me their ideas here and there. Almost every time they're pitching me the idea, it is something that the, that ClaudeCode and Codex

Starting point is 00:11:06 can do themselves pretty well today, like just in one chat thread. Or the app. Like the apps can do them. And that's what I'm like, what's blowing my mind now is that, is that in many ways. That's what I'm saying. Yeah. So like they're using, they're using vibe coding tools to vibe code something that doesn't

Starting point is 00:11:25 necessarily need to exist. Yeah. Because you could just use the app itself to do the thing. Yeah. And they're already widely available. So it's, it's been, it's been interesting. Yeah. So I think there's, I think there's two things.

Starting point is 00:11:36 Like, one is that if you, if you've been, like, hesitant to jump into vibe coding, uh, because, like, it's, It's just a little bit too much an hassle. Like, Andyra Carpathie's like, obviously very, very comfortable being like, oh, yeah, let me deploy to Versailles and do all this. Like, you can figure all that out, but that leads to this world where it's like, oh, I was staying up all night. I was really, really like burning the midnight oil to get this app deployed and like do all

Starting point is 00:11:59 this stuff. Well, like, a lot of that's going to go away. And like, you're not going to need to do that. And frontier models are, but then there's also the this question, what you had, which is, like there needs to be this higher order loop of thinking around, okay, you have a problem. Should you actually vibe code an app or should you just try and one shot it with the current model capabilities? Because for a lot of things, like-

Starting point is 00:12:24 You mean, and within a lot of your favorite? Yeah, within Chachapit, within Gemini, within Claude, the actual apps. Like, you can like take a picture of your food and say, hey, start tracking my calories. Like, there's a lot of things that the, that the apps can just do in one chat thread that people are doing, but I think that there's like this tension between when you actually need to for sure go and vibe code something versus when you can just do it in like a one-shoted LLM context. And so like frontier models are already able to, in basically 90% of situations I feel like instantiate exactly whatever's required to solve the actual problem under the hood entirely abstracting away like code and tools.

Starting point is 00:13:06 Like you will just not be aware of what's happening and it doesn't matter. Yeah. And then the second... Yeah, I would add to my previous statement by saying that doesn't mean that there's not necessarily a business there because sometimes taking a raw capability and presenting it to people in a way that's very easy for them to digest. You can still deliver value and you can get customers and people will pay you money. Yep. But it has been fascinating to see, like, does this actually need to be an app?

Starting point is 00:13:33 Yeah, yeah. I mean, there's a ton of apps and software that will still be valuable, whether it has a liquidity pool or some sort of unique source of strength or some differentiation point that the existing chat apps can't hack at all. And then there's also just like marketing arms effectively where it's like, okay, yes, any frontier model in a chat app could do this, but you weren't aware of it. And this company was really good at running ads to actually get awareness. going and then drive downloads of the specific thing. And so we see those in the app store all the time. So the other thing that I was reminded of was, did you ever read Union Square Ventures 2016 blog post

Starting point is 00:14:22 Fat Protocols? Are you familiar with this? So Fat Protocols was this concept around how in the web, like web, I guess 1.0, 2.0, there were protocol layers, which are like TCP IP, HGTP, HGTP, S, M. like file transfer protocols, HTTP. And for like a couple of years, the crypto community was like the group that developed and maintained like HTTP, they basically created the standard that the web ran on and yet very little value accrued to the creators and the maintainers of that protocol. And crypto would be different because the Bitcoin protocol had the value capture component

Starting point is 00:15:05 like baked into it. And so there was this idea of, like, like the application layer in blockchain would accrue very little value, and the protocol layer would capture the vast majority of value. So this is on the web, the applications on top of HTTP. You can think of like Facebook as a beneficiary of the protocol of HTTP, because that's how the actual information in the photos and the text gets transferred to you. But the HTTP standard does not accrue the value. The value accrues to the application on top. And if you scroll down, you'll see the blockchain example, which was sort of borne out that the application layer was pretty thin on top and most of the value went to like the tokens and the protocol below.

Starting point is 00:15:47 Yeah, exactly. Ethereum's a good example. Salon is a good example. Of course, there's value in the application layer and there's some companies that are being built. But this was basically this thesis that the, he says, we see this very early, we see this very clearly in two dominant blockchain networks, Bitcoin and Ethereum. The Bitcoin network has a 10 billion market cap. Wow. I think it's like a trillion now, right? Isn't it $700 billion?

Starting point is 00:16:12 Yet the largest companies built on top are worth a few hundred million at best. Now we have Coinbase in the tens of billions. So both sides of the protocol application layer did very well. But the point is still true. Similarly, Ethereum has a $1 billion market cap even before the emergence of real breakout applications on top and only a year after its public release. And so that was sort of the core thesis on this fat protocols thing.

Starting point is 00:16:40 And I think there's something similar happening in the AI value chain. Of course, there's like a bunch of other dynamics going on in the AI value chain, and there's a lot of capture and complicated market dynamics. But the models feel like they're getting fatter every month, and they're sort of eating away

Starting point is 00:16:56 at the edges of what you can do with them. And so increasingly you can just get more and more out of the core model, which is an interesting dynamic. And then third, there's still like this huge question of like walled garden jumping. We've talked about this before, but it's almost, we need like a different term for like the dead internet theory. It's like the walled garden internet theory. Like the internet's not dead.

Starting point is 00:17:21 Like there's great, there's great information in Substack on certain legacy web, legacy media websites, and on Facebook and on X and on YouTube. Like, all of those companies don't want to interact with each other. And so that's where you get something like, oh, well, like, if you write code, you do get access to it loosely. Or if you're puppeteering a browser on a Mac Mini, you get access to that. Or if you're digging through iMessage locally, like, that can require a different workflow. But that's more of like a legal and business discussion than a technical one. Like, there's no reason technically that a single LLM wouldn't be able to just query every single web resource, except for the fact that,

Starting point is 00:18:03 the various tech companies don't want each other to talk to each other. And so the models, I think, will continue to find a way under the walled garden, over the walled garden, through the walls. Like, they'll seep everywhere. And it's more of a question of just inference cost, how long it takes to actually grind through the wall. But they're already figuring out a way around. And Open Claw is a good example of that where a lot of the walled gardens were sort of brought down by running. Yeah, except I think SAP came out and said no, no unauthorized agents here.

Starting point is 00:18:40 They're trying to, they're trying to put up the walls. They're trying to build a moat. They're trying to get some alligators to scare off the agents. Yes, but I would be very, I would, I would be very surprised if they're able to stop me from if I have SAP and I'm running it locally, for me to take a screenshot of my computer and then tell the mouse to go where it wants. It's very hard to fight back against these, like the compute. I mean, I've seen this guy on, on YouTube who uses like ever more contrived aim bots. Like he has one that's, it's like a robotic mouse. So it's using the actual mouse and keyboard, but it's robotic fingers on it.

Starting point is 00:19:23 And then it's controlling, I think he plays Rainbow Sixthage or Counterstrike. And he's cheating, but it's look, it's a camera looking at the computer. So until you get to like, you know, world coin eyeball scanning, making sure that you're not using this. He's also done one where he, I think he put like electro stimulation on his arm that would do the aim bot for him. And so it was his arm physically moving, but it was puppeteered by an AI essentially. And, you know, it's very, very demo phase at this point. But imagine that everything is so locked down that no AI agent can interact with SAP. but then I have the electro-stimulation

Starting point is 00:20:01 and I can just type super fast because, and it just does whatever it needs and I'm the, I'm the, the human instantiation actually pushing the keys. Rune had a good tweet where he was like, you know, people are now just like vessels for the AI where they just like, the AI

Starting point is 00:20:15 tells them what to do and then they just act exactly like the model says. Wasn't John Collison talking about you? John Collison was saying, like the humans get the thing off the high shelf. It was just like every time I have to like go and like export a PDF and upload it to chat GPD because I can't get it in there by default, even though I could just give them a web URL. I have to export it, print or whatever.

Starting point is 00:20:39 This is the modern era. I think we should talk about GameStop. Let's do it. What's up with GameStop? What's going on with GameStop? Yesterday, it came out through the Wall Street Journal. GameStop was preparing to make an offer for eBay as part of Ryan, CEO, Ryan Cohen's plan to turn the retailer into $100 billion-plus juggernaut. GameStop has been quietly building a stake in eBay's shares ahead of a potential offer

Starting point is 00:21:02 and could submit an offer as soon as later this month, which they did this morning. If eBay isn't receptive, Cohen could decide to take the offer directly to eBay's shareholders. And they released a letter yesterday to Paul Pressler, who's the chairman of the board over at eBay, happens to be a friend of mine. Yeah. Wait, really? Yeah. I, I, I, um, are, his daughter and my wife are good friends. Oh, cool. So I end up, we end up hanging out a decent amount. And we're neighbors. Does he do a lot of, does he do a lot of, like, podcasts or press appearances? It can, it can, we can, it can, it can be discussed. I'm just saying, like, determinism of Pressler would be like somebody who dominates the press. The press circle. I think you would be doing a podcast.

Starting point is 00:21:56 Press circuit all the time. Anyway, Paul, Paul is fantastic. Yeah. And Ryan wrote this letter to him yesterday. GameStop is proposing to acquire all common stock of eBay at $125 per share. We have accumulated a 5% economic stake in eBay through derivatives and beneficial ownership of common stock and are filing a Schedule 13D and HSR notification tomorrow. Our offer is $125 per share comprising 50% cash and 50% GameStop common stock, which we will get to in just

Starting point is 00:22:26 a little bit because Ryan Cohen discussed this on CNBC this morning. Yep. That represents a 46% premium to eBay's ineffective closing price on February 4th, 2026. The day GameStop started accumulating its position in eBay and blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah, but let's go straight to the CNBC. So quickly, there was a question. So, Bislett said, can someone please tell me how GameStop has $56 billion? It's not a $56 billion company.

Starting point is 00:22:57 There were questions. And Ryan Cohen went in the ring with Aaron Sorkin over at CNBC. Or Andrew Ross. I put the gloves on. Andrew Ross Sorkin on CNBC Squawk Box to interview about the GameStop eBay acquisition. We can play this. Walk us through how you can get to that price and how it would work. It's on our website.

Starting point is 00:23:20 It's half cash, half stock. but the details are on our website. Can you help? I've read them, but can you help our audience understand them? Yeah, which part exactly? Well, I think we can start with the idea that the market cap of GameStop is call it $11 billion. You have $9 billion on your balance sheet. Arguably, if you're providing effectively,

Starting point is 00:23:57 all of your stock, and then the cash that gets you to 20. You have this letter from TD, that's another 20. We're now at 40, but we're still off by, call it 16. And the 20, as far as I understand, while it's considered a highly confident letter, meaning TD's saying they're highly confident that they would provide the financing, it's not locked financing. Yeah, we'll see what happens. Founder, no.

Starting point is 00:24:34 Never doubt. I hear you. I understand that. I'm just trying to understand where the rest of the money would come from. It's half cash, half stock. I hear you. I'm just saying that that math doesn't get you to the, to the price that you're offering.

Starting point is 00:25:00 Ryan, that's a pretty straightforward question. I don't get it. Where's the rest of the money coming from? Andrew laid it up pretty clearly. I don't understand your question. We're offering half cash, half stock, and we have the ability to issue stock in order to get the deal done.

Starting point is 00:25:19 But the full details of the offer are on our website. But you're on our air. We thought we'd get Oh So I don't understand Your question Where's the money coming from That's the question

Starting point is 00:25:40 You're wondering Record scratch, freeze friend You're wondering how I tried to buy a $55 billion Company with $40 billion Earmarked Yeah so I don't Do you think he was expecting This to go out on Sunday, Monday

Starting point is 00:25:56 GameStop stock to pop Like crazy Yeah and it's actually down today. I guess that's possible. I don't know. I also don't understand why he can't just say like, hey, we're in the process. Like, we got a highly likely letter from a bank for 20.

Starting point is 00:26:12 Yeah, we need 16 more, but we're going to go get more letters from other banks. We're going to go get other equity investors. Like, this is a whole process. We're excited to announce this. And this is like our first close. Like, we're not like we're fully ready. Yeah. But maybe you can't do that.

Starting point is 00:26:28 I don't know. He took this offer to Paul, the chairman of the board. Yeah. Now, Paul has to look at this and be like, okay, is this a real offer? Yeah. And I imagine Paul will watch CNBC. And that's not going to give anyone a lot of confidence. Yeah.

Starting point is 00:26:42 It's, yeah, I mean, you can imagine it in the context of like buying a house. You know, you show up and someone says, like, I have 80% of the money. And like, my bank will underwrite me for 50% and I have 30% in cash. And you're just like, look, I need the full amount. I need the full amount from someone. I'm wondering what Ryan, I'm, you know, eBay is an incredible business. It's been remarkably durable.

Starting point is 00:27:09 They've faced an onslaught of competition for every single category from sneakers to watches, to art, to name any cars, right? Any category there is like a vertical competitor to eBay. And yet the business has done, has been remarkably strong. Yeah. It's up pretty meaningfully.

Starting point is 00:27:28 this year, right? Like, management is executing. And, and again, it's unfortunate for Ryan and his bid that the most viral sort of video clip out of all of this is just him failing to answer, like, you know, a pretty straightforward question and not having an opportunity to talk about, okay, why do you, you know, presumably if you're buying this company,

Starting point is 00:27:51 you think it can and should be worth a lot more. Why, what's your plan? what's your plan for the business? Why are you better suited to run it than Jamie, who's been in the seat since 2020, worked at eBay from 2001 to 2009? So he's a veteran, knows the business very well, and you're coming in $16 billion short. At least that's what it looks like. And anyways, this bid could be over before it really started. Yeah. Or, I mean, it could attract a bunch of investors who want to line up and fall in line and wind up producing the 55 or so required. But it does feel like it's a ways away. Anyway, we have our first guest of the show in the waiting room. We have Matt from Alpha School. And here he is in the TDPN Ultram. Nat, how are you doing? I'm doing great. How are you guys? We're doing fantastic.

Starting point is 00:28:51 Incredible to finally have you on the show. We enjoy, we love when your name. pops up in the chat and we've both loved following your career the last I appreciate that guys as long as I've been on the internet it's awesome to be here I mean I've been following you guys since it feels like it must have been the first week or two wow love the show and I love seeing you guys crush it it's brought a lot of joy to my my weekly commutes so thank you guys for doing this amazing and yeah we're we're super excited I personally am excited that that you're at Alpha school now because yeah Alpha School is moving into my town, it looks like. It's early, early stages, and I've just been curious to learn more

Starting point is 00:29:32 about it. And, yeah, especially, like, what you're working on specifically. Here to answer any of it. Should we go back further? Open up all over. Yeah. Yeah. Why don't you give us an overview of, like, everything you've worked on the last few years? Because, like, every, every, I think you've, you have done a better job than almost anyone at identifying these sort of, like, big macro. technology changes in cycles and then like just like experimenting and innovating in a bunch of cool ways so um so yeah take it take us back a little bit yeah so i i got the entrepreneurship bug in high

Starting point is 00:30:10 school didn't have many ways to get into it then and then in college i went to business school and figured business school teach me how to be an entrepreneur realized really quickly that that wasn't the case and ended up just trying to go figure it out of my own so got super deep into SEO ran a successful SEO focused marketing agency for a number of years until 2021 basically handed that off and eventually got acquired. But that was kind of my first taste of seeing this cool thing in technology, trying to figure it out, trying to build a business around it, and just feeling incredibly fulfilled by getting to do that.

Starting point is 00:30:47 And so when I stepped out of SEO, I got pretty heavy into the personal knowledge management space. So I was doing all of the Rome research and Notion stuff, got deep into crypto during that 2021, 2021 era and was writing smart contracts and working with a couple teams through that. And then when building with AI really became a thing, a few years ago, I was using cursor right at the beginning well over a year before we had these terms like vibe coding and whatnot. It was just completely obsessed with it and was trying to figure out what I wanted to do with it for a while. I had a course that was really early on how other people could start learning how to vibe code. And at the end of last year, I said, okay, I actually want to stop doing the Solopreneur thing. I want to kind of like go after something much bigger. And while I was in the process of figuring that out and tinkering with a few bigger things I wanted to work on,

Starting point is 00:31:44 Alpha School reached out to me. And they told me that they were working on this wild idea for a new high school. and they were looking for a kind of scrappy AI native entrepreneur to come help them build the curriculum and get the whole thing launched and built. And it was kind of one of those like, do you know anybody else who would be good at these kinds of things like you? And I'd never considered taking a job, certainly never considered working at a high school. But when they told me they wanted to build this and that they were looking for somebody like that, I was just immediate like I had to come do this. Okay, before we get into Alpha School, why you're witnessing the explosion of LMs, why didn't you get into like answer engine optimization or AI observability? I'm sure you thought about it and given your background building the SEO agency to exit.

