Pod Insights Equity Podcast The AI industry's 'next big bet,' and should we just buy Chrome? About this episode Today, on TechCrunch’s Equity podcast, hosts Kirsten Korosec, Devin Coldewey and Margaux MacColl dig into the week’s tech and startup news, including what a new scaling method could mean for AI startups if it works and for chip providers if it fails.
Listen to the full episode to hear about:
The DOJ’s latest proposal for Google: it should divest its Chrome browser. While we wait for a break in the monopoly case and a better search alternative to arise, the Equity crew thinks that Wiz might’ve been on to something when they said no to Alphabet.
Dual-use drone startup Tekever’s $74 million raise and its part in the European defense tech boom
How Converge Bio plans to build an ’everything store’ for biotech LLMs. While on the subject of LLMs, Kirsten took us deeper into ServiceTitan’s S-1, including its boilerplate warning about AI.
Network states 2.0. Margaux had the latest on Praxis’s unusual $525 million raise and the regions its founder is eyeing for a new city.
Equity will be back next week, so stay tuned!
Equity is TechCrunch’s flagship podcast, produced by Theresa Loconsolo, and posts every Wednesday and Friday.
Subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. You also can follow Equity on X and Threads, at @EquityPod. For the full episode transcript, for those who prefer reading over listening, check out our full archive of episodes over at Simplecast.
Credits: Equity is produced by Theresa Loconsolo with editing by Kell. Bryce Durbin is our Illustrator. We'd also like to thank the audience development team and Henry Pickavet, who manages TechCrunch audio products.
Learn more about your ad choices. Visit megaphone.fm/adchoices
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:11 Kirsten Korosec? Hello and welcome back to Equity TechCrunch's flagship podcast about the business of startups. Today is Friday, November 22, 2024. I'm Kirsten Korosec and joining me today are my fellow TCers, Devin Coldewey and Margot McCall. What's going on, you two?
0:29 Speaker 01 I saw some otters today.
0:31 Kirsten Korosec? Interesting. Not anything to do with tech, but appreciate it. I didn't realize it had to be tech. I was gonna say I walked my cat. Okay, two amazing animal moments, but we should probably get into the show. Today, we have deals of the week covering surveillance drones and everything store of biotech LLMs. I cannot wait for Devin to give us the lowdown on that and how the age of AI is giving us a new risk factor in IPO S1 docs. Okay, insider-y, but relevant. And then we're going to get into why current AI scaling laws are forcing labs to change course. And what I'm really looking forward to is network states 2.0.
1:13 Speaker 01 Are we already on 2.0?
1:14 Kirsten Korosec? Oh, yeah. Oh, yeah, we are. Before we get into the first deal, which is a story that I believe Margot wrote about, I did want to mention that there had been some reports about this and now we have final confirmation. The Department of Justice has argued that Google should, in fact, divest of its Chrome browser as part of a remedy to break up the company's illegal monopoly in online search. A couple of things. This does not mean it's happening. It needs to be resolved. We have plenty of court hearings in our future to be following. I did want to mention the best suggestion on Blue Sky.
1:54 Kirsten Korosec? After we posted the story, one of my followers suggested that the Onion buy Chrome, which I thought was amazing.
2:03 Speaker 01 I support it. We can lend them the money. The press is really, really rich. So we can just lend Onion the money to buy Chrome.
2:10 Kirsten Korosec? Should we just buy Chrome? We should. We should definitely buy Chrome and then it will, you know, we'll solve all of media's problems by actually boosting quality journalism.
2:20 Speaker 01 Pure consolidation. I do think this is, it's interesting seeing this actually happen in the courts and Amit Mehta is a very smart judge, as far as I can tell. I seriously doubt that Chrome is going to be completely disconnected from Google, but this is the kind of thing where you put it on the table and it's so scary to Google and whoever that they're like, oh, what if we did this instead? And they're like, oh, Chrome can exist in a siloed thing and we will disconnect our tracking cookie like, you know, technology. They do all this stuff in a panic to like, prevent it from happening. So I think it's more of a, let's put our power on the table here and be like, we could, but do we need to?
