DOJ vs. Fed Chair, Apple Repositions Vision Pro, Ben Thompson Joins | Tyler Cowen, Harley Finkelstein, Andrew Feldman, Nathan Nwachuku, Anastasios Angelopoulos

27:48This year's red carpet made a strong case for statement watches with bold dials. Did we kind of call this originally with the show? Did we? The joke early on was like, quiet luxury is over. Yes, loud opulence. Not loud opulence. A lot of these are screaming loud opulence. In fact, I need to return to my, let's start with the beginning of the slideshow. We need to return to our original statements about the value of loud opulence because I see these and I'm like, I couldn't pull these off with my life dependent on it. But The Rock was spotted watching a Chopard, much like Maduro. But this one is the Alpine Eagle Frozen Summit.

28:32Look at all those diamonds. I like it, but I'd like to see. I mean, could we get at least a couple more diamonds on here? I think it's a little understated. He could have gone with, you know, the rainbow one, the gem set. It's the Rolex with all the different colored diamonds. Who else is wearing fine luxury watches at the Golden Globes? Adam Scott was wearing a Vacheron Constantin Traditionel Perpetual Calendar Ultra. This is nice. That's a very nice watch. I like that one a lot. I also like this Omega Seamaster Aqua Terra that Glenn Powell was seen wearing. Mark Andreessen, notoriously an Omega guy.

29:09Omega guy. That's right. New fund. Maybe this was a nod from Glenn saying, like, I salute you. Hat tip. Yes, he celebrated the$15 billion. He probably read the Packy piece. Yeah. And he says, you know what? it's time to put on the Omega Seamaster Aqua Terra. In terms of Omega Seamasters, this one stands out to me. Gold is a choice, but I think it's working very well here. And you know who else is wearing an Omega Seamaster? George Clooney, also an Aqua Terra. I would love to know the details of Omega and Rolex and the other brands fighting over people like Clooney, right? Because I don't think Clooney's putting on a watch without getting paid.

29:49Okay, well, we have Ben Thompson in the Restream Waiting Room. Let's bring him in to the TV fan, Ultra Delphi. Fantastic. Ben, how are you doing? Here he is. Good. Thanks so much for hopping on the show on such short notice. I loved your piece. I have takes to drop. Please. Let's go. Let's go. Okay, so, I mean, drop your first takes, and then I want to know, is there anything to what I was saying, that Apple actually does read your work, and they do want to do it, but they just can't because of some contract? Oh, I thought you were talking about watches. Oh, no. Let's talk about watches. What's on the wrist, Ben?

30:24I have a Tudor GMT. Very nice. Very basic. But highly recommended. Tudor's a great place. And it's covered in diamonds, I assume, like the rocks? It is not. It is very basic with a rubber band because rubber bands rule. Ooh, okay. Rubber band. That's good. Yeah. So I disagree with your take. Please. So my overall view, NBA was on the Vision Pro on Friday. They had an NBA clip when they watched the Vision Pro. So I was there when they watched it. So I had to try it that day. They took that clip out for the demo that shipped. And that was in Apple stores. So you only saw that clip if you were there the first day.

31:05And so I don't know if it was a rights issues, which maybe lends itself to your argument or whatever it might be. But that was for sure. It was like three seconds. It was the clip that sold me on the Vision Pro more than anything because you felt like you were there. It was amazing. Yeah. So they have this, finally have a live sports thing, which is a big deal. They demonstrated that they can show stuff live. And it worked. It worked well. I don't think there was any technical issues. The cameras are quite small now compared to what they used to be. You could see them on the sideline table and underneath the baskets.

31:36So this technology works. You can watch things live in the Vision Pro. and it should be amazing and it sucked. And it sucked because you're getting jerked around all the time. You're not, you're there. You feel like you're there. It is immersive. The unimmersive part is the Apple part, which is some producer in a truck is moving you around. There's a pregame show. I don't need a pregame show. I'm sitting courtside. I can watch the players warm up. And the reason why I disagree with your take that Apple knows this and wants to do it, is that every Vision Pro video has the same problem. Okay. The Metallica video, super cool.

32:17You walk in, the opening scene of that is amazing. Well, there's the little bit where they're in like the green room or whatever. But you're walking in, you're walking in on James Hatfield, and you're in the crowd. And it's like he's walking through and people are reaching, and it's like it's so amazing. And then suddenly he's going up the stairs and boom, you're jerked to somewhere else. And you're seeing him come on the stage. You go back, they cut this MLS video, which they have the rights for. They're overpaying. You want to talk about the F1 deal? Look how much Apple's paying MLS. Like, this is one of the most absurd overpays in the history of sports.

32:47I think they can do whatever they want. They cut a – I love lap tracks, but it's very distracting to hear. I'm sorry. They cut, like, a season-in-review video that had, like, 56 cuts in it. It's like done by a YouTuber on crack. Do you think it's production teams that are just trying to create work for themselves? because they go in there and they're like, hey, boss, actually the people just want to put their goggles on and hang out. It's like, well, what are you doing here? I feel like there's just a lack of confidence here. Like what I – it feels very like – it feels when you say it, it's like half-assed that we're going to go out, we're just going to set a camera there, and you can sit there and watch the game.

33:28It's like, no, there's this pressure in part because it's not popular. It has to be a big production. We have to have a pregame show. We have to have dedicated announcers. The end result is you get six games, which are physically uncomfortable to watch because you're getting jerked around. And they could have just the Vision Pro there for every game, and I would pay a lot of money to watch that. They could have it at every concert. And what's amazing is that, number one, it would make it better. I'm fully convinced of this. And number two, it would solve their content problem because you could suddenly have tickets to every single concert in the world, to every single game in the world.

34:09It is a classic tech issue where you put in the upfront costs, you buy the cameras, you install them, and you have marginal upside forever, everywhere. And it's latching onto a major trend, which is live, which is this idea that when everything's commoditized online, everything's digital, live experiences are worth more and more and more. You go back to the Ares tour in Taylor Swift, It was a big topic of discussion at that point because this is something people are craving. And sports specifically are pretty immune to AI because nobody wants to watch the AI box play. No one wants to watch AI basketball, right?

34:46Yeah. It's the human aspect of it that's compelling. The thing that was most crazy to me is that it was geolocked to Southern California. So I get that. I'm not going to beg on them for that. But part of it, the reason why it was geolocked is because Spectrum spent all the money to do all this production. And so, of course, they get the benefit. It's in their catchment area. But you don't need to do all that. They're overthinking this. Like, literally, set a camera on the sideline and do nothing else. I will be happy as a clam. Okay, okay. It is expensive to do immersive production. At the same time, Blackmagic has the Ursa Cine Immersive that, have you seen this thing?

35:30Two fisheye lenses on the front, 16K. It costs$30 ,000. And it feels like if this is the case, there has to be an opportunity for someone who's a little more agile. You know, maybe it's Red Bull. Maybe it's some league that's not front and center to just go and do this. They have to pay 30K up front and then stream it. Yeah, 30K is nothing, okay? Like we're all in the tech industry. We can laugh at figures like that. Okay. So, so third, but that's an upfront investment. Like the problem is they tacked on all these marginal costs to this production. So when you logged on for the game, there was a dedicated pregame host and show for the Apple vision pro viewers.

36:11That's almost certainly not going to be as good as the main pregame. I know that where they're paying like hundreds of millions of dollars is sitting courtside and watching NBA players warm up and make 57 threes in a row because they're incredible yeah like i don't know if you guys have ever sat courtside but like it's it's um it's amazing it's amazing speaking of watches this is where i spend my watch money okay okay okay it's it's it's an unbelievable experience and and you and they have all like they have like uh for example they have like replays in there yeah which okay yes you can see a bit of replays you know where else they show replays on the scoreboard how can i scoreboard i can look up because i'm wearing a vision pro that's immersive and has this fisheye camera where I can see all around it.

36:55I don't need a scoreboard bug down at the bottom. I can look at the scoreboard. Like I actually, during the game while watching it, it was actually kind of hard to see the scoreboard bug because it was way down at the bottom. So I was looking at the scoreboard in the arena the whole time. Yeah. That's the whole, that's the whole thing. You're there. It's, it's, and so all that stuff, the, the pregame show, they had dedicated play-by-play announcers and analysts for the vision pro they had these multiple cameras they had a production truck who's choosing how can i upset ben this time and they'll be like the other cameras you didn't need any of that and and and and you thirty thousand dollars is a one-time cost yeah you spend that money once and you put these cameras in every arena for every game everywhere and then you you suddenly have this exclusive selling tickets to live events in a way no one else can i think it's a huge like it seems so clear to me this is the vision pro use case this is your portal to every live event in the world what do you what do you think about meta meta's efforts on uh nba uh partnership as well specifically with oculus division division pro is better i mean like the the like the oculus has advantages relative to the Vision Pro.

38:15Actually, I love Mark Zuckerberg's post when the Vision Pro came out. It's like, oh, we're not surprised. We could have done that, but we didn't, which was valid. I'm also not particularly interested in the use cases that are good for that. I don't care about gaming. I don't care about a lot of the other stuff. This is, for me, the killer use case, for sure. Being able to... It seems like the next Quest, if they stay in VR, will be better. We talked to James Cameron, and it seemed like he was pretty excited that they might have gotten the same screen. from the Apple Vision Pro. Like, yeah, they're two years behind, but you bring that screen into the next quest for - Which is fine, but there's trade-offs for that as far as like number of pixels and the field of view and all those sorts of things.

38:54But like the whole point is that$3 ,500 feels overwhelming, like for what you get. $3 ,500 to be able to attend any sporting event in person is one of the greatest deals of all time. Like that - How do you think they should - I used to be in Taiwan. I used to be in Taiwan I would fly over for specific games like and guess what I would spend more than $35 on my ticket but I'm not flying in the back of the bus at this point sorry like so like this would this is it is like it's like when the Apple Watch they realize oh people it's fitness remember they watched with all these lists of different things and what you could like remember you could like draw on your Apple Watch and like draw yeah heartbeat texting to send someone a heartbeat that never went who is it Which is fine.

39:40Who's the Apple exec that's the most hardcore NBA fan? Because I feel like you got a... Eddie Q. So I know how much Golden State tickets cost because I sat... Well, I actually... I was poor. So I was in the second row. Behind the seat that was inscribed with Eddie Q's name on it. Oh, okay. There we go. Yeah, he would be the guy that needs to argue for this. Okay, so do you think this is a next season they might get it right? Are they already pre-baked on the next, on the next, like, cause they have like four more games. It feels like they could take this. They have the footage. Like they could just not edit the next game, not do that broadcast and do exactly what you're prescribing.

40:19I hope so. Look, free advice Apple. That's why I made, I've written this. So number one, the reason I disagree with your take is because every video they make suffers from this problem. They are produced in a TV style for a device that is fundamentally different than TV. That's the core problem. But is there at least a possibility that Metallica also said, hey, we don't want a substitutive effect. We don't want people teleporting. It's so funny to think. And I'm less than the same thing. There's a meme behind the scenes. They've made lots of other videos. Every video they have suffers from over the top.

40:52Yeah, even the Alicia Keys thing. Can you imagine if the NBA actually sold a ticket where every 30 seconds you had to stand up and go watch from a different view? They're like, it's amazing. You see every angle. You're like, please, sir, I just want to sit. I turned it off. I watched half the game on TV because it was uncomfortable. What do you think is the long-term ideal viewing experience? Is it you have it on TV and you have your Apple Vision and you're kind of like, you know, based on how closely you're paying attention, you're either like locked in with the Vision Pro on or you have it on the whole time?

41:31I think that a big thing for sure, the biggest missing experience is, of course, I want to be on my phone half the time. Like any person. I do have the M5 Vision Pro. It's the first time I ever took a demo unit. I've always like declined to do that. But I'm like, there's no way on earth I'm paying for this. But I do actually want to see if there's any difference. My actually biggest takeaway is the pass-through is much better. This might be, I didn't do a direct compare. That was just sort of my perception. And to be honest, I haven't used a Vizit Pro in a long time. So who knows how good my memory is.

42:04The pass through was really good. I had no issue using my phone at all. It did seem better than it used to be. So, but maybe that might be placebo just to be clear. Getting screen time through another screen. It's a long trick. We got levels. Wait, so your screen time app could potentially show more than like 160 hours a week. I have 26 hours of screen time to be. Because I was in the Vizit Pro using my phone, my laptop. That's right. What a world, what a world. That's funny. Well, I mean, are you optimistic about any other Apple Vision Pro, like, opportunities? Because I was looking at this$30 ,000 immersive camera, and I was thinking, like, I was actually advising some friends who, they shoot a much more, like, evergreen podcast, like these really definitive interviews.

42:53And I was saying, maybe you should start shooting. They were shooting an 8K because they were like, hey, maybe one day we'll license these to Netflix. We want the extra resolution. we're not distributing in 8K, but we have it now. And I was saying, well, you want to go further? Why not 16K? Why not immersive? Maybe you should get this thing, start recording now. Now this particular device can only record for like 45 minutes, I think. And then you have to swap the cards or whatever. But I was like, maybe you should be doing the flow. Maybe not this year, but maybe next year. Is there some opportunity where someone who's independent comes in?

43:22There's lots of opportunities. And I think there's real enterprise opportunity, like for like especially with the pass through and things that you can do on those lines there's productivity sort of possibilities but you need like what is the anchor thing what is it everyone knows that this does and again i think the apple watch is a good example here it took them a while to get to that they launched with the physical fitness stuff but it wasn't clear that that was the thing for a few years and now every vision every apple watch thing is fitness fitness fitness fitness and saves your life and you know uh all those sorts of things and to me this is it it is you get live in your living room that's the vision pro pitch that's gonna be fun we'll just have to wait another five years for them to back down from their their opinions it's but we're gonna find we're gonna find out that they have like you did write this two years ago like no i wrote this so long the day i wrote this the day it launched i said what's amazing about this is you don't need production.

44:21All I want is to feel like I'm there. And it delivers. Of course you don't know you're there. The resolution isn't perfect. There is a bit of tearing if you go super fast. The peripheral vision's not perfect. But it is amazing. It's an incredible experience that no one else can imagine. Quest can a little bit, but the vision project, in part because Apple invented these cameras. Blackmagic is making them, but it's Apple's whole format. It's Apple created the whole thing. And yeah, it just feels like this is a company that, it's like a lack of confidence. You let the device do what the device can do.