Starting point is 00:32:41 I'm sure you would have been well positioned to do well there. Did you consider that at any point? I thought about it and to be honest, I was just kind of tired of the SEO game. I'd been so deep in that for so long that I just kind of didn't want to go back to that. And I just loved how fun building stuff with these AI tools was like on the software side. And that was kind of the route that I was starting to go down. And, you know, thankfully, like, this still kind of heavily scratches that itch because, like, even though this is a school and we're learning how, you know, we're teaching and trying to build all of that out, so much of the internal systems and how we figure out to teach this requires getting really deep with AI and building a lot of our own internal custom software. And it's really cool to, like, be a part of a school that's also investing so heavily in that.

Starting point is 00:33:39 And so it's been fun getting to bring those skills in that energy here. Awesome. Introduce, reintroduce Alpha School. We've talked about it on the show. But for anyone in the audience that isn't familiar, I think it would be helpful. And then I want to talk about the program that you're building out. For sure. And just to double check, because I had to get a message that my mic wasn't sounding right.

Starting point is 00:34:01 Is it not plugged in? I'm looking at it. It looks like there's no cable. Oh, my God. I was just like looking at Wow Okay and then I think you probably need to select it And change the other thing over

Starting point is 00:34:14 But I was looking at it I was like wait that's not plugged in at all It's hilarious There we go Works New studio you know Oh yeah this is Oh this is nice

Starting point is 00:34:28 Much nicer silky smooth now All right yeah Sounds awesome Okay Okay so yeah Reintrodu Alpha School for those who who don't know the model because I think it was many people learned about it through Invest Like the Best I guess I mean that's how I learned about it.

Starting point is 00:34:47 Well there was like there was like you know yeah rumor mill around it right there was like there was a lot of there was a lot of lore a lot of people were curious and then invest like the best I felt like was like this first mainstream kind of exploration of the vision and what have been done to date and how you guys were approaching it. But obviously, that was before you joined. Yeah. So Alpha School actually started 12 years ago with McKenzie Price here in Austin, and it was a smaller, smaller micro school, but still built on that fundamental idea that you could compress the entire academic day to about two hours if you leaned on what we know about learning science, which, you know, the teacher in front of a room of 25 students trying to teach everyone at the same time

Starting point is 00:35:34 is just not the best way to learn. And if you were able to kind of customize the education to each student based on where they were, you could get that day a lot shorter and then give them a lot more time for learning life skills, doing these interactive workshops and making them just love going to school, right? Because no student feels like they're being held back or not getting a chance to get caught up. And so when generative AI came on the scene in 2022 or so, Joe Lamont, who he and McKenzie had known each other for, forever and his daughters were already going to alpha school he kind of went to her and said hey

Starting point is 00:36:09 with AI we can now take this model that you have figured out and we can start to we can bring this to a billion students we can make this the way that like every student gets to learn and that's when alpha school that he brought um he brought big investment to alpha schools started like building it up expanding it and that's why a lot of people feel like it just started in the last few years because that's when it really started to grow and take off And Joe had been kind of, you know, he wasn't very public-facing figure for a long time. He wasn't doing podcasts, wasn't doing interviews. And then a year ago, he would have loved the podcast circuit back in like the 90s.

Starting point is 00:36:49 He would have crashed it. Yeah, yeah. But then a year ago, he did that invest, like the best interview. He did the Colossus piece that I know a lot of people have read. And that's when he started really sharing more about what they've been up to at Alpha School. They'd had a few years of working on the new model, building out the software. And now it's just, I mean, it's growing like crazy.

Starting point is 00:37:11 We were at an info session in Boca Raton two weeks ago, and they thought maybe 20, 30 families were going to come, and there were over 100. And the school that they were planning on starting there got filled up just almost immediately, right? It's been very cool to see how much excitement there is, especially in the crowd of younger, more tech forward parents who have been thinking for a long, time that like, hey, there's got to be kind of a better way to do this. And it feels like we really have the technology now to make it happen. Is the, is the model like start with kindergarten

Starting point is 00:37:47 and then just grow with the class? Or is there actually like, let's get a bunch of juniors in high school to jump in so that there's like continuity from day one and it feels like a full K through 12 experience? Or is it like grow iteratively? We do both. So a lot of students do come in and we go down to pre-K four right now. So pre-K-4, kindergarten, first grade, a lot of students do come in at that age range. Again, just because there are more younger parents who are excited about the way the school day works. But you do have a lot of kids who come in starting in high school. We have people, I'm at the high school in Austin right now and we have students who transferred in as 10th graders, 11th graders. It's kind of if they're really

Starting point is 00:38:33 interested in something or if they feel like their current school isn't working for that and they want to try this model instead will take them basically any time of year and the benefit of starting earlier for a lot of students is that they don't have as much catching up to do a lot of students transfer in from even really good private schools and it turns out that they're a year or two behind in math or reading or one of these subjects and so they do go through that catch-up period but again it goes back to just like the nice thing about how having software that can personalize the learning experience to each student. If you're a seventh grader who's actually in fourth grade math, there's no seventh grade math class that will help you get caught up. But with time back, with our learning software, it can identify, oh, okay, you actually need to go back and really memorize your times tables because that's becoming a big bottleneck for you in trying to learn algebra. So we're going to go back and do that first. Maybe Ryan Cohen. If you guys could, uh, Ryan, there was being asked some,

Starting point is 00:39:33 math questions about the eBay take over this. Oh, yeah, the eBay for the way. And there was some like adding and subtracting that they were doing that, that, um, wasn't adding out. It wasn't, things weren't adding out. Sign them up. Yeah, extension program. Um, talk about the founder school initiative specifically. How does that fit in? It feels like, uh, like a different track or how, how separate is this from sort of the, the typical alpha school experience? Yeah. So, So basically, you know, we have our high school in Austin, and we've been looking to expand the high school offering for Alpha. And so Joe went to all of the high school students last fall, and he said, hey, you know, what would be the just best, most incredible version of high school that you can imagine? And we have a really good mix of what the high schoolers are working on in their afternoon time after they finish their academics.

Starting point is 00:40:28 We have a girl who's putting on a Broadway play. We have a few students writing books. One student opened the biggest bike park in Texas, but we have a lot of students who are really interested in entrepreneurship. And so they said, you know, it would be incredible if we had a highly dedicated entrepreneurship program so that we could just go after this as much as possible because we have all this time in the afternoons. And we, you know, we just want more mentorship and more guidance and more support around

Starting point is 00:40:56 trying to like build big businesses. And that got the idea going in his head to, say, hey, what if we opened a high school where we were essentially saying that we're going to get the best mentor network possible? We're going to get the best resources possible for whatever bottlenecks challenges you're hitting. We're going to build the best curriculum possible so that we can take what might be years and years of kind of fumbling around in the dark like a lot of people do when they first start that entrepreneurial journey. And we try to teach all those course skills as quickly as possible, give you the ability to develop expertise in whatever areas you're most interested

Starting point is 00:41:38 in, and then build a big business by the time you graduate high school. Because nobody really doubts that a sophomore at Stanford can drop out and build a million or billion dollar business. And so, like, why can't a 17 or 18 year old? And kind of like working back from that said, okay, well, you get this incredible network at a place like Stanford of other students who can be your co-founder so you can work with and be motivated by. You get these incredible mentors who come in, talk to you, you know, give you some advice, or just inspire you around what's possible. And then there is some stuff around the academics and the classes. But there's functionally no reason we can't pull that back down to the high school level, especially considering

Starting point is 00:42:23 that these kids get like five hours in the afternoon every day after they finish their academics. So that's 4,000 plus hours over four years. And then if you just cut out all of the things that people usually end up wasting time on or don't have the resources to figure out, it is eminently doable to get somebody to a million dollars by the time they graduate high school. Yeah. At the very least, like teaching, teaching some of these fundamentals, I think this has been the thing that YC has always done so well, which I think is amazing because it's all open source, which is like, okay, like, what are the, what are the, you know, they can probably, teach like the foundation of like a Y, what what the YC method to making a startup is in about an hour.

Starting point is 00:43:05 And it's like simple rules like talk to your customers that oftentimes people don't do early in their founder journey where they're just like, oh, I have to spend like six months like making everything perfect before I can talk to a customer. And it's like, no, you actually can just go have the conversation first before you even start any of that. And so there's basic things where if you can get somebody to like start the clock earlier, start taking shots on goal, start going through the motions, but doing it with like some general guidance that you, that again, like the you can get on the internet, but you have to kind of piece it together. And there's something about having having that in person experience that I think is very important too. Yeah, having somebody in the room with you to say, hey, you are wasting time on this thing or you. you need to go try to sell this to someone, that five-minute conversation might save you months of your life. But a lot of people when they're starting out don't have someone to do that.

Starting point is 00:44:06 And it's like just by providing more of those touch points and having that expertise in the room with you, I think you can really, you can accelerate it quite a bit. And YC is a great example where they do an incredible amount of work with people in three or four months. And we get four years. So I'm, I think we're going to be. The other thing that's very, a ton of progress.

Starting point is 00:44:25 Very cool right now. John and I were at dinner last night and one of John's friends is, will likely get to like a million of ARR building a business that historically he would have always needed to raise venture capital for, but given all the AI tools, like he just doesn't need to raise any money. And so he has all this. One, he has optionality. He's not like raising money, locking himself into this thing for a really long period

Starting point is 00:44:51 of time. But it's very cool because historically, capital would have been a huge constraint for like a high school student where, you know, even if they can go to a private school, they don't necessarily have like another, you know, a huge amount of money to invest into some idea. And now they basically don't need any capital for a huge number of different types of businesses. No, it's totally true. It's when the, like, AI is the big thing that removes the bottleneck around like, why can't a 17-year-old build a million-dollar business? where you don't need to spend as long developing deep expertise in programming or software development

Starting point is 00:45:30 because, you know, you can certainly for a V1, you can prompt a lot of it. You don't need to raise a ton of money for hiring or design or build out or any of these things because, again, you can get like an initial version going with AI. And also, if you're a 15-year-old trying to get started and you're trying to hire a marketing person or whatnot, like a content marketer might not want to go work for a 15-year-old. Like that'd be kind of an awkward thing, right? And so if you can just use AI to get started and get your business going, yeah, you take out a lot of the capital requirements, take out a lot of the expertise buildup requirements. And again, it just shortens that timeline.

Starting point is 00:46:07 It makes a lot more things possible for them. How are you thinking about the, or how are you wrestling with the idea that like so many great founders did not study entrepreneurship or business? They usually, like, when you give the other examples of like the person that's putting on a Broadway, I don't think that that's not an entrepreneurship path. I see that person and I'm like, oh, they could wind up owning like a media company or they could wind up buying and, you know, consolidating the Broadway industry.

Starting point is 00:46:37 I don't know. There's like a bunch of different ways. And when I look at entrepreneurship in America, I think it's one thing to look at like Silicon Valley tech startups. And there's another to think about like the millions of small businesses that are like a dentist office or a law firm or something. or some sort of business that's built around a unique expertise with a completely different path. And then entrepreneurship is just like the cherry on top at the last second where that individual

Starting point is 00:47:03 realizes that they just want to build their own company around their expertise or their discipline. And it feels like there's always this wrestling with like if you only learn the entrepreneurship piece, like what industry are you actually an expert in to go and offer an improvement in? people always go back to like Mark Zuckerberg. I think he took some computer science class, but he studied psychology, and that makes a lot of sense in the context. So Sundar has an MBA. Satya has an MBA. Those aren't the entrepreneurs, though.

Starting point is 00:47:33 No, no, I know. But this is notable, right? They're like the professional CEOs. So Sundar, Satya, Andy, and Tim Cook, they all have MBAs. Jensen, Elon and Zuck do not have. Yeah. Based on. Yeah, that being said, though, I think, I think, yeah, it's a good point, which like clearly, clearly you,

Starting point is 00:47:51 if you have a vision and you work very, very hard and you get some luck along the way, you don't need to be that good at business and you can like, you just end up being successful. You learn it on the fly. Yeah, you learn it on the fly. But at the same time, I think a lot of entrepreneurs, like people that want to be entrepreneurs will spend five years

Starting point is 00:48:08 wasting a bunch of time, kind of like learning these foundational, like, truths about entrepreneurship. And so, I mean, when I think back, I would have loved to learn all the kind of like the YC way in high school. right and just take some shots on goal even if it didn't even if I ended up going to college getting into something totally different and then coming back to entrepreneurship later yeah and to be

Starting point is 00:48:31 clear you know the the students aren't going to be like studying entrepreneurship and there's going to be this very heavy focus on expertise building in a domain that you are passionate about because that is going to lead to a lot of these opportunities and that's something that alpha school already does really well. The high school students are heavily encouraged to, or actually they're required, to pick a topic that they really want to develop deep expertise in, and then we're teaching them how to use AI, how to use writing, how to use producing, just like any kind of content media to develop that expertise. And that's going to be a core part of this program as well. We also have this whole additional aspect of it that's really important to us that we're calling

Starting point is 00:49:13 kind of the philosopher builder canon drawing on Cosmos Institute and working with them a bit, where we don't just want people who are like slinging mobile apps to try to hit a revenue goal and then driving around in Ferraris, right? That's absolutely not the kind of like ethos that we're trying to inspire here. Yeah. I want to see that I want to see the Alpha School parking lot in Austin just Uris wall to wall. Yeah. No.

Starting point is 00:49:46 So it's like, yeah, every, yeah, by the end, they all started. econ business and then six months in they're ripping a course and they're like here's how I got a Uris at 16. This is not going to be a school on how to sling peptides or anything like that. That's Andrew Tate's school.

Starting point is 00:50:03 No, I mean we actually have a really rigorous freshman year reading list that's basically grounded in philosophy and kind of like trying to develop a broader sense of the world and a lot of it goes back to one of the big challenges with being an entrepreneur

Starting point is 00:50:18 or kind of doing anything in high school is getting out of the local high school mindset. Everybody wants to build an app or something for their peers, right? They're not thinking about the bigger problems in the world they could be going after. And so by spending a good amount of our time, encouraging them to pick areas they want to become experts in and really deeply understand and encouraging them to think and read and write about like the bigger challenges in the world that they might want to attack, then we can kind of encourage more of that thinking where it's not just, again, we're like, we're not trying to build hustler school here. And so, like, I'd completely agree that you couldn't, like, just teach entrepreneurship and expect

Starting point is 00:50:58 that to lead to something interesting. It needs to be balanced for, for it to work. In terms of the best way to learn, do you prefer the Stanford method where it's just two people, some microphones having a conversation? Or do you like the Dorcasch Patel podcast way where there's an expert at a chalkboard sort of breaking things down step by step? Because it seems like there's two sort of ways to learn these days. John is doing a bit because people have been doing teaching classes at Stanford that are effectively podcast. It's just like doing Stanford classes. It's just a hilarious inversion.

Starting point is 00:51:34 But I am interested in like, you know, what is the world for just like hearing from an expert in an unstructured way, more conversational versus like, you know, having information delivered and more structured pattern? Is it like, you know, right tool for the job, right person for the right flow? Or is there one way that you think is actually better? I think it really depends on what you're trying to provide. So if we were doing a workshop or a session on, you know, earned media or on TikTok or something, I think that having very short informational sessions followed by implementation, practice, you know, finding issues and then getting very structured feedback to, you know, the challenges that you're running into, that's going to help accelerate your learning the most on that, like, very tactical

Starting point is 00:52:21 skill. But where the conversational longer form stuff becomes super valuable is hearing stories of how other people got started or what other people did, where it's not so much, like, here's exactly how to do this thing, because it'd be really hard to hold an hour-long lecture on how to do Facebook ads in your head. But hearing how somebody thought about developing expertise, finding business opportunities, getting started, especially hearing those starting stories

Starting point is 00:52:50 from people who are now really successful. I guess something that I think those Stanford classes do so well is they make these incredibly successful people feel very human, very approachable, where you realize like, oh, they were just like me at one point, just getting started. They didn't have all these resources. And those kinds of sessions are extremely valuable as well.