2:59 Kirsten Korosec? Yeah. My goal in this, I'm not in the room negotiating, obviously, but if I had a wish fulfilled, it would just be that the search gets better because it's been in a downward spiral for a long time.
3:12 Speaker 01 The worse it gets, it's no longer a monopoly. If everybody is just like, this is terrible, they will no longer have a monopoly. So it's kind of solves itself. Maybe that's what they're doing. It's some 7D chess.
3:22 Kirsten Korosec? Yeah, it is 7D chess. Speaking of which, I did want to mention Wiz, you know, they turned away from a possible acquisition from Alphabet. And by the way, Ingrid just had a great scoop, which is that Wiz has now gone on to acquire Daz for 450 million to expand its cybersecurity platform. Just, you have to check out that story. But man, Wiz just seems smarter and smarter for avoiding that acquisition now because a lot of the complaints around Google and Alphabet in general is a sort of monopolistic behavior.
3:55 Speaker 01 Well, if you have enough money to buy another company ending in Z for 450 million dollars, I don't think you need to worry about getting bought.
4:02 Kirsten Korosec? Yeah, absolutely. So other deals, Margo, I want to toss it to you. And this is all about surveillance drones. And big surprise, you're writing about dual use once again.
4:12 Speaker 02 I actually did not write this piece.
4:14 Kirsten Korosec? Oh, you didn't? Okay.
4:15 Speaker 02 I want to give a shout out to Mike Butcher in Anaheim, who did write this piece. But it's a company called Together. They raised 74 million for dual use drones. For me, what's interesting about this deal is that they're based in Portugal. They were invested in by the NATO. Innovation Fund. And one thing that one of the investors said, and part of the appeal was that it was quote, fully European. I've been following the European defense tech scene pretty closely because US-based companies have had a really hard time breaking into it. Andrel and Palantir execs have both really lamented the fact that European countries only seem to want to give contracts to European companies. So when I first heard this deal, my first thought was, oh my God, another drone company. But at this point, I think being based in Europe is kind of its own competitive advantage.
4:59 Speaker 01 No, I agree. I mean, we've had that, oh, Europe only wants to buy from Europe. It's like, okay, yeah, but US only wants to buy from US. Asia likes to buy from Asia.
5:06 Speaker 02 For sure.
5:07 Speaker 01 Everything's getting kind of Balkanized or like regionalized for, I think, good reason. It makes sense as far as like defense contracts and stuff. So I'm interested in seeing how these things develop individually and whether they kind of converge, you know, like in evolution, everything eventually becomes a crab, right? I wonder if all of these drones will eventually get the same type of features. They'll use the same protocols. They'll use the same little landing prongs or whatever. I'm curious about how they'll sort of co-evolve.
5:34 Kirsten Korosec? By the way, thank you for the visual of the hellscape of like drones all coming together into some sort of transformer crab, which is exactly what just popped in my head while I was listening to you.
5:43 Speaker 01 I hope that happens too. Well, one from every, every region, seven different, you know, continental drones all joining together Voltron like.
5:52 Kirsten Korosec? Oh, amazing.
5:53 Speaker 02 Well, one of the thing TechOver is developing is a big drone that can carry a bunch of tiny drones and then release a swarm.
5:59 Speaker 01 It's already happening.
6:00 Kirsten Korosec? Yeah. Also, total hellscape. Can't wait for that movie. But I did want to bring up something actually important and not, you know, fictional. I find it interesting that the NATO Innovation Fund is, if I heard you correctly, mostly investing in European companies, but NATO includes the US. So what is going on there?
6:22 Speaker 02 Yeah, I mean, I can't confess to have a ton of background on the NATO Innovation Fund. I assume the partner that did the investing was Europe-based.
6:29 Speaker 01 I mean, it's the North Atlantic Trade Organization, right? But like, I mean, US is just one of the, I don't know what they call them, partners or members or whatever. There's kind of a share and share alike. And like, does the US really need investment from like, hey, everybody in NATO, why don't you chip in a little bit so the US can finally develop a drone system? And they're like, are you kidding me? You're the richest country in the world. Get out of here.