45:02Stop trying to like overdo it. No, step back. Just step back and let it do its thing. And I'm more convinced than ever. Again, I've been having this take a few times. I wrote about the MLS thing. I wrote after the Metallica thing. I put it on the front page of Checkery today. I'm like, look, no one in the world cares about this device other than me. But I'm desperate for someone to read this. So there's no paywall. You don't need to get afforded to you. Just read this. And they had to do the bucks too. It's so funny. I know. It's like they're kissing up to me. It was a royal flush for you. I have a meta question about just being in a position to review new technology and having to deal with like the Pepsi challenge.

45:48Are you familiar with the Pepsi Challenge story? I am. I'm Gen X. I'm very old. Okay, yeah. So the Pepsi Challenge, for those who don't know, they had people taste a little, a one-ounce cup of Coca-Cola, one-ounce cup of Pepsi. And overwhelmingly, the random people that came and tested said the Pepsi tasted better. And what the result was was that Pepsi was sweeter, so it tasted better over one ounce. But over 16 ounces, a full can, it was more like 50-50, and people actually did prefer the Coca-Cola. And I feel like with a lot of hardware devices, you put it on for 10 minutes, it's amazing. You put it on for half an hour.

46:18It's incredible. But then pretty quickly, the big tech executives say, oh, that's enough of the demo. Let's talk to you about something else. Now write your review based on 30 minutes. And it's a very different experience than two weeks, a month, seeing if it collects dust, seeing if you churned. And I'm wondering just how you deal with that mentally where you have to talk about something, but at the same time, you're not getting the – you haven't been able to spend a year with a product. No, it's a good point. So my way to deal with that is I generally don't do product reviews. Sure, sure. So that makes it much easier.

46:49I think with the Vision Pro, though, is a great example of this phenomenon. It is like the first 30 minutes of using a Vision Pro is one of the most mind-blowing experiences in your life. And it continues to be. I actually, I remember when I first got it, I actually had, I felt perfect follow-on from the watch segment. I felt like a total douche, this thing. I had my assistant, I had it shipped to his house and had him fly to Taiwan to bring it to me. Oh, yeah. because it was only available in the US. And I remember actually I was going, I think, this is awful, this is so delicious. I was going on a ski trip like a week later.

47:26So I'm up there, and I remember I'm sitting in this bar up in Niseko, and I'm like, all the ski instructors, because I was friends with, I had the same guy for many years. He brought all his friends over, and they're all trying this. I honed my whole demo script. And we still talk about this afternoon with the Vision Pro in this bar in Niseko, try it. And it's amazing. It's absolutely incredible. But then what do you do with it, right? Like that's the big question. What I will say in the Vision Pro's defense is even now, I don't use it very often. Every time I put it on though, I'm like, this is so cool.

48:06Like it retains that aspect of there's something magical about this experience. It's just searching for that ongoing reason to come back that use case. And that comeback use case needs to be something that only it can deliver. The productivity things are cool. I would, I would. Yeah, the other thing that's, the other thing I think is really real is like, maybe from Apple's side, like they have this device. It has so much potential. It's so exciting. It's so cool. And then it's almost disappointing that like, okay, the use case is just like live sports, right? Because they're like, we want to create the next computing paradigm.

48:41But the reason that it actually is an incredible niche is that like what's more mainstream than bread and circuses, right? What's more mainstream than sports? It's a chicken and egg problem. Like the way you get those productivity experiences and those new things that weren't possible before is you have a large install base that draws developers in to create those experiences. But to get there, it needs to be a market. so you need sort of the initial use case to create like this is why the iphone is the greatest platform ever everyone needed a phone yeah like so like so you you got that built-in sort of advantage so that developers like they didn't have to do anything to earn developers they just shipped a better phone that's what i'm saying like if you create this amazing live sports experience you get a million people that are using it multiple times a week and then and then they're using it they're watching they're watching together it's a social thing then they can be asking.

49:37I don't know if it's a social thing. If you see what I look like, I don't grab, I'd put it on. I don't look very social. No, but I'm talking about more like, maybe you have a FaceTime call open with a buddy in the headset, and you're watching it together, or you're hanging out watching the game, or something like that. Well, I mean, I would just say, I think it's, the Vision Pro is more of a single player experience. But, what is the biggest single player market in the world? Productivity. Like, when I'm doing work, like, I want, I'm on my own working on it. Right now, I have four monitors. Like, I'm one of those crazy people.

50:11Like, a Vision Pro, I wish I would have brought it with me last summer when I went back to Taiwan because I was actually doing work, and I was working on my laptop, and it was terrible. Like, oh, I can't believe I didn't bring this. This is the perfect use case for me to use it. But by and large, I don't need it. But if there were particular apps or use cases that were uniquely enabled by it, maybe that would be a reason. but you're only going to get those by drawing in developers. You're only going to get developers by having an addressable market that is worth putting the investment into, which means if this has always been, this is one of my big theses on strategy.

50:45It's always been a question on tech, chicken or egg. What comes first, the device, the platform, or the developers? I was at Microsoft in with Windows 8. I spent a lot of time and spent a lot of Microsoft's money getting developers. that doesn't work. What matters is demand. Demand pulls in developers. You need the users, which means you need the core use case to sell a bunch of devices. And then you can start the virtuous cycle. To me, the core use case for the Vision Pro, if anyone remembers it exists, and I'm not sure that Apple executives do, is live. Live in your living room. You do that, you sell a bunch of devices, and then you pull people in.

51:23I have a buddy, TJ, who worked at Apple at one point. When the Vision Pro was announced, he got it. He was so excited. He was like, I'm going to spend the next three years building products for the Vision Pro. How long did he last? He lasted like three months. Yeah. Like, no, no, no, like less. He was like, what's the point of building? What's the point of building for a super talented, but what's the point of building if there's nobody there? You mentioned the four monitors. Are you going in on the 52-inch Dell 6K monitor? No. Why not? Way too low resolution. Way too low resolution. Actually, I lied.

51:57No, I'm actually on the fifth monitor right now just for podcasts. So I have two 4K monitors on my laptop. I have another. I live with that LG square screen. Oh, yeah, yeah. It is relatively low resolution. But then for podcasts, I have a 55-inch TV here. Okay, staying on Apple, Gemini. The Gemini Siri news dropped this morning. It had been kind of previously reported. So in some ways, it's old news. What are you expecting out of this new version of Siri? Well, the bar is 55 feet underground. So I expect it to be a lot better. No, I think it makes a lot of sense. I mean, now that the federal courts have approved Google and Apple combining to rule the world, Google has the infrastructure to support the scale of Apple.

52:48They know how to work together. They've worked together for years. There are very natural partners in that respect. Google can think big picture about this. Like I'm sure Apple, the report that Mark Gurman had last fall is Apple's going to pay like a billion dollars, which again, we're in tech, that's no money at all. But it's tied into the search deal. Like they can massage it. The problem with working with an open AI or an Anthropic is they need to make money. And so like Apple doesn't, I think there's a more sort of I scratch your back, you scratch my back sort of thing here. Google's talking about working with Apple's chips, adapting it, whatever it might be.

53:23Apple does bespoke stuff like that. So I think it makes a ton of sense for both sides. Do you think it ever flips? And do you think Google will be paying Apple? Because there's this news also that you're going to be able to do shopping through Gemini. And so you could imagine a world where you go to Siri, you ask for a product and there's ads in that Gemini result. and Google's the one that's monetizing that. So they're passing some of that revenue back to Apple. It's a good question. I mean, I think when Apple talked about like the next generation Siri and Apple intelligence, I was pretty optimistic about this idea of Apple basically replaying the search.

54:00That's my whole thing. I'm like, in some way, if I'm Apple and I'm paying for an LLM to use it to power a product that can basically do search really well, that could eventually have an ads business attached to it, that eventually could have like commerce built into it. Like it's this weird, it's this weird situation where like Gemini is effectively helping. Like that's, that's what I don't fully understand yet. Cause I'm going to be like, I think the market structure was so perfect for Apple in that they need, like, it's just a default search engine. It's there's no sort of deep integration into the product.

54:32It's just like when you type in your search bar, what engine is used. And so it's super easy. Like there, there's like a concept when it comes to like figure out who's going to win value chain, super easy substituability is good for whoever can plug whatever in. So they could choose whatever they want. So it made sense that the value flowed that direction. The difference with this AI stuff is it needs to be integrated deeply. And Apple is more on the defensive here. They need to have quality AI features built into the operating system. And so I think the need is more on Apple's part. And the benefit Google is getting, my suspicion, is less that there is less about the ad thing and more they're getting some incremental revenue and they're probably getting a lot more.

55:21I mean, Apple's going to be precious about the data. We'll see how that sort of works. they also don't need, it's all incremental. I would imagine. Even like the internal reasoning rollouts and all of that that happens in Gemini, Apple's not going to be able to claw all that back. So you're going to have all this reinforcement learning training on, okay, maybe you're obfuscating the privacy data, what the person asked for. But all those interim steps of, I went to a website, I interacted with it, I figured out how to click this button. That's going to accrue to Gemini, you would imagine. And what it speaks to, I think, is what would be the red flags that would come up to this deal from an Apple perspective?

55:58It's like, well, they're not yet. that maybe they need to tough it out and build their own crappy LM so that in the future, they control their own LM such that if you get to a world with, say, AR, right, where I'm actually, lots of dispute about this, but I'm pretty optimistic on just-in-time UIs, where the idea that something pops up that is nothing but the decision you need to make in that moment for whatever it might be, and then it goes away. Like, to me, this was the best part of the Orion experience was the Facebook sort of AR glasses was when I received a call because the OS I used there was kind of like Quest just sort of dumped in there.

56:36And it didn't really make sense. You have like blocky stuff and there's like an Instagram and stuff. It's like, I could just look at my phone. This is going to be better here. But you could be, I got a conversation right now. I could have a notification popping right now saying so-and-so is calling and I could use my little wrist thingy and dismiss it. And it's just there when I need it. It only has one option. accept or decline, and then it's gone. And I could see that being a future of UI and important for Apple to control, but by not shipping their own, by depending on Google, they might say pretty words about, oh, we're going to simultaneously develop our own.

57:12No, that, whatever. That's not going to happen. You're going to be so invested in this other one. And that speaks to the value is accruing to Google because they're the ones actually developing and pushing the technology and Apple is sort of trailing along. Given that I have a hard time seeing if Apple's going to get more and more deeper into this using Gemini, are they going to be able to credibly go to Google, say pay up or else we're going to switch to someone else? I think that's probably unlikely. So I don't think it's going to play out like search, but we'll see. I was obviously wrong about this one, so I could certainly be wrong again.

57:46Sure. What was your reaction to the Manus acquisition from my point of view is exciting, specifically because Zuck has just has been spending all this money on talent, but if you look historically, he's been very good at like buying something, scaling it. It's sort of unclear to me so far, whether he plans to take Manus from$100 million run rate to multiple billions, or just leverage the team's ability to build great agentic, you know, effectively workflows product. How are you thinking about it? Yeah, I just, I have a hard time seeing the meta in the enterprise sort of angle. So no, but, but, but even, but even like talents, but even it's like, like figure out the best product in this category and buy it for me.

58:31Like that, that kind of workflow, like that's what I can imagine the Manus team being. That's my sort of understanding. Yeah. It's like, this is actually a really excellent product team that is doing very good work and is worth having on board. And does that mean growing their business into something larger? That's a material source to our business. I think that would be a mistake. but I think the idea of agentic workflows is obviously a very compelling one. Actually, one of my favorite things, Mark Zuckerberg has been all over the place on AI. I did an interview with him like nine months ago.

59:01That was kind of bizarre. It was right when he was clearly thinking through, maybe I need to reset everything. And I thought that sort of came across in the interview at the time. But one thing he did say is actually, what is the largest sort of most profitable at scale agent right now? Facebook advertising, which I think is a very, or Google Advertising, a very astute point. You go in, you tell Facebook, I want, I'm willing to pay$1.49 for a customer or$9.99 for a customer, go get them. And it goes out and gets them. Now, is that a full LM, denominated, probabilistic workflow? No, it's not really LM's.

59:38It is more probabilistic, particularly post-ATT. But the idea that you ask for a job, the focus is on the job to be done, as opposed to specifying how you do it. and that is just by and large for most advertisers, particularly in the long tail, probably at the fat part of the tail too, they just don't want to give up control, the better way to do it. And I think there's going to be lots of things like this. I think Google's announcement with Shopify and this idea of ads that are perfectly targeted to user, very compelling, makes a ton of sense. That's where something like Rock, by the way, fits in like super fast inference so you can generate, like you think about how advertising works, these auctions that are run and you get ads in the time it takes a little web page is absolutely incredible.

1:00:21Where the next step is to be insert generative ads into there, which is going to require very high inference speeds. So I think, and this has always been the most compelling short-term AI monetization opportunity, is basically Google and Facebook ads. That's why I've been super bullish on both of them. I think it's why Facebook needed to do this reset and why Google, you know, it's justifiably been sort of going to the moon recently. Well, what do you think of Eric Suford's point that you might not actually need to put the ads in the LLM responses in the sense that you could be going to an LLM and saying, tell me the history of the Roman Empire.

1:00:58It knows that you're shopping for shoes, for new shoes, and it shows you ads for shoes in the middle of your Roman Empire deep dive. Just like on Instagram, you can be scrolling one thing and your algorithm can be recommending you a certain type of content, maybe like dogs, showing you dog videos. But then it also knows that you need luggage for your next trip and it's showing you ads for luggage or something. No, I think it's a great point. I think Eric makes some good points about like you're running a real risk of conflict of interest, even the perception of conflict of interest. Totally. Otherwise.

1:01:29And so I think that's a very good point. And the reality is that's a much larger business. Like just be – what you're looking for is inventory. The number one way to predict sort of the upside for meta over the last several years, the stock market has consistently had it totally backwards. Every time their sort of their price per ad would plummet because the stock market would freak out. This happened with stories in like 2018, 2019. It happened with reels a couple of years ago. And the issue is that the reason why price per ad plummets is because there's a massive increase in inventory. And when there's a massive increase in inventory, it's just more places to show ads.

1:02:12That is a huge opportunity. You saw huge run-ups both times as they figured out how to monetize stories, as they figured out how to monetize reels. And this should be an opportunity of how to, like, it's going to take a while to figure out how to monetize. You know, maybe it's just an aspect of when they're reasoning, when they're thinking or image generation, that might be an ad opportunity. And it's a big opportunity. A big problem is these folks, I think, the OpenAI has hired so many meta people, it confuses me why they haven't been on board with this. Like, it's a win-win-win. You need money to fund your operation.

1:02:50The best way to make money is to show ads. Why? Because you get to deliver a better product to more people. This idea that we're going to commit to a world where if poor people get a worse product, that's not how tech works. The reason why tech is amazing is it generates a ton of consumer surplus. You do this upfront investment. And because it's monetized by ads, everyone gets the best product. It's great. And OpenAI, like having this religious devotion for so long about not doing this, if they had launched, they could have launched the world's crappiest ads in 2023. By today in 2026, they'd be good.

