China's 2035 Tech Master Plan

You're watching TVPN. Today is Monday, April 6th, 2026. We are live from the TVPN Ultra Velm, the Temple of Technology, the Fortress of Finance, the Capital. Capital to everyone. Celebrated yesterday. How was your weekend, Jordy? Good. I saw Project Hail Mary. Oh, you did? Yeah, I did. I finally saw it. Did you see it I did not see it. That was it? Ben, you saw it, right? Yeah. What was your review? It was great. I was telling you. I think it was uh it was fresh. It was fun. People put it on the water Will the Pew didn't like it. He didn't. He said it was interstellar for Chuds, right? Isn't that what he said? He said something. You liked it, right? I thought it was great. It was very enjoyable watch. I don't know. Should I see it? Yes, I think you should see it. In theater. In theater, for sure. Okay. It's definitely a good thing. It's not too long, it's like two and a half hours. But it's uh it's an Andy Weir. It's uh have you seen the Martian? Brutal. It's basically just like endless stream of problems and then like quick solutions. So there's some sort of problem they need to figure it out, problem they need to figure it out. All of it's I don't know, like loosely like hard sci-fi, like somewhat believable. It does have aliens in it and stuff, but it's a fun time. Uh anyway. Well speaking of real space, what's going on with Artemis 2? Artemis II is uh is live streaming right now to almost a million people on YouTube on the NASA YouTube channel. I believe it's also on Netflix. Um and the stream title is just We Are About to Fly Around the Moon with Authority from NASA. Uh we can actually pull this up and see . I would love to hear what is going on right now. Because I think it's happening like as we speak. Like they are they are up. I also uh wrote a little bit of a uh r retro active on the company. I have a spider on my Yeah, okay, so one minute ago NASA posted that the crew uh are now the farthest any human has ever traveled. Two hundred and fifty thousand miles from the earth. Let's see if we can get this. John is dealing with a little spider here on his microphone. There we go. There we go. Still living too. What do we got going on here? Yeah, but can we get the audio on that? Oh there's no audio? Oh weird. I think there's there is audio on YouTube, but uh I it might not be letting us like stream it through. I don't know sometimes there's like DRM stuff. Anyway, um Matthew Gallagher, he's the founder of Med V. And uh we were supposed to have him on the show today. He unfortunately cannot make it. We do have uh uh John Slotkin coming on, the chief medical officer from Gelsinger to or Geissinger to uh talk about claims around oral GLP1 drugs. Uh we read the New York Times story on Thursday , and then over the weekend there was a whole much a whole bunch more analysis about the company. And uh there were blending. one billion, one the one person one billion dollar totally company. Totally. And then the New York Times would take it down. Yep. But in this case it happened in reverse. So uh last week the New York Times broke down the story of Med V, a telehealth provider of GLP1 weight loss drugs. And the framing was basically that this was the first time in history a single person had built quote a billion dollar company. And um that's what went viral. Dgiging in there, were some small caveats and some big caveats. The small caveats were that uh you know, the founder, Matthew Gallagher, had hired his younger brother, so technically it was a two-person startup. Most people were like, yeah, okay, we'll count that, it doesn't matter. It's a it's a family member. Right? And then also uh the other one. It's only twice as many people as the original challenge. But but they had almost twice the revenue, 1.8 billion. So but the question was the the valuation. So the headline stat in the article was that they are on track to do 1.8 billion in sales this year. And that the on track is doing a lot of work, of course, because that's clearly an ARR number that's that's extrapolated out. And the GLP1 market's moving very quickly. It's not like Roe and HIMS and all the other providers are going to be, you know, just sitting down and letting this company run away with whatever secret sauce they've discovered around customer acquisition, although they might not go into this, which we'll go into. And it was insane scale. But uh companies don't unilaterally trade at 1x revenue. Like that's that that's sort of a given in this article. It's sort of presupposed that if you're doing one point eight billion in revenue, uh you would trade above a billion dollars. And so this counts as a billion dollar company. I always thought that the billion dollar company would be on market cap, not valuation. Not on revenue. Because uh you could do, you know, if you're if you're gonna say, you know, okay, any any any just try and say the biggest number, you would wanna set up some sort of like payment network where basically you could report GMV really, really high, and you actually have a very, very small business because you're taking like one percent as your true net revenue. Um and so so there was always a question about the margins, and then also revenue durability matters. Like the the the ongoing question around building these one-person high growth companies is is their durability because if you can do it and then I can do it and then Tyler can do it and then and then any one with uh you know access to a coding model can build it. Very quickly we're going to erode each other's markets and trade market cap very quickly. And I think any VC that would look at this. Uh they might get to a point where they're like, yeah, this is a unicorn company, look at the growth, I think there's something here. But at the same time you could see someone like a private equity guy valuing this and saying, uh well, I I need to see risk. I need to see durability. There's some lawsuits. Exactly. And we'll get into that. So uh MedV uses uh two companies, CareValidate and OpenLooth Health Open Loop Health to handle the doctors, pharmacies, shipping, and compliance. And we'll get into the compliance thing, but uh that's a lot of outsourcing, and of course that outsourcing uh takes a toll on margin because you have to pay the doctors, you have to pay the pharmacies, you have to pay the shipping, you have to pay the compliance. And they're paying care validate and open loop health So it's still it's still a case of the thing. I started asking what does this company actually do itself? Just marketing and maybe they're too aggressive about it. We'll get into it. So uh these GLP1 drugs are so also aren't cheap. And when you sell them even when you sell them legally, uh even as a telehealth wrapper, a lot of the value is gonna accrue to the pharmaceutical companies that own the intellectual property. At least that's how it should flow when things are functioning properly. So you can effectively build a telehealth company on top of a pharmaceutical company like Novo , uh that uh you where you are taking a thin margin on top, you can get to big numbers, but the the pharmaceutical companies are gonna wanna be paid because they did all of the expensive FDA trials, all the expensive R D, and they need to recoup that. So then the other