6:52 Kirsten Korosec? True. And also, you know, the Department of Defense has quite a large budget, so they have plenty to be able to, on the dual-use side of things, certainly invest in all sorts of different technologies from autonomous vehicles and drones to all sorts of fun stuff. Do you see this, before we move on, as sort of like a European defense tech boom, or is this just a smattering of deals?
7:16 Speaker 02 There's definitely a European defense tech boom. I think you're seeing it, obviously, in parallel with the US defense tech boom, but also the fact that, like, Ukraine and Israel are like much closer to Europe. They're involved in those wars to varying degrees. It's closer to European borders, and they want European companies to be deployed in those wars.
7:36 Speaker 01 Yeah, if you're an EU member state and NATO state, like, this is a very, this is happening in your backyard. And there's a tremendous amount of investment and innovation in the defense space in Europe, because they can't rely on US drones. No, they need to have, like, European-built ones, and, I don't know, have it all be local.
7:56 Kirsten Korosec? All right. And I think we're going to probably see more of that happening throughout the world. Completely moving into an entirely different direction is a story about Converge Bio, which has built an everything store, which I love that phrase, for biotech LLMs. They raised 5.5 million in a seed round. But Devin, I want to understand, please break down what exactly is Converge Bio doing here?
8:22 Speaker 01 Well, the everything store, I think that's the way they put it. But the issue is, you know, you have all these large language models, ChatGPT and stuff like that. Those are all trained on the English language, literature, the internet. But the same model type, they just find patterns in long strings of data. So the biotech world has been looking at these models as ways to understand and interpret and generate interesting data, like genomes, like protein sequences. There's been lots of really interesting work with it. But this is increasingly a problem, that the models come out, they get trained, they have these huge foundation models that get trained, but they're kind of stupid. They're like general purpose. They're not trained to do anything in particular. They're just giant.
9:06 Speaker 01 Pattern identification, statistics, things. And it's really hard to make them do something in particular. Like with ChatGPT, you need to make it polite. You need to make it helpful. You need to make it not, you know, repeat conspiracy theories, stuff like that. That takes a lot of post-training. And so what Converge Bio is doing, the post-training work for the foundation models in biotech that already exist. So if you have a foundation model, it's all about DNA, RNA, protein, protein interactions, whatever. But you need it to be more specific to this antigen or this system or, you know, like this pathology. They provide a way to further train the LLMs and also let them explain themselves a little better.
9:49 Speaker 01 So they're trying to put together that service for biotech.
9:52 Kirsten Korosec? All right. So if I'm hearing it correctly, then how would their business model work? Would biotech company X come to them and say, hey, we could, we would love a customized tool to achieve this purpose? And then Converge would do this sort of bespoke LLM for them? Or is it more automated than that?
10:14 Speaker 01 It's a little bit of both. It's more like if I'm a, you know, pharmaceutical research company and I, I work in antibodies because I make vaccines or whatever. Right? There are large language models out there that are focused on antibodies and stuff, but I don't know how to use them because they are very general and weird. And I need to train it on also my internal data. So what Converge does is uses the proprietary data to enrich the large language models and fine tune it. And making it something that you specifically as a company can find useful and that your researchers can just use like a lab tool. At least that's, and that's the idea. So yeah, the pharmaceutical company would come to them and say, here's what we need. Here's where we work.
10:56 Speaker 01 Here's the data and Converge says, great, give us a million dollars and we will make this available to you.
11:03 Kirsten Korosec? Interesting.
11:03 Speaker 01 That's the general idea. There are companies that are working in this space already, lots of them, but.
11:08 Kirsten Korosec? That was gonna be my last question. Like there are a few phrases in the venture world that I am not a fan of. And one of them is, what is the moat? But if there's a lot of competitors out there, like how briefly, how are they different here? How do they have a chance to compete?
11:24 Speaker 01 I don't, you know, work directly in this space, but when you think of like Ginkgo Bioworks, whatever, like these are already working in the machine learning and large language model biotech space, but they're a little more full service and they're a little more like we're building an entire foundation model that you can use on our platform, whereas Converge is like, hey, there are foundation models out there that anybody can download and use, but we will let you use it as your own on your own data and you own everything about it. It's a little more of a like, you know, roll your own solution rather than like providing a full service thing.