Starting point is 00:53:12 So I think it's just kind of like what problem you're trying to solve at that point in time. Jordan anything else? Any predictions on end of this year? Was Joe inspired by PMF or die for the structure? Because it's a hundred, isn't it 150K a year? But if you don't hit a million dollars in profit, you get your money back. Isn't that it?

Starting point is 00:53:35 That's correct. There we go. That's PMF or die right there. We got to lock these kids in a room, stream it. This is the way. Yeah. No, I was going to say, yeah, yeah, predictions on, on, highest run rate

Starting point is 00:53:47 12 months from now from a student how big do you think they can go? Because in today's world like if they're not getting to nine figures in 12 months like they should probably just jump down. Yeah I mean obviously you can't hold me to this

Starting point is 00:54:06 but I would not be shocked if by the end of year one we have at least one student who's well past the 100K mark. I wouldn't even be horribly surprised if you've got a student who's like halfway to the million or even there, right? Just you find that right thing, you lock in, and we're going to have some students coming in who are already making some progress on their own things.

Starting point is 00:54:27 Yeah, whenever we talk to YC founders, Teal Fellows, are always like, they were always making money in late middle school, high school, and it was, they were picking ideas that had kind of a cap to it, right? Or they were competitive market. Or super competitive, but if they just applied themselves to it, to a category that was made. Maybe a bit more boring than Minecraft stuff.

Starting point is 00:54:49 Yeah. Minecraft servers, hacks and there's no limit. But I'll tell you what, guys, you know, it would be a lot of fun is I think these students are going to crush it. And so if we want to do like a TBPN Founder School demo day next June, I think that'd be pretty sick. I'm so excited to see what these kids do. Yeah. Let's do it. Yeah, we have to set the record for the youngest TVPN guest.

Starting point is 00:55:12 We had that kid who applied to YC, right? remember from Australia he came in for a couple minutes how old was he was he 16 I think he was 13 I thought he was 13 yeah he's really yeah he came in with his dad yeah and then if any of them build products that we can use let us know well we'll be happy to be beta customers yeah yeah I mean I'm going to be shouting from the rooftops everything that they're working on so I'm I'm very excited about that amazing uh well congrats I personally I don't I don't think of myself as like a favorite customer. I'm more of a Sigma customer, but you can do it. Okay. I thought you were going to buy a product from Alpha school. I want to, I want to be the Sigma customer. Anyway, thank you so much

Starting point is 00:55:53 for taking the time to come chat with us. Appreciate you guys. We'll talk to you. Great to get chapp. We'll talk to you. We'll talk to you soon. There's some back and forth on the timeline. XAI's GPU fleet is running at about 11% utilization exposing how hard it is for AI labs to fully use expensive Nvidia hardware. This is May 2nd from the information. Trace Cohen says, How is this possible? There's infinite demand. We're going to talk to Anjny in a little bit about the infinite demand because it is real, but there's some extra context here.

Starting point is 00:56:22 It says 11% almost certainly refers to Flops utilization, not 89% of GPUs sitting idle. If it were the latter, they would just throw more GPUs at training. It's notoriously hard to maximally utilize GPUs even when saturating inference capacity due to things like uneven mixture of expert demand and memory stalls, the 11% figure. means their inference stack sucks with a large contribution from lacklustre architecture, contrast, deep seek, who do an amazing job maximizing utilization at every level, including innovative MOE load balancing. And I mean, even if the utilization is low, it's like, well, the team seems ahead of the curve

Starting point is 00:57:00 solving the problem by acquiring cursor or partnering with cursor, because the whole thesis of the cursor deal was XAI has compute capacity. But Tyler, what are you thinking about this? 11% does not mean only 11% of the GPUs are on. It's just that like, you know, the maximum like flops that you could get, say it's like a thousand. You only have like 110 like flops basically actually being used just because of like what their like inference stack is. Yeah. So the GPs are still on.

Starting point is 00:57:26 It's like they're just like sitting idle. But you want it at 100% right or close to it. Yes, but that's like extremely hard because you have to like. You have to be using it all the time. Yeah, because you know, you're constantly loading, you know, memory on and off or whatever. And you might just see uneven demand. I mean, that happens all the time with these AI products where there's a boom. Like I bet you that you that, uh, you.

Starting point is 00:57:42 utilization spikes during like big moments on X because a lot of people are tagging like at GROC is this real and a lot of people are are using GROC in X when there's a particular you know there's those moments where it's like an election or something and everyone's on X talking about the thing the raid on Osama bin Laden or something like that and everyone's like querying at the same time to get more information that's probably lighten the GPUs on fire during around big model releases and whatnot anyway Jordie where do you want to go next. We can bring in the waiting room.

Starting point is 00:58:14 We can do that. Let's do that. From CASA. Michael is in the waiting room. He's the co-founder and CEO of CASA. He's here to talk about subscription-based home ownership platforms. How are you doing, Michael? Good to meet you.

Starting point is 00:58:25 Good. Good to be with you guys. Thanks for having me. Thanks for having me. Great to have you on. You announced a new round last week. Couldn't get you on on Friday, but you're here today. Hit us with the news.

Starting point is 00:58:36 We raised a $20 million series A, led by four rudder. bunch of great folks around the table as well, including a recent guest, Mr. Travis Kalanick himself. Whoa. How did that happen? Well, I give you a little bit of backstory. I left school when I was 18 to join Uber. Uber was 50 people at the time and started out as an unpaid intern working on the supply side getting L.A. up and running. Got a job off a couple of weeks later and decided to drop out.

Starting point is 00:59:09 and a year later moved up to SF to work on Product & Eng, which was really my passion. Did you know, did you in all of your 18-year-old wisdom know that Uber was a special company prior to joining? Or were you just happy to have an internship? Maybe a little bit of both. I mean, it was obviously something that spoke to me. But I think for me it was a way to actually just get immersed in the culture I really loved. But from afar, I mean, I know you guys are in L.A. I'm from L.A.

Starting point is 00:59:37 and certainly it wasn't what it even is now with exposure to startups and tech and all of that. Totally. Yeah, it was very fortunate. It actually happened on the backs of a tweet, which is a fun story for maybe later. Was the whole team built off of tweets? Basically, yeah.

Starting point is 00:59:56 Basically, there was a TechCrunch article that they were coming to L.A. I followed it a few months later. Someone tweets from the account says, we have a press lunch, email us here if we missed an invite. and I just hit up this email as an 18-year-old kid who was not in the press, but just wanted to meet these guys, and they sent me an invite and was like, come meet our CEO, Travis Kalanick, and it was this restaurant in West Hollywood,

Starting point is 01:00:17 where I was like half the age and height of anyone there and shmozed my way into an internship with the GM that they just hired for L.A., and that was kind of the start of it. That's amazing. How long were you there? Was there for almost six years. I left the day that TK. left. I knew that, you know, I didn't want to be there. He wasn't there. I knew I was only going to work for him or myself after that. And actually, I ended up working for him again a few months after he, he left in all that madness.

Starting point is 01:00:48 He was calling for round two with cloud kitchens. And so I joined him there. We were about 30 at the time and was working on product and design for software side of our business until 24. That's incredible. It's quitting in solidarity with your boss. This is underrated. Underrated. All right.

Starting point is 01:01:08 Talk about Kasa. Yeah. So not sure what you guys already saw, but Kasa is a personal property manager for your own single family home. You know, I guess the origin story is coming off with Uber experience myself. I was really fortunate enough to consider buying my own first home. And I think, you know, like any first time home buyer, you're super excited about the emotional aspects of having a home and making it your own and sharing with your friends and

Starting point is 01:01:31 family. It's also this huge financial endeavor, probably the most expensive thing you ever buy. And pretty much everyone realizes once they haven't for the first time that, you know, at least a part of the ownership experience actually sucks. It's basically another part-time job. And, you know, the home is basically a thousand products in one. There's always stuff that's going wrong, always stuff you want to make better. And you're really left to deal with it on your own. And right, most people don't have the time, the expertise, negotiating skills, sometimes even interest to deal with all this stuff. And they'd rather be spending that time with their

Starting point is 01:02:01 loved ones and go on vacation and doing fun things. So, yeah, makes total sense. I've, I, there's so many things around my house where I'm like, months go by where I'm like, I should, I should do that or, you know, whatever. I just, I don't get around to it. And then, and then stuff comes up, which is like, oh, do you have, like, where's the floor plan? And I'm like, I don't know.

Starting point is 01:02:23 It's like an image email. Yeah. What's pink color? When did we redo the, when did we change the HVAC, Phil? or like all this stuff and it just gets lost in emails and tech. So I totally get the pain point. Has there been a lot of shots on goal on this concept in the way that you're positioning it? Like it seems like one of those ideas that I'm surprised that I can't think of like a name brand company that does this.

Starting point is 01:02:51 Yeah. So I'm sure you've looked for it. Yeah, it's a good question. I mean, if you look at the companies, you know, at the intersection of home services and tech right now and the few that remain, It's like the Yelps, the thumbtacks of the world. And, you know, they might market themselves as homeowner products. I don't actually believe them to be. And you don't have to look farther than their P&L to understand that, right?

Starting point is 01:03:10 100% of the money they make comes from the vendor side of the business. So these guys are happy to send you shitty plumbers or super expensive plumbers if those guys are paying the platforms $20 a click. And to be honest, the vendors hate them too for the same reason. So I think this is prime for disruption. You know, to answer your question, I would say I don't really think is possible to build this in what I would say is the right way to build it. until the last couple years with the multimodal models. Basically, how Kasa works under the hood is you sign up, and the first thing we do is we come to your house

Starting point is 01:03:37 with a bunch of specialized hardware and software. And in the span of a couple hours for a standard-sized home, we will develop an extremely intricate understanding about everything about the home that matters from a servicing perspective. So it could be a full 3D model, LiD scan of your house, every appliance, every electronic,

Starting point is 01:03:53 every exact paint color, the light bulbs inside of all your light fixtures. So to do that in the span of a couple hours with one or two humans was impossible a couple years ago and is now possible. And that's really what powers everything that we do. So we're actually really fortunate we got to take this like very first principles look and approach to all these homeowner problems. And it starts with the physical day-to-day pain points, the things that a handyman might

Starting point is 01:04:15 help with or another type of physical vendor. But there are also these other kind of meta-concepts around the home, anything from your property taxes, to utilities, to all these things that we now kind of monitor end to end and we'll do a more interesting job of helping you out with over time. And I'm assuming you have the, this is an account that theoretically should stay with the home because I remember when I bought my house, I felt terrible because I would hit up the seller, like the previous owner all the time like, hey, do you know where this thing? I just was like, it was fair to ask all these questions and they were happy to answer,

Starting point is 01:04:53 but it was still like I felt like it was wildly annoying. and if there was just basically like a database that all this stuff had been like dumped in and a single source of truth, they could have passed that to me. I would have been happy to pay to have access to it. So theoretically, I'm sure you're building out your model for the business, but you're hoping that, you know,

Starting point is 01:05:12 a house could change hands multiple times and the CASA would be kind of like continuous through the whole experience. Yeah, absolutely. I mean, not to bring in a legacy name, but you can kind of think about it like a Carfax for your home too, or certainly that's the direction. We would move in, right?

Starting point is 01:05:30 So you have this neutral entity. Obviously, we have no particular affiliation to any of the owners or whatever, and we're able to just provide this very unbiased view at how has the house been maintained, what are the general costs to maintain the home? And that might even become interesting in other contexts, right? Like to lower your home insurance costs, if we can provide the insurance entity with like a pardoned understanding of these are the proactive maintenance plans that have been happening for a home,

Starting point is 01:05:53 there's no reason why you shouldn't be able to save money. there. And so these are really all the opportunities when you're taking this homeowner approach, homeowner-focused approach that you can kind of pursue that don't make sense for anyone else in the picture right now. Very cool. And have you been in stealth the last couple years or like what's the history of the company? Yeah, yeah. So we've been in a private beta since Q3, 2024. We basically put out this very, very simple app. It was basically a two-way messaging client for a few of our friends or family in the bay. And we basically said, hey, whatever problem you have with your home, send it to us. And we'll try to figure it out. And for us,

Starting point is 01:06:25 us, you know, we weren't charging anything. We just wanted to understand what were the bounds of requests that we were going to get. Where do people want to have a concierge for their own home up with? And on the other end of that, that was basically my co-founder, Michael and I just responding to these issues, calling up vendors, figuring out handymen on the fly. And, you know, I'd say in the span of a few months, we understood where the sweet spots were, what people were looking for, and the beta itself took on a life of its own. So what started as a handful of folks ended up with hundreds of homes all in the bay,

Starting point is 01:06:54 and eventually in LA. On the backs, actually, most notably on the backs of a tweet from one of our beta members, Lenny Richitsky, I know you guys know. Nice. And yeah, he shared it. You know, he was like, hey, you know, on a few extra homes. And we were like, sure. And then he tweeted about it. It was a very sweet and generous suite.

Starting point is 01:07:13 And the way it just went through the roof. And honestly, we've been working through that one for many months. So that was the first time a lot of folks had heard about us. And, yeah, that's what we're right now. Yeah, what does it take to launch a new region? You probably know what you've launched a few regions, I'm sure, back in your Uber days. Yes. Actually, the first thing I was doing at Uber was standing around at LAX knocking on a black car, town cars, recruiting drivers.

Starting point is 01:07:40 And by the way, at the time, it was like primarily 50 to 60 year old men who had not seen an iPhone before. And we're like, hey, do you want to, you know, change your whole life and use this phone and drive people around? And yeah, I think for us is very similar in that, you know, I see. this as a parallelizable playbook. So we, you know, at Uber, we spent the first couple of years really perfecting what was the playbook for landing in the city and going from zero to 100. And once that really was understood and stabilized, you saw us launching, you know, 50, 100, 150 cities a year. So, you know, we'll do the same thing here, which is we're understanding what that looks like. L.A. is the second test of that. We got a handful of more markets coming this year.

Starting point is 01:08:17 but I think you'll see us kind of do that exponential growth curve on the number of markets we're supporting. You know, also the benefit of others. How do you think about disintermediation? Obviously, Uber did fantastically, even though you meet a great driver. No one really jumps from, you know, the driver that took you to the airport is now a full-time employee. Very, very rare. At the same time, like dog walking had a much harder go where people found great dog walkers and said, how about I pay you cash and we cut out the middle man and disintermediate.

Starting point is 01:08:50 It feels like there's some sort of interesting unlock here with the data around the house that is valuable and provides enough value that both parties would want to stay on the platform. But walk me through the thesis of avoiding disintermediation at scale. Yeah, it's a great question, John. Actually, I think those are two perfect analogies, right? So for Uber, it didn't make sense because a huge part of the value for Uber was that you can get in a car in two minutes. even if you get a driver's number, forget about all the insurance and the other things you get. But it just doesn't make sense to call a guy and wait an hour from to come get you. It doesn't even make sense for the driver.

Starting point is 01:09:23 The Dog Walker one's a bit different, right? Because the dog walker presumably lives in your city and, you know, once you have that connection, you're good to go. Yeah. And it's like every week, the exact same pattern. Exactly. Exactly. Exactly same service and all of that. We're very much in the Uber camp.

Starting point is 01:09:37 And I tell you that honestly just from experience, which is that we've never had a single one of our vendors or handymen poached by a homeowner. And the reason for that is obvious to me from the inside, and I think we'll become obvious to folks as a service gets bigger and more prevalent, which is that all of the things we do to surround the visit, right? So in the CASA app, you can book a handyman visit, and you can say, hey, I want these three things done for my home. What actually happens at that point is a combination of a bunch of fun AI stuff and our own kind of concierge human team.

Starting point is 01:10:04 We'll look at those tasks, we'll look at the history of your home and all the information we have about your home to do that thing really well for you, right? So you might say, hey, I want to hang up a bunch of family artwork in the hallway. And we'll come back to me, we'll say, hey, do you already have the frames? Do you want us to find them for you? If you do, we'll come back to you with a bunch of options based on your past preferences, or we'll start to understand your preferences. And we'll actually have those frames ready for you or in the hands of the handyman before they even show up. And all of that context about your home and the preparation for this work is what really makes it valuable. And also getting the right person to your home.