question is you add in you add in the CAC from digital ads, uh they were apparently spending a lot on Facebook to acquire customers, and you quickly get uh you quickly wind up an estimate of pretty thin margins. I think the estimate uh Shield shared, SHIEL MONAT shared was like 15% or something like that. Um but uh uh it's it it's totally possible to get to a valuation that's lower than a billion dollars. Um depending on how everything flows through the episode at the same time though, if they're outsourcing all of the kind of key areas of the business to these other players, open loop, et cetera, it's possible that of that th those very thin margins, right, the fifteen-ish percent, the margins on that are very, very extreme, right? Sure. Yeah. So you could still get to a number that was in the hundreds of millions of dollars of of uh of EBITDA. Yeah, uh yeah. I mean the fifteen percent margin looks like a hundred and fifty million, maybe two hundred million in like basically profit. That is crazy. Um but we'll see what the FDA has to say long term about that, whether there will be lawsuits or settlements. There was already a warning letter, and this turned into a big drama about the company. And was that was there any mention of this in the original story? I don't think so. Uh and so there were a lot of people that were s it's and and I I believe I pulled up the the New York Times piece and there was no like correction or anything yet. Maybe there will be more follow on. I feel like the uh the founder does need to, you know, respond and have their opportunity to kind of correct the record because these things can kind of run away in either direction. Um but uh even with all the drama, uh there were a ton of you know super legitimate questions about how valuable the company really is and how long they'll be able to continue their current business model without pretty serious changes. And that's around the marketing stuff. So BedVee received an FDA warning letter uh just two months ago for misbranding violations. So warning letter, if you're not familiar with the FDA's language, uh it can be pretty wide-ranging. Like sometimes it requires just a small tank a small change to marketing materials And sometimes it's basically a like a shutdown the company moment. This happened a lot with the nicotine category with uh illegal uh vape vapor products. And sometimes like the warning letter would be like you are warned in that if you continue to sell these like we will put you out of business. And some of them are okay, we're warning you about this particular claim on your website, you need to change this language and it's pretty pretty m minimal. Uh I went through this uh ten years ago we started Lucy 2016 and uh while I was there we were extremely nervous about warning letters because if we got a warning letter, we thought it would mess up the FDA applications that we had in progress, it would mess up distributors, like anyone who you're partnering with would say, Well, I don't want to work with somebody that has a warning letter, have you dealt with this? Um we were able to avoid all this, but uh one one kind of concrete example is that um we like Lucy cannot make quick claims about the nicotine gum product. So uh even though most people think oh this is like nicerette, it's actually regulated in a different category. It's regulated as a tobacco product, not as a pharmaceutical product. And so you can't run a digital ad that says like, hey, you're a smoker, quit smoking with Lucy. That's a violation because it's not regulated as a pharmaceutical , as a smoking cessation aid is the term. And so in the nicotine category, you would have companies create a product. Yep. Usually there would be no founder associated with said product because people didn't want to put their face on it. Yeah. And they would do a lot of things that would they would uh make sales grow incredibly quickly, like you know, making marketing claims, just where and how they decided to sell, etc. etc. That the ad channels that they use, yeah. And so they would they would go from zero to hundreds of millions or billions of revenue very, very quickly. But the enterprise value of the company would be near zero. Right? Because no nobody would want to buy a company that had all that kind of baggage. Uh and so if you were gonna value it, it'd be like what is the like very, very, very near term, you know, uh sort of like revenue opportunity. Yeah. Yeah. I mean the famous the famous comparison is Joule versus Puffbar . Uh so Joule uh went, you know, raised a bunch of money, got very big, uh put into FDA applications, got those applications denied, uh marketing denial orders, then got them stayed in the courts, they set up a I think their whole headquarters was in DC, they were working very closely with the FDA, truly engaging to sort of clean up the, you know, all of the problems around marketing and sales and formulation, just really try and get to a clean bill of health with regard to the government relations. Uh they were successful and they got approved as uh not as a smoking cessation aid, so they still can't make the claim that Joule will help you quit smoking. Um they they but they did get approved as a tobacco product, uh you know, which I think the FDA says it has to be suitable for the protection of public health, which is a very vague way of saying like it's a net benefit, net good, that that that that it's on the market. Which is good, you know, people can debate that back and forth. But um but back in the Joule days, they also could not say quit smoking with Joule . They wanted to put say they wanted like a lot of people that were you know were smoking cigarettes did switch to Joule and they were pushing for that. But they couldn't actually make quick claims. So they used the word switch. They said switch to Joule instead of quit with Joule. And that seemed fine for a long time. I think they they they eventually trademarked it and then I think they had to pull away from it at a at a various points of time. But there's clearly like this gray area in your marketing. Um can you say you want smokers to upgrade to nicotine gum? You know, all of these things need to be sort of litigated with the FDI. Um what has MedVee been up to? Trevor Burrus We wound up playing it very safe. Now MedV appears to have taken a much more aggressive approach. Um they are apparently running uh 800 fake doctor accounts on Facebook to sell compounded GLP1. Shield Monot verified that the accounts are not actually doctors. Some even have cartoon ishly fake names like Tucker Tucker Dr. Tucker Carl Zin MD . If you like I feel this is one of those things where I feel bad for someone who clicked on an ad that was deceptive. From Dr. Tucker Carlson MD. But if you're getting your GLP1s from Dr. Carl Tuckerzen , like d you probably are in on the joke to some extent. I don't know. Uh is clearly not above board and they have to clean this up and deal with this. Uh they were also sued in a class action lawsuit last month for violating California's anti-spam law. That stuff can be crazy expensive because a lot of the fines are on like a per instance basis. So it'll be like, okay, yeah, $10 fine for every text message you sent. And it could be like you sent like, you know, 50 million