1:03:24They'd be making money and people wouldn't rebel against it. Now they're going to have to launch ads. They're going to suck. And people are like, this sucks. I'll just go to Gemini or whatever it might be. just, it drives me bonkers. It does seem like there's a little bit of like, not to go back to chicken and egg problem, but like the first mover disadvantage, like the first LLM that has ads, where there'll be a whole press cycle about, oh, Gemini's the ad one. Would have been a lot easier if you were the only LLM. I know.

1:03:53It risks the entire company. Like they need to get their opportunities in the consumer space first and foremost, because the consumer space needs to monetize via ads. And the fact they didn't get there or start to get there, still haven't started to get there, it's a company imperiling poor decision. Last question from my side at least. Predictions. Do you think there's anything to the prediction that OpenAI will buy Pinterest or partner up with a social network at a more deeper level than they are yet? Well, so it's really interesting. This is one of my, I was actually, I was talking to, I met when I was in New York a few months ago.

1:04:33And I was at this office. Actually, I was being the semi-analysis guys. Great guys. But they were sharing an office with like a hedge fund. And one of the guys is like, look, you're responsible for one of our worst all-time decisions. And I'm like, what's that? He's like, we bought Twitter. I'm like, I never said to buy Twitter. That's a terrible stock. Unless anyone's going to overpay for it. And he's like, but then I remembered what it was. It was, I think, 2017, 2018 when Twitter bought Mopo. And my theory at the time was Twitter is just a very poor inventory for advertising. Part of it is it's text-based, especially then.

1:05:08There were fewer images. There's also a mindset. When you're on Twitter, you're, like, girded up for battle. Oh, yeah. And you're, like, you're trying to engage. Like, as opposed to Instagram. Like, Instagram, the ads might as well be Instagram. Yeah, yeah. Axe is fight or flight. Yeah. No, exactly. Exactly. Exactly. And so my theory then, but Twitter has the potential to really understand your interests in a really sort of deep way. And so what they could do is they could harvest signal on Twitter and they could manifest it with using Mopub and inventory sort of across a bunch of apps. And that would be sort of a very compelling model.

1:05:43Again, who knows if I was right, Twitter's executives or Twitter was incompetent for years and years and years. At one point, I said, I'm never covering this company again because this is pointless. us. That's a bit what sort of Apple Evan did, who did buy Mopub from Twitter and has made a ton of money doing that. But I think that would be the thesis there, which is because we're people are dumping everything to this LM, we can get all this signal and understanding what we need is inventory to monetize that signal. And could that be inventory to do so? The theory makes sense. It feels like they have a lot more important things to spend money on.

1:06:21I think probably at the end of the day, particularly post-ATT, like, oh, no, like your own and operated properties are always going to be the most valuable. I think figuring out how to monetize in-chat GPT is probably the best. But it's not an insane idea for that reason. I like that. Well, thank you so much for having me on the show on short notice. We're huge fans here. And congratulations. Like, I am on a app. No one cares about the Vision Pro but me. Therefore, I take it upon myself. Correct the record. The strongest soldier. Well, if you are. Please. Yeah, if you want the deepest analysis, the hottest takes on the Vision Pro, of course, sign up for Stratechery.

1:06:58Get on the Pro plan. This one's free. You don't need to look. This one's free. But you know there's going to be podcasts. We should hire. We should get a plane, fly over, and drop this new one as leaflets. Leaflets over Cupertino. We'll do it. I'm down. Let's do it. Let's do it. We need to. We need to. Well, thank you so much. Have a great rest of your week, and I hope your 2026 is off to a great start. Yeah, great to see you, Ben. All right. We'll talk to you soon. Have a good one. Label Box, delivering you the highest quality data for Frontier AI. Get in the box. The Label Box. Every time. They didn't give us that.

1:07:37They didn't give us that tagline, but it's just simply too good. Get in the box. Get in the box. Walmart partners with Alphabet's Google to allow shoppers to purchase products through Gemini. So Walmart is jumping in with Google. We're having Harley from Shopify on the stream later to talk about agentic commerce, what's happening there. There's a lot of other news. And I have the audio on my phone. Google is posting a video of Wing. The future of retail is landing. They're taking shots at our boy Keller launching a drone delivery service. This one hit me pretty hard. I know. I know. We love Zipline.

1:08:18We love Keller here. We love Google. Obviously, they're a sponsor. Can Google just leave one future of X thing for someone else? Yeah. Right? I didn't even know about Wing until today. Was this an acquisition? Wing.com, one of the best domains. You're going to be texting Keller like Sam Altman and Elon were texting each other about the future of AI. You're going to be like, the future of drone delivery is in our hands, brother. We got to beat Wing. No, Google's been working on this for a long time. Obviously, this is going not to be, you know, the only, there's only going to be one company with the technology.

1:08:52There's going to be variety. Tyler, what do you have for us? Just on the point of Elon and Sam. Yeah. Elon just said on the Apple and Google collaboration, he said, seems like an unreasonable, unreasonable concentration of power for Google. Given that they also have Android and Chrome. Oh, interesting. So he's still on the monopoly. He's taking shots. He doesn't like, he's not a fan. Well, we are fans of Plaid here. And so let me tell you about Plaid. Plaid Plaid powers the apps you use to spend, save, borrow, and invest, securely connecting bank accounts to move money, fight fraud, and improve lending, now with AI.

1:09:21We talked about Apple confirming, Gemini, very excited for that. I want them to roll this out immediately. I know that it's probably going to be some normal release cycle with very polished ads and onstage keynote and a developer preview, and there'll be a whole cycle to updating. But we are in the age of AI, Apple. Just ship it today. Just replace Siri with Gemini today. I'm sure a lot of people would be fans of that. But, you know, they operate the way they do. Eric Topol. I pulled a little history on wing.com. Please. So it started. Wait, they went wing.com? Wing.com. That's what I'm saying.

1:09:59It's one of those. It's not only this is an amazing partnership, but a fantastic domain. Yeah. So it started with an X, Google X, the Moonshot factory. It really is a factory. The original mission was focused on emergency medical response. Okay. So they wanted to deliver defibrillators to heart attack victims. Basically, they pivoted away from emergency services to last mile commercial delivery. They started doing their first real world trials back in Queensland as early as 2014. It graduated from X in 2018. They later became the first drone delivery company to receive a Part 135 air carrier certificate from the FAA.

1:10:47And they've just been scaling the network since then. So, yeah, I guess they're going to be able to serve 40 million people by 2027. Yeah. I mean, as an American, as a human, as a technologist, I want more and I want competition. But as a big fan of Keller and ZipLine, I want him to dominate. No, I think Google maybe did this. They knew Keller had the potential to be one of the greatest in history, but they realized if he didn't have a viable competitor, he would never live up to his potential. So they're inspiring him to grind harder. Exactly. That's what's going on. Okay, now we understand it.

1:11:21Well, Figma. Figma Make isn't your average vibe coding tool. It lives in Figma, so outputs look good, feel real, and stay connected to how teams build, create code-backed prototypes and apps fast. Open AI launched ChatGPT Health. Now Anthropic has Claude Code for heart attacks or something like that. Healthcare and life sciences. Claude Code. I'm dying. Give me blood transfer. Don't make mistakes. I used Claude Code this weekend in a funny way. I was having slow Wi-Fi, which, of course, is a weird thing to go to Claude Code for because it uses the Internet, so you're going to have slow interaction.

1:12:01I had it diagnose my – I gave it a prompt. I said – what did I say? It's somewhere here. Maybe I lost it. But I basically told it, like, hey, I'm having problems with my internet. Can you just go fix it? It ran all these different diagnostics, pinged Google, pinged all the different DNS servers, ran through everything, ran speed tests. And it came back and told me to turn it off and turn it back on. And it actually worked. and I could have saved myself like 45 minutes of sitting there being like, yes, I'm okay with you using curl. Yes, I'm okay with you using WGED. The entire time superintelligence was just turning it on, turning it off and then back on.

1:12:38It's Lindy. It's Lindy. It should have just preempted me and just been like, look, dude, have you at least turned it off and turned it back on? Anyway, Tyler, what do you think? If Claude was being slow because of the Wi-Fi, then that's an example of like, you know, self-improvement. Self-improvement. Oh, it improved itself. Yeah. Yes, yes, yes, yes, yes. It was a weird issue. Really high ping. The bandwidth was fine, but the ping was really bad. It was very annoying. It did resolve it. At the end of the day, Claude Code did a very thorough job of telling me some time-honored advice. Thank you to Claude Code.

1:13:15Interesting that everyone's pushing into healthcare. I'm still waiting for the push into legal. I'm wondering if that'll happen or if that's more complicated than healthcare. I'm also wondering maybe healthcare is more lucrative, more viable, more... I would love to be in the meetings where they have prioritization of whose lunch they're trying to eat off. Whose plate should we eat off of? The lunch meeting. The lunch meeting. Well, New York Stock Exchange. If you want to change the world, raise capital at the New York Stock Exchange. We love the nice seat. Can't wait to be out there. Again, we're planning our next trip, so stay tuned.

1:13:54Cannot wait. Meta CEO Zuckerberg is launching MetaCompute, planning tens of gigawatts this decade and hundreds of gigawatts longer term. This effort will be led by Santosh, Jannard Han, and Daniel Gross, DG, back in the mix. Santosh will continue to lead our technical architecture, software stack, silicon program, developer productivity, and building and operating our global data center fleet and network. Daniel will lead a new group responsible for long-term capacity strategy, supplier relationships, industry analysis. Let's give it up for industry analysis. I love it. Planning and business. Let's give it up for planning and business modeling as well.

1:14:32DG, DG, this is not his first rodeo. Do you remember the Andromeda cluster? Yes. NFDG? That was really, really cool. Early and right on that. So they were doing, was it AI Grant was the name of the program? So they backed a bunch of companies, some crazy companies in AI Grant too, sort of an incubator, accelerator model, a bunch of early stage companies, smaller checks. Give me the details. Let me just read this. So in the first batch, they had Perplexity, Cursor, Replicate, Chroma. Chroma, too. I think they had, didn't they have Julius? Julius was in maybe a second one. It might have been in the second one.

1:15:07But there's so many bangers. If you just go to the site, it lists all of them. It's amazing. And so what they did was they went and bought a whole bunch of GPUs. They built a cluster. They had a feeling that GPUs would be important. And it was, so basically anyone in their portfolio was sort of default GPU rich, at least for a little bit. Although of course the resources were shared. And I have heard about another accelerator doing this recently. I need to confirm with the head of that accelerator if it's okay to share, because it's very exciting as well. But I was asking them like, what does it mean to build a cluster at this level?

1:15:41Did you buy land? Did you build a data center? Or did you just go to Microsoft and say, hey, cordon off this little rack of GPUs. We want those to be dedicated, at least instances. And I believe it was more of the latter, but still not their first rodeo here. So I think this project's in good hands with DG. So let's hear it for the Meta. Yeah, Meta also brought on Dina McCormick. Oh, yeah. Joined Meta as president and vice chairman to work on partnering with governments and sovereigns to build, deploy, invest, and finance Meta's infrastructure. This is making a wave on Truth Social, I saw. Government getting involved in financing Meta's infrastructure.

1:16:22So he's saying, he's like, Sam, they said it. I'm going to say it now. It's not a backstop. It's the front door. Dina was a former deputy national security advisor. Well, we have a new sponsor of the show, Sentry. Sentry shows developers what's broken and helps them fix it fast. That's why 150 ,000 organizations use it to keep their apps working. Welcome. Century. Century. So welcome to them. Time to have you. NVIDIA is investing a billion dollars in an AI drug lab with Eli Lilly over five years. This will be very interesting. We're excited for the Eli Lilly progress. Drug lab sounds. It's AI Ozempic.

1:17:03It's the two biggest super trends of the last five years, weight loss and AI. Could it get any better? Well, now it will. We'll have to dig into more of this partnership. Tom Brady is now the face of the former CEO of X's new company, E-Med. Oh, that's right. Did you see this? I did. So Tom Brady, if you're not familiar, he was the former face of FTX. That's a rough one. And also an NFL player. Was he actually the face? Because there were a lot of celebrities that partnered with FTX. Larry Day was in there. He was part of some of their bigger campaigns. He did a bigger campaign. He did a TV campaign.

1:17:40But again, you know, it's hard if you're a celebrity to... But this is an interesting one because... So he's joining as the chief wellness officer of eMed. Yeah. And it's interesting because this kind of just makes him the face of GLP-1s, right? Yeah. Which is kind of a beneficiary, like is beneficial to the entire industry, right? Sure, sure, sure. If you sell GLP-1s, you can be like, well, Tom Brady is... Yeah, he's down. Yeah. I wonder if he's on them. That'd be interesting. Anyways, so... regimen is. Maybe he's just super AGI pilled. Maybe Sam said, hey, I'm an investor in Anthropic. And he said, well, I think Anthropic's going to win.

1:18:19I'm partnering up with you. I don't care about the structure of your hedge fund. I don't care if there's a backdoor out of your trading platform. I'm in on you because of your investing track record. What about that? If Anthropic gets out at a trillion, there's going to be a debate at least about Sam Begman-Fried's legacy. He, of course, invested, what,$10 million for 10 % of the company? He owned 8%. 8 % of the company. So that would be maybe like an$80 billion position today, something like that,$50 billion,$10 billion dilution. I don't know. It was a good investment. Just like getting your business on Gusto is a good investment, the unified platform for payroll, benefits, and HR, built to evolve with modern, small, and medium-sized businesses.

1:19:02Like us. Like us. uh paramount skydance has now initiated what insiders are calling plan d they're running out of letters as they look to upend netflix's winning bid for warner brothers discovery hey maybe the d just stands for discovery so you know maybe this is actually one of the earlier plans but maybe they're maybe they're saving the best for last plan w never take the l skip plan l go straight to plan w get the w we're rooting for you ellison uh it involves bringing home to investors the immense amount of regulatory uncertainty involved in the Netflix deal and how that could spell trouble, not just for the transaction, but for Netflix itself.

1:19:45And so if Netflix finds itself in a quagmire trying to acquire Warner Brothers Discovery, that could be bad news. And David Ellison wants to make that clear to all of the shareholders that going with a Paramount Skydance might be a little bit less bumpy of a road. We will see. From the photo shoots, seems like they're pretty happy together, but we will see what happens. It will be a proxy fight, and I'm sure we will be following it closely. Some more news from Axios. Besant told Trump that investigation of Fed chair creates a, quote, mess. And so basically, last night, Besant was saying the secretary isn't happy.