Starting point is 01:10:35 So one handyman might be really great at carpentry and another might be great at a different skill. Setting the right person at the right time is super important. and also the fact that you can just get someone at your house, usually within 24 to 48 hours at any time, and you can't do that with one person. So handyman exhibit spiky intelligence, and you solve that. That's what's going on.

Starting point is 01:10:53 Yes. Perfect. I love it. Well, congratulations. I mean, this seems like, you guys are based in SF? We're based in SF. We launched with SF in L.A.,

Starting point is 01:11:03 so it's SF Bay Area plus L.A., a bunch of other stuff to come, and yeah. Oh, and Fourrunner earlier the round. Fantastic. Well, very cool. Thank you so much for the time. I'm going to sign up. I'm definitely.

Starting point is 01:11:12 I'm going to sign up. I'm going to sign up. I love it. Have a great rest of your day. We'll talk to you soon. Great stuff, dude. Thanks, John. Goodbye. Up next, we have Maddie Hall from Living Carbon with a absolutely massive plan to work on

Starting point is 01:11:25 United States reforestation. Maddie is in the waiting room. But we'll bring her in to the TV panel show. Maddie, how are you doing? What's going on? Oh, thank you all for having me. Thanks for hopping on on such a momentous occasion. Take us back, though.

Starting point is 01:11:40 Introduce yourself and the company, and then I want to get to this. a half a billion dollar deal. Absolutely insane scale. Congratulations already. I know. It's very exciting. I think largest amount of money that's been raised for aforestation

Starting point is 01:11:51 of degraded land in the U.S. Absolutely crazy. Prior to starting the company, though, I was actually working at Open AI. Oh, no way. Cool. Yeah. And explain Open AI for the audience.

Starting point is 01:12:04 No, I'm kidding. Very, very cool. Offa, you may or may not have heard of it. Yes, yes. Yeah, but I think it was pretty clear that the biggest blocker, between where we are today and super intelligence is energy, right? And the math just continues to get to get more compelling.

Starting point is 01:12:21 Like the hyperscale emissions have been up more than 50% since 2020. And at the same time, like a lot of data center projects are being blocked, like $64 billion, something crazy. So living carbon, our focus is on transforming old mine land and abandoned farmland into forests that either can produce carbon credits or sustainable forest products. And then we sell to the world's biggest companies, Microsoft, Google, Meta McKinsey to mitigate their increased emissions from AI and data centers. And then how vertically integrated is the business? Because I can imagine this being something where, like, you're working with a whole bunch of contractors in different forestry organizations and it's a lot of financial work to, you know, bridge the gap between the hyperscalers,

Starting point is 01:13:07 which you have connections to and the different forestry organizations. But are you planning on like building machinery, hiring people, planting trees, planting seeds, like how in the weeds will you get? Yeah, so I think of living carbon is like the general contractor of the project, overseeing all of the different pieces. That's largely what this octopus deal enables us to do. It is them covering the cost of all of the site prep, the planting, the land lease, all of the above. And then living carbon axes, really the overseer and the manager, of those projects, but then also shares in the revenue with octopus from their investment. So we don't own nurseries. We don't own land, but we have programmatic tools to identify

Starting point is 01:13:52 the sites that are best suited for our projects. And we do all of the end-to-end negotiation financing, off-take agreements, getting all of those pieces of the puzzle lined up. Are there labor shortages and reforestation? Like if somebody wants to get in on the AI boom, should they be learning how to plant a lot of trees quickly? I mean, I think it's possible. I think, you know, what you have when you're doing these land-based projects is value that is going to persist, right? Regardless of, you know, how much of the software space, the foundational model companies end up eating. So, you know, it's a great long-term cash-flowing business.

Starting point is 01:14:36 Yeah. The goal is a quarter million acres in the short. term, you've already done 25,000 acres. Walk me through lessons learned from that, what that actually looked like, how that initial first project came together. Yeah, so our flagship project right now is about 25,000 acres that were in the process of planting with our customers, Microsoft, Google, all of the above. And that's taking degraded land in Appalachia, so abandoned mine land and reinforcing it. And what's interesting about this project is it actually has the potential to offset the entirety of all of the emissions of San Francisco on an annual basis.

Starting point is 01:15:20 So our plan with Octopus is to hopefully 10x that, which would remove all of the emissions of New York City on an basis. Got it. Walk me through why mines cause so much deforestation. I think of a mine as like a hole in the ground. and, you know, it digs under, but it's not exactly clear-cutting, like, miles and miles, or is it? Like, what actually happens in the process of mining that affects the forest? Yeah, so a lot of these sites that were working on, they actually have been, like, stagnant since the 90s and just really sitting for decades. It's not just the mining pit, but all of the land in close proximity to it.

Starting point is 01:16:00 When coal went bankrupt in the 90s in the region, there was very little effort put into a lot of the remediation and restoration that would actually allow for that land to be put back into production. So these sites have been really liabilities on the balance sheet of a lot of the mining companies and private landowners for 20 plus years. Can you get me up to speed on some of the other like reforestation tradeoffs that are happening right now around carbon credits? Like I've heard one thing about like some carbon credits that are basically just designed where it's like, okay, we were going to cut these trees down, but we didn't. And so we want the credit and that feels like little bit edgy. And then also I'm wondering about international

Starting point is 01:16:43 because I imagine just on a cost basis, you could probably plant way more trees internationally because the land's cheaper. But does that offset carbon emissions like globally, potentially, but do the hypers get credit for it if it's happening halfway across the world? So how have you unpack some of like the hot topics in reforestation? Yeah. So I think our focus has really been on like an active intervention that wouldn't have otherwise happened without funding from carbon credits and the context that we have there. It's hard to say with certainty whether or not a tree would have would or would not have been cut down, but we can say with certainty that these sites would not have been restored

Starting point is 01:17:20 without our work. And so you're looking at these areas, and the natural rate of regeneration is very low. It is cheaper to do it in other parts of the world, and I think there are some amazing projects that are being developed in Latin America and globally, I think for us being in close proximity to the areas where data centers are being developed and having that regional focus such that our projects, you know, will offset carbon within the same region

Starting point is 01:17:49 where large-scale renewable projects are being built out and data centers being built out. And that's been desirable. Yeah, so it's not just abstract to the local voter who's making a decision around environmental impact. Yeah, it's one of the solutions to, how ugly data centers are to some people to just put a like forest fully surrounding it is that could we see that in the future oh I sure hope so I sure hope so that yeah yeah a magical data

Starting point is 01:18:16 center in the center of a forest I would love that I would love that too I mean that's what we're doing so very cool that's great awesome well tell us about the the new deal I want to hit the gong how big is the new deal up to 500 million Congratulations. And thank you so much for taking the time to come chat with us. And we will talk to you soon. Yeah, very happy you're doing what you're doing. Have a good rest of your day.

Starting point is 01:18:43 Sounds good. You guys too. Cheers. Go bye. Up next, we have Ajny. Is this the first time on the show? No, he's been on the show before. Second time on the show. Well, let's bring him in from the waiting room.

Starting point is 01:18:55 As a free man. First time with his new fund, AMPPBC. He's here in the waiting room. He's here with us in the TV panel. that how you doing? How you doing? How you doing? Generational run.

Starting point is 01:19:06 Yes. Generational run. It's getting started, guys. No, generational run is for people who are retiring. Oh, yeah. That's true. That's true. I think it's appropriate to, I think it's important to recognize when you're on a generational

Starting point is 01:19:19 runs that you realize you have to actually level up even further if you want to stay on the same trajectory. Yeah, yeah. Because if you just get complacent, then. Yeah, it's important to like count the chickens before they hatch. Like that's what you're saying? No, no, no. The opposite.

Starting point is 01:19:33 Anyway, great to have you back on the show. It's been fun to, but yeah, catch us up to speed on. Yeah, when did the fund launch? What, you know, what's the strategy? And, yeah, like, like, walk us through, like, the current thesis for how you want to actually develop the firm. No, I think we should talk about something more interesting. Let's talk about eBay. Let's talk about eBay.

Starting point is 01:19:56 Okay. Yeah, what you got. Okay. So here's what I'm, and I want you guys to kind of spar with me, right? I looked at, Ryan did this CNBC interview and everybody's like pinging me and saying, oh my God, this makes no sense, blah, blah. And so here's my take on it, okay, which is that like if you read the Gamestock deck like carefully for eBay, most of what's been said about the deal in the last 48 is

Starting point is 01:20:19 basically totally wrong. I read, I was just before jumping on, I was reading Michael Burry's piece on it, which you guys should check out. And he is right that the leverage is pretty tight. but I think he's answering the wrong question and so is Ryan on CNBC where they keep asking him like where's the cat

Starting point is 01:20:38 how are you going to fund it? Yeah yeah. 50% cash, 50% stock. Yeah, yeah. I don't think the question the question isn't can game stop afford eBay? The question is whether the underlying

Starting point is 01:20:49 business actually works and I think it does but not for the reason I was expecting Ryan to pitch. Okay. So if you pull up eBay's 10K from February yeah, fiscal 25

Starting point is 01:21:00 and I did not understand this on Reddit. eBay spent $2.4 billion on marketing. How many new users did they get for that? I mean, you guys are marketers, so you understand. One million. Whoa. That's crazy. The user base went from 134 million users to 135 million users

Starting point is 01:21:20 after spending $2.4 billion on marketing. And that's basically, you would have to imagine they're just having to reacquire all their old users, people that have been on the platform before or maybe even lost their account and they're coming back. I don't know, but that's $2,400 of marketing per new user on a site that every American already knows exists. Sure. Yeah. So where's all that money going?

Starting point is 01:21:45 Right? That's rough. So I don't think Cohen is, I don't think he's buying eBay. Like just watching Ryan, I don't think he's buying eBay because he thinks he's smarter than eBay's product team. I think he's buying eBay because he can see $2 billion of fat. that Wall Street has been pricing as fixed cost. And so you go, okay, let me cut that. And the interest on the debt just pays for itself.

Starting point is 01:22:08 Interesting. But he doesn't necessarily want to say that because he could kind of give that idea to the management team. But he still has to put the money together, right? But is your thesis that like the deal is coming together. He has investors that he's talking to. But it's too early to say, oh yeah, I actually do have a fund that's going to give me another five over here. I got seven over here.

Starting point is 01:22:28 And it will math out, but just give me a week. or is there something else going on? It depends on which investors he's talking to, but if he was pitching me, here's what I would underwrite. I'd say, okay, that's the floor. The floor is Ryan's going to cut $2 billion from this thing of that.

Starting point is 01:22:41 Put that into treasuries and we're going to make more money than it's currently yielding. Okay, so that's the floor. Now the ceiling, because I'm a technology investor, right? Yeah, the opportunity. The bulk case.

Starting point is 01:22:52 So Amazon's used and collectibles business has been flat for six years. they tried renewed, they tried collectibles, they tried trade in, none of it's still. You cannot put a 1962 Mickey Mantle card through the same warehouse as a phone charger. That category, that category, collectibles is structurally defensible against Amazon. Amazon is for phone chargers, Mickey Mantle for eBay, right? eBay has the marketplace. Gamestock has like 1600 stores that could physically verify the good.

Starting point is 01:23:23 That's a real mobile. And it's worth more in the AI era than the human era, right? Right? Why? Because when AI agents, I collect, I love, like, rare pens. Okay, this is a mom blah. Sure. As I get old. I love it. I like that. I'm a pen guy. And I love old vintage glasses. I don't know if you can see my Jacques Marie Maj box in the back. But cool. Nice. But when I'm out of time, what do I do? I have Claude go look for rare pens and glasses for me online. The biggest problem with used rare asset purchases is fraud. Yeah. So I often will tell, So Claude will be like, Andre, I found this amazing pen, this momblown pen. And I say, okay, can you triple check that it's real? It's not fake. And that's where things go up real because there's no way for him to verify that without messaging the agent and so on.

Starting point is 01:24:11 Yeah, yeah. But if you have 1,600 stores where people who have momblown pens can go and physically verify those assets at GameStop. Sure. Now, Claude just says, yeah, I checked. It has the physical verification stamp from. Somebody brought it in. Exactly. So you can't, you need physical verification built into the system for agentic commerce.

Starting point is 01:24:32 And look, the reason I know this is real is because a few years ago, I think you guys, we talked about this last time, but I sold my last startup to a company called Discord in 2020, peak pandemic. So I come on as the head of platform. My job is supposed to be, you know, Antigop built SDKs, APIs and so on for gaming. We helped this company called Mid Journey get going, AI, you know, generate. But 12 months later, suddenly, I think. find myself running without realizing an e-commerce business because it was the summer of NFTs.

Starting point is 01:25:02 And the board Aves are blowing up. And suddenly, we have more than $10 billion of GMV flowing through Discord buy cell paid channels. And Jason and Stan are like, hey, brother, your job is to capture a piece of that pie. Yeah. Okay. Homework accept it. So we start doing a deep dive.

Starting point is 01:25:21 And we realize ultimately what these users need and pay for. So you have liquidity, right? Ultimately, that's what a marketplace like eBay and Discord provide in sort of community commerce is liquidity. But you cannot provide liquidity if you don't have physical verification. And for Discord, that was just out of scope. You know, you sneakers. And it worked. You're saying it worked for NFTs because you had the on-chain.

Starting point is 01:25:44 Yeah, you don't need me physical. There's nothing physical to verify. We owned it and you can transact. Yep. Exactly. And so we had, we are bootstrapping the e-commerce platform at Discord with NFTs. But, of course, everyone on the board is like, well, how long are NFTs is going to last, it's a fad. So Ange, what else is coming? We go look at rare sneakers or keyboards, like

Starting point is 01:26:00 bends, all this stuff that nerds like me love. But for those you need physical verification, and once we realize physical verification is out of scope, we next the project. Yeah, you're not going to go have the 1,400 like retail locations where people can drop things off. Yeah, that's, that's quite, that's quite interesting. So that's when I realized, okay, eBay is this undervalued asset. And I hope that Ryan has figured it out as well. Because if he hasn't, he's giving me ideas. Yeah, yeah. Have you, have you tried to walk through what, what, you know, given that you're probably more AGI-I-pilled than, then I would say, 90% of VCs, have you tried to play out? What is it? How would you, how would you build eBay from the ground up today with an agent-first

Starting point is 01:26:48 approach? Is that even the right question to ask? Look, I have not, guys, because right now I'm a computer infrastructure guy, right? We start an amp as this public benefit corporation whose job is to be an independent system operator of the compute grid. We think about, we think we're roughly in like 1885 Industrial England where the steam engine's been invented. Everybody knows that you can make cool new products like, you know, steel and notebooks and pens and cars.

Starting point is 01:27:14 And there's this very scarce input called coal that everybody is hoarding. In this case, it's compute. And if you fly over industrial era England, you'll see all these factories getting set up. And everyone's running a generator in their backyard at half capacity. And I'm going, this makes no sense. If I'm looking at all my portfolio companies, you know, these clusters are running at like half utilization. In fact, Elon's like got 500,000 GB300 in Memphis at running at 11% MFU and less than 60%

Starting point is 01:27:41 node allocation. I mean, this is $12 billion of compute being wasted. So I set up AMP as an AI infrastructure organization where we buy a bunch of compute. We do long-term leases. We pull that all those clusters on the grid. We coordinate capacity, drive up utilization. And by the end of this year, I think we'll have, you know, several billion dollars of compute coming online. But that's what I've been focused on, night and day since like I left in recent Horowitz in January.

Starting point is 01:28:03 And so, no, I have not had time to look at how to redo eBay. But if Ryan called, I'd probably help them out. But right now, it's wartime on compute, guys. So, yeah, I want to talk about AMP, but I also want to talk about just last question on the combination of eBay and GameStop. Like I get the thesis, the bulk case. GameStop is $10 billion. eBay is like $48 billion right now. you put them together.

Starting point is 01:28:25 Maybe you get to $100 billion. I'm in for the bull case. The question is like, what's going on with like the plan? Because it feels like Ryan just doesn't have the capital. But then he announced it. Like what, like what do you think is happening behind the scenes? Because there's one thing where you could throw it out as like, oh, like these two companies should work together. Here's a bull case.