text messages or something. So you could have some massive, massive liability. Of course, uh, that will be litigated in the court of law, and there might be a settlement, and they can figure out what the right damage is. Uh, if they are even guilty, they're still early in the process. Uh and so the end result is more of a story of pushing over aggressive marketing tactics. Yeah, I mean so so pull up pull up the actual picture uh of uh of Shield's post because I think that you can just see how confusing this would be to somebody maybe uh a little bit a little bit older. Yeah, zoom in. You can just see you see somebody that looks like a doctor. Oh yeah, they get the stuff. You're seventy years old on Facebook, you're not necessarily putting it together. It's duck Dr. Tucker Carlson MD. So my co-founder, Lucy, is uh uh he he has a PhD from Caltech, but it's in bio physics, and so he's not a medical doctor, but he is a doctor of science from Caltech. Like he has a really good uh the you know background, but he's not a medical doctor, and I would always joke with him, like, we gotta put you in a stethoscope, we gotta take some photos of you in a stethoscope, like the aura is so high. And he was like, No, that's like such bad practice. Like, we can't do that. Like, you doctor is associated with medical doctor, even if you have a doctorate in something. It's like the famous like, I'm dying. Does anyone have a doctorate? It's like yeah I have a doctor in uh a jurisdictor or whatever. The other interesting thing, medvi uses the dot or g what? Domain. Isn't that for nonprofits typically? That's very I would think so. Yeah, because you land on that site and then you think it looks like oh a doctor advertised this to me and now I'm on a nonprofit medical website. Yes. So uh yeah, in general I think the story is uh pushing over aggressive marketing tactics to the limit, more than AI allowing low headcount scaling. Uh but and and this has happened for a long time. Back in the old days, I want to say like 2014 Facebook era, uh, but the early days of online marketing, there were countless stories of questionable supplement sales or telehealth operations scaling on the back of insane ads. The classic formula for for like mega scale was if you could get an ad approved , uh, if you could get it like sort of like whitelisted or or through the approval process, it wouldn't automatically get reviewed that often. So once you were approved, you could say you could spend a thousand dollars or you could spend fifty million dollars and there was there wasn't early on there wasn't a natural trigger to like re you know uh like ref act check it was like if you could get approved for a small ad you could scale it up and so people would do all these crazy things to try and get the ads approved. There were all these uh tricky things where they would route to one website and then after the ad was approved, they changed the website, and then the you know the all the ad platforms had to eventually figure out like, okay, we need to be scroll scraping scraping a layer deeper like consistently or like every day. But the canonical example of the uh the health supplement or telehealth operation that would uh always go viral and always just pr int was if you could get an ad approved that had Harvard scientists, Brain Pill, and Johnny Depp all in one call to action. I'm not kidding about this. This is real. So it was like always this picture of Johnny De pp coming out of the ocean. That was the one that was like really scroll stopping. No, no, it was like a paparazzi photo. Which they didn't have the rights to use. They would not partner with Johnny Depp. They should not be running that ad at all. Um and then something about brain boosting or like making like limitless pill make you a genius, that type of marketing would all like it's very general. Like every who doesn't want to be smarter, so everyone would click on that. And then like Harvard scientists would like like lend its credibility because oh if it's from Harvard like it's good. And so if you could get those three terms approved, there were a whole bunch of like sketchy operations that would get ads approved and then just pump like a hundred million dollars of sales behind them. And it seems like maybe there's a little bit of that going on here where a lot of these a lot of these ads should not have been approved. Maybe there needs to be a validation. I know that a lot of companies that partner with uh doctors, even just to sell like skincare supplements, sunscreens, really anything, uh the the rate to uh to advertise with an influencer who's a who actually has some sort of medical credential is like way higher.. Yeah Because it's way more effective. And that yeah, and it's way more effective. But if you can just like fake that, then uh you're basically arbitraging that, but you shouldn't be. So we have we have two employees but eight hundred fake doctors on Facebook. Not great. Yeah, so we'll we'll see how this works out. I I think the press tours probably cancel like uh ended uh for for the near term. Okay. Well, I I think people have a lot of questions for uh the founder. Uh Matthew Gallagher, we'd love to have him on the show and uh dig into all this because maybe there are some good uh good context that is missing from all of the all of the criticism that's been uh levied over the weekend. But uh we'll see. So I was reflecting on like we're still in pursuit broadly of this one person, one billion dollar company. It's still an interesting question. Will it happen why might it happen why might it not yeah in my in my head it was always new startup is created yeah into like solo founder creates a hit product yeah raises a hundred on a billion and that is the mo that is like the c leanest path. Because for so many other paths, you just run into , yeah, maybe I want to be the one b one person, one billion dollar company, but I want to win more. Yeah. So if adding incremental people helps me build a better company. Yeah, those random pieces point right. Yeah. Why would I not add add at least a handful of people? Yeah. Yeah, totally. Um so yeah, I mean I think I think software is an interesting category, especially if it was open source software somehow, where there was some sort of flywheel where you could verify what was actually built and then you could verify what the sales were because it was done on some you know sort of open platform or some sort of you, know , uh some uh some sort of uh thing where uh the the actual reporting and the analytics. I mean apparently the the New York Times did verify the run rate that they that they used. So there the their money was flowing through the business. Um but having having a much cleaner representation of what the financial picture is, I think would would help. I was personally excited about the prospect of a video game. Um and and just bit more video game developers seeing breakout success. So um couple years ago, uh before in the in the olden days when I had free time, uh I got woefully addicted to uh poker themed ro guelike deck building game called Bellatro , which I don't think anyone here has played. But this game is amazing. So much fun. So it's it you're basically playing poker, but there's no financial stakes, there's no money, there's