1:20:25Or no, someone else said the secretary isn't happy and he left the president. No, that's kind of a white pill, right? Somebody in the admin that seems to have just kind of a steady view on things is pushing back a little bit. Nothing ever happens. Yeah, you need a secretary of nothing. Did anything happen? No, nothing happens. But Vanta happens and compliance happens. So again on Vanta, automate compliance and security. The leading AI trust management platform is, of course, Vanta. There was a massive debate that raged in a Google Doc, which was published on Substack. The battle between Jack Clark, who's at Anthropic, of course, the co-founder of Anthropic, and Dwarkesh Patel.

1:21:12They were going up against Michael Burry, Cassandra Unchained. They had a good discussion, and it's been published free to all on Substack, where you can go and read it. highlights very cool Tyler yeah I have some highlights I mean I wouldn't so one thing I it's not really a debate at all it's discussion it's like there's a bunch of questions and they go and they basically they're just like AI is amazing I agree yeah no it's funny like all almost all the quotes I pulled were from uh Burry because the Dorkesh and Jack Clark I disagreed with everything they said yeah it's like oh it's like not that interesting like I already I already know this boo so one thing your own opinions bro I agree with everything you said I have no You're absolutely right.

1:21:49Yeah. So one thing I was saying, Jack Clark, he writes, I've been doing research on the cost curves of various things recently. And the examples he gives are dollars of mass to orbit and dollars per watt from solar. Okay. So it's like, okay, Anthropic is looking at space data centers. Sure. Right? Sure. Some other stuff, I thought Burry said his one policy proposal. So just the fact that Jack Clark is doing that research indicates to you that they might be thinking about orbital data centers in the future. Yeah, I mean, massive orbit, mass orbit, and yeah, I don't know what else. Yeah, no, it makes sense.

1:22:23I mean, maybe he's just interested in space. He might just be interested. Yeah. He might just be like, I want to go to the moon. I'm bringing Claude with me. Yeah. Yeah, I don't know. Claude, go to the moon. The other interesting Burry take was that he thinks that blue-collar work might be more susceptible to AI disruption than many people think. He made this point that he's been doing his own plumbing and electrical work, apparently, I don't know how true that is, but he said that. Yeah, yeah, yeah. Can you read it? I have the quote, yeah. He says, given how much I can now do in electrical work in other areas around the house, just with Claude on my side, I'm not so sure about jobs going away.

1:22:58If I'm middle class and am facing an$800 plumber or electrician call, I might just use Claude. Take a picture. I love that I can take a picture and figure everything out I need to do to fix it. That was a big moment for me. I forget who was sharing this, but somebody was like, yeah, they just took a picture of the most insane wiring diagram. No, it was just a bunch of wires going everywhere, and it just one shot of it. You take a picture of your car. It needs engine out service. You're like, where do I screw? I'm doing it myself. Well, we have the perfect guest to talk about AI diffusion and its impact on the job market with economist Tyler Cowen joining in just a minute.

1:23:36First, let me tell you about MongoDB. choose a database build for flexibility and scale with best-in-class embedding models and re-rankers. MongoDB has what you need to build what's next. And without further ado, we will bring Tyler Cowen in from the Restream Waiting Room. Tyler, how are you doing? Good to see you. Welcome back. I'm fine. Good to see you. Congratulations on the Vanity Fair piece. I loved it. Thank you so much. I'm so glad that you were able to check that out. It was a lot of fun. Very funny getting dressed up all fancy. When are you going to be in Vanity Fair? They really have not been doing their work.

1:24:08I think I would love to see a Vanity Fair photo shoot with you as a subject. I think it would be fantastic. We should make it happen. A style icon, right? Yes, definitely. Well, I mean, they really should do more reflecting on Conversations with Tyler. The team's grown a lot. Did you just hit a recent milestone? I saw that there was an event. Can you just get me up to speed on the organization and how long you've been working on the show? conversations with Tyler is now 10 years it's a few hundred episodes

1:24:42an episode a week that is fantastic well uh it's one of my favorite shows it always has been for I feel like I might have been a decade-long listener uh maybe maybe not the first year but I got on early and I've been very pleased the whole time um but anyway thank you so much for joining us. We first wanted to have you reflect on this point from Michael Burry that artificial intelligence might displace more blue collar jobs than people are expecting because you can now take a picture of your toilet and learn how to do plumbing or you can take a picture of an electrical panel and you can do your own electrical work.

1:25:18Does that resonate with you at all? Do you think that there might be some substitutive effect in the DIY community that might have an effect on those trade jobs that previously have been rumored to be very AI resistant. Well, that's true, but there's still a net job boost coming in those areas because there'll be all these new projects, new data centers, new sources of electricity, and they need plumbers. They're not all going to use chat GPT to take the photo and try to figure out how to fix the thing under the water or whatever. So the house call plumber, maybe that goes down by 20%. but so many other new things happening, terraforming the earth, whatever, other countries, Africa developing, those jobs will be doing great.

1:26:04But that's it, I agree with his example. Sure, sure. How are you tracking the overall data center boom? There's been this narrative that data centers are the only thing that are propping up the US economy. It feels like we hear about new data center projects every day. The numbers are getting bigger. There's a new zero every time there's a new press release. And at the same time, the overall power generation, the big, big US national number is not yet started to move, but it feels like this might be the year where we see more electricity generated in America than before. How are you tracking the overall impact of AI on the broader economy?

1:26:45We need to do much better, but I think it's wrong to feel that without the data centers, the economy would collapse. Those resources would be used to do something else and different. It might be less risky. It might be more immediate. Actually, consumption would be higher in the short run, if not for AI in the data centers. So what we're doing is investing longer term, taking some risk, boosting medium term productivity. I'm all for doing that. But again, without that, we'd make more toys or more hot dogs or more something else. So no big deal. How often are you listening to AI music? I try it reasonably often, but I don't yet listen to it because I want to.

1:27:27It gets closer every time. I would say within two years, I think I'll be listening to it on purpose. Right now, it's experiment and tracking the field, not for enjoyment. Sure. Do you find the act of playing the game of trying to generate a song that you like more entertaining than the actual resulting music? No, I don't find either entertaining. For me, it's a pain. You know, I have a very large collection of albums and compact discs. I can hear literally the world's best music at my fingertips when I want to. And the other is a distraction. I do it as my own investment in seeing where things are headed.

1:28:09Do you think there's a particular... I had someone share an interesting anecdote recently, which is that one example, a professional singer who's sort of top of their field within country music. They made something like$400 ,000 last year, basically going into the studio and recording sample tracks. You know, basically a songwriter would make a song. They would sing it. And then they go out, or this person would sing it. And then it would get pitched to other artists to be made into like a hit, right, and actually be produced. And his work literally went to zero because of Suno this year. And he doesn't have the ability.

1:28:51Nobody hires like a backup singer to go on tour. Like, you can be like a studio guitarist and get live jobs. And so their work went from, because now people can just prompt with Suno and say, hey, use these lyrics, apply it to this track, and make it sound like this artist. And so that's just been kind of an interesting space. And I was talking to another friend yesterday that says, basically, everybody is using Suno at scale, and very few people in the industry are willing to actually talk about it. Interesting. You know, I played some for my wife recently, and she said, I hate it. And I said, why did you hate it?

1:29:28And she said, I hated it because it was good. And that's where we're at right now. That is, that is. But yeah, it's notable, too, that it's maybe not, it's not great for, like, the end listening stage yet, but it's totally good enough for, like, the sampling for the professional kind of workflow side of things. Yeah, yeah. Do you think there'll be sort of a legal reckoning, figuring out how all the royalties flow through these models and saying that, okay, there's 1 % Taylor Swift in this generative sample and 2 % someone else, and then all of the money from sort of like a more complex rev share that might flow through these generative models?

1:30:09I don't know if that's practical. I mean, the Beatles took from Chuck Berry, Buddy Holly, Carl Perkins, everyone. Yeah. Without paying and the world went on, even whether or not it was fair. And I think we're going to redo that experiment. I don't think it will be all that different. Most musicians don't make from recording anyway. Yeah, that makes sense. Have you tried Claude Code? No, I have not. I've been meaning to, but I've been traveling and I'm finally back home. And I will be trying it. Everyone tells me it's amazing. Yes, yes. People are maybe one click behind you. I remember you declared 03 AGI, and now people were maybe hesitant to call the game there, but many more people have jumped in with affirmations that Claude Code is AGI because it can do so much more when it actually has access to your full computer.

1:31:02But it is a little bit more cumbersome to get set up. How did you process Powell's video yesterday? Well, it was terrible what Trump did. Fortunately, most of the markets did not react very much, but gold and silver went crazy. To me, that's a sign that the dollar is much less of a safe haven, and Trump is going a bit, you know, what I call the Captain Queeg route, and he's just not reliable. And that's very bad, but I don't think it's really going to change inflation or interest rates very much. Is that just because Powell holds steady? No, because we already wrecked the independence of the Fed, which I'm not happy about.

1:31:39But that's the ugly little truth behind this story. That's why it's not been worse what Trump did, because it was already wrecked. Why is Fed independence important? Sometimes the central bank needs to do things that are not politically popular, and they need that shield to do it. But the basic problem is our debt and deficits are so high that over time we will monetize them to some extent and have higher inflation, because we prefer that over higher taxes, no matter what we might say. And so the Fed independence has taken away through that mechanism, apart from whatever Trump said yesterday. Oh, so just by running - The word may have shape.

1:32:17Like, I'm not telling you not to worry, but I'm telling you, you should have been worried to begin with. Black bill. So yeah, you're saying that just by running high deficits, that reduces Fed independence. And that's an old story, but it's now closer to a tipping point. And Trump is making that much worse. So I would say his actions on the fiscal front do more to hurt the Fed's independence than his words. But they're both bad. Can you walk me through your mental model for thinking through what the right level of debt for a nation is? Some of the numbers get so big, it's trillions of dollars.

1:32:55It feels very abstractly bad. But then if you look at it more like a mortgage relative to debt to income, it feels maybe more reasonable. How should nations think about the level of indebtedness that's appropriate? The United States is a special nation, and we can get away with more debt, which is great for our living standards. But it's bad for our political responsibility because our leaders all know we can get away with more debt. I think what we'll need to do at some point is have half a dozen years of something like 7 % inflation, get the debt down. It won't at all solve the problem. The problem will never go away, but it will give us some breathing room.

1:33:37We'll then lower inflation, maybe have a recession, and start all over again. And it's highly unpleasant, and a lot of people will be thrown out of work, and living standards will be lower. But we've already spent that money. We can't default. and that's facing us over the course of the next 10 to 15 years. But I don't think the world will end. Is America's position in AI the biggest counterbalance to all that sort of negative view? It is. So if AI would help our economy grow one percentage point more a year, we could just afford the whole thing and we wouldn't even need to inflate. Now, that would not be my best prediction, but it's not impossible that it could happen.

1:34:17And I've been predicting half a percentage point a year, which still puts us right on that border of can we, can't we, we don't know. I wouldn't want to stake the whole house on that. But again, there's some chance we'll squeak by due to AI and assorted productivity gains. Yeah, but why not 1 % growth? Why not 2 % growth? These tools, they feel amazing. It feels like... Or triple digit, as Elon alluded to. But it does, you use these tools and it feels like you're more productive. And there's some studies that say, oh, maybe you can generate documents that don't get read. Basically. Is that what's going on?

1:34:52Like, like, like what is holding back AI from actually moving the needle on economic growth significantly? Well, first, about half of our economy right now is just totally perpetually sluggish. Look at government. Look at most of the nonprofit sector. Look at higher education, parts of our health care sector. Add that up. You're at 50 percent of GDP. To some extent, they use AI already, but in pretty trivial ways. Like, oh, it saves people some time. They don't work as hard. They take more leisure. They hang out more at the water cooler. That's fine. It's fun. But it's not going to get rid of our debt problem.

1:35:26And then you have some startups. You have programming. You have the dynamic parts of our biomedical sector that are already using AI a lot. But the more they use AI, the more efficient they become. And the more the inefficient parts of your economy are left over. And it's just really hard to grow much faster. Unless you're like China playing catch up. Yeah. So the number is hard to estimate. But GDP is a huge mound of stuff, most of which is produced by people who are not neither white nor black-pilled on AI. Sure. And it will take things a long time to change. Yeah. How do you think about the legacy of Thomas Piketty and the new debate over Piketty in the 22nd century, I believe?

1:36:09Was it 21st century, 22nd century that Dwarkesh wrote about? 20th century. 21st. It's in the future. Talking about economic inequality, increasing gains to capital and increasing returns to capital. Dworkache did 22nd. 22nd century. Piketty did 21st. He's too focused on the labor share. I think real wages will go up a lot. If real wages go up, workers are happy. They might be upset that Elon or someone else is a trillionaire, but I don't think that will drive our politics. I think if real wages go up, we'll be fine. People will feel good. So my view is very different. But, I mean, real wages will go up.

1:36:49They'll feel good. But then they'll also be upset that there's now multiple trillionaires walking around. And so won't those emotions actually drive the politics? It feels like they do today. I don't think people are that envious. They're envious about the people they went to high school with or maybe their brother-in-law. Most people are not that envious about Elon or Bill Gates or whomever else. Yeah, yeah. I was, I was, envy is local for the most part. Yeah. I was running to envy each other, right? Yeah. Maybe you envy, you know, the chic, the richest chic in Saudi Arabia. I bet you don't even think about him that much.

1:37:20He does have a great car collection. Or if you do, you think about his role in the AI world, which is fine. But you're not upset that he gets however much hummus and you have less hummus than he does. It's all about the hummus. I was doing a thought Don't you think envy is a factor in the current wealth tax sort of discourse? I think California is a crazy state, but I think it's just more a money grab than envy. And I think there's a very good chance they ruin the greatest engine of wealth creation, maybe in human history. So I hope it fails. And soon, even the risk of it, as you know, is inducing many people to leave or get ready to leave.

1:37:59Yeah. What's the retrospective view on the Laffer curve in light of what's happening in California today? Well, at some tax rate, the Laffer curve is true. We're not usually in that range, but California is playing with fire and experimenting with that range. So I dearly hope they don't do it for their own sake. It'd be great for Austin, great for Miami, maybe good for New York. Yeah, yeah, interesting. New York has its own problem. What's your personal investing strategy? Buy and hold. Focus on other things. Yeah, but buy what? Because if you buy individual names, it can end up being wildly distracting, and then you can't focus on the other things that are maybe more productive.