12:00 Kirsten Korosec? Got it.
12:01 Speaker 01 And they think that they can acquire earlier customers in that way, ones that aren't going to go and spend, you know, a million dollars a month, but they can afford a hundred grand once to fine tune one model. I don't know, it's there, but they're trying to get in early with the like, you know, more than half of these biotech companies have no experience whatsoever with machine learning and foundation models. So they're like, look, we'll get you in there early and then lock you in as a customer.
12:24 Kirsten Korosec? Awesome. Well, I have a question for Margot that is not about biotech. And I'm just wondering if, like myself, when I'm looking to relax at night, I cozy up by the fire and I open up an S1 and I read the risk factors of whatever recent IPO.
12:42 Speaker 02 They're honestly really fun. I get excited when a company files.
12:45 Kirsten Korosec? Oh, I'm not being sarcastic, although like the fire part, I live in the desert. It's not that cold here. But I will say this, one of my favorite things about an S1, which is, you know, the prospectus that a company that's going to go public files and it finally reveals all the secrets and all the things you've always wanted to know, hopefully is the risk factor section. And Venture Editor Julie Bort found a little fun thing when cloud service startup Service Titan filed for its IPO. In the risk factors, and Devin, you'll enjoy this, the company included some boilerplate warning that LLMs that help grow the business could lead to damage. And they specifically named.
13:29 Kirsten Korosec? Microsoft and OpenAI. First of all, Margot, I'm wondering if you've come across this as a risk factor for any other IPO.
13:38 Speaker 02 I haven't seen it. I don't know if you have. It was interesting. It was interesting to read and it definitely made me think that this is going to be something that a lot of S1s have in the future.
13:46 Kirsten Korosec? Yeah, I wonder if it's just going to be this like standard language that there is a lot of standard language in a lot of these risk factors. And then you sort of, the goal of course is to root through and find the really unique things for that particular business. It's like, ooh. But as far as I can recall, I have never seen this. And I'm wondering what the risk factor is here, or what the exposure to business is here, beyond the fact that an LLM could, you know, create a hallucination or something incorrect. Is there anything else that you can think of, Margot or Devin, beyond the hallucination part, that would be an obvious like risk?
14:19 Speaker 02 Yeah, they mentioned sharing like customers' private data with third-party companies and not the risk of security breaches for that.
14:26 Speaker 01 Yeah, that's almost, it's standard in a way with like when you deal with third parties. If you're like, oh, we have to use a cloud provider and the cloud provider could get breached. I think it's a step beyond that with the AI stuff because it's not just hallucinations. It's also like, there is really unknown unknowns when it comes to AI security. There's all these novel ways of attacking LLMs and generative AI that people really don't know about. So it makes sense to to step carefully when you're implementing it at scale. You want to say out front, like, this is a new and weird technology and we don't really understand what the risks are, but we know that they probably exist. It ends up sounding pretty general, but I would expect
15:09 Speaker 01 to see that a lot more in these kinds of filings.
15:12 Speaker 02 Can I ask, because I don't cover IPOs a lot. I tend to cover earlier-stage startups. Could this be a sort of like forcing function for startups to use safer AI models or to use like, and we were talking about Sierra, which is Brett Taylor's, which is AI like made for companies to not hallucinate and to be safer so that they don't have to put in language like this in their S1? Or is it kind of like, if it becomes standard, it becomes standard?
15:35 Kirsten Korosec? I think it's a little bit wishful thinking because I mean, the risk factors is imagine a team of lawyers going over every possible risk. And so even if there are solutions out there like Brett Taylor's, you know, new company is trying to do, they're still going to put it in there because it's the what if. Yeah.
15:51 Speaker 01 But I think that, Margot, I think that you're right in that there will be, if the risk is stated, then and it becomes standard to state it, then I think a solution will also become standard. And whether we have a regulation that requires it, or investors require it, or lawyers require it, someone is going to say, oh, well, there are these unknown risks. And that's why we retain a pen testing startup that does specifically this kind of testing. And they do, they give us an independent report every quarter on whether these things are vulnerable to this and that. And so you have a kind of built in, they say, well, we know there are risks. So we're taking this mitigation strategy that they might not otherwise have put in.