no microtransactions at all. Interestingly, the game actually got banned, I think in Jap an or some other country for being like poker and being banned for being anti-gambling. No, no, for No, no. Banned for for being like gambling aesthetics because you are playing with cards. There's no real money. You pay like $10, $15 for the game and then you can just play unlimited. But it's done extremely well. So uh it sold over five million units. I saw some reporting that it might be seven million and maybe even more now because they've gone multi-platform. Uh and I think at fifteen, twenty dollars a uh uh a download, that might be close to $10 milli0on in revenue. And it was made by a solar solo developer who goes by local thunk. Over a two and a half year period, he originally wanted the game to be a side project that he could put on his resume, but it wound up being a massive success. And so depending on your valuation methodology, you could probably underwrite Bellatro close to a billion dollars if it's generated $100 million of basically free cash flow. You know, there's obviously two and a half years of RD, but that was just this developer's time, the future revenue streams, the offshoots, the merch, like you you get to a number that's close, and uh and you have to imagine that that his two and a half year development uh like life cycle would have been pulled forward by the help of AI. So he basically developed the the game in in entirely in the pre-AI period. So it came out of the Does it still have momentum? I think it's probably trailed off a little bit because people sort of you get to the end of the game after I don't know ten, twenty, fifty hours or something. And then there are some people on YouTube that will do like tons and tons of like like try and break the game because you can get really, really crazy with like combinations of different things. But at a certain point, like you do kind of finish the game and then you're like, okay, I'm I'm moving on. And he never added microtransactions. He never added like any sort of live service model subscription, anything like that. It was sort of just like it did really well and then we'll see. Um and so uh yeah, this came out as a demo in 2023 and officially launched in January 2024. So fully like pre-AI development. And there's lots of there there is lots of hesitation in the video game community around AI, specifically around the the art generation, like pixel art generation, all of that. But if you think about AI as a tool for writing game logic and particularly like replatforming, it took him I want to say like a full year and I think he partnered with a developer or a publisher to actually distribute the game because it was originally a PC game, and then it went to Nintendo Switch and then PlayStation and then Xbox. And it's a perfect game to be everywhere, but that takes a lot of time and you sort of have to rewrite a bunch of the software. And that's something where uh AI should be able to pull that forward or make it even easier to launch everywhere on day one. And so the interesting thing is that Bellatro broke through because it paired like this familiar design language with this wildly engagement, engaging progression system. And so bringing fun ideas like that to market faster is something that like I'm really looking forward to. And we heard Shol to from Anthropic talking about like building his own RTS game. And obviously he has a full-time job, so this would be certainly uh something that would be like extra. We're we we've been working on a bunch of simulators that are sort of feeling like games. Some of the mechanics are getting pretty good now. Um game like but but you have to imagine that uh that in the future there'll be more and more game developers. Yeah, I mean some of the stuff that I've seen internally at TPN, it's like gonna blow already vague post. Are you the vague post gang? Uh have you stared into the abyss the abyss yet? Because everyone talks about at the AI labs they they all have abysses. And I feel like I feel like abyss 101 is like don't do not stare into it. Yeah. Like maybe glance. Yeah, I notice the abyss. I think a glance is okay. And or maybe like like saddling up to the abyss and just kind of peeking your head in is fine. But I would not stare into it. I I I think that's something you can do. Let's go over to the app store. Yeah. The app store saw an eighty five percent increase in new apps So uh it was just ticking up uh quarter over quarter and then jumped massively. Uh no surprises here. Uh feel like everybody everybody that I know outside of tech uh has uh has an app now. They're on unconstrained. They have an app, but at the same time I have yet to find an app that's on my home screen that was solo developer vibe coded, something like that. So I'm waiting for the flappy bird moment or the Bellatro moment, this like solo thing. Even if it's uh like a productivity tool, um, there's clearly more apps, but uh actually going viral, like what is the Harry Potter Balenciaga moment? What is the viral moment that uh breaks through and actually makes it to the top of the app charts? Because right now it just feels like the the long tail of the app store is getting potentially fatter because there's more single use apps, there's more small apps, but will we see one of those break out and become you know fantastically successful? That's sort of like the next the next uh you know critical moment that I'm I'm you know eagerly awaiting, but we'll see. Uh what's going on in China, John? Uh the pork industry is a victim of its own success. This story is incredible. They got swine scrapers over there. Skyscrapers filled with swine. I'm not kidding about this. Uh so this is from the The Economist and I read this yesterday and really enjoyed it. Pork holds a unique place in the Chinese diet. It was once a symbol of good the good life. The Chinese character for home is a pictogram of a pig under a roof. It is so important that the government has a strategic frozen pork reserve. And the news media are always full of the ups and downs of the pork industry. It's like that you they gotta cover it like it's a horse race. Like what's going on? It's a pork race. No, it's serious. Uh uh TBPN for the pork industry in China. That's the real, that's the real opportunity. There have been plenty of peaks and troughs in the past decade. The trough is active over there, literally. In 2018 to 2019, when African swine fever ravaged pig herds, many smaller backyard farms were wiped out. Prices went through the roof before long, the industry came trotting back. The big worry now is that it is doing too well. More efficient farming methods, producing ever more pork are colliding with slow consumption of the meat. Now all the news is about overcapacity. She who uh sells cuts of pork at a market in Beijing says she only got to eat the meat on special occasions with a child. These days she says it's so cheap, people can have it whenever they fancy. The oversupply pushed live pig prices to a 15-year low in March. Some farmers are losing over $40 per animal. Part of the problem is that some pig farmers have been aggressively expand have aggressively expanded production in an attempt to gain a bigger share of a shrinking market. On top of that, big