Starting point is 01:28:47 Let me know if you want to be on the team that does this. And then there's the other one, which is like make the offer before the capital is lined up. but I just haven't been through enough of these stories to actually understand, like, why it's playing out this particular way. To be honest, I think he's sitting there with, I think he has $9 billion of cash. Yep. He's in a $10 billion company. I think when he announced this, I think he expected the stock to pop like crazy and he'd be going on CNBC being like this merger could make sense. Sure.

Starting point is 01:29:16 And I think that we'll issue another $20 billion of equity and then we'll merge or something. And I think if you looked at how kind of frothy some things in the market could be, you could have imagined that playing out. I mean, the Allbirds thing was I'm sure you appreciated that from the meme standpoint. Direct competitor to you. Yeah, direct competitor. You got to be careful. Amp versus Allbirds will be the new horse race.

Starting point is 01:29:41 Anyways, that was my read. Because GameStop is basically valued at like the brand and all the retail locations and everything is valued at like a billion. He's getting no credit for all the cash. So to be clear, AMP is not a cloud business. We are, so I started AMP as a holdings business. Yeah. And I've got an infrastructure business and a capital business.

Starting point is 01:30:02 And the infrastructure business secures compute and passes on at cost to our portfolio companies. We have more than $1.3 billion in commits for our first fund. I've been added eight weeks. And so we do to venture capital investments. We put $300 million into an profit. Yeah. Oh, okay, cool.

Starting point is 01:30:15 Yeah. But we need to raise another roughly, you know, $6.5 billion this year. And more is getting committed by the day. But we give away the compute at cost to the independent ecosystem because my belief is that, you know, that they're like sort of, that the optimal unit of research today is a like a focused talent dance team outside of the hypers, you know, Anthropic encoding, which was I was the first one of the first, I'm certainly, I think, I'm the first angel investor, if not the first investor in the round. They're saying you're the Jason Calacanus of Anthropic. Unfortunately, JCal had, I could never talk JCal, but like, if I call intern or something, fine, I'll take the, win. But I think more importantly, like, I think compute is this strategic asset, which I've been yelling about for four years, and it's a primary bottleneck on these teams. And if you're not at the

Starting point is 01:31:00 hyperscalers, you just can't get access. So we buy up that compute. We give it at cost to the portfolio companies. And then we reinvest the profits of carry and fees to buy more compute and so on and so forth. And so I'll take as much capacity as I can get from Albers. I love it when new people go into the business because that gives us more supply. So if you're the Albert CEO, listen to this, please send us your compute. We'll take it all. That makes, that makes a lot of sense. What are you excited to invest in? You're investing in teams that need a lot of compute. You're trying to find things that aren't going to get steamrolled by Ampropik,

Starting point is 01:31:32 who's another big portfolio company. There's, you know. There's application and stuff. There's real labs. There's a lot of other different things that need compute. But what do you think? Because I feel like a lot of ECs would never say this out loud. But a lot of, I get the sense from a lot of ECs,

Starting point is 01:31:51 they're kind of like paralyzed where they're like they really don't they don't have a clear sort of understanding of where things will be in five years and they feel like they need to be active and so it's a mix of like doing new neo lab doing some neo neolabs maybe doing some application layer stuff and just kind of praying but i would hope that you have a given given your background and how you're approaching this you have like a stronger thesis on on where the opportunities to invest at the early stages. Look, in some sense, it's back to the future. I started my career at Kleiner Perkins when I was 19.

Starting point is 01:32:27 My first board seat was as an observer with John Doer when I was 20. I wasn't old enough to drive. Oh, dream, sorry. I didn't have a driver's license. I wasn't old enough to drink. And I got the chance to apprentice with like the grades like Brooke Byers. And look, that's the vintage of venture capitalists I grew up admiring, like Arthur Rock. And that's my, you know, our thesis at AMP on the venture capital side,

Starting point is 01:32:48 our business is called the AMP Foundry where we help create co-design, you know, new labs one at a time. My current one is called periodic labs, and we just decided to lead the series. I led the seat round last year with Liam, who was the co-creative chat GPT, and Doge, who was there, who led some of the quantum physics teams at DeepMine, and we're trying to find new high-temperature superconductors there with physical. We have a 30,000 square foot facility in Menlo Park. I spent three days a week there. We do a stand-up every morning from 8 a.m. to 8.30 a.m. And then we make our priorities and then go execute.

Starting point is 01:33:14 And, you know, basically we have AI's, predict new materials. We then have robots synthesize the new materials. We then have an x-ray diffraction machine that tests whether the material has the properties that robots, the AI set. And then we pipe that verification loop back into the training around like as many times as we need for the agents to continue predicting new superconductors. And in the last 90 days, we've had more material verifications than I think in the last decade in the field. And so I'm a huge believer in unblocking frontier progress in domains where the verification sort of loop is clearly just like, we know it's going to work, but execution is the bottleneck. and then I like getting these the best teams, the best scientists, the best engineers, the capital, the compute, the commercial help they need.

Starting point is 01:33:51 Now, I think that the beauty about having Anthropic around is that it's made this idea of the bitter lesson and scaling legible to the capital market. So now instead of me having to call up 22 friends, I'm getting 21 knows, which was the case with the C-drawn of Anthropic. Now I make two calls. Instead, we get like, you know, we get three times oversubscribed. So capital is no longer the bottleneck, which is phenomenal. You know, again, we've been out of eight weeks, have more than a billion dollars committed

Starting point is 01:34:13 from our first fund. I'm a solo keyman GP on the fund. And there's lots of institutions, pension funds, sovereign funds who are like, how much more can you put to work, especially in public's, privates, buyouts? It's a bonanza for people who want to be true partners, who want to be the Arthur Rock of this era. I think if you believe in the bitter lesson, it's not new. It's been around for ages. The three of us talked about this over a year ago at the last recent Horwitz AGM.

Starting point is 01:34:35 Yeah. Still better. I'm more zen than I am bitter. How are you thinking about building out the team? on both sides. Trust is the moat. So there's five of us on the team. My full-time engineering co-founder was Sebastian Lobo.

Starting point is 01:34:50 We were roommates 14 years ago at Stanford, and then he went on to build a grid. Overnight Success. There we go. What was that one? Overnight Success. Overnight Success. Thank you. 12-year overnight success in California.

Starting point is 01:35:02 Exactly. So Seb and Mihai built the Borg-X-Worg GQM scheduler, which kept the Google internal capacity pool at more than 95% utilization there. If it was at 94% utilization that was considered a major outage globally. Andrew Erskine is my partner on the operation side. He was a partner at Oric, which was the outside council for Anthropic. And he was my GC at ubiquity 6, which was the company I sold to Discord. And then Rosie, who's my chief of staff, ran comms for me from Edelman when I was at Andreson Horowitz, you know, when I was a GP there.

Starting point is 01:35:31 So you got the band back together. It's the Andrenion, basically. And, you know, we called AMP not after my initials. Sometimes people think that. It's not Antrimeta partners or anything like that's about energy. It's about ampere's, the unit of energy. And we want to amp things up because we think we're entering like the great Renaissance and technology.

Starting point is 01:35:50 And, you know, if you can have a small team that trusts each other across context, you know, compute capital sports teams, buyouts, leverage technology, all of this stuff. These are all buckets and categories that we've all put, you know, traditional capital allocators have put around these asset classes that shouldn't exist. And I think if you have the flexibility to go back to first principles with a small team that you trust, you can execute, you know, with orders of magnitude, less, a size of a team as a firm in this new era with the right tools. I don't know if that makes sense. No, totally.

Starting point is 01:36:18 You mentioned taking positions in public companies. The fund structure is a PBC. Are you also an RIA? Like, how are you thinking about navigating both of those asset classes since that's a little bit of traditional? We are in process. Okay. Yes, we are in process of becoming an RIA because, you know, we founded the firm barely. Yeah.

Starting point is 01:36:35 Yeah, 90 days ago. But I'm used to that cadence because in recent Horwitz was a RIA. I was a general partner in the AI infrastructure fund for several years, as you guys know. And we were in RAA. I'm used to the compliance, the regulatory sort of guardrails we got to follow. And I think LPs trust us to, you know, have that cadence from day one. And so we're going to make sure that we, you know, Zach, if you remember this, like, you know, 12 years ago, Zach went on TV to say, move fast and break things.

Starting point is 01:36:57 And then you have to update the thing to be like, move fast with stable infrastructure. And I think we move fast with stable infrastructure from day one, essentially, because we are an AI infrastructure team. Yeah. Talk about the PBC. Like, if I'm playing back, like, well, what? what year were you referencing 1850 or something like that? 1885.

Starting point is 01:37:13 1885. So if I go back to 1885 and I think about the financiers that, you know, created the industrial build out, they were not public benefit corporations. They were personal benefit. Yeah. And I mean, there was a lot of good that was created. We got railroads and trains and, you know, machinery and cars and all sorts of things out of the Industrial Revolution.

Starting point is 01:37:34 There were also things that were rough and there was unionization and battles and back and forth. Like, what is the PBC in service of solving? What, why PBC? Yeah, great question. So there's the, I'll tell you the substantive answer and then the vibes answer, right? Sure. So from a substantive perspective, we do two things, right? We have a venture capital business and we have an infrastructure business. Both things have this very unique property called positive externalities. When implemented correctly, venture capital can unlock massive positive externalities for the ecosystem and for the world because you end up funding innovation when done correctly. And then infrastructure is the same, right? When you have compute that's used by small focused town-dense teams like Anthropic that's able to produce

Starting point is 01:38:10 10 times more soda capabilities than like DeepMind, which is 60,000 or 160,000 people, then you're generating positive externalities for the world by being much more efficient per unit of input with the output they create. And so I was like, huh, well, what happens when, as an economist, you look at positive externalities? Usually you have market failure. You have underconsumption of that good. Well, how do you correct the market failure?

Starting point is 01:38:31 Usually you get the government to intervene, but if you don't have the government intervening in time, what do you do? you become a private sector participant. And then if you look at the arc of 1885, you know, private sector businesses that ended up correcting market failure by producing public benefits, they ended up getting regulated as utilities. That's what AMP is. AMP is a self-regulated utility that provides venture capital and infrastructure to the world's leading scientific teams, the next Dario, the next, you know, Guillaume at Mistral, who created Lama, the next Robin who created stable diffusion. These are my generation's smartest minds. I'm not smart enough to be them, but I can be their intern.

Starting point is 01:39:09 And instead of waiting for the rest of the space to come up with standards and institutions to enforce this, we're like, dude, let's just do it ourselves and show the world you can have fun while doing it with a small team. You don't need to be, you know, something called a, you know, with these words like RIA, multi-stage asset class firm doesn't matter. Just let's skip ahead to the part, the good part, and like, you know, use all the proceeds that we get from management fees and carry to keep the space like innovating at the base

Starting point is 01:39:34 that we were promised, you know, 12, 13 years ago and instead we got tweets and not flying cars. You know, when I was at Stanford, I got the chance, as an undergrad, I had the chance to take Peter Thiel's class, zero to one. And he was, you know, his whole moniker was, we wanted flying cars and we got, you know,

Starting point is 01:39:50 140 characters. Well, 240 characters. Thank you. And now I'm back at Stanford teaching CS-153, which is the largest class on campus. It's called AI Coachella. We've got thousands of people following along. Coachella is a good one.

Starting point is 01:40:01 And it's got frontier systems because it's all possible now. We're literally in our lifetime, we're going to have flying cars. We're going to have room temperature superconductors. We are going to solve cancer. We just want to do it in a way that's stable, predictable. We want to skip all the boom and bus cycles. And the way to do that is to lead by example and say, hey, guys, the public benefit, you know, is to provide goods and services that are utilities and make sure that we don't, like,

Starting point is 01:40:24 let's be the adults in the room and not do the stuff where we try to be robber bands and monopolists and got greedy along the way. And so that's the substantive answer. The Vives answer is, look, I don't want to get sued by shareholders for whom it's not legible, why I'm giving away billions of dollars of compute at cost of portfolio companies, right? Because that's what we're doing. And that shareholder, you could argue that shareholder value that we're destroying, but I would argue in the long term, we're actually creating orders of magnitude more value.

Starting point is 01:40:47 And if you look at Ben and Jerry's, you look at REI, they've become stable enduring businesses in categories that are fairly crowded. And eventually, I do think technology and AI will get crowded. It will get commoditized because technology is never the most trust is. community is a moat, culture is a mode, execution is a mode. And so we're trying to skip ahead to that part, but it takes time for people to get aligned. So until then, we'll see an AMP joint venture with private equity to help distribute. Diffusion.

Starting point is 01:41:16 Yeah, you know, if you go to our website, it's called AMPPublic.com, because I do think if you look back to the, like, vulture era of private equity, you remember like R.J.R. Nabisco and Barbarians at the gate, we should just learn from, from their mistakes and go, can we do private equity but done right in an aligned way? Let's not rush to lay off hundreds of thousands of people and then not re-educate them and prepare them for their new opportunities. I'm an optimist, as you guys know, I came from Andreessen Horwitz. So I'm a sort of a rational optimist. I believe the transition can be done in a positive way. Is that me getting kicked off?

Starting point is 01:41:48 There's like a bell ringing, but that might be here. No, I don't know. Oh, okay, cool. No, I'm looking at the only bell we have is this. But no, we're. We got plenty of sounds for you. Continue. We'll do everything we do is governed by a public benefit charter.

Starting point is 01:42:06 So if we do private equity, you'll be governed in the public benefit. If we do education stuff, that'll be in the public benefit. Look, I've made more money than I know what to do in life. I'm 34 and I'm just getting started. So my goal is I'll be remembered for having been a net positive influence in the space. I just got tired of telling people I told you so because after a while they started looking in my returns and going, why didn't you give me a call? And I said, I did look at your email.

Starting point is 01:42:26 I introduced you to Anthropic in the series A. series B and the Series C. And so at some point, I was like, you know what, I'm just going to go direct, talk to the LPs, set up a platform, build infrastructure, and hopefully be known as a generally like sane, common sense, rational point of view on stuff that can often be, is not legible to people from different parts of the stack. And that's what the class is about. So CS-153. That Stanford.edu, I would recommend anybody watching, go check it out. The lectures are all online and the first one went on on Stanford's official page on Thursday. Was that with Scott Nolan? So Scott, no, actually, Scott was Lecture 8. We put up lecture one, which is mine as a kind of the opening act. Because Scott is one of the mainliners, the headliners of AI Coachella. Scott will be up soon as well.

Starting point is 01:43:12 And then Jensen was last week. So I think he followed Scott. Well, thank you. Last question for me. Do you think the world is prepared for it not to be a bubble? That's a good question. The world, if the world prepared not for it to be a bubble. Oh, yeah.

Starting point is 01:43:30 Yeah. Inertia is, guys, inertia is a powerful thing. Most of the world still has no idea what AI is. It is crazy. And I've been flying to places where I thought there would be diffusion of AI by now. And they just are barely using Claude, chat GPT. I mean, these things are still alien to most of the world. And so if we stopped capabilities today and half of us in the AI ecosystem vanished off the planet, nothing would change.

Starting point is 01:43:53 It's still so early. Yeah, it is really. Very cool. Well, great to catch up. Congratulations on a very impressive fundraise. and a very unique approach and looking forward to the next conversation. Thank you much.

Starting point is 01:44:05 You guys, on a generational run, and I hope the acquisition does nothing but give you guys more steroids and more fuel for the fire. We need four of you every day. Fantastic. We'll talk you soon. Have a good one.

Starting point is 01:44:17 Thank you, guys. Bye. Cheers. Up next, we have Ben Lamb from Colossil Biosciences. I made a YouTube video about Colossil years ago. I think this is the first time I'm talking to Ben actually live,

Starting point is 01:44:28 but fantastic. company. Super interesting, Ben. Great to meet you. How are you doing? Hey, great to meet you. Wait, what YouTube video was it? Something about the Woolly Mammoth coming back. Right after your announcement, you went viral for the first time. And I sort of, like, made a video essay talking about the company, a little bit about your background. And then also just some of George Church's work, trying to give more context on the science, like the more incremental steps. Like I think the media generally wanted to jump to conclusions about like bringing back to the T-Rex, right? That first year was a lot of interesting fielded calls for sure.