1:38:49Diversified portfolio. I've thought now of buying some more gold just as a hedge. It's not that I think it will do well, but I see higher risk. and Bitcoin's not really a hedge, so maybe gold and silver. Again, I'm not saying it will make you rich. I'm just saying if everything else falls apart, you'll have something. Yeah. Have you pushed any AI labs to develop taste or smell technology to help with the judging of food? It feels like culinary aspects are particularly AI durable or resistant, but I don't know if that's something that can be overcome with better technology. Maybe it's cheaper for now just to use humans and have the AI record what the humans say and judge the food that way.

1:39:39That seems to work well. Judging food is not hard. It's one of the easiest problems for humans to solve. Humans will do it basically for free. How many food bloggers get paid? Well, they might get some free meals, whatever, But so many people do it for free. I just wouldn't put AI money into that for a long time. I'd work on almost every other problem first. You know, Mercore, they wanted to hire some basketball analysts to make basketball commentary better by AI. I'd rather do that than food. It's harder. Sure. That's very funny. What about this call for new aesthetics? Do you believe that we're in some sort of local minima for the development of new aesthetics?

1:40:24Do you attribute this to just broad stagnation? What are the key sources that led us to this point where it feels like aesthetically we're stunted? I think poor taste is a big problem around the world, but especially in America. You look at the older parts of San Francisco, the Victorian homes, they're beautiful. You look at the new buildings. I like some of them. but a lot of them are just awful. Or you look at parking structures, or you look at a new bank branch that goes up somewhere. Awful, awful, awful. Are we really so poor a society that we cannot afford to invest in more beauty? So Patrick and I are trying to get people to think more systematically.

1:41:03What can we do to try to make much more of our world just plain, flat, outright lovely? It absolutely can be done. is there an is there an element of the financialization or or uh i don't know like multi-party like multi-pronged uh stakeholders that go into the development of a campus or a building now that maybe didn't exist a hundred years ago where one person could just decide that they have some crazy vision for a building and they go and build hearse castle and it kind of destroys them over their life, but it's a singular decision. And now a company that builds a property or something, it doesn't have, there's not a single person that can be maybe authoritarian about the decision and the taste.

1:41:53Even if they have the taste, they get overruled by a committee of investors and stakeholders that all kind of put the kibosh on whatever they want to do. That's one problem, and there are too many veto points. But if you go back earlier in time to the 1920s, say, you look at a neighborhood like Shaker Heights near Cleveland, where the homes are being built by different families for the most part, they're still far more beautiful than homes being built today. So I don't think that's the sole main root of the issue. I think just bad taste is. Do you think there's a technology angle where it feels like in the modern era, a lot of young people lament high housing prices.

1:42:34But when I look at their choices, they'll often choose to live in a small studio apartment in a very dense city to be around other people. And when they go to their apartment, it is small. They don't have a library, but they have a Kindle. And they don't have a room to dance or something, but they have a TV where they can watch dancing. And they don't need a movie theater because they just have their phone to watch a show on or something like that. And so the number of places, you know, if you were rich, you know, a century ago, you needed a lot of spaces to do different activities. Now every activity can happen on the couch with a TV.

1:43:13Do you think that there's some effect there? A black mirror of you, John? I don't know. Well, if you do all that and you live in the West Village, you're surrounded by beautiful buildings. Sure. I'd like there to be more places in the country where you have that choice to be surrounded by beauty, that you don't just have to be all scrunched in, that you can have reasonable living space and afford it, and what's around you looks nice. Again, people have done this in the past, even the distant past, even sometimes in medieval times. So to claim we can't do it now, it's simply a failure of will. There are laws that need to be changed, procedures that need to be changed.

1:43:46But I think the first step is just to wake people up and increase awareness. Yeah, it feels like the YIMBY movement has a serious amount of energy behind it. And yet, in my lifetime, I haven't seen that much movement on it. Maybe the ADU law in California is moving a little bit, but it still feels incredibly slow to get permits. Is there a dynamic where landowners, property owners are actually pushing back against that? My feeling is that the ADU, new support for ADUs doesn't actually really increase the housing supply. Because a lot of people are like, you'll let me put another structure on my property?

1:44:23Great. It doesn't mean I'm going to suddenly, it's a new single family home, right? Yeah, I don't know. What do you think? I think the IMB movement could go much further if it could promise it would boost housing and make neighborhoods prettier. Right now, it's like, well, we're going to boost housing. Cost will fall. That's wonderful. but maybe your neighborhood will get uglier. Sure. So you do it in Buffalo. You replace the old with the new. Yeah. The new is probably uglier. Yeah. I don't like that. I think you would have a lot more successes when it can promise beauty. Yeah. Sorry. On the new aesthetics, I heard a critique or an idea that new aesthetics come from problem solving.

1:45:04And so it's less about the taste and opinion and will and more of there's a specific problem in society in a particular neighborhood, and then a design emerges to address that particular problem. And so the critique was that we won't get the new aesthetics until we identify particular problems, or maybe the bland aesthetics are solving for a particular set of problems that are in front of us. Does that resonate with you at all? Not that much. I mean, I would say the problem is ugliness. If you look at older structures around the world, especially in Europe, but often in the U.S., They can be very ornate and have all kinds of flourishes and small details that are lovely.

1:45:44Those are not solving problems. They're put there because people think it will make the thing look better. So it's not mainly about problem solving, but I'm not against solving problems hardly. Has there been any studies on the effect, the local effect of brutalist architecture? Like, does it make people stay at the office longer or anything? Because I can actually, it's funny. I can appreciate brutalist architecture when it is surrounded by nature in a big way. It stands out as contrast. Yeah, it's a nice contrast. And it feels like a triumph of man in some way. But it oftentimes will make a neighborhood much less warm.

1:46:27A shipping container home in a forest sometimes can look beautiful if you see it as a getaway. But yeah, is there any, like, what made brutalist architecture, you know, be such a force in the world? People don't like it once they live in it. Personally, as a tourist, I often enjoy it. I think it's interesting or sometimes creative. But it's not the model I would want to seek to spread through the world. Because to live with it all the time, I think it wears thin on you. Why it ever got as far as it did? maybe England is the best example. The 60s, the 70s, it's just cheaper. They're in a hurry.

1:47:08Concrete is easy to manage. You have a bunch of planners who think they know better. And a lot of it was a big mistake. And I would say maybe England got the worst brutalism. Maybe parts of communist society got the best brutalism because they really needed the structures after World War II. But I don't think it should be our emphasis moving forward. We want to move away from that. Do you think there's 8 billion people on earth? Because there's a popular exact number of people on earth. No, no, no. So, so tell me the back, no, the backstory is that there's, there's a growing conspiracy theory specifically on access this weekend, uh, movement of people that are just saying there's no possible way.

1:47:49Google says it's 8.2 billion. Yes. But this is many people, many people are, are disagreeing with China's official data. One of the arguments is that in certain countries, China, among them, there are incentives to inflate your local population number. So you get more resources from the state. You do that. You game theory that out across the entire country. And there's a homegrown effort in China from people that are trying to prove this. And they're going and they're finding these cities that are just barren. It's a city that has, you know, a thousand homes and thousands, you know, thousands of structures.

1:48:27Well, yeah, yeah, yeah. But there's like a homegrown movement to prove that China does not, in fact, have as many people as the official numbers claim. What do you think? I think there's a modest inflation there. If there's a betting market on this, I'd love to get in on it. Okay, that's good. That's good. I'm going to bet on, you know,$8 billion or more. That's my bet. Yeah, yeah, yeah. Okay. But if it's$8.1 billion and not$8.2, that I can believe. Sure, sure. Yeah, that makes sense. Back on the architecture question and the aesthetic question, it feels like when people are dissatisfied with the current status quo of architecture or any aesthetic, there's often some sort of emotional response to return to a previous era, to just go back in time.

1:49:13And I'm wondering if you think that that is the solution. We need more Art Deco or mid-century or all sorts of different, I don't know, Gothic structures or go back in time to just build new buildings that look like they're hundreds of years old. Or you think we actually need something that's entirely new and never before conceived? Everyone draws from the past. Mostly I think we need freedom to experiment. It's not that I want everyone to follow my preferred path. The city where I love what they've done with new things and somewhat older things is Helsinki, Finland. So personally, that's what I want.

1:49:52But do I want every city to look like Helsinki? Of course not. Or the modern parts of Copenhagen, I think, are quite striking and beautiful. So those would be my preferred directions. But most of all, it's about simply the ability to raise your hand and say, this is ugly. We don't want to do ugly things anymore. And we all look around for different ways of avoiding that outcome. Yeah. I mean, the risk is that you just have NIMBYs that say, I think everything looks ugly and I don't want anything to build, I guess. If you have the option to raise your hand and say, that's ugly, we shouldn't build it.

1:50:22I don't know. Earlier on the show, John was sharing how in 1947, there was a very, very small number of televisions in the US. 16 ,000. 16 ,000. Eight years later, it was 32 million. So we were talking about how that was effectively a fast takeoff, a hardware fast takeoff. Do you think we can see something similar in robotics? There's a bunch of different exciting projects, a lot of which we've covered on the show. I saw a video of a robotic hand yesterday spinning a screw perfectly. And it looks like it's sped up like 10x, but it's actually in real time. So it's like spinning a screw 20 times faster than a human can.

1:51:02Interesting. Um, but, uh, but yeah, so, so I think people, you know, see the current state of humanoids today, they can do cool demos and other robotic form factors, but, uh, it doesn't quite feel like, you know, we could have a population of a hundred thousand, a hundred million of these in three years, but, uh, who knows? I don't think we yet see the killer app in the home. So if I ask myself, what do I want? If I could have a fully functioning robot that would be like a maid or a butler, that would be useful to me. But most things short of that don't quite seem worth it. Do I have a Roomba?

1:51:39No. I don't know. What if there's a mechanical vacuum cleaner that goes around on its own when I'm at work? Fine. It just doesn't seem that necessary. I do my own laundry. It's a welcome break sometimes. So I think we're still waiting to see the killer app. Yeah. Any other technologies that you're excited about this year? Well, everything in biomedical. I mean, even without AI, progress against cancer. I think youngish people today will die of old age and maybe live to 97 or whenever the time is. And that's already in the cards. You don't even have to be a big AI optimist. I don't think it will be soon.

1:52:20To me, it seems like a 40-year process where you have some gains coming now, but it's mostly complete within 40 years, that you just won't die of most of the things that kill people. Cancer, heart attacks, there'll be accidents and there'll be deaths by old age. Yeah. That's an extreme white pill. I love it. And that's likely. And again, it doesn't have to rely on AGI though. Of course, AGI can help and accelerate that. How do you think about your information diet, your information consumption. It's obviously AI augmented, but I imagine you don't start the day with the prompt to an LLM and maybe the LLM prompts come in once you've explored or read some source text and want more context.

1:53:10Is that how you're using them? Oh, you imagine wrong. I wake up in the morning. I had questions when I was lying in bed, questions when I was falling asleep. And to ask an LLM something can easily happen in the first 10 minutes. Okay. Because I've been thinking and wondering, and I don't get up just to ask the LLM. But once I'm up, it's like, hey, I'm here. Why not? And if the question takes a while to answer, then I scroll through Twitter, email, whatever. It's perfect. You want to set your queries in motion early in the day. Yeah. I actually do find that often I'll kick off a deep research report before bed because I know it's cooking for me when I wake up.

1:53:47I love that. That's actually a great problem. Iran is in flux. Very hard to tell what the outcome will be there. But if you have a free Iran, how should people be thinking about how that would impact world markets? I think it's unrealistic to expect a free Iran. I would like to see a stable Iran, which has been rare in world history, though Iranians are often very successful. There are so many ethnic groups in the territory, and they have expanded or contracted so many times. What they need is stability and some breathing space and a government that's not one of the very worst in the world. And I think the chances of getting that now are pretty high, maybe 50%.

1:54:34But we should set our sights a bit low and not actually reach for too much and let the Iranian people over time make it as good as they can and not have it be a question of outside influence. So, oh, we're giving you this. We're making you do that. I think that's counterproductive. What about the looking back, the postmortem on the tariff and the trade war from last year? Obviously, the markets did very well, even though they were tumultuous during the process. What's the economic view on the impact of the tariffs in 2025? Not as bad as we thought, but no gain, no upside. manufacturing in this country, manufacturing employment, they're still falling.

1:55:15Our allies are mad at us. People pay higher prices. When I buy my favorite Rainier cherries from Chile, they're 9.99 a pound instead of 6.99 the year before. What's the point of that? So I'm against it. Against it. Well, we are certainly not against your appearances here. We love having you on the show, and we appreciate you taking the time to come chat with us. Thank you so much for coming on the show. Always a pleasure. Always a pleasure. I'm glad your 2026 is off to a great start. I hope you have a great rest of the week. We'll talk to you soon. Take care. Great to see you. Have a great rest of your day.

1:55:49Fin.ai, the number one AI agent for customer service. If you want AI to handle your customer support, go to fin.ai. And I'm also going to tell you about Cognition. They're the makers of Devin, the AI software engineer. Crush your backlog with your personal AI engineering team. And up next, we have Cerebrus. in the Restream Waiting Room. Andrew Feldman is the co-founder and CEO, and we will have him join us in the TVPN Ultradome. Andrew, how are you guys? He's back. Good. How are you guys doing? We're doing great, and we're thrilled to not just have five minutes today. We have a last time you came on.

1:56:27We really botched your last appearance. So thank you for taking a second shot. Giving us a second shot. Giving us a second chance. There aren't many second chances in life, but we appreciate you taking one with us. But anyway, how are you doing? How's your new year going? I'd love to just hear the state of the business and the level of optimism going into 2026. I think it's an extraordinary time to be in. Well, first, thank you for inviting me back. I appreciate it. It's always good to talk to you guys. Thanks. Look, it's an extraordinary time to be in AI hardware. I think we have seen exceptional interest.

1:57:06we've seen validation in the market we've been telling people for years that fast inference would be a separate category and so important was this new category that Nvidia spent 20 billion dollars to buy the number two player in it and so what an exciting time to see I mean you were just talking about the guys of Cognition big customer bars, and what a great product, blisteringly fast, smart as can be, great coding tools. It is what we're seeing. Maybe a way to think about it is this, is that we had a GPT moment in 2023 where people first realized that AI was interesting. And in the second half of 2025, people demonstrated it was useful.

1:58:01Yeah. Right? and that that is exploding. They're finding different uses for it. These aren't demos. These are production use cases, and they're demanding vast amounts of AI computes underneath that. Yeah, the fast inference thing is so obvious that it's become a meme. I'm sure you've seen these where people will be joking about, well, I'm coding, and while I wait for my coding agent to come back to me. I open TikTok and I scroll vertical videos. And so I saw one guy made a script that as soon as he sends his prompt, it automatically opens them and then it closes them as soon as it responds so that he doesn't get sucked into a rabbit hole.