companies save their bacon in the downturn by concentrating their pigs into mechanized modern facilities where they could be kept isolated from the swine fever. Modern farms are a marvel of a industrial agriculture, though not of animal welfare. One in Hubei province has twenty-six floors. It's a swine scraper. They it's twenty-six floors of pigs. That's the same thing. Do we know why why consumption is declining? Just oversupply. So they got hit with this uh they got hit with the swine flu and they were worried about all the pigs getting sick and dying. And so they instead of having them cross pollinate in like open fields, they pushed them into these literal skyscrapers that are twenty six floors tall. And but I would assume over over over consumption over produ ction you have over supply, prices come down, but they're saying that just overall consumption is dropping too. Trevor Burrus Well, yes, so people are uh moving to chicken and seafood. Increasingly middle class, the middle class Chinese see uh pork as less healthy than chicken and side seafood, but it's really like a supply a supply-side store uh story. Uh that may be more pork than even China can eat. The average Chinese person guzzled 28 kilograms of it in 2024, but that was two kilograms less than in 2023. Uh to boost prices, officials have slashed subsidies, ordered farms to cull droves and told the pork reserve to buy more meat to little effect. Some big firms are looking elsewhere. said it would bring the first high rise, uh it would build the first high rise in Vietnam uh capable of rearing 1.6 million swine a year. Pig farms are going vertical as profits flatline. And so they are moving the pigs inside, which is what which is you know winding up with law with like increased yields because uh the pigs aren't getting sick. Anyway. Um America's new best new weapon in Iran is a drone inspired by Iran. Uh you had asked about this on one interview we did with some . Yeah, I forget who it was with, but I was I was intri you know, I was surprised to see the US copying the show No, it it feels like uh entirely the smart move. It's battle tested, they're cheap. uh we we have a good understanding of their capability because American allies have faced off against them. So there's a lot of reason that it that it makes sense but uh I I feel like America had to swallow its pride to some degree. Yeah. Uh to copy you know something that uh you know the the enemy uh made. So they're calling it the Toyota Corolla of drones. Uh the powerful low cost attack drone the US is using in its war with Iran doesn't come from one of America's more than 400 venture-backed drone startups, and it isn't the product of Silicon Valley Ingenuity. Instead, the drone having its moment in the Middle East conflict was designed by the U.S. military itself using reverse engine ered Iranian technology. From the earliest days of the war, the FLM 136, or LUCIS, as it is known, has been wiping out Iranian military targets while better funded hardware systems and drones from defense startups have had little involvement. It is a victory for the U.S. military, which went from blueprint to battle-ready drone in less than two years, jettisoning its tradition of slowly buying very expensive equipment. The creation of Lucas is an early proof point. That's a narrative violation. New strategy of making cheap drones quickly, and a sign that the Pentagon can change the way it does business to better prepare for modern conflict. So I I want to know so much more about like so they're not subcontracting this at all? Is this all like is the entire supply chain uh the the U.S. military itself? Are they milling the parts and factories that they built? During the Biden administration, a small group in the Defense Department seized on the idea of America building its own version of the Iranian Shaheed, a fearsome attack drone that militaries and proxy militias across the globe have sought to duplicate. Rush Russia, which is lobbying about 4,000 modified Shaheads at Ukraine every month, according to Ukrainian government data, did more than any other country to demonstrate the drone's capabilities. A small team in the U.S. military's research and engineering office put together plans to build an attack drone. Bedas on deconstructing a Shahed, the military had recovered from Ukraine. It was the first known occasion in around half a century that the U.S. had reverse engineered another country's military technology for its own use. Former defense officials said the last time it was a Soviet-made pontoon bridge. That's interesting. A former senior defense official described Lucas, which stands for Low Cost Unmanned Combat Attack System as the Toyota Corolla of Drones. It may not have all the features or top-end components, but it was built to be affordable and plentiful. The cost of Lucas ranges from 10,000 to 55,000. Uh that's a big range, according to a Pentagon spokesman, in line with the Iranian model, Tomahawk long-range cruise missiles, several hundred of which have been used in the war with Iran, cut at least two million, cost at least two million apiece. The Defense Department is committed to scaling cost effective autonomous solutions for the joint force, and Lucas continues to be a prime example, the spokesman said. Former senior Pentagon official Michael Horowitz, who was one of the leaders of the team that developed Lucas, said that other militaries had the ability to make their own low-cost, long-range precision strike weapons. The issue was the U.S. was spending nothing, zero dollars on that kind of system. Uh who now who he's now on the Council of Foreign Relations. Lucas was given a spot in uh a in 2024 in a Biden era initiative to field thousands of autonomous weapons by last August. Its inclusion was controversial, said mo former military officials. Lucas just wasn't just a mock-up, but it beat out more mature systems on the offering. Interesting. Uh because the government owns Lucas's intellectual property, it is using the same approach it employed in building ships during World War II, enlisting a cross-section of second and third tier manufacturers, okay, who crank out the drones on demand during wartime. Little known Scottsdale, Arizona-based Spectre Works and Huntsville, Alabama, Alabama-based integration innovation were tapped to build the drones. So they are partnering with private companies to actually do uh build and assembly of these drones. Uh a total of five manufacturers will be selected, each set up to produce three hundred drones a month, a former senior defense official familiar with the plan said. Spectre works in integration innovation The Marine Corps was first to use the drones and then ordered around 6,000 dusted for the Indo-Pacific. Uh but then the war with Iran began. The drones were handed over to U.S. CENTCOM in February and made their first appearance in combat The Trump administration has enacted sweeping reforms in defense procurement, making it easier for the military to quickly buy weapons and emphasizing commercial technology to modernize the U.S. arsenal. In particular, in August, decision by Defense Senate Still, other changes will take longer to trickle through Pentagon's bureaucracy, and it will take time to reorient America's to a new way of fighting