Starting point is 01:45:05 Yeah, yeah. And I think people were sort of like, oh, the, you know, the Jurassic Park analogy is so fun. They were on a jump straight to like the T-Rex running loose in Times Square or something. And I was like, no, like there's a world where this business works, even if it's just like some cool new animals for zoos. Like you don't even have to get that crazy just looking at the price of like different zoo animals and stuff. I know that the, obviously, the mission's a lot broader. We talked about some of the, not deforestation, but dealing with those Siberia. I'm blanking on this, but, you can tell me about all the applications.

Starting point is 01:45:35 All the carbon modeling, for sure. Yeah, we had, our original thesis was, you know, synthetic, this is like the hardest synthetic biology challenge. And kind of access to compute AI and synthetic biology all paired together will create kind of this unique opportunity to, you know, build a lot of different tools and technology. And so our view is like if we start with the hardest problem and we couple it with kind of like an existential problem, which is, you know, losing biodiversity, you know, I think it forces us to, you know, build a platform that's pretty robust because working with DNA that's a million years old or 73,000 years old is a lot harder than just working with something like right out of a lab or, you know, bought off the shelf from some like XYZ DNA provider. Yeah, yeah. So where's the company today?

Starting point is 01:46:24 Like what, what are the, I mean, it was, this was always like a walk, crawl, run project. There was going to be interdural development. It wasn't going to jump straight to bringing back the oldest creatures in history. But where of the success has been, where of the setbacks been, where is the strategy pivoted? Take me through some of the recent developments. Yeah. So the biggest developments recently are, you know, last year we had kind of a couple like watershed moments.

Starting point is 01:46:48 We showed the world the objectively cute woolly mice, which at the time were the most genetically modified multicellular organisms out there, right, where we took the mouse equivalent of the mammoth genes that we're targeting in our Asian elephant cells made woolly mice. They went crazy viral. And I do remember sitting at South by being very concerned because it's like, wow, people lose their mind over these mice. Wait until they see giant wolves in a month. And then we showed the world that the dire wolves where we took about 73,000-year-old skull

Starting point is 01:47:20 and a 12,000-year-old tooth and made puppies. And so those were kind of two really interesting data points that shows kind of the end-to-end pipeline. Not only could we identify extinct variants and we could replicate them, do the ancestral state reconstruction, model them in and actually put them successfully into living cells, and then go through the entire quality control process to deliver actual living animals to Earth, but do it in a way that's completely humane, that passes all of our IACUC and Humane Global Certifications, and do it, you know, get exactly what was predicted.

Starting point is 01:47:56 And one of the biggest things that we've learned from that is all of the AI systems coming online are just accelerating it. Like you're seeing a lot of different markets and industries be affected in a lot of different ways and, you know, a lot of doom and clume is being sold around AI. But synthetic biology and all of the modeling and taking the data sets that all of our teams are generating and pushing them all forward together is something that, you know, we're really seeing a mass acceleration here. What's on your short lists for animals to work on?

Starting point is 01:48:29 So we feel like it's like, you know, there's a big long list. And I imagine the list in your head is a little bit longer than what's on the website. And so I want to dig into some of the deep cuts, the B sides. Yeah. So we've definitely announced Woley, Mamet, Tasmanian, Tiger, Dodo, Moa, and then as of last week, Blue Buck. But even with that last week, you know, we didn't say, you know, oh, we're starting the Blue Buck. we're going to be on this 10-year journey. We're pretty far.

Starting point is 01:48:54 We'd already done all the ancient DNA work, all the comparative genomes work, all the stem cell work, all the animal repo work of cloning in antelopes. And so now we're really just in the editing phase. And two years ago, we were doing three to five edits at a time. We're now doing over 200 edits at a time. So that scale function has gone really quickly for us. And we haven't seen an upper end of the delivery. So I think that, I think it's highly likely you can see a couple more dire wolf-like moments of species that haven't been announced, but, you know, just, you know, show up. And I think we'll show you some additional progress on some of our big projects.

Starting point is 01:49:36 And then what's the team like? I imagine that you have a lot of scientists on staff at this point. What's the shape of the business? Yeah, so we've got 260 full-time scientists. Wow. We have 17 academic partners around the world. 80 postdocs in academia, 75 global conservation partners, and now we have five government partners around the world.

Starting point is 01:49:56 So we've closed a little under $650 million, and the teams are going quite well. Labs in Boston, Dallas, and not in. Amazing. And then in terms of like commercialization, is it still like, you know, just focus on the science, like the business opportunities will come? Is it experimental little test projects?

Starting point is 01:50:19 Like how, or are you already starting to think about like, you know, the, the SpaceX Starlink? Like there is a real commercial enterprise that will be self-perpetuating, not science funding. I'm sure you're bringing in revenue from a bunch of different sources. But have you started thinking about long-term? Yeah. And effectively living museums. Yeah. I think about all the things that you listed off I read about with my four-year-old all the time because these, you know, children's book.

Starting point is 01:50:46 are not like, you know, they don't discriminate between living or dead species, right? I've read, I know too many dinosaurs at this point. But if you were telling me like, okay, we have a zoo and you can go see all these animals throughout history, that would be pretty compelling, and you would have effectively a monopoly over said animals. Sounds kind of bad to say monopoly over a species. That sounds a little bad and evil. But ultimately, you know, we love zoos.

Starting point is 01:51:19 We work with a lot of zoos. We're not anti-a-a-lott-a-lose stamp on me because I say, you know, zoos are a little bit transactional. If you do look at all the data and the studies that, the scientific studies have shown up around kids seeing animals in zoos actually gives them a higher appreciation nature, a higher appreciation for biodiversity. So I'm not anti-Zoo, but it does feel like a little antiquated. It does feel a little transactional.

Starting point is 01:51:42 Like I pay money. And I have young kids. I've taken them to zoos, right? You pay money, you go see zoos. It feels a little transactional still. So I think we have grander ambitions on that. And so while, you know, we have no intention to make zoos, I do think there's highly likely that we all have partnerships with ecotourism with the animals,

Starting point is 01:51:59 with countries where you can see the animals back in their natural habitats. Yeah, more like a national park where... It's like Kruger National Park and all these different locations, right? And I think that also making that accessible and putting science on display, versus animals on display, it's something that we're really excited about. Like, we think it's as important that you understand the conservation impact,

Starting point is 01:52:23 the ecological impact, in the science of how you make a dodo as seeing a dodo. So I do think that there'll be educational and media experiences that's a large portion of the company that focuses on that. But I still think that's going to be dwarfed

Starting point is 01:52:36 by the government work and just the synthetic biology pipeline. We've already spun out four businesses from the company in four years, two of which we've announced, of which we haven't. One of them kind of got leaked, which is Astromac, which we raised. Last round was at $2 billion valuation, nine months old. So we are building fundamental technologies that I think have broad applications to government, you know, conservation, as well as

Starting point is 01:53:03 that of human health care and disease modeling. So synthetic biology is kind of this end-to-end pipeline. I think it's pretty interesting. But then separately, governments are now coming to us and we're helping them understand the assets in their biodiversity, how they can actually data mine that and protect their species and kind of help them think about biodiversity in a different way. So not as far as the long-term applications to nature credits and biodiversity credits, but how can we help governments underwrite the protection of their biodiversity of looking at, you know, massive non-model species wins like the GILA monster

Starting point is 01:53:38 and how the venom from that actually led to a trillion-dollar GLP-1 market, right? And so helping them understand these assets that they have should be protected. If we can't get them to protect it because they should protect animals and they should protect their environment. It's like, okay, if you can't love the animals, can you love the environment? Figure out of it. Figure out how to make money. No, it's a good. Yeah, too, too.

Starting point is 01:53:59 You got to line all these things. It's a pragmatic look at conservation, right? It's like, hopefully we can get them to care about the animals, if not the ecosystem. They've not just themselves and if not the economics from treating human disease. So it is working, and we now have five government partners online and giving them the tools to really understand what they have and why they need to protect it. So I do think that we can really have this next-gen conservation narrative while also helping countries monetize it in a way that's good for them and their people. Well, congratulations to the progress. Do you anything else?

Starting point is 01:54:32 No, this is cool. I've been hearing about the company forever. It's great to meet you and understand the vision. and I'm sure you'll be come back on when you have your next. Yeah, we'll let you know. As soon as we drop something else crazy, we'll let you know. Perfect. We'll talk to you soon.

Starting point is 01:54:46 Great to me, Ben. Have a good rest of your day. Cheers. Goodbye. I'm next. We have Jake from Serval. We're running a couple minutes behind. We've got to catch up.

Starting point is 01:54:53 He's the founder and CEO of Serval launching the future founders program to train next gen operators. How you doing? I'm doing great. How are you? We're doing fantastic. Welcome to the show. Introduce yourself in the company and the project. I'm Jake, I'm the co-founder and CEO here at Serval.

Starting point is 01:55:11 We're an AI platform for employee support. So you've heard of AI for customer support. We support it internal employees when they ask for new laptops and access to applications and all that stuff. And we've got this new program called Serval Start to help deploy Serval in these large enterprises and give future founders the opportunity to get inside an enterprise and deploy AI, which we saw with the news today, is becoming really the gap from technology to actual implementation to impact. Yeah. Talk about the integration points with the systems that we know.

Starting point is 01:55:48 Is the point that you can work across different point solutions, be sort of a meta solution? Or will you wind up building point solutions and sort of becoming your own compound startup at some point? I think the idea is actually the latter, that you become the own compound startup. We're increasingly ripping out systems of record like service now and actually taking over that entire service area. Yeah. Where has adoption been the strongest in terms of like massive enterprise Fortune 100, down to SMB, down to startups? Like where are you seeing the best traction? Yeah, we started like a lot of companies in that kind of tech startup world with great companies like Notion and perplexity as early adopters.

Starting point is 01:56:29 But we've worked with some of the largest companies in the world, Fortune 20 companies, Fox Corporation. and increasingly, I would say, more excitement and more interest from the largest companies of the world because that's where they have the most pain. They've got thousands and thousands of employees asking for password resides, asking for access to applications, all these things that can be automated. Yeah. Talk about the Future Founders Program specifically and like how that's structured, why this particular model, how everything plugs together.

Starting point is 01:56:58 You know, what we noticed was that the biggest pain point is increasingly not the strength of the models. It was actually the implementation in a large enterprise, understanding the change management and the stakeholders and approvals and all the things that need to happen. And you actually need really talented people to go in and run that process that have to be able to, one, build relationships, sell the product, but also build product because you're going to find out all these feature gaps. And when you think about who is really good at understanding customers and selling and also building product, it's all people that are either former founders or future founders. And we started hiring that profile for this role and realized, hey, this is also a great. training ground for them because you get to go into these massive enterprises and really go through the motions of being a founder, the same things that me and my co-founder did when we started the company.

Starting point is 01:57:41 And so why not really formalize this and focus on hiring the next generation of founders to give them a training opportunity. So we bring them in. We give them opportunity to build enterprise automations, deploy AI, large companies. We accelerate their vesting so they vest after six months. And we have the expectation that, like, do this for a little while and then go start your company. And we'll connect you to our investors like Sequoia and first round and Red Point and General Catalyst.

Starting point is 01:58:06 And we'll set you on your way and give a great reference and give you a great experience. Yeah, that makes sense. How big do you want the program to be? We're going to start with a class about 12. I think we'll expand that over time. We'll do probably two classes this year. And then we'll see where it goes. We might make it bigger or smaller over time.

Starting point is 01:58:23 I think there's a big question across the board on how big a lot of teams get. But we feel like given our growth, we need as many people as possible. in the shorter term. Yeah. Sorry, Jerry. No, it's cool. It's cool. I mean, a lot of, there's a lot of people out there that want to be founders, but in order,

Starting point is 01:58:39 but haven't necessarily been inside an organization to discover the right opportunity to go and build. And so I can see this being a win, win, win. Yeah. What's the, what's the, like, the winning background for someone in this program? Like, in order to, I imagine that working at a big company can accelerate you becoming a forward deployed engineer. at a different company because you can immediately plug back

Starting point is 01:59:04 into your former organization as long as you left on good terms. Is that the correct model or is it more like, oh, come out of college and jump into FDE work? Yeah, I think it's actually closer to the latter. I think you could be a new grad. You could also be someone with a couple years of experience. I think general intelligence ends up being the biggest predictor and a technical background.

Starting point is 01:59:25 So someone who's CS degree or has worked as software engineer because you're going to be asked to write code. And it's going to be production code. It's not going to be just some bespoke automation. It's actually going to be impacting our product. And then you've also got to be somebody who can sell a large enterprise. So we are hiring ranges of experience. And some are new grads and some are 20 years of experience and have more experience to the large enterprise.

Starting point is 01:59:46 We think that there's a fit for everybody. But I think it's going to be folks that are incredibly technical. And that could be starting a company instead. That would be their alternative path. Yeah. Makes a lot of sense. Well, thank you so much for coming on the show and breaking it down for us. Have a great rest of your day.

Starting point is 02:00:00 Crush this. We'll talk to you soon. Let them now. Thank you. Goodbye. Appreciate it. Cheers. Up next, we have Garth coming back on the show from Panthalasa.

Starting point is 02:00:07 One of the coolest companies. Farming energy from the ocean with mega machines, mega projects. We've got to pull up some video of Panthalasa and what Garth has built over there because it is a magnificent structure when you see it. We will figure out the waiting room in just a minute. People are showing. deep seek progress versus China's falling behind labs. China's actually falling

Starting point is 02:00:36 even more behind. Noah Smith is saying export controls work. Export controls work. Export controls work. Yeah. So the map, the chart is that if you look across GPT40, opening I, 01, 03 mini,

Starting point is 02:00:53 03, then opus four, then GPT5, 5.2, opus 46, to 5.4, and then opening AI is GPT 5.5. America and the United States AI lab seem to be on a steeper curve compared to DeepSeek, Alibaba, K-Wan, Deep-Seek R1, V3, Kimi, K2, K2.5, and then Deep-Seek V4 Pro. Yeah, I mean, if you go, if you rebuild this line just between Kimi K2.5 and Deepseek V4 Pro, it is looking steeper.

Starting point is 02:01:27 Maybe they'll figure it out, but at least for the for the current. moment. The AI gap is bigger than you think and potentially export controls work. We'll see where they go, whether they stick around. But we have our next guest in the waiting room, Garth from Panfala on the show last year. But welcome back. How are you doing? Hey guys. How you doing? Long time. Too long. But you've had, you've made massive progress in the interim. You brought us back a big number. Yeah, you brought us back a big number. How much did you raise? Tell us what you raised. We did what we could. Oh, are we going to do a gong? Let's do it right away.

Starting point is 02:02:04 It was 140. Let's do it. Fantastic. Okay. Talk about the progress. What specifically unlocked such a huge fundraise? Break it down. I mean, let me tell you, congratulations to you guys too.

Starting point is 02:02:18 Thank you. Like very happy for you. Yeah. And thank you for jumping on the show so early when we were a tiny show with just, you know, making phone calls to random founders. That was awesome. It was a lot of fun having you on. Yeah. So sorry, what was the question? Progress.

Starting point is 02:02:34 Yeah, the question is like, yeah, why this round right now? I mean, I can imagine a bunch of reasons. I can imagine like you built the thing successfully and it's working. Or I can imagine just it's very, it's way more clear what cheap energy means. Because with LLM training, you load some GPUs on this thing, you send it up to a satellite. Like a lot of the pieces of the puzzle to underwrite this deal feel like they've fallen into place. But yeah, there were a lot of business. businesses that we're pitching at the point that we first talked that have made less and less sense, right?

Starting point is 02:03:07 Even things like application layer companies would be one. And it feels like your business makes more and more sense as the sort of. Especially on the back of like data center bands and energy expenses and natural gas flaring and G.E. Vernova's out of stock. And even if you get it, it's going to be, you know, natural gas powered like going to the ocean, going to some way. far away, seems to make a lot of sense. Yeah. Well, I think there's two things happening at the same time.

Starting point is 02:03:37 Number one is this is a deep tech play that we started working on a long time ago, and we had to develop the whole new technology to do it. No one has built an autonomous system to go to the middle of the ocean, capture energy, turn it into anything. You know, computing fuels is one of the things we're going to be doing too. So, you know, that took time, and now we're at the point where we can actually start scaling these systems. So that's a big unlock.