1:58:47But it's brain rot, but it exposes a real behavior, which is that there's a lot of waiting involved in programming these days. How cool is it that other people are making TikTok videos about your value proposition? Yeah, exactly. It's like royal flush. How often does that happen? Never. An entire ecosystem is being brought to bear on how shitty the competition is to use. Yeah. And what a cool thing. Yeah. Well, since we have more time, I'd love to go back and actually get more of your full story, more of the story of the company, and sort of take us back in time to the initial idea, the first things that you did, all of that.

1:59:31So if you could sort of reintroduce yourself and the company, I think that that would be a really good level-setting experience. Great. I'm Andrew. I'm one of the founders. And with Sean and JP, Michael, and Gary, we founded Cerebrus in early 2016. what we saw on the horizon then was right there's still going to be some soundboard here the overnight success is real the overnight success it really has been right i mean people people were skeptical at very various amounts of times and i think people are finally getting it now yeah i think that when you build real things yeah right you don't get the same sort of rate of growth that you have in a viral software.

2:00:18When you actually build physical things, there's components, there's factory, there's real manufacturing work that needs to happen. Especially when you do deep tech. What we set out to do, what we saw on the horizon, was a new workload. AI wasn't the same. It presented fundamentally different challenges to the computer. And that if we were able to build a new design, we would not be a little bit faster, not one or two or three or five times faster, that we could get to 10, 20, 30, 50x faster on this work. And the way to do it was to solve a problem that had been unsolved in the computer industry for 75 years.

2:01:07And that was how to build a really big chip. Most chips are the size of postage stamps and our chip is is the size of a dinner plate right so yeah and and so um now it it took an enormous amount of work yeah to build yeah what walk me through exactly like you you incorporate the company and then you're doing design and you're calling a manufacturer and saying yeah this crazy idea and then i imagine like a year years go by until the first one comes off the line, you can actually test it. What were those early days like? They were unbelievably challenging. I think when you do really hard work, your first one is like that first pancake.

2:01:54It's sort of messed up to make the pancake and the pan isn't hot enough and the pancake is all nasty. It's like that. We had an idea that we could solve this class of problems that had broken in every other effort to build a big chip. And we spoke to backers that had vision and believed in us. and we put together one of the truly world-class chip teams in the industry. There are only about half a dozen world-class chip teams, and we have one of them. And then we laid out a plan, and we flew out to TSMC and said, we think we can do something with your process that nobody else has ever been able to do.

2:02:45and they listened and and to their enormous credit they listened and said shit that's a good idea and in the meeting they greenlit it that's amazing it was amazing what was their what what risks were they identifying what was their pushback i mean even if they say yeah we'll take your money and build this thing what are they cautioning you about at that moment well also yeah also i feel like if you're a founder you get a lot of no's sometimes it's your customers sometimes it's with investors or talent, but if you're building a chip and TSMC says no, do you just go back to the drawing board? That carries quite a bit more weight than just an investor saying, I'm going to pass it.

2:03:32TSMC is the best in the business. And so we heard no's. When you're interested in pioneering work, when you're interested in fearless engineering, you're going to hear no a lot because everybody's afraid of the problem that you're trying to attack. And when everybody's failed at a problem, then those who are medium, lack vision, are always like, it'll never work. It can't be done. Others fail. Yeah. It takes a special type of person to say, well, just because others fail doesn't mean we need to fail. The world has changed. The tools are better. We have better insight. We can use architectural techniques to avoid some of the things that broke previous efforts.

2:04:19What we did is we went out and we studied the previous failures. We came to believe that we could work around everything that had broken previous efforts. and we brought this back to our investors and they're like, wow, this could be really big. We tuned it, we tuned it, we flew out to TSMC and they agreed. I think even when you do pioneering work, you encounter problems that nobody thought of. I like to tell the story that, imagine before Everest was summited, you get to base camp and there's a team there having tea that just failed to summit and you talk to them and you go hey guys and they say there's this part about halfway up that's really really hard it beat no one's ever gotten past it right and you go up you go up and you come down and you're having tea with the same guys and you lean forward and say that wasn't the hard part right and when you do things that nobody else had ever done you encounter things that nobody else has ever encountered.

2:05:25And we had to invent materials. We had to invent packaging techniques. The problem that caused, for example, the B200 to be 18 months late was a problem of the coefficient of thermal expansion that we'd solved in 2017. We saw it. We knew it was coming for them, and we'd solved it. The problem that caused the Dojo project at Tesla to fail, we'd predicted we'd solved it in 2018 um we we had already encountered these problems we'd found ways around them and uh you know what fun is it to be an engineer if if you're doing the same stuff everybody else has done already i mean it's only fun when you're doing what were the early kind of key customers that uh because it's great if you can have a partner like tsmc that says hey you guys are onto something.

2:06:13We want to be a part of this. 2016, there's no LLMs, right? There are no LLMs yet. Who's doing AI that might be a potential customer? That's right. Early customers included a visionary group at Galaxia SmithKline, not who you think of as likely to bet on us. Truly visionary group and one of the, I think, truly great minds in the application of AI to pharma is a guy named Kim Branson, who runs AI for them. the military and the national labs, they were accustomed to understanding what could be done with blisteringly fast hardware. And so our first, our serial number 01 went to Argonne National Labs, part of the DOE infrastructure.

2:06:59And we have projects today with Argonne and with Sandia and the, I guess, Department of War, it's now called, that measure in the hundreds of millions of dollars. And so they were early customers, and they were willing to take early machines that were a little rough around the edges. And that was our first generation. Then we delivered the second generation a couple years later, and it got better and better. Then we delivered the third generation, and then it took off. And we had the rise of inference as a meaningful workload. And suddenly, performance was everything. Talk about that early critique that I heard that part of what was going to be difficult about this was the fact that when you're operating on the wafer scale, if there's one defect, you throw the whole chip out as opposed to with smaller chips.

2:07:55If there's a defect over here, well, I still get 80 % of the chips. That was the received wisdom. Remember how this works. I mean, imagine a wafer is like a big circle. Imagine it's a cookie sheet your mother rolled out into a circle. She takes a handful of M &Ms and throws it up in the air, and they land throughout this. We'll call those the flaws. They're randomly distributed. And say you can't have a cookie with M &Ms. Your mom goes like this and does a cookie cut through the entire thing. And if there's an M &M in it, she has to throw away the cookie. This is exactly how it works. Now, historically, the bigger her cookie cutter, the higher the probability she'd hit an M &M.

2:08:42And the more good silicon around it, good cookie dough, would need to be thrown away. This was one of the problems everybody said could never be resolved. And we solved it in 15 months with$12 million. It wasn't even a hard call. That's incredible. Wow. Everybody. To this day. Everybody said this, yeah. Everybody. And it worked like this. There is another way to solve that problem, and that's the way memory has solved it. Memory has lots of identical tiles. They're called bit cells. And in the array of bit cells on a chip, they have redundant rows and columns. And if there's a flaw, they map it out and use one of the redundant ones.

2:09:30Interesting. And that's it. Yeah. They're built to withstand flaws, not avoid them. Okay. And so we looked at memory and said, nobody's ever done this in compute. But if we built a compute architecture with a million identical tiles, and we took 5 % of them, and we held them aside for redundancy, we could withstand almost all failure patterns. And so we thought about the problem differently. and it was top of everybody's mind. They've just been taught that big chips have lower yield. They don't really understand the alternatives. In the same way, when Google first built their big data centers, they said, we don't want servers that are redundant.

2:10:20If they fail, we'll route around them. We don't want to pay the extra cost of making them high reliability. If they fail, we'll shut them down. We'll get new ones. We'll route around it. So does that mean that from chip to chip, there might be a slight difference in the number of flops or power that can come out of the chip? Is that like a real thing? Not that it matters practically, but I'm just wondering. We have to invent a way to communicate across that little bit of cookie dough between the two cookie cuts that your mother makes, right? She puts the cookie here and she puts the cookie cutter here for her Christmas cookie and there's a tiny little bit of dough.

2:11:02And usually what she does is she lifts all that bit of dough up and rolls it and makes more cookies later. But in the chip world, that's called the scribe line. And they run a laser across it and that's how they dice the chips. That's how they cut them. They obliterate that tiny little bit of distance. We have to invent a technique to run communication across those. And we had to use the tools that were already being operated at TSMC. And so we benefited from the fact that we had this extraordinary chip expertise, not just how to write logic to make the chip work, but the back-end design, what we call the physical design and timing.

2:11:47We knew EDA tools and we knew manufacturing of chips. And so we were able to think very differently than most companies. How does intellectual property play into this category? In so many tech startups, they say, you know, don't worry about patents, worry about your network effect. This feels very different. What's your IP strategy broadly? Deep technology companies have to worry about technology. Yeah. Why would statement of the century? You've been talking to SaaS guys a lot. Right. They worry about virality and they worry about network effects. We worry about the fact that we can build things nobody else can build.

2:12:32And that is a combination of protecting your IP with patents, with trade secrets, with segmenting your manufacturing so your manufacturers can't see. Oh, interesting. the interaction between steps. It's using all the tools and all the wisdom in your toolbox to defend your invention. Sometimes patenting is not the right. When you patent, you have to disclose. If you have to disclose something that is then very difficult to tell if somebody else used, you might want to think about not patenting it. Yeah. And so generally you want to patent things that you can then measure if someone stole. And so we have a very aggressive patent strategy, but we also have a very aggressive trade secret strategy and a collection of other tools that we use to defend our inventions.

2:13:29Talk to me about space data centers. It seems like if it does work, if the heating question is solved, you probably don't want a server rack with a ton of chips in space. That seems like more to manage, more networking. Wafer scale compute in space could be a logical extension of that. Are you excited about that? Have you dug into it? Has it been something that's been on your roadmap from day one? Actually, I was digging into it this weekend with some friends. I think the following. I think one of the real weaknesses of today's GPUs is by being little tiny chips, it has put a lot of pressure on how you tie them together.

2:14:13And this is why NVIDIA bought Mellanox. Why did it make sense? Because they knew that the individual chip would be far less powerful than a collection of chips. And if you're going to bring a collection of chips to a problem, you have to tie them together. Now, that problem is made more complicated in space, right? You would like bigger blocks up in space. I think there are a lot of hard problems yet to be solved with data centers in space. I think we ought to be working on them, but I don't think it's something that you're going to see in production in the three - to five-year time frame. Sure.

2:14:54What about on the topic of cooling? We were just talking to Jeremy from Semi Analysis about how Meta changed their data center design from this H structure that was air-cooled, very efficient, took them two years. They couldn't do water cooling. Now they're doing water cooling in their new data centers, the tents. What have you learned about various ways to cool? What's special about your product, specifically with regard to cooling? and energy management. This is not a complicated problem. I love it. I'm sure. You want to employ me? You want me to handle it? It's not complicated, right? I got it.

2:15:35I can do it. In Buffalo, at a Bills game, there are always three idiots who have their shirt off, right? It's 10 below, and they've got their shirt off. They've been drinking hard, but they're not dead. Why aren't they dead? because if you're in the water and the water's 45 degrees and you're there for eight minutes, you're dead. Why is that? It's because water's really good at sucking heat off of heat sources, right? We say in the technology, the thermal density of water is really high. It rips heat off you, whereas the air doesn't do that. So they can stand there with their shirt off and not be dead.

2:16:20it's the exact same thing when you're cooling the chip that water has an ability to pull heat off a heat source that is vastly better than air so you have two choices, you can blow cold air over your chip and you don't get the same cooling effect as if you ran cold water against the back of your chip or against the back of a cold plate And so it is more efficient by an order of magnitude to use liquid, water in particular is great and low cost, to pull heat off the back of chips. And we were among the first to do it. Google started doing it with their TPUs in 2017. It had been done previously in the supercomputer world exclusively.

2:17:12And now, you know, the new WeDoIt, been doing it for years. Gamers. I remember you build a gaming rig. You get liquid cooling if you want that extra juice. That's exactly right. Gamers have been doing it forever because they want to run those GPUs hot. Yeah, totally. And Facebook was, they learned a lesson and they jumped on it. Well, they were optimizing for one thing, which was this energy coefficient. And now they're optimizing for speed and new scale. and so they have a different set of optimization parameters and they adjust to their strategy. What else in the supply chain have you had to find partners for or develop in-house?

2:17:55I know you have experience actually building servers, but how much else changes when I'm racking your product versus an NVIDIA product from the energy that's outside to the land, to the building, to the, we talked about the cooling. What else is different? How do you solve those to actually stand up, whole data centers? Right. So we designed to fit in the exact same footprint as an NVIDIA L72. There you go. Right. And so we are underneath that envelope. We will fit in any data center designed for that. There are fewer little boxes to stuff in because we have a bigger box, but we fit in the standard racks.

2:18:39You just roll them in. They're exactly the same as you bought for your Dell servers or your Supermicro servers or your Arista switches, we fit in exactly the same. We use the same power. We require the same cooling that all high-end AI chips now use. And so we were able to sort of fit in the envelope that they carved out to make it really easy for data center operators and owners to roll us in. And then on the demand side, are customers who want to use your chips asking, hey, we'd love for you to rack these with the hyperscalers or set up a deep relationship with a bunch of neoclouds or they want to buy and operate themselves or they want you to buy to build the data center and just offer basically like an API for them or VM?

2:19:39All of them. All of the above. We have an API service. It's ripping right now. We have customers who buy the hardware and ask us to operate and manage it for them in their facilities. We have customers who buy the hardware and they operate it and manage it in their facilities. We have customers that, the whole gamut. You can buy our hardware. You can rent it by the week, the month, the token from various clouds. So the customers have sort of consolidated around either they have demand to buy hardware and operate it on one side or they're sort of under 30 and have grown up in the cloud world and don't want to think about hardware.

2:20:30They want to think about tokens and then they pay by the token. What use cases outside of the coding agents you mentioned? Cognition working with you, that one seems super logical in the sense that people are waiting for these agents to complete their really long rollouts, really long, like tons of tokens generated to build a whole app or build a new feature. Where else are you excited about the potential of not just AI this year, but faster AI, faster inference, having an impact that might be tangible to someone like a software developer or even a consumer? I think some of the use cases we've seen in pharma are extraordinarily interesting.

2:21:13I think so. That feels weird to me because if I'm developing a drug, it's going to take me three months to get mouse data. It doesn't feel like a couple of minutes matters. What's going on in pharma? I think it is exactly because the process is so long and painful and expensive that if you can rip a year out of an 18-year process where the whole patent is only good for 27 years, you've made a huge difference. And so you can run more experiments in less time. Interesting. And the probability that once you run these experiments, you have to go to the wet lab, which is much slower. The probability that the one you select out of the simulation, out of the AI, is a good one and makes sense to go to the wet lab and take to the next level increases dramatically.

2:22:03I think the act of being a researcher is really one that AI is particularly well-suited to help. You want to scan the literature. You want to make sure that whether it's in Chinese or French or English, that if there's a publication about the gene you're studying, that you're on top of it. You want to be able to scan the literature. Basically, every question is a deep research report. That's right. If you can go from half an hour to two minutes, that's a huge speed up in your workflow. That's right. What you're speeding up is the researcher. Yeah, that makes sense. You're increasing the number of questions they can ask per unit time.

2:22:40Yep, yeah. And that is a very, very exciting application. Yeah. Jordy? Out of your sweet spot, but how do you expect the memory and CPU market to evolve this year? I think, first, memory has historically been an extraordinarily cyclical market, right? I mean, it has had painful troughs and crazy high prices both. and I think right now what you're seeing is a tremendous amount of demand for HBM which is a form of DRAM I think those fabs aren't easy to build so additional capacity isn't easy to bring up and so you've had this steep upswing of GPUs which are heavily dependent on this HBM and as a result all the capacity for DRAM has sort of moved towards HBM.

2:23:42That's made traditional DRAM and servers more expensive. When it gets more expensive, the hyperscalers panic. Panic may be the wrong word, but their response is to place orders for the full year, which causes the Dell and the server bills to go, holy crap, I need to buy more. and you just watch this sort of crescendo of price exploding. You just see this huge movement in prices as everybody worries that they're not going to get it, so they place all their orders for next year early. It looks like the year is going to be way bigger than it actually is going to be. That's what's happening right now.

2:24:28How are you thinking about the financial management of the business? You raised the$1.1 billion Series G last year. Are you going to be building the company for a really long time in the private markets, taking it public? Thank you for the sound. That's perfect. Yeah, we need to hit the gong again. We need to hit the gong probably. There you go. I appreciate that. But where do you see the company going? How do you want to operate? What makes sense in 2026? Yeah, unlike some of our competitors, we sought to run a business that could be measured on traditional financial metrics. I love just taking shots on the way into that answer.

2:25:15Look, I mean, I think we do. We seek to run a business that people can look at and understand. and our gross margins are better than all our startup competitors. We're growing faster and we're bigger. And so I think that's really... Give him the foghorn. You guys got to tell me like in a pre-briefing what the various sounds mean so I know what... They're all possible. This is the heavy tanker ship. You'll know if it's the wah, wah. The happy takership is backing up the truck for big business. We love it. I think we raised the money because we had a ton of opportunity to continue to invest in the business to produce extraordinary growth.

2:26:06That's the right reason to raise money. You raise money because the opportunities in front of you are large and many and maybe have a time element to them, so attacking them quickly is of value. And so that's why everybody, I mean, if you go out and raise money from other people, that's why you should do it is because your opportunity set is ripe for pursuit. And so that's what we did. I think we will deploy it in expanding manufacturing, which is a real value in the acquisition of additional data center capacity, a collection of other things, expansion internationally. These are all things that the business is asking for right now.

2:26:54Let's hear it for opportunities that are ripe for pursuit. I like that phrase. And traditional financial metrics. I like pursuits and traditional financial metrics. I see those a lot. When in the past has it been worth a clap to say that you like positive gross margins, that you'd like to— Only on TV. You'd be surprised. We've had some people— Only on TV. We have people come on. They're trying to bait people by saying, I don't care about gross margins. I don't care about gross margins. Yeah, they do. They do. We care here. And we thank you for caring as well. What can people expect out of any big announcements in the pipeline?

2:27:29And how can people work with you? specifics. I think you can go to cerebrus.ai and you can see how fast it is yourself. I think the beautiful thing about the cloud is that it's dead simple to demo and to try. We've got a free tier. Jump on it. Play around. Ask a GLM model, which is a great coding model. Just ask it to make a video game for you. All right. And you will see in three seconds, it'll make Space Invaders for you. That will be real code. Ask it to play pool, to make a two-player pool game. Or if you're a physics nut, a three-ball interaction according to laws of physics. I mean, just ask it to do interesting things.

2:28:17And you will see right away the joy, how fun it is to engage with an AI that is truly interactive. That's amazing. And I think that's the best marketing. If you go there and you enjoy it, send us a note. We'll follow up. But play with the AI. I mean, it's screamingly fun. Just use it. Well, thank you so much for taking the time to come chat with us. We hope you have a great rest of your day. And congrats on all the progress. Enough time to really chat. And yeah, congrats on all the progress. Guys, thank you for having me back. It's very much appreciated. Anytime. We'll see you soon. We'll see you soon.

2:28:51All right, guys. Goodbye. Vibe.co, where D2C brands, B2B startups, and AI companies advertise on streaming TV, pick channels, target audiences, and measure sales just like on Meta. Well said, John. And I'm also going to tell you about Console. You see it here on the sticker. Console builds AI agents that automate 70 % of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. Let's go. We are now entering our lightning round. There you go. The Lambda lightning round. That is real. The cloud is real. It is lightning in the Ultra Dome. Nathan from Terra coming in.

2:29:33Nathan is in the restroom waiting room. Let's bring him into the TV pin Ultra Dome. Nathan, how are you doing? Very good. Thanks so much for taking the time to hop on the show. Love the shirt. Fantastic shirt. It almost looked, oh, I thought it was a jacket for a second. Very, very cool. Looks great. Please introduce yourself. Introduce the company. Yeah. Awesome. I am Nathan, co-founder and CEO of Terra Industries. I'm a 22-year-old software engineer. Essentially, we're building Africa's first modern defense prime. Wow. What got you into the industry? What were you doing before? How did you recognize this opportunity?

2:30:11Well, I began my journey in tech when I was 15. I had an accident that made me lose my sight. Overnight success. And at some point during the pandemic, I was 17 at this time, I went on to launch my first startup that was solving the education crisis, especially in emerging markets. And that allowed me to understand that basically, if we truly want to help industrialize the continent, if we truly want to help push Africa towards a massive industrial growth, we have to solve much more foundational infrastructural problems. And I started to think around, you know, what is the single problem that we could solve today that could basically cut across every industry?

2:30:57And that's, you know, what that common denominator was security. So sometime in 2004, I and my co-founder, Maxwell Maduka, we decided that we want to start terror industries to essentially give Africa the technological edge needed to defend itself, its resources, and its critical infrastructure. Sure. How are you thinking about the first products? There's a lot of defense technology companies that focus on drones or maritime or ground autonomy. There's a whole bunch of different products. I imagine you want to build them all eventually, but do you have a foothold that you want to dominate in first?

2:31:40Yeah, so for us, we're thinking quite multi-domain. The question for us is how can you protect these critical assets from all sorts of threats? And right now, these evolving threats are coming from the air, from land, and from sea. So in the area domain, we have our long-range and short-range drones. In the ground domain, we have our sentry towers. and soon in the maritime domain we'll be deploying some usvs that will be carrying out large-scale maritime surveillance all these different systems are being powered by one unified software platform that carries out data intelligence collecting all this data from all the different systems synthesizing that data and allowing these security agencies to identify and track threats in real time so this is basically what we're trying to do So you could almost think of it as kind of like the same strategy that Palantir did in the West.

2:32:30But the major difference is we also have to build out the hardware infrastructure that is collecting this data because there is no previous surveillance infrastructure that allows us to basically plug in our software platform. How does the procurement process work in Nigeria? How is there a program of record concept? Are there grants for smaller projects? What's the go-to market? You guys already have a couple million dollars in revenue. It sounds like that's growing quickly. Yeah. How did that come together? Yeah. When we first started the company, we focused mostly on privately owned infrastructure assets.

2:33:12You know, lower barrier to entry, shorter sales cycle. That allowed us to rapidly scale. So mostly owned by corporations and pyloneers. But over time, we're able to build an impressive enough portfolio that allowed us to then sell to governments and militaries. So the defense contracting model in Africa is fundamentally very different from how it's done in the West. In the West, it's quite bureaucratic. It could take you one to two years before you get to a program of record. In Africa, it's quite very network-based and also very solution-first. So essentially, no one is really expecting you to write dozens of proposals and come to office and present them like PowerPoint.

2:33:58They're expecting you already have a solution to your problem. So in a very funny way, the Android approach of taking R &D head-on and having the actual products to then go carry out live demos on the field works perfectly for the model of defense contracting on the continent. And that has allowed us to scale very rapidly. And that is something that most Western companies don't understand. But Western Chinese, they come here with papers. and they end up wasting a lot of time and not really being very productive. Do you guys have a dual-use opportunity? I can imagine people that have various assets on the continent, that could be oil and gas.

2:34:41I can imagine maybe your sentry towers are applicable, drones, etc. Even for non-kinetic solutions, just surveillance from a security standpoint, What kind of opportunities are you guys pursuing there? Yeah, we don't do kinetic at all. So we are entirely focused on building surveillance. Because, you know, essentially the Nigerian army today has enough firepower to, you know, basically destroy any terrorist group in West Africa. The problem with insecurity in Africa today is not the lack of firepower. It's the lack of intelligence, the lack of visibility. understanding where these attacks are coming from, how it's being perpetrated, looking at previous attacks to sort of like put out a pattern recognition that can help you predict future attacks.

2:35:33This is the major issue. So what Terra is trying to do is build that intelligence and surveillance layer across the continents, not necessarily building kinetics. We don't ever intend to get into kinetics. I think the sort of struggle and war that African governments are fighting against terror does not require us getting into kinetics, you know, what the religious negative community. And that kind of takes me into, like, my second point of, like, data sovereignty. Because today when we announced the round, there was a lot of, you know, viral tweets of how Palantir is now in Africa. We are sellouts and stuff like that.

2:36:16So I want to just make it very clear, you know, this is an investment from, yes, a Palantir co-founder and several other VCs. But this is the technology that is entirely owned by Nigerians, is built by us, is managed by us, is operated by us. We have no direct relationship with Palantir. We don't share any data with Palantir. So that's just something to just make very clear. That's a good point to make. Obviously, as you scale the business, you're going to need to hire a lot of people. Where are the great talent pools in Nigeria or Africa broadly? What's the Stanford? What's the Waterloo? What's the Y Combinator?

2:36:56Where will you be sourcing candidates from? Yeah. First off, you know, Nigerians are some of the smartest people you could ever meet. You know, very underrated talent base. there's already a lot of software talent on the continent, mostly because of that fintech boom of M-Pesa exactly, M-Pesa, Pasta so there's a lot of existing software talent and in terms of hardware there was basically a hidden hardware base that have been working mostly with the military, they have been doing a lot of side projects, they have never been able to commercialise their skills at scale. And, you know, what we essentially went did was we went to these communities and we hired the entire communities.

2:37:49Community acquisition. It's great. So for us, really, in scaling talent, it goes beyond Nigeria because part of our strategy is we want to build a decentralized manufacturing network. So having a factory in each sub-region. Our manufacturing base right now in Nigeria would serve West Africa. In the next couple of months, we'll be opening a factory in East Africa to serve the East Africa region. And the reason that we're doing this is, first of all, it's much easier in terms of supply chain and logistical networks. But second, the threats that are being faced in each of these regions are actually quite different.

2:38:29In West Africa, you have mostly in the Saharan, very hot climates, very desert climates. You're mostly deploying mostly aerial and ground systems. But when you come to East Africa, you have issues, for example, mostly in the maritime, with the Somalian pirates, where you would mostly have to deploy maritime systems. So you can't essentially have one centralized factory building for the continent. Africa is too big for that, too big, too complex. You have to basically use this decentralized manufacturing network. So each factory we open would tap the local base of the region it is located at. Give us the information on the round.

2:39:11You raised some money. We want to hit the gong for you. How much did you raise? $11.75 million. Congratulations. There we go. And thank you for taking the time to come chat with us here. Yeah, great to meet you. I'm sure you'll be back on very soon. and congrats to the whole team on a very cool milestone and getting a great group of people involved. You got 8BC, of course, Valor Equity, Lux Capital, SV Angel, Silent, Mickey Malka, Alex Moore. Let's go. Let's go. Well, thank you so much for coming on the show and have a great rest of your week. We'll talk to you soon. Great hanging. Goodbye. Phantom Cash.

2:39:51Fund your wallet without exchanges or middlemen and spend with the Phantom card. You see the ticker with the crypto prices in the bottom of the stream. That's from Phantom. And before we bring in our next guest, I'll also tell you about Lambda. Lambda is the super intelligence cloud, building AI super computers for training and inference that scale from one GPU to hundreds of thousands. So up next, we're going over to LM Arena. Anastasios, welcome to the stream. How are you doing? Hey, brother. How's it going? Nice to see you. It's going fantastically. It's going extremely well for you. You raised some money recently.

2:40:27Let's kick it off with a big gong smash. How much did you raise? $150 million. $1.7 plus. Pretty good stuff. Great stuff. Great stuff. What do you do? How do you make money? Our friend was just joking around. This account near. They were saying LM Arena raised at a$1.7 billion valuation, and they were using the Michael Burry graphic. Very rude. Very rude. with friends with you. They're just joking around. But it's okay because you're here to tell us why that's the deal of a lifetime. Yeah, explain. Listen, evaluation's a big problem. Evaluation's a big problem. People don't know how to solve it.

2:41:07You have all these different AIs, the space becoming more complicated. You have different modalities, different models that are doing, all these people competing in coding, competing in software engineering, competing in the general chat. You have models from China, models from the US. Who are you going to pick? Let's say you're a consumer. You're a developer. Even a business, you're going to make a decision. You're going to make a decision. You're going to spend millions and millions and millions and millions of dollars on that. You've got to make sure you make the right decision. Yeah, procurement.

2:41:36Procurement's a big market. And all these problems are huge. And the ultimate problem is how do you help people find the best solution for them? And as a neutral platform, of course, it needs to be separate from the labs. It needs to be a third party that does this. We're positioned well to attack it. So I imagine, I mean, independent. We were joking around that if you ever wanted to return the capital, you could get one of the labs to pay you$10 billion just to put them in permanently hard-coded at the top of the rankings. You'd be out of business in 12 months, but you'd walk away with a pretty penny.

2:42:07That's the thing. I don't think that's what's happening. How do you make money? It loses the value if you do that. Yeah, exactly. So how do you make money then? Well, what we do is we help labs and enterprises get analytics on how well their models are doing. Help them understand the strengths and weaknesses of their models in different domains, like math, coding, instruction, following, multi-turn. Think about it like a full-body scan of your model. And what distinguishes us from the standard benchmark is that it's all live, so you can't overfit it because there's constantly new data coming in.