wars, even as China is developing advanced ways to strike the U.S. While Lucas has been a success against Iran's degraded air defenses, that isn't a guarantee. It will be a battlefield star in more complex environments, uh, an electronic warfare expert said. Um in the Middle East, there is no meaningful jamming of GPS, which can cause drones to crash or fly off course. Every technology gets defeated at some point, he said. There's still a worrisome there's still there's also still a worrisome lack of cheap U.S. counter drone technology, which has allowed Iranianed-back militias to continue using small drones to menace U.S. military bases in the Middle East. The small number of unmanned surface vessels in the region are still years away from being the autonomous fighting machines they're manuf acturers have promised. The absence of a broader supply of modern US cheap US systems in the Iran War has served as a wake-up call. We're not ready, said Julie Bush, co-founder of the defense tech firm Valinor Enterprises and a former Palantir Technologies executive uh who has been on the show. Uh the government doesn't have what it needs to scale that they uh what it needs at the scale that they need. Yeah, uh did you see that uh uh AGM post Yeah. I I mean I think the the estimate is around ten ten ships a day are going through down from a hundred pre-war . Yeah. Uh but there could be more that just aren't that that again have their transponder turns off. Yeah, he was saying something about like people will turn off the transponders even when they find like good fishing spots because they don't want other people to find their good fishing spots. And so it's done more than uh more than people think, I guess. Uh but uh Sutrini is over there. Citrini research analyst made it over to the Strait of Hormuz in an absolutely insane move and some real crazy investigative journalism going on. I'm hoping to have Citrini on. I obviously want to be uh s uh secure about whatever is shared because it does seem like uh a risky proposition to be over there right now. Um but uh love seeing them. Uh love seeing the reporting so far. Uh another tech news. Uh Anthropic is making its own splash with an acquisition. This was on Thursday, April 2nd. Uh $400 million for Coefficient Bio started last fall, developing an AI drug R <unk>D uh platform. The startups team will be joining Anthropics Healthcare Life Sciences Group, which is aimed at developing tools for biotech workflows ranging from drug discovery to clinical tests. Very exciting. Dimension Capital owned over 50%. Who is Dimension Capital? Have we had them on the show? I don't think so. We probably should. That feels like a uh multi-stage research-oriented investment firm partnering with companies across the frontiers of science and compute. Well, huge win for them. Congrats to the . Not bad at all for six months. There's also an article in the Wall Street Journal about OpenAI anthropic 's finances reveal challenge from Berber Gin, Scoop God. Uh OpenAI and Anthropic are racing toward potentially record-breaking IPOs by the end of the year. An inside look at the financials of both companies prior to funding rounds completed earlier this year, show their Achilles heel, the soaring costs needed to train new AI models. $1 billion on computing power for AI research in 2028. That means the company anticipates burning $85 billion that year, even it even after almost doubling sales from the years prior. Such losses would would dwarf that of virtually any other public company in history. Anthropic doesn't expect to spend nearly as much, but its rosyest forecasts tell a similar story of mounting computing costs. Both companies are releasing new versions of their AI models at faster at a faster cadence than ever before while pouring more resources into the training runs that create them. The arms race is showing no signs of slowing. Yeah, so pull pull up this chart from Andrew Current, this is from the Wall Street Journal piece. Uh projected open AI and anthropic model training spend for the remainder of this decade in billions. The Wall Street Journal says they got the data from financial documents shared with investors. So yeah, I think I mean obviously uh absolutely wild numbers here. I I think uh this is for training specifically, which is a cost that uh that that can be This is a question of like uh yeah, like like each model by itself is profitable, but you keep having to spend more on the bigger model. Tyler, what do you think about that? This is like um emblematic of like the the Dorcash critique of Dario, right, where that he's like if you basically compare open anthropic, like Dario's being way more conservative about spending, like he's like, Oh well we don't wanna you know, go broke or whatever. But yeah, I mean w when you see the the chart like this, it's like pretty crazy. Uh it does seem like there's uh there's a debate internal uh at OpenAI according to the information around uh the uh IPO timing. This is from the the information from Anisa G ardizzi and Amir Afrati. OpenAI's CEO and CFO diverge on IPO timing. Sam Aldman has committed OpenAI to spend six hundred billion dollars in the next five years and privately said he wants to go public as soon as the fourth quarter, despite expectations his company will burn more than two hundred billion before it starts generating cash. Behind the scenes, according to the information, Sarah Fryer, his chief financial officer, has voiced concerns that the reflect tensions and risk inherent in the CEO's extraordinarily ambitious plans. Opening asked committing billions of dollars a year in advance to help finance data centers. Fryer last year began reporting to the head of applications instead of Altman. Altman has excluded her from some conversations related to financial plans. And so you can go to the information to read more about that story. But it is it is extremely high stakes. There's i in general the uh the AI companies have like i i are they railroads, are they SaaS companies? Is there's this incredible CapEx, are they both uh the electricity? How you yeah, no one really has consensus here yet and it feels like there's a there's an an ongoing divergence in how do you how do you model these financially how much risk should you uh should you take there's also this question of like the end of the curve, the end of this of this of the exponential. We're seeing this exponential growth in model capability relative to investment dollars. The scaling laws have held for a long time, but at the same time , you can't just throw endless money because at a certain point you don't want to spend a trillion dollars to get like one more IQ point out of the model. Like there is a limit to this. And so nailing it out of the way. If you're thinking about like a hiring an employee, it's like okay, so this this one individual is you know, their their their base salary is gonna be two hundred K a year, or you could you could you could hire this other person uh and they're they're one IQ point smarter, but you have to you have it it's gonna cost you five million dollars. Five million dollars a year. The the choice is is obvious. So we'll see we'll see when we get to that we'll see when we get to that point. Uh overall I think like tensions between uh this is