Starting point is 02:04:01 That means that we can start building our manufacturing plants. It means we can start getting all the bugs out of that, getting the bugs out of the first fleet that we're putting out starting later this year. So that's the sort of inflection point on the company. But then you've also got, I think, everyone figuring out that there aren't that many ways to get energy on the planet. It's the really hard thing.

Starting point is 02:04:21 It's like there's gas, which you can scale. There's solar, but it takes a lot of people and a lot of land. And so what are your options? people are saying, well, you know, you guys were onto something when you said middle of the ocean is a pretty good play. So it's been a lot of fun to gather that coalition of investors who are like, wow, this is a totally orthogonal play. And it makes a lot of sense. We can go and do this. How much energy does one, what are we calling it, device ship?

Starting point is 02:04:49 Node. We call them a node. Okay. N-O-D-E. Yeah. How much energy? Because we talk about an average metacampus might be half a gigawatt or something. And, you know, we're moving towards a gigawatt a month.

Starting point is 02:05:02 And there's different clusters and different sizes of campuses. And I imagine you can create a fabric of these nodes that work together. But a single node, how much energy are we generating? How can we think about it when we compare it to just the broader, like, data center campus world? Yeah. The typical node will be on the order of 500 kilowatts. Okay. That's what we're thinking it'll be.

Starting point is 02:05:25 Yeah. It's 100 kilowatt up to a megawatt. Yeah. So it's like one rack to mold. multiple racks depending on density. Okay. And yeah, they can talk to each other sideways. Yeah.

Starting point is 02:05:34 So we'll have radio. Of course, they're talking to satellite. And so we're building this grid, you know, and it's not for synchronous training, right? This is not fiber between everything. But if you want tons of embarrassingly parallel inference, you know, running all the same models, running future models that get bigger or smaller, we think we're going to be perfect for lots of, lots of that. So the way that we see this going is like, once the energy starts to crash through, what's available in the grid, it all just becomes about who can actually build in an elastic fashion to make, you know, 10 gigawatts, 100 gigawatts, 100 gigawatts per year is the kind of thing

Starting point is 02:06:10 that we're getting asked to do. So that's the kind of framework that we want to build up, just pure manufacturing to get energy capacity out there. Yeah, yeah. So you build a thousand them. So I imagine a huge industrial project now to actually scale the manufacturing process. What are you, is there anything unique that you need to do? do around battery storage to maintain generation capabilities? Or would you just like take a node offline if it's in still water? Because I imagine that there's a sometimes it's generating more than others depending on what the conditions of the ocean are. Yeah. So the reason we chose this resource, and it's not just any ocean that we're going to. We're not going to, you know,

Starting point is 02:06:52 200 miles off the coast of San Francisco or something. We're going to the southern hemisphere oceans, which are really power dense. The wind is blowing all the time. The waves are on even more than the wind, because it's this big battery for wind energy, big battery for solar energy. So when we're there, we're on almost all the time. There's like a couple times per year where we dip down a little, and we do have some battery on the system for that, but it's not like solar where you're using the battery every day. It's just a couple times a year. We don't need as much battery. We can cycle that battery far less frequently. And so the level of uptime that we want, We can go to four nines.

Starting point is 02:07:28 We can go to like 98% uptime if that's the economic optimum. And so we do a lot of that optimization to figure out for the chips we're running, for the workloads we're running, what is the exact right amount of battery? What is the exact right size of the payload and all of those things? Where are you thinking about placing a network of nodes? Around Antarctica. Antarctica. Yeah. I mean, essentially, right?

Starting point is 02:07:52 It's the southern oceans. Yeah. Yeah, it's like way, way north of Antarctica. It's like, you know, Southern Hemisphere, south of Australia, south of New Zealand, all the way around the planet. But like, yeah, we don't go, you know, we don't go down into those parts of the Southern Ocean.

Starting point is 02:08:07 Okay. Yeah. We can also do some in the North Pacific and so forth. That's where we're doing our pilot fleet. Yeah, yeah. But the real energy resource that we think is optimum for this is that Southern Hemisphere belt. So do you, I mean, if you're, like,

Starting point is 02:08:21 I'm just thinking about if you, if the, if the road is like road to a, a gigawatt. That's like 2,000 of these units that will happen over years, I'm sure. Do you have to build the factory nearby to cut costs or if you make them in America, can you ship them down there effectively? How does that work? The best deployment story is one, yeah, where you have your factory or factories pretty close to the resource. Sure. And so we're working with folks in some of those countries that are nearby. So, you know, identify the right sites. Go build there. right now we're building our first pilot line though near Portland where we are because we want to like dial in all the manufacturing get it really good then we can go and start carbon copying that unit to the right places yeah that's fascinating uh joy do anything else absolutely wild stuff it's such a cool project such a cool project uh well i cannot wait for the for a future video where you are jet skiing between the different nodes it's going to be it's going to be incredible it'll fit a lot of i don't know if the nodes would even see each other i mean you put a thousand

Starting point is 02:09:23 down there. I know there are going to be jets. Yeah, you can spread them out pretty far. Yeah, but anyway, we'll get you guys down there on a yacht. Can't wait. So you can't, so you can check out the fleet. Yeah, can't wait. Incredible progress. Thank you so much for taking the time to come shout with us. We'll talk to you soon, Garth. Goodbye. Thank you, guys. Cheers. Up next, we have Katie Hahn from Hahn, from Hahn Ventures, announcing a new fund. This is very exciting.

Starting point is 02:09:46 This is the biggest number, no. Andre, beat her up. Yeah, but. You almost had the biggest number on the show. show today. We were catching up with the launch. But anyway, thank you so much for taking the time. Welcome back to the show. How are you doing? Thanks. I'm doing well. How are you guys? Thanks for having me back. Of course. It's always going to talk to you. I mean, there's a lot of different directions we could go. But let's talk about the fund, the thesis, expansion,

Starting point is 02:10:12 interactions between different theses, like how you're seeing the market, how you're seeing opportunity and venture broadly right now. Yeah. So it's been four years since we launched Ton Ventures. And, you know, the world looks a lot different four years later, and we're excited to have a billion dollars in fresh funds to deploy behind founders. Oh, thank you. Even though it wasn't the biggest of today, I appreciate that. Wild times that we are in in the venture world, for sure. Yeah. Well, we're taking this fresh funds of billion dollars and backing founders who are building what we're calling the new economy. And by the new economy, I'm thinking about three structural shifts we're seeing right now.

Starting point is 02:10:55 The first is new financial rails and infrastructure. I mean, think companies like Aribor, right? Gone our bankers' hours, gone our wire cutoffs. I mean, you're talking about global from day one, 24-7. This is a bank that opened on a Sunday. The second structural shift we're seeing are new assets in markets. It started as stable coins, but it's quickly kind of we're now talking about tokenizing the stock market. And you see giant like Black Rock and you see Coinbase and Robin Hood doing that.

Starting point is 02:11:24 But it's also opened up new markets. I mean, think prediction markets. Think perpetual markets. You've had a lot of those guys on the show. And then the third structural shift we're putting this new set of funds behind. And this is the earliest of the category. But of course, we have an early stage fund too is what we're calling the agentic future. And it's not just AI broadly.

Starting point is 02:11:42 It's where AI and crypto intersect. And I think that's a lot more areas than people realize. I mean, we're talking about building for a world. world where the end user is not necessarily a human of a financial product or service, but as a computer or an agent. So how does that impact things like privacy? How does it impact things like provenance and trust and what form will they use? How will agents pay for things or subscribe to services?

Starting point is 02:12:06 And blockchains aren't good for everything, but they are really good for some of those things I just mentioned as our other cryptographic tools. Yeah. Yeah. I mean, I remember that being like back in 2015, 2016, the original machine to machine payment. thesis around Ethereum and even some Bitcoin folks were talking about it. Tons to dive into there. Just in general, we were talking to Collison Brothers last week about how stable coins are maybe in some sort of a winter. Like there was a big surge of 30% growth in the market and then

Starting point is 02:12:38 it's been a little bit flat. Do you think that this is a regulatory story? Is this a technology story? Like, what is the next breakout adoption? Yeah, yeah, the next adoption hinge point for just crypto technology to diffuse further. Yeah, sure. Well, I mean,

Starting point is 02:12:58 I think actually, I take the other side of that. I think you have now, MasterCard, for example, just paid $1.8 billion for one of our stable coin investments, BNK. And that was the third largest ever in history acquisition by MasterCard. That's crazy. And so I don't, I don't think we're

Starting point is 02:13:14 in a stable coin winter. I think actually you have double-digit trillions of dollars in transaction volume now flowing through stable coins. And this is a technology that didn't even exist guys 10 years ago. And now it's about to surpass the combined transaction volume

Starting point is 02:13:28 of Visa and MasterCard. And we think that'll continue, but the story doesn't stop at tokenizing dollars because now, of course, a lot of economies want stable coins for their currency. So I think that's a tipping point. Another one is tokenizing

Starting point is 02:13:41 other financial products and services. is like the stocks I'm talking about. Yeah, yeah, yeah. Yeah. Is that a disruptive innovation or sustaining innovation? Because I can imagine that the black rocks, the coin bases, like the Robin Hoods of the world, like do very well in that world. They're already set up for them. They have the audiences and the customers.

Starting point is 02:13:59 And they, you know, expand into that category. At the same time, you could imagine we've seen upstarts like polymarket Kalshi, like, you know, slightly new twist on an old idea. and it's just a huge new unicorn or decicorn emerges, and sometimes multiple in a single category. How are you viewing the idea of like on-chain equities as a venture opportunity? Well, and it's not just on-chain equities. I think it's starting there, right?

Starting point is 02:14:26 It started with Fiat and now starting with, for example, securities or stocks, but I think it won't end there. I mean, I think it will be all kinds of financial products and services. And I'm not one of these people who is going to tell you all assets will be on chain. Every single physical asset will be in chain. Maybe eventually that's an end state, but I don't think it's an end state necessarily in our lifetimes. But I do believe that financial services and products will end up on chain.

Starting point is 02:14:52 And we're just starting to see that. You mentioned prediction markets. It's a great example. Again, and by the way, you talk about the original prediction market was auger, right? And that was just ahead of time years and years ago. I mean, when was that, 2016? But that was before you had stable coins. And it turns out you need some of the infrastructure to really have prediction markets find true product market fit at scale.

Starting point is 02:15:15 And that's what they're undergoing right now. And I think we're still early in prediction markets. It's like sports and politics. Last time you were on, last time you're on, I think you made the call, if I remember correctly, that, yeah, it was still early in stable points. Sorry, not stables, but prediction markets. And you were correct. There was a boom. There was a bunch of new kind of like vertical.

Starting point is 02:15:37 approaches, how are you processing it? Do you expect to make like a net new prediction markets bet out of this fund or have you made your bets at this point and you're just going to watch it evolve? Well, in Fund 1, we made a bet and Coinbase bought that bet. And that was the clearing company. So I certainly hope my prediction is that we have a prediction market bet in Fund 2. We don't have one now. It is some an area where and but the prediction markets that we invest in might not look like what they look like today. Remember, we're a venture fund. We're not a hedge fund, so we're looking at over the next 10 years. And I think an area I'm particularly interested in, and we as a fund are interested in for prediction markets, is some enterprise use cases.

Starting point is 02:16:18 Think about insurance. Think about litigation predictions. Think about drug trials. Think about risk hedging and all kinds of other business uses, aside from the fun of sports betting, aside from the fun, if you can call it fun of betting on politics, outcomes. There are a lot of business use cases and institutional use cases, and that's why you see folks like the New York Stock Exchange and ICE getting involved. But I think we're really early scratching the surface. So the question is, is it going to, is value going to accrue there to one of the established entrance already? Or will some new upstarts come about?

Starting point is 02:16:54 And so we're keeping an eye on both. And as I said, there's huge opportunity, but with huge opportunity is going to come a lot of legislation, we think, and regulation. And yeah, how do you think the whole how do you think the whole insider, you know, given that you don't have an active bet, it sounds like, how do you think the whole insider trading? Oh, yeah. We do not, we do not have an active bet. Yeah, yeah, yeah. That's what I'm saying. So I just feel like you can.

Starting point is 02:17:21 But I have a huge bet on her taking a bet in the next couple of years. Okay, okay. Because she just said that, so I'm going super long on that. No, but my question is like, it's like, maybe I should read rice. I hope we have a bet. I just wanted to ask around like how you see the insider trading debate evolving because clearly if you're somebody that is looking to prediction markets for alpha, for data theoretically, like prediction markets.

Starting point is 02:17:50 You mean specifically like right now it's handled on terms of service level and it might eventually be handled at like the government regulatory level. Yeah. Right. I think now you have a little bit of what we'll call self policing. and, you know, industry best practices, but the industry is still early and what are best practices today might not be.

Starting point is 02:18:08 I mean, and I think a lot of these platforms, look, a lot of responsibility is going to fall to these platforms. This is a novel issue with any nascent technology. You're going to have novel issues of first impression from a statutory point of view, from a constitutional point of view. You're talking about insider trading.

Starting point is 02:18:22 That's a criminal point of view. And as you guys know, I spent over a decade as a federal prosecutor at the Justice Department, and I've seen a lot of, a lot of situations emerged that, frankly, companies weren't really contemplating and that just come about. And I've seen a lot of that in my career as a prosecutor. And I think that these platforms are really going to have to be thinking very deeply about these issues. And I think they

Starting point is 02:18:44 know that. I mean, they're only going to be heightened because of the opportunity. And I think that's really exciting, but it's the flip side of the coin. Right. And it's not just insider trading. Right now, these platforms are asking for federal preemption, which I think is really interesting because you might want federal preemption today or you have a CFTC chair who is not hostile to prediction markets. But don't forget that Gary Gensler himself was once the CFTC chair. And so I think it's a really interesting question of do you go state by state? Do you ask for federal preemption? And so there are a lot of questions beyond just insider trading.

Starting point is 02:19:20 So unpacking that risk, there's a world where you get federal preemption, but then as different rules and laws are written, if there's a different, less friendly CFTC chair, the downstream, like, the implementation of that oversight could be disadvantageous to the industry. Is that basically the risk? You're exactly right. You could win the battle but lose the war. Yeah, that's interesting. Yeah, but that's, I do see why a lot of those platforms are looking for federal preemption right now. But again, you've got to look at this as a multi-trillion dollar asset class and you've got to look at it over the next decades. And I think, you know, social media was a huge category, a huge market opportunity for investors. And yet also these platforms had to get really sophisticated over the last decade with how they police content, for example, and how they work with regulators.

Starting point is 02:20:12 And it's not just in the U.S., as I said, this structural shift of new assets and new markets means it's global from day one. So you're not just thinking of the U.S. You have to think about globally what do regulators across the world think about this. They don't always just follow what the U.S. and the CFTC think. On the intersection of AI and crypto, are you equally excited about bringing crypto to existing AI agents? Someone has an open claw and they wanted to buy something. And so stable coins speed that up, machine to machine payments, micro payments, all of that.

Starting point is 02:20:51 Or is there actually more of an opportunity or maybe an equally or maybe less discussed opportunity around bringing AI to crypto, thinking about like a cursor, but it's really good with writing smart contracts or something like that, where you're still primarily selling to the crypto community, but you're bringing AI tools to bear for that world in the way that there were several big crypto winners that were sort of Web 2.0 SaaS products, but they applied their strengths to the crypto world very successfully and built huge businesses. Are both of these Equally exciting is one of them more hype than the other? How are you processing those two opportunities?

Starting point is 02:21:31 I think they're both really early, and I think they will naturally, we're interested in both. I think there's a case to be made for both, and we can talk a little bit about that. But I do think that it's more of an opportunity for our early stage fund right now, because we are very early at this. I think to the first point of, you know, you mentioned micro payments, this is something that those of us in the crypto community, over 10 years ago, we're talking about with companies like change chip and other things. And it turns out that fast forward, now there is a use case. And I think we've heard John Collison talk about how he's tickled about this. And I think that's the word of use, tickled about the case for micropayment. And, you know, we don't think agents who work 24-7 around the globe, again, they don't work

Starting point is 02:22:14 bankers' hours. We don't think they'll necessarily just be using credit cards for payments. I mean, think about the need for instant settlement. And if you think about the need for finality and these transactions that are, you're, again, around the glow, but also micropayments. And we think crypto rails are actually perfectly suited for that. So that's the first thing. But in terms of bringing blockchain-based solutions or cryptographic tools,

Starting point is 02:22:38 we also think there's a lot of exciting areas there where there's an intersection. And again, it's very early. But you can imagine provenance. You know, the blockchains are real, blockchain-based systems are really great at proving provenance. And I think in a world of AI where you're wondering the provenance, you're going to be wondering ever more about the provenance of things. We've already been wondering about the provenance of things where most things were created by humans.