16:29 Kirsten Korosec? Well, we'll have to see how this plays out. And I will certainly be looking at every S one, although there haven't been a ton of them lately, to see if there are L L Ms. After the break, we are going to talk about two really interesting things. One about current AI scaling laws and also network states 2.0 as I'm calling them. And I can't wait to hear what Margot has to say about that.
16:55 Kirsten Korosec? So I'm again going to turn to Devon to help me with this. But basically we had a story about how current AI scaling laws are showing diminishing returns. And as a result, it's forcing AI labs to change course. So before we get into that, what are we talking about here?
17:14 Speaker 01 What is laws?
17:15 Kirsten Korosec? What is laws? What is laws?
17:18 Speaker 01 It's, it's, it really is very basic. Like it makes it sound really technical. I'm going to try to keep this brief because I want to talk about network states, which are so ridiculous. But like, think of like Moore's law, you know, Moore's law was basically like every, every however, long 18 months or something, you double the amount of transistors on a chip. And that led to computers becoming faster and faster in an exponential way for many years. And every time that happened, it enabled new applications and so on. That's an example of a scaling law, and it was very reliable for a long. Time. So in AI, the scaling law that we've been working with is nowhere near as formal as Moore's law. But it's basically that the more parameters, which is the sort of like the amount of data that a an AI model has to work with.
18:00 Speaker 01 The amount of parameters and the amount of like compute power that you put into training it, the better it is. And it was really unclear, let's say, a year, a year and a half ago, it was unclear whether as you go from like, you know, a million to a billion to a trillion parameters and like a week of compute time to months and months of compute time on a thousand times more computers. It wasn't really clear whether you would get exponentially better AI. Or sort of like, at some point, it doesn't matter how much compute you throw at it, you'll only get like 1% better AI. And that's what actually people are sort of experiencing now. This is still very preliminary. It's not like, oh, AI is done.
18:40 Speaker 01 It's more like, we've been scaling AI, we've been making AI models better by making them bigger and by training them with more data. But A, we're running out of data. They already used all the data in the world and stole whatever else they could find and threw that in there too. And the compute is getting so expensive, it's ridiculous to try to say like, oh, we need 100 times more to get 10% more performance. So what they're finding is that there's other places that you can gain capability in the AI models. And it's not just by making the model bigger. The one that everybody is focused on right now is called test time compute, which is a fancy way of saying letting the model, for lack of a better word, think about the answer before answering.
19:26 Speaker 01 And this has been shown in various ways for a couple of years that if you if you ask one of these models to just answer a question, like a complex question, it'll answer. But if you say, break down the the question into multiple like steps and show your work, then the model will take 10 times as long, but the answer will be way more reliably good. So it's like, if you give it this room to move and calculate in a slightly more human-like way, and I can't stress enough that these are not thinking like people, it's just a matter of breaking down a prompt into subprompt. And so that's what OpenAI's O1 model is. But basically,
20:09 Speaker 01 Making the models think more about the answer before they give you the answer. And that could be like, it thinks for five seconds. It could be it thinks for a week and does like research, you know, to send this little agent out to subscribe to the Financial Times and skim a few articles from the archive. I don't know.
20:26 Kirsten Korosec? All right, so I'm going to cut off the lesson.
20:28 Speaker 01 All right, all right, all right.
20:29 Kirsten Korosec? No, I mean, it's just a lot to take in, but I liked the lesson. I know, I know, but we don't have an hour long podcast.
20:36 Speaker 01 No, I know, I know.
20:38 Kirsten Korosec? But, but I would say this, like, first of all, very interesting. And it just seems like this test time compute is starting to shape up, or at least being maybe marketed as such as like AI industry's next big bet. And I'm wondering, Margot, although I know you're not like as deep in the weeds necessarily on AI, if you're seeing anything percolating in what you're writing on the venture side.
21:01 Speaker 02 I am going to take this answer in a tangent direction because
21:06 Kirsten Korosec? Okay.