not the first time there's been tension between a CFO and like an an ambitious uh CEO. And so uh have you ever hired a CFO at any of your companies? No. Wow. Go dude. No. Uh we hired a CFO at our f at my first company and it was a it was an awesome experience actually. But the tension was insane. But but I was not I was not the CEO. And so the CFO came in and she had taken multiple companies public and was like, this is just a widgets business. This is so simple. Like, we should just be like running this sufficiently. And of course, there was like a big debate about that. But uh she's uh she was very, very formative in my uh early business care er. Uh yeah, it's interesting because that was that was you know consumer package goods company, but with a lot of Silicon Valley venture capital jammed into it. Yep. So the pressure from the board experience was one thing. Yep. And then you have the CFO who's saying, hey, we're selling the thing. Yeah, and so much organic growth, but the the the the nature of uh like soil and and consumer packaged goods, like it it's not a fundamentally different business at the end of the day. Like it does just become a marketing manufacturing margin game. And you eventually trade on Ebita . And so the what what do they call it? The reality distortion field doesn't quite apply, but when you're dealing with an entirely new technology, uh it's a whole different discussion. Anyway, there's another article in the Wall Street Journal about the new jobs that are being created by AI. Uh let's go through this. So AI is raising big fears about uh employment losses, but it's also giving rise to new engineering and training jobs. And uh we're going to have uh Liz Hoffman from Semaphor on the show later to talk about the March jobs data. Uh there was a massive jobs beat and uh a whole bunch and more jobs added to the economy, but the chart is crazy. It's like up and down every quarter there's uh Yeah the tough thing that you know the every single jobs report has been revised downward so massively that it's hard to look at a new jobs report and and actually get ex print like that and say, Yeah, this time it's different. This time it's different. And then now you have a war and major geopolitical tensions and this weird dynamic where uh the oil prices are driving gas prices, which is driving inflation, which is maybe gonna cause less driving-based vacations over the summer, less tourism, like you could see inflation plus some economic weakness regardless of what the technology industry is doing. And so you could wind up with a situation where rates have to stay high and so that creates unemployment and it's a it's a rough time to be a Fed governor, I'm sure. But let's dig into what the Wall Street Journal is talking about, the where different jobs are playing a part here. So artificial intelligence has sparked fears it will become Uh there's also uh a research paper, I don't know, a PDF from OpenAI today around uh the the social contract and how the government should be thinking about uh redistribution and the safety net in the age of AGI. Uh it's a long document and honestly I haven't had a chance to read it yet, but uh we will go through some of it later this yeah. I saw one repon response. Someone just said just put the just put the tokens in the bag., bro But uh we'll we'll we'll read it and we'll we'll we'll cover it. Um and uh yeah it'll be interesting to see how it maps onto like previous political platforms, like who picks it up? Because like there's obviously a collection of political ideas and uh and and various politicians might say, hey, I've been saying that for a long time, I like that, or hey like I don't I I completely disagree with that. I I have a different philosophy, but we'll we'll we'll work through it. So it's also AI is also fueling a crop of new careers. AI created 6 40,000 jobs between uh 2023 and 2025 in the US, according to an analysis by LinkedIn of job posting data, including new white-collar positions such as head of AI and AI engineer, that tally doesn't include the huge number of temporary construction jobs tied to building the mammoth data centers AI relies on. And I'm wondering how temporary will they be if the ramp, I mean if the if the CapEx ramp continues and you know you're going to build 10 times as many data centers, you would imagine that Many construction jobs are temporary, right? You build you build a house, job is over, you move and build the next one. Yeah. Uh so LinkedIn's head of economics, Corey uh Kanten ga says uh we're not talking enough about uh we're not talking about enough jobs to change the direction of the labor market, but for AI roles growth has been pretty much straight up. And Citadel has something to do with the thing for roles like AI engineer, to me that was just you were going to hire an engineer and you're slapping AI on the front of it because you want somebody that is excited about AI. Yeah and maybe more greenfield projects, maybe something that's uh you know the it it's it's like the new full stack. Like no longer are you hiring like a back end engineer, front end engineer necessarily. Like you you you you might be hiring someone who's more flexible. What do you laugh at that? I love how that the first example, this guy, uh Zach, who they're who they're profiling in this, the first the first picture of him is just him with a surfboard running into the ocean. Hey, he's surfing after work. He's he's got good work life balance. Let's read that. That looks like the middle of the day to me. Look at look at the you know, just look at look at where the light is, John. That's not after work. You uh you unless he's wrapping up. Not another word about surfing. I don't celled again for being anti-surfing. Post uh What were you you were anti surfing before working after work surfing in really cold water before work? Oh is bad. It has not worked well for me. Okay, for you. Apparently there's people out there. That uh death threats. Yeah. Did you get death threats? No. You just got it dunked on. It's fine. It happens. Uh Zach Kinsler, who's twenty five years old, got his master's degree in business analytics less than two years ago. Uh he in February he assumed the title of head of human AI solutions at Boodle Box, an AI education startup based in Colorado Springs, Color ado. His his job involves client training and using AI to speed up tasks for coworkers. That is a good gig. And says he feels lucky to land in a growing field. The conversation about whether artificial intelligence is good or bad needs to go away because it's not going to go away, he said. One of the biggest questions facing a shaky U.S. labor market has been that has been especially hard on white-collar workers is whether AI's ability to mimic human skills like research, writing, and coding will cause significant layoffs. Big cuts at Oracle this week, while the company invests heavily in AI data centers underscored some of these worries. Corporate leaders, including some in the tech world, have issued grim forecasts about AI's potential to replace white-collar workers at a massive scale. According to a recent Goldman Sachs research report, AI could automate tasks that account for a quarter of all working hours in the United States, especially in fields such as