Starting point is 02:23:02 And now that's starting to shift, and more and more is being created by computers, and we think more and more will be created by agents. And so what does that really mean for provenance? What does it mean for privacy, by the way, when all of a sudden all of your data, we have things, you know, GDPR and the California equivalents that are out there, but in a world where you have different AI agents

Starting point is 02:23:23 that can really quickly, much more than humans can, kind of undo that privacy, do you look for zero knowledge-based proof solutions for that? And that's, again, a cryptographic tool. So I would say reputation systems, provenance, privacy, agentic finance. They're all fair game and right within the wheelhouse of what we've been doing guys for the last four years and we're excited for the next four years. And I think we are probably, it's for a billion dollar fund, We have the trust of 35 or so global institutional investors from around the globe who very much believe in this future of the intersection of these technologies. Yeah. No, it makes a ton of sense.

Starting point is 02:24:04 Yeah, talking to the Collison brothers really, I don't know, just very much like reset me on the actual need for micro payments, the value that comes from that, talking about token theft and basically setting up an account, getting a. on 30-day billing, not being good for it at the end of the 30 days. Like, that's solved if you're paying in a income stream. Streaming payments actually makes a lot of sense there. And if agents have something to go look at, that's immutable, this leg of submittable. And so you start to see. And, you know, it's not just the Collison Brothers and Stripes that are doing this. Obviously, Coinbase and companies like Robin Hood are really leaning heavily into these areas.

Starting point is 02:24:45 And then you have, again, the giants like MasterCard or Visa who, I mean, I think Visa's chief product officer was saying this next period over the next decade for agentic finance is one of the most interesting. So they're keeping their eye on it closely too. And so are we. Amazing. Jordy, anything else? Tremendous progress. Congratulations.

Starting point is 02:25:07 Thank you so much. Let's not let it go another year for your next appearance. Absolutely not. We'll talk to you soon. Have a good one. Thank you. Goodbye. Thank you all.

Starting point is 02:25:16 Bye. Up next, we have Nick from Rivett. He's the co-founder and CEO. He's been on the show. Has he not been on the show before? He's been on the show before. Right? Nick, welcome the show.

Starting point is 02:25:29 You've been on the show before, right? Yes, quickly, for a seat around. It was a great time. Welcome back to the show. We need to update our CRM, our database. Can we have you down here's first appearance? I know that that's fake news. Anyway, a lot of people have been worried about artificial intelligence.

Starting point is 02:25:47 They say, is it worth the water, isn't worth the energy. But I think after they see these results from tax bench, they're going to change their tune. Tell us about it. Yeah. Well, before, yeah, yeah, introduce a company first. Yeah, Rivett is an AI enabled accounting firm. We do tax returns and tax advisory work for thousands of companies. When we implement these models, there you go, yeah.

Starting point is 02:26:12 When we implement these models, right, we need a way to test them. So we've had an internal benchmark that we've used for. a long time to see whether these models are actually doing our job as accountants effectively. It was time to publish that benchmark. And so that came out today, the results are pretty striking. The top models don't get most questions right if you ask it the same question five times in a row. Interesting. A lot of them do incredibly well when you ask it once and cross your fingers that you're going to get the right question or the right answer.

Starting point is 02:26:39 Yeah, yeah. But you ask the same question five times in a row, are these models actually reliable? Turns out they actually really aren't, especially at scale. interesting and and why are they not like doing enough RL like you would think that like these these these Carpathia has this take there's verifiable kind of like results right like this should be something that the models could do need lean well so maybe it hasn't been prioritized enough so most of the answers require five six seven eight steps and you have to get the right answer every time. If you get one step wrong, the ultimate result is wrong. And what we've found when you ask

Starting point is 02:27:21 it these types of questions is that it'll get a lot of it right. But that single big skip, you know, hey, my client lives in New York City. They're about to sell their company. Does their, you know, does their stock qualify for QSBS? It'll check if they held the stock for five years. It'll check if they were under the gross asset limit when they acquired the stock. But then it'll pull a brand new article saying, hey, New York's considering running rid of QSPS. Oh. Oops. It must not recognize QSPS.

Starting point is 02:27:55 No, they don't qualify. And ultimately, the client is left out, you know, in the dust. Yeah. It's like two news heavy, two, like, web heavy, not like law. Yeah, and it's an interesting, it's a very interesting domain because these are, these, this is not like doing, like, research. for like a project where like, you know, the consequences of getting like a fact wrong or like somewhat defined, like maybe you just like came to the wrong conclusion. But this is like you, if you don't work with a human at all and you just work with a model

Starting point is 02:28:26 and then you get something wrong, it ends up being like maybe it costs you $100,000. Maybe it costs you a million dollars. Maybe maybe more, right? Maybe it's even five grand. And then still. And so I think like this, I expect a long period of human in the loop on. and potentially forever on this domain specifically because for a number of other reasons. Anyways, full employment for Nick.

Starting point is 02:28:52 The final boss. Have you have any of the labs like reached out and like are they actively working on this? Or is it not a priority because they know people like you will will be putting them the models through the paces and kind of like verifying the work? So we'd love to try all the new secret stealth models that no one's, you know, that are not available online quite yet. We do have an email entry on the website. If you want to try your own model against our benchmark, we'd love to run it against it. Nobody's reached out quite yet, right? And I'm not super surprised.

Starting point is 02:29:26 Researchers have very different priorities than real businesses running these workflows. I don't think anthropic cares about my benchmark, which is fine, right? That's not their business. we're going to pick the best model for our purpose and move on. Funny enough, even the new models don't do as well as some of the old ones. We've actually found a regressions as they launch new models. They perform worse and worse. 47 performs worse than 4-6 from Anthropic.

Starting point is 02:29:52 5.4 Pro performs better than 5.5. The 4.1 model from Grok beat their 4.2 model. So it's not always newest as best, flashiest as best. you really have to test the specific model that's come out to make sure that when you're implementing it, it actually does improve your results as a business, you know, the person calling the model. Interesting. Yeah. Really fascinating.

Starting point is 02:30:14 How you started this business not, my understanding is you didn't start this business because you saw like models getting much better, but really that you just thought you could build a better service around tax for startups. and you now, I guess, get the benefit of the intelligence explosion. Where are you, like, even with the flaws, are you still getting a lot of leverage out of them? Or is it not really, because of the lack of reliability, is it not giving you quite the leverage that you'd like yet? No, it's a totally fair question. We get a ton of leverage as long as they are paired with a human who really knows what they're doing. You know, you talked about the stakes for some of these questions.

Starting point is 02:31:03 If you're just playing with Klaude to write a poem or make a website, you as the end user know what you like and don't like, and you can steer it towards what you want to see and don't want to see. If you aren't an accountant, you have no idea what to steer it towards. You have no idea to prompt it. Hey, please double check that New York recognizes QSBS. Hey, please double check that you pulled the new tax rules under O triple B. You need to have somebody skilled who knows how to talk to it and engage with it

Starting point is 02:31:28 and structure the prompt correctly. And make sure that when it tells you something, it actually smells and sounds right. So we do get a ton of leverage out of them. We have quite a few workloads that are powered by them internally. But it requires a ton of work on top of that call to make sure that the answer that's ultimately shown of the client is correct.

Starting point is 02:31:48 And not just hallucinated figures or hallucinated rules that ultimately, you know, nobody's going to jail, but they're going to get a lot of letters in the mail and owe quite a bit of money. That's not good. Yeah, André Carpathie was talking about how there was a huge jump in chess capability between GPT 3.5 and GPT4. And what he attributes it to is just like the researchers were interested in chess during that training run. So they fed it a bunch of chess data and it got better at that.

Starting point is 02:32:17 And it feels like there might be a moment in the future. We're like, oh, like this lab on this model date got really interested in chess and went and bought like all the, or in tax data. and bought all the relevant training data, and now it's jumping up on tax bench. If you were to, I mean, I'm looking at some of the stats here, and it shows like GROC 4.1 fast reasoning at 4.2%. You know, some others are in like the 10%, 12%, 13% range.

Starting point is 02:32:48 Is this like Arc AGI where we're just at very, very low passing rates and you're sort of waiting for them to start climbing up exponentially, but we haven't seen that takeoff yet? So to the models credit, the scores that you're referencing are mostly from data retrieval, which is the most operationally difficult

Starting point is 02:33:09 category of questions that we ask the models. They're given access to a client's entire data packets. This could be hundreds of pages of PDFs and Slack messages and emails. And asked to answer a question. A question could be something like, find their 2023 tax return, find the carry forward capital losses that go under a 24 return.

Starting point is 02:33:30 If you ask a junior CPA straight out of college to do this, look at it right 100 times out of 100. It's not difficult. It's in the exact same spot and the exact same form every time. These models really struggle. They really struggle. They'll just make a number up. They will get frustrated when they can't find the 1040 off the bat.

Starting point is 02:33:48 Maybe they have trouble OCR and the client took actually pictures of the tax return versus a real raw PDF. They really struggle with the combination of searching and then analysis on top of that. A lot of them skip. They'll pull the wrong carry-forward figure and ignore that there's a $3,000 allowance against current income for 23 to subtract $3,000. It's small, tiki-tacky things, but if you try to deploy these models into a real production workflow, they're going to get it wrong. You have to have a human review it, right? You have to have a neck to choke.

Starting point is 02:34:17 Something goes wrong. And so we've hired a great team of next-to-choke, so to speak, to actually make sure that the work that's delivered to the client actually. does what it says it's going to do. Have you thought about building a harness? It feels like a lot of the unlock in software was on the back of Codex and ClaudeCode code. Like something like the harness is somewhat simple sometimes, but it clearly unblocks the thinking model in a very important way. And it feels like for some of those rules-based or tool usages, like just something, a thin wrapper around these these models could potentially lead to much higher scores.

Starting point is 02:34:59 Have you looked at that? You are hired as our next product manager. Welcome to the team. You start on Monday. No, it's absolutely on our list. One of the most impressive tools in the market that we've seen today is Thompson and Reuters co-counsel, which is the UI looks like any other model, but it's built on top of their tax library. Oh, interesting. They don't get anybody API access today.

Starting point is 02:35:22 It's hundreds of dollars per seat per month. team. And so I would love that as an API. We'll be building on it when they launch it, I think, coming later this year. That's what we were told. But in the meantime, we'll be building our own and jumping down the Rapportle, so to speak. Good luck. Yeah, good to get the update. And we will, every researcher that comes on, tax bench, this is the most point one. Forget Eric AGI. It's all about tax. AGI. I don't care about gaming. I think as the IPOs start happening, they'll start caring more about performance on tax. Probably, yeah. I mean, it's a huge, huge market, huge uplift. You've seen what happened in the coding market.

Starting point is 02:36:05 If you add accounting and finance to that, you can see the next leg up on all the revenue charts. Well, thank you so much for coming on the show. Great to see you, Nick. Have a great day. Thanks to you, guys. We'll talk to you, see you. Goodbye. Bye.

Starting point is 02:36:18 Breaking news. Breaking news for gamers out there. What's that? I know a lot of. gamers out there. They have laptops. They're worried about the price of these laptops, the batteries, oh, electricity is getting expensive. What are you going to do? How are you going to charge your laptop? We got the solution for you. It's a gas, it's a gasoline powered laptop. So it's pretty simple. It's a one-of-kind gasoline-powered laptop. It's offered for just $850. Looks like it's

Starting point is 02:36:47 running Windows XP. You might not be able to play the latest and greatest games. Will it run crisis, maybe, maybe not. With a full tank, you can get an hour and a half of run time. It runs a two-stroke engine, and it's perfect for off-grid computing. That's hot right now. And so the laptop specs, let's take you through it. It's got an Intel Court 2 duo. Two gigs of RAM. Ram's going up in price. This is valuable. This is an appreciating asset. 120 gigs of hard drive space. That's going to hold a lot of games when you're off-grid. You only have a little bit of gasoline. If you want a game, this is the laptop for you to good working. Running Windows XP.

Starting point is 02:37:25 And it says that it starts easy. The two-stroke engine on the gasoline-powered laptop, yes, it does start. Colin in the X-Chat says it gets 300 tweets to the gallon. 300 tweets. Oh, no, I accidentally said GTA5 to max settings. Well, at least it'll serve as a benchmark test. How many Chrome tabs can it open before it crashes? People are having fun with this.

Starting point is 02:37:52 The gasoline-powered laptop, this is true hacker mindset. Whoever built this is an incredible engineer and did something. They did be impossible. They built a gasoline-powered laptop. I've seen a couple other of these, like gasoline-powered projects, people making all sorts of different things. It's always a funny gag. The actual breaking news is that the White House is considering vetting AI models before they are released. Trump Admin, which took a non-interventionist approach.

Starting point is 02:38:20 AI is now discussing imposing oversight on AI models before they are made publicly available. Well, FDA for AI. We'll see. It would be potentially an executive order to create an AI working group that would bring together tech executives and government officials to examine potential oversight procedures. Okay, so this would be an executive order to create a working group that could potentially create an oversight body. Okay, so we're a couple steps away, but it seems reasonable. I don't know. Depends on what the what the, what the, what the, what the, what the, what the, we're just a couple of steps away. Okay. So we're just a couple steps away. But it seems reasonable. I don't know. It depends on what the, what the, what the. what the benchmarks are, but you certainly don't want. And Trump says, we're going to make this industry absolutely the top because right now it's a beautiful baby that's born. Interesting way to put it. We have to grow that baby and let that baby thrive. Is this real? It's a real quote.

Starting point is 02:39:06 Are you messing with me? This is a real quote that Trump said about AI. He said, we have to grow that baby and let that baby thrive. We can't stop it. We can't stop it with politics. We can't stop it with foolish rules and even stupid rules. It's not a baby. It's a $10 trillion industry. It's like the engine of the global economy. Anyway. Dean Ball's got a quote in here. What does he say? The technology is moving extremely fast and there are few formal procedures, but they don't want to overregulate. He said it's a tricky balance.

Starting point is 02:39:36 I say don't release it unless it's acing tax bench. It's got to be able to do the taxes before it gets out into the wild. No, obviously you want these models to be safe. You want them to be reliable. You want them to avoid negative external. and anything that gets us in that direction is probably good, but everything comes with tradeoffs. Final post of the day. Yeah. I'm Tommy. Hi, PhD in Hammerology here. All right. So what we're looking at is a nail.

Starting point is 02:40:06 That is the correct mindset. When all you have is a hammer, everything looks like a nail. Also, go check out Riley Walls' new project. He's shipping stuff every week. This one got a million views. You probably already saw it, 27,000 likes. 10% of AMC movie showing sell no tickets at all. So if you want to go see a movie in a private theater with no one else, he made a site

Starting point is 02:40:27 that finds empty theaters and tells you exactly when you should go and book. You can go see Project Hail Mary at 12.30 p.m. today in New York. If you don't have work, you can go see Project Hell Mary in your private theater. It's available at walser.com slash empty screenings, W-A-L-Z-R. dot com empty screenings you can search by zip code let's see what's around us is there anything good there's 10.45 p.m. devilware's part of two okay okay so zero seats enjoy it enjoy being allowed got it this is very funny yeah there's he he does he does uh surface some that have one seat or two seats interesting way to make a new friend me and you because you think so you think oh i got the

Starting point is 02:41:14 zero seat theater you're you're in the one seat theater and Somebody's like, I want to meet the psycho that went to the empty theater. And then they're talking your ear off. Who knows? Well, maybe you can enjoy it. Well, thank you for tuning in. We'll see you tomorrow at 11 a.m. Sharp Pacific Time.

Starting point is 02:41:31 Leave us five stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. And we will see you tomorrow. Love you. Goodbye. A little bit missed time there. See you tomorrow.

TBPN Live - GameStop + eBay, Neural Computers | Nat Eliason, Michael York, Maddie Hall, Anjney Midha, Ben Lamm, Jake Stauch, Garth Sheldon-Coulson, Katie Haun, Nick Abouzeid

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.