21:07 Speaker 02 Well, the only conversations I really have about AI are in relation to foreign policy and defense. When I read this piece, the first thing I thought about was that I talked to a lot of big Silicon Valley China hawks, and they have really made a good living off of selling this idea that we are going to be in like an AI arms race with China. And when I read this and I'm like, okay, if, if the improvements in AI become a lot more incremental and we hit some sort of ceiling, what happens to that Cold War? We can't outpace China just flooding money and building data centers. What does that look like? If there is sort of a ceiling where after that, it's pretty incremental improvements and our adversaries can relatively quickly catch up to that.
21:45 Speaker 01 You know what I mean? That's absolutely, it's not tangential at all. It's, it's super like central to the discussion. And yeah, that's, that's absolutely what everybody's worried about. They're like, was the success in AI just a matter of how much money and how much H100s you could buy? If that's the case, then it's just a capital race. But if there is some level of research and ingenuity that has to be done, that's a different problem. And there's no guarantee that you can beat, you know, China, for example, on any particular problem, but it supposedly they're like six months behind. That's at least what that report from the China Hawk, what's his name, the guy you?
22:19 Kirsten Korosec? Jacob Helberg.
22:19 Speaker 01 Yes, Helberg. They were saying is about six months, China is about six months behind. Well, whatever, maybe they are, maybe they aren't. But if you can stay ahead of the game and be like, oh, we're reaching the end of this scaling law and we're figuring it out before the adversary, that's a huge advantage.
22:35 Kirsten Korosec? Well, we could obviously have a whole entire show about this. I will say that it's right now, it seems pretty unclear how this test time compute is going to scale. And I did want to know before we move on to network states, that apparently if it fails, there is one A16 partner who says that AI chips that specialize in high-speed inference could go up dramatically. We're not going to even let Devin comment on this because we could be here all day. He's laughing just for the folks who are listening.
23:03 Speaker 01 It is potentially true. It could be because you need test time inference. That needs to be fast, needs to be low latency. You can't wait a month. You have to, you have to get it as fast as possible. So that is cerebrus and stuff are going to go up.
23:14 Kirsten Korosec? Okay. Well, now finally, to finish the show with network states, I think we're all like super interested in this, but Margot, please, first of all, what's a network state?
23:25 Speaker 01 God.
23:26 Speaker 02 Great question. A network state is basically like an internet community in search of a physical community. So it's these digital groups that are in search of buying land. The way it's worked out is that those end up being like special economic zones. The dream is to like have their own country and they've made a weird amount of progress, which you, you were the most famous ones. And I wrote about how last week he announced that he had like gone to Greenland to try to buy it. Really, he did what he's been doing for years now, which is he goes to countries that he thinks might be amenable to an influx of capital and entrepreneurs and try to convince them to let him build
24:04 Speaker 02 a city. And he's raised over 500 million in a weird way. It's basically a loan that he can take out at certain milestones. But theoretically, if he hits these milestones of, of building a real city, he's got half a billion dollars that he can draw upon.
24:23 Kirsten Korosec? Okay. So backing up for a minute, first of all, I want to say that like the reason why I was saying network states 2.0 is because not that long ago, I forget which wave of cryptocurrency hype we were in, but at that point, there was a lot of activity around. Creating these like, quote unquote, utopian states centered around crypto. And so this seems like the second wave, but maybe we're more in like the third or fourth wave. I'm not really sure. My question for you is you mentioned a moment ago that this isn't the first attempt of Dryden Brown to, you know, try to buy uninhabited land. So has he tried with Greenland before? Are you talking about like other areas?
25:07 Speaker 02 For the past few years, he's been to like 30 countries. He told me, he was like, I just spam all the local government people on like Din. And then I show up and I, I say, I will buy, I will build a city here if you let me. I think right now they're looking at originally, he was looking at the Mediterranean. I think now they're leaning towards the Caribbean. Unclear where, but they had a big event in Dominican Republic. So that might be a hint.
25:29 Speaker 01 These things are so weird to me. I mean, I understand why countries want to take, they're like, oh, some rich guy with, you know, has a hundred million bucks that he wants to drop on a, some sort of utopia. Sure, come to the Dominican Republic. We'll buy you dinner and you can try to spend your hundred million dollars here. I mean, it's just like getting a resort investment because these projects are basically resorts for rich people that, you know, have a Bitcoin vending machine in the corner. That's what I've seen so far. Like, there's nothing utopian about it. I have not personally built a utopia, but I, as far as I know, you need to start at the bottom and work your way up. You don't start with a bunch of rich people and put them in one place and then say, okay, it's utopia.