administrative support, legal work, and architecture and engineering. One one interesting pitch I was hearing, I think it was from the open AI doc was uh go like, you know, th there w I think it was during the industrial revolution the unions lobbied for the the weekend, the idea of the weekend, and that just became a thing and stuck around forever, uh basically. And uh and and so there's a question about like, you know, do we pitch a four day work week somehow nationally, some like, you know, bigger thing? I don't know. There's all these interesting like competitive dynamics and obviously people wind up working on the weekends because they're like, I want to outcompete the case. David Center is like I want all my enemies to move to a four-day work week immediately so they can properly relax and recover. Mandate that for everyone except me. Still, analysts say it's hard to judge how many job cuts thus far are genuinely attributable to AI and which are being made for financial or operational reasons. This is something we've been going back and forth on forever. A new survey of uh 750 chief financial officers found that AI had essentially no negative employment effect in 202 5 . Uh but data show employers are increasingly hiring for AI talent. In 2023, AI related roles made up only one point six percent of all job postings, according to an AI job tracker co-led by Anil Gupta, professor at the University of Maryland's Robert H. Smith School of Business, and job market company, data companyink, l up. Never heard of that before. Two years later, that figure had more than doubled to 3.4% of uh all AI of all job postings being AI related. Uh one rising job is head of AI. That's a good gig. Uh in the three years from twenty twenty-three to twenty twenty-five, companies sought to fill two hundred and twenty-five thousand such jobs. There are now two hundred and twenty-five thousand heads of AI. You need 20 years of of experience prior for most of these jobs, which is maybe tough. Tough to put that together. Uh no, I'm joking. China's uh what what what's happening in the Chinese AI talent race over in the economists. Yes. Uh Dave Frieber was talking about this on stage at Hill and Valley about how uh there's been this like big flipping flippining from where research is coming from. The the headline stat is something like uh 50% of where is it? 50% uh where okay. Um uh this was from Jensen Wong at NVIDIA. He said, is it possible that the United States falls behind China? Uh he asked himself a question uh during a question and answer session about artificial intelligence late last year. He says the a the answer is absolutely yes. Pools of capital as well as a regular supply of NVIDIA's cutting-edge chips. But Mr. Wong's uh concern uh related to an equal equally important ingredient of innovation, human talent. Until recently, most leading AI research was produced by experts based in the West. That is changing. In 2025, for the first time, more studies presented at the world's top AI conference had lead authors based in China. Tyler, what you you're you have more complicated? Yeah, I think this is just misleading, right? Because researchers in the US are are studying, but they're not putting that work into study. Scroll down on that article a little bit more, you can see a chart where the US uh like tw between twenty fifteen and twenty twenty it like looks like it's starting to go exponential and then basically plateaus, right? Because all the best researchers basically go to the labs, then they you don't publish anymore. Wait, uh so which chart are you referring to? The one in economist. Is it called active AI researchers in the thousands? Yes. Student early career. And so in the U in the United States it's growing and then in uh up until twenty twenty one and then it stagnates because all the labs went closed. And uh all the Chinese labs are still open source. So yeah, I I I think that's a uh I don't know. That that's the the the That's a very reasonable mitigating point here. There is just the question of like, do you still need a pipeline of open source researchers, of uh a robust research uh practice? Also like if if you're in the Ilya Sutzkever camp of like the age of research, maybe you do need a bunch of bizarre ideas that are happening and you get like more. Oh true. He's like even more secretive. Yeah, yeah, but I I I think you could like expand from that into uh into like the the n like there might be interesting um like knock on research that gets pulled from random places and like the cross pollination is valuable, and even if he's not a practitioner of that, you can still make the argument that uh research is valuable and cross pollination is also valuable and China does have that advantage. Yeah, I just think like if you look at China, like there's the anecdotal evidence of like what happened to the Quen researchers, right? Like you know, the the people running that company are just like nothing. Do we ever find out that they go to XAI or what's going on? I I don't think we don't know yet. But like clearly like something is like different about how how they think about researchers than in the US, right? Because those guys are like extremely good. Yeah. And somehow there was some reorg happening. The chat is demanding an ad break. We'll we'll figure it out. Uh a lot of people were saying that you know the ads were were part of the program. We will figure out a way to do something ad related. Uh until now you have to put up with uh some random soundboard. And that's the best I can do for you. But there is a different article in The Economist. People who are just watching for the ads, I guess. Um China's new master plan for its tech economy in uh 2020 and beyond. And so uh they have uh this is the 15th five-year plan for China's economic development. This was adopted in March and the economist is breaking it down. Uh it talks of industrial upgrading, new quality productive forces, and the like. Yet in plain language it translates into Elon Musk's fever dream. So China wants to build skies dotted with flying taxis, fusion power fueling factories manned by humanoid robots, unstoppable quantum computers, six G mobile devices plugged directly into people's brains. It's a dream come true for you, Geordie. Are you gonna you're gonna renounce your citizenship? Yeah You're like yeah no capital. Absolutely. Yeah sign me up. Sign me up. But there's I mean obviously they're taking they're taking like frontier sci-fi development very, very seriously, and it has a lot of knock-on effects. So uh past plans also displayed gumption. In 2015, the most high-profile plan in years, dubbed China Made in 20 uh made in China 2025, set the goal of catching up with America and ending reliance on foreign technology. But catching up while China uh which China has pulled off in areas like electric cars, clean energy, and even artificial intelligence is one thing. Dominating technology of the future is another. Can they do it? One reason for this breathtaking ambition is the desire of Xi Jinping, China's leader, to rusher to usher in a modernized socialist state by 2035. Remember, he thinks he's going to live to I think 150, he was caught on, maybe some hot mic saying. So he's thinking in centuries still. A modernized s uh Chinese socialist