2:42:34Oh, interesting. And it's on real users. So we have tens of millions of real users that are coming to Alamarina using Gaf for this huge diversity of different tasks. And that's, of course, the data that's used as the basis for these analytics. Yep. Talk to me about the proliferation of different LM arena categories. I imagine there's an exponential curve there as different models have different capabilities. We're not just testing math and science reasoning and chatting ability. There's so many of these. So how fast is it growing? How are you setting up your system to handle potentially tens of thousands of different benchmarks and models?

2:43:14How is that growing and scaling? That is a great question, and I think that we're honestly still developing what that big roadmap looks like. But that being said, here's where we're starting. We have a bunch of different categories. I mentioned a few earlier, like math coding, instruction following. But we've recently also dug in a little bit more to look at different types of users, basically almost like user research. Expert users. What are experts saying about all these different models? We have a leaderboard for that. Occupational categories, law, medicine, business. which models are best for these different categories of usage marketing.

2:43:49We have that information on Almarina so that it's a one-stop shop for you to understand the different industry, economically valuable industry-level performance of these models. But if you really think about it, the long-term vision of this sort of thing, it's almost like every individual should have their own evaluation. Which model is best for you? The technology brothers. Technology brothers eval. It needs to have its own kind of thing because you're doing your own tasks and they might be very different from what I'm doing. And so eventually the future you can imagine of building a technology like this, we're building the infrastructure to do this, is to have evals for every individual person on earth.

2:44:24How do you incentivize people to participate on the voting side, the ranking side, that human element's really important in terms of grading different LLMs? What's the current incentive structure? How's that evolving? What are the challenges that come up with running a multi-sided marketplace like this? very good question uh incentives are really really important if you think about incentive is almost like a little hack into the human brain that tells you how to move how to operate yeah right so that's why games are so powerful games can get you into an incentive system that's just like hey i'm flappy bird or i'm doing my candy crush and i'm just swiping swiping swiping we do not really incentivize users on elm arena to vote and that's part of the power of the platform.

2:45:10Unlike, let's say we were to pay people to vote, we don't do that. The reason we don't do that is because we want them only coming to do their real job. They come because they get value out of the platform and they vote only because they want to, because they're intrinsically motivated, because they've got something to say. You can imagine if you incorporate incentives that it might hurt the leaderboard. So we're very careful about the way that we go about that. However, we are exploring how can we design the incentives in sort of incentive-compatible way in order to preserve the integrity of the leaderboard while also rewarding people to vote.

2:45:44That's on our minds, but not yet done. How confident are you that you can encode big model smell into a – you know, the taste, the vibes into a ranking? Well, if you look at the leaderboard, it's there. It's there for you to see. Yeah, I mean, it does reflect it. And it goes to show how powerful human preference can actually be, that people are seeing things that you wouldn't necessarily anticipate. And that's why when, you know, model developers develop their model, the big problem is when I put it out into the real world in the wild, what's going to be the performance like? I don't know unless I see it.

2:46:24And that's because people react in these, you know, strange ways when they see a model for the first time. That's part of the power of this platform is that it puts it in front of those people. And then we give analytics to try to understand how different individuals, you know, what are the different usage patterns and who's voting for what. So you can code things like big model smell, identify, you know, but it's not uncommon that model providers will find out through us that their models are really great at math or really great at coding. And they didn't even actually ever know. What does it take to get a model on LM Arena?

2:46:56Is this something that the lab is giving you preview access to? Is there an application form? Do you already have relationships with these folks? So we have a contact form on our website that you can use if you want to, you know, submit a model. Of course, we have capacity constraints, but we try to be judicious and let, you know, everybody participate in the battle. And to, yeah, in terms of paying. Someone was asking DeepSeek for. Pay to get on the leaderboard. DeepSeek R4. What does it take to get a model like that on the platform? We'd love to get DeepSeek R4 on the platform. I don't know.

2:47:35Maybe it's already in progress. Okay, yeah. How much does hardware matter here? We were just talking to Andrew from Cerebris about the speed that comes from that. There's an element where plenty of people would rather use a dumber model if it's 10 times as fast or vice versa. And so you have to create an apples-to-apples situation, most likely, or at least pre-cash the results so then you're not noticing speed. But do you think that speed tradeoffs will become a bigger piece of what you do? For sure. Yeah, there's no question about it. Really, the ultimate question that people have is how do I, for myself or my organization or whatever, choose along the period of frontier of speed, performance, and cost?

2:48:23Yeah. Those are the three things. And usually it's performance and speed, number one and number two. Because everybody's spending so much money right now that it's like, you know, ring the gong again. So much money. There you go. Go to gong. Money being spent, baby. Now, the issue with that is there's gong twice. Let's go. The speed, if you don't equalize it, you can get bias in the ratings. Yeah. So we want to be able to disentangle speed versus performance, and that requires us equalizing. But if you go to direct chat mode, you can use the Model C just as it is, and we're planning on expanding the leaderboard to get more of the speed in there, the latency, as well as costs so that people can make all those trade-offs right on our platform.

2:49:06Yep, that makes sense. Would you ever do anything at the application layer, or does it just get too chaotic because UI is a new variable and there's all these other factors? I mean, this is application. Elmarine is an application, no? Yeah, sure. But I'm talking about like, you know, two legal AI tools and trying to, because if you go down the procurement route and that becomes, I don't know, I don't know if that becomes a big business. Yeah, you could see like a G2 cloud type of business here as well, although I have no idea if that makes any sense for you. Totally understood. So the thing is that it's hard to do that for our users, right?

2:49:44Because then we need to build 10 different product services or 100 different product services. And that's one of the exciting things about the ecosystem right now is that it's actually the fundamental product surface where things are evolving very quickly and also a lot of value is being aggregated. There's a reason why number one revenue stream of open AI is consumer. It's a consumer application. And other companies have different strategies, but ours simply cannot be to replicate every application. so the value that we hope to provide again is to help people understand the different trade-offs of different models evaluate them for their use cases procurement so on and so forth and that might mean helping enterprises link together with their feedback understand their users better perhaps warm starting from the large user base that we have and giving them those analytics and tools to help them make decisions that's how I can imagine us moving into the application workflow Will you ever create a romance benchmark?

2:50:41I think it's a good idea. How quickly, what models make people fall in love? Comedy benchmark would be hard. Let me ask you, are you asking that for any specific reason? Got him. Got me. We talk about comedy, comedy bench. It's difficult. It's something we try to test internally. They usually end up being funny because they're so not funny. Yeah, it's pretty brutal. It's pretty brutal. Or if it is good, it's like recycling. Like clearly it went to Reddit, copied the top joke, and then just regurgitated it. Well, did it ever make you laugh? Has it made you laugh before? Yes, but accidentally. But when it's accidental, it's amazing.

2:51:20It's the best. I see, yeah. Where did the idea for this come from? Like what was the inciting moment to start the company? Well, so there's to start the company and then there's the idea. I've been personally working on Almarina for almost three years now, along with Wei Lin and Yon. And there was actually a big group of students at Berkeley. This came from an academic project. I was doing my PhD on like theoretical statistics, theoretical machine learning, you know, kind of abstract. I was in a basement proving theorems. I had no idea that I would be. He built this in a basement with scraps. He built it in a cave with scraps.

2:51:54Seriously, there were rats in the ceiling. No way. Not kidding. I'm glad you have$150 million for a new office. Hopefully it's nice. Go Bears, baby. And so what ended up happening is that I got looped into this project that was happening at Skylab, which is, you know, Jan and Joey Gonzalez and Luca and all these people at Skylab were working on it. That's the lab where Databricks came out of and AnyScale and Ray and so on. And they were working on, like, how do we evaluate? Early days of ChatGPT, you know, there was Model A and Model B. They were doing, one was doing better than the other on the benchmarks.

2:52:27Then you would chat with them. And it's like, hey, these benchmarks are not reflective of how good it is at talking to me. So how are we going to measure that? It was in the context of one of their early open source models called Vicuña that they had developed. And so the sort of pairwise preference, hey, chat with the models and see which one you like better, that strategy emerged from there, but it started with Amazon gift cards. It didn't start from organic usage. It started from passing around Amazon gift cards to people at Berkeley. And then it sort of popped and then it went down and kind of sat at 30 users a day for a little while and then it grew and grew and grew in large part due to Wei Lin's sort of consistent effort in Twitter and so on.

2:53:05Awesome. Up until the point where it became this juggernaut. A true overnight success. Juggernaut mode. We'd love to see it over here. And thank you so much for taking the time to come chat with us. Yeah, great to have you on. Congrats on all the progress. Hey, great to see you guys. Thanks so much. Have a great day. Yeah, you too. Have a good one. We'll talk to you soon. Goodbye. Turbo Puffer, serverless vector and full-text search built from first principles and object storage, fast, 10x cheaper, and extremely scalable. Team out of Shopify. Yeah. Which is relevant to our next guest. It is. It is.

2:53:34We will bring them in in a second. But DeepSeek founder is running a hedge fund as well, in addition to building DeepSeek. Guess how much they're up. 57 % last year. Massive year. That's high flyer. Boasting the potential. War chest for a company that's already shaken up the global tech landscape. Is this good for the AI? It's good for the open source community. It's good for academics. What do you think, Tyler? Oh, I have some other breaking news. Give me some breaking news. Okay, so Anthropic, they have Cloud Code. They launched Cowork, which is Cloud Code for the rest of your work. No way. So it's like everything else.

2:54:11That's crazy. Yeah, it's exciting. It's basically a local app. Oh, it's a local app, so you don't need to do the terminal stuff anymore. You can just use it in an app with a prompt box. Yes, and then it can interact with all your local files, whatever. This is going to be really, really big. I don't know if you guys have used the Cloud Chrome extension, but it's super good. I have. The computer use is super good. So yeah, this is very exciting. Okay, yeah, yeah. Get ready for some threads, people. People are going to be breaking it down on all the fun things they had to do. I fixed my Wi-Fi in under an hour by rebooting it.

2:54:44But no, seriously, I was listening to Doug O 'Loughlin from Semi Analysis talk about how he uses Cloud Code in a knowledge work setting, and it's fascinating. So he'll kick off one deep research report about one company that he's researching, then a few more, and then he'll do a deep research report on top of that. But instead of it all living in the clawed web UI, it's just creating markdown files that then he stores in Obsidian, and then he can run these meta deep research reports on the other deep research reports that he's put together, interact with whatever's going on in the semi-analysis, private data world, all the data that they've collected, interact with their Slack.

2:55:21They have a Slack bot that interacts with it. And so he was talking, he was very, very one-shot by Claude Code and was saying, everything is a skill issue now. Everything is a skill issue now. Much like generating AI voices. But thankfully, we have 11 labs. Build intelligent, real-time conversational agents. Reimagine human technology interaction with 11 labs. Well, Star Wars has already begun in reality. Ask a ski resort in China. there is a robot that's skiing. This feels like it should be AI, but I believe it's not. I believe this is real. This is remarkable. You don't have to go skiing anymore, thanks to AI.

2:56:03Cancel the skiing trip. I have a ski trip coming up at the end of February. Canceled. Canceled. I'm sending this guy instead. Yeah, just watch the video stream. You don't have to put on the boots. Realistically. You don't have to get cold. You don't have to sit on the lift. You don't have to be in lift lines. Yep. You've seen the drone that will follow you as you ski and takes really cinematic video. Yeah. How could it be more cinematic? Get the camera at waist height. And you can go way harder. Hit the double black. Hit the cliffs, right? You don't even have to worry about rocks, right? No, he's just tumbling down on the skis.

2:56:38This is fantastic. I love the idea of having a robot follow me with a camera. I don't know what else you would use this for. Ski patrol? Surveillance? I don't know why they built this other than it's really cool and fun. But the obvious use case is Red Bull viral videos, probably. Yeah, so the use case here is obviously this is the new ski patrol. There's thousands of these on the mountain, and they monitor your speed in real time. And if you're skiing recklessly, it just skis into you and explodes. and so instead of ski and really instead of uh instead of getting your past pulled you just get your life pulled that's over sense um so anyways that is that is a dark uh sci-fi future i have a prediction from guillermo from versell but first let me tell you about railway railway simplifies software deployment web apps servers and databases run in one place with scaling monitoring and security built in uh guillermo says there's definitely a non-zero chance that we get quote generate your own GTA 6 in a few minutes before GTA 6.

2:57:47I think that mechanically it's possible, but the humor is what makes GTA so special. the writing and the character development. I mean, also the mechanics are very unique. It's not just a walking simulator. I don't know. Have you played GTA? Yeah, I mean, but Sholto's game looks pretty good. He put out a screenshot of it. I trust Sholto to make GTA. Are you sure it's not just an image gen outfit? He's like, look at this crazy game I made. I mean, the Grok folks are known to put up some viral videos espousing video game development. But clearly video game development is here. It's going to happen.

2:58:22but we have our next guest and guess what ad I have to read before he comes on Shopify Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store on mobile, on social and on marketplaces and now with AI agents and that's exactly what we're going to talk about we have Harley from Shopify in the TV How'd we do? How'd we do? Was that planned? That was amazing No, it's somewhat random but we timed it up perfectly but we're very excited Speaking of random I've never done an interview in my life I'm at NRF, the biggest retail show in the world. There are 40 ,000 people here.

2:58:55This is a Shopify booth behind me. There's going to be random people walking by. Amazing. I see that snowboard there, that little Easter egg. That's how it all started. That's how it all started right there. Wow, I love it. I love it. So you're in person, but today we're talking about - Where's NRF? Is that Vegas? Javits Center. Javits Center, New York City. Cool. That's fine. It's sort of the largest conference actually on the planet for retail and commerce. It used to be more about the traditional retailers, but we've sort of hijacked it and made it about modern retail, direct-to-consumer technology companies.

2:59:25You're kind of behind enemy lines there. I mean, you do have a POS system, so it makes sense. We have a POS system. But actually, and they've given me the keynote four years in a row, which is pretty cool. Not bad. And today for my keynote, I brought up Emma from Skims, which is amazing. And Ben Francis from Gymshark, we talked about basically this idea that there'll be more billion-dollar brands built in the next 10 years than the last 100 years. And both Emma, I mean, think about Jim Shark and Skims. They're like 10 years old, and they're dominating. So it was really fun. And we made some big announcements.

2:59:55Yeah, I mean, what's the key to building a billion-dollar brand if you're starting in 2026? It feels like it's incredibly crowded, and yet there's a new breakout brand every year. What advice are you giving to people to get started other than just hop on Shopify and let that take out a lot? There's a lot. I