26:09 Speaker 01 You need to start by essentially like overpaying all the ground level, like workers, farmers, cooks, dishwashers, like making sure that everybody is being overcompensated because you at the top have too much money. And all these guys do is like, go buy a like little plot of luxury land and say like, here we shall have our network state where all the rich people shall come once a year. And it's like, that's just a resort. It's just a hotel.
26:40 Kirsten Korosec? I like Devon's communist manifesto. Okay, I was gonna say, it does not sound at all very capitalist. You are fired from equity, by the way.
26:47 Speaker 02 They, they really, what I found really interesting about my conversation with Dryden, I've talked to him, I think, once before, but this time around, it was like right after Trump got elected and he felt very emboldened and he felt like having a Trump presidency was really going to help these network states, which is interesting because they're all abroad. And yeah, these guys, they view themselves as like cowboys going abroad and. Find a new land.
27:12 Kirsten Korosec? Before we wrap up the show, are there any successful ones? And if there are, what do they look like? I mean, what, what does that utopia look like? Is it just capitalism on steroids? Like, what is this utopia?
27:25 Speaker 02 There is one in Honduras called Prospera, which is, they've got one kind of like resort building completed.
27:32 Speaker 01 I just read about this. It's like a total, it's a total scam.
27:35 Speaker 02 It's not a scam in that they let you register your business there and you can pick from like 36 countries that you want to be regulated by. What do you mean? For rich people, it rocks.
27:46 Speaker 01 I should say, it's not, not a scam as much, you know, I mean, anything, everything is a scam in capitalism, of course, but like, this is like the ultimate. But seriously, I mean, like you have like Canary Islands and places like that that are tax havens. That's, you know, just to have a tax haven, call it a tax haven. That's what it is.
28:03 Kirsten Korosec? So first of all, Margot, I think that, you know, we should note for the audience that Devon has been planted here as the token contrarian. And, and actually, he's very pro-capitalist. You should see his house. It's insane. He's just like playing the part and it's actually not true to his ideals at all.
28:19 Speaker 01 That's a total act.
28:20 Kirsten Korosec? It's a total act. But please, finish your thought. And then I want to mention one thing about crypto.
28:26 Speaker 02 I was gonna say, the craziest part of Prospera is that they do like medical testing there. Like they had one guy that like got his Tesla key like embedded in his hand. Yeah, that's like not FDA approved. So rich businesses go there and you can, you can get whatever implanted. So maybe that's a big plus for Devon.
28:45 Kirsten Korosec? Yeah, that's, that's funny. The whole Tesla beat, I have encountered people who have embedded the key in their arms. But I'm really curious to see with the rise of crypto, especially those who have been emboldened by the new Trump administration, if we're going to see even more of these network states, you know, it just seems like crypto is kind of having a moment. And we'll see how this evolves. In the near term, Rebecca Bellan had a conversation recently with the founder of blockchain analysis startup Chainalysis about crypto post-election. So please check that out. I think there's some interesting insights in there. We are out of here. Thank you so much for listening to Margot educate
29:27 Kirsten Korosec? us on network states and Devon's extreme bias on capitalist society. Thank you too.
29:33 Speaker 02 Thanks for having us.
29:35 Speaker 01 Oh, thanks comrades.
29:37 Kirsten Korosec? And on that note, Equity will be back next week. Until then, you can find Equity under the handle @equitypod on X and Threads.
29:49 Speaker 03 Equity is produced by Teresa Loconsolo with editing by Kell. Bryce Durbin is our illustrator, and we'd like to give a big thanks to our audience development team and Henry Pickavet, who manages TechCrunch Audio Products. Thanks so much for listening, and we'll talk to you next time.
Previous Building trust in crypto with Jonathan Levin of Chainalysis Next Aerospace Corp's CEO talks literal moonshots and Space Agenda 2025 Page updated October 26, 2026 · transcript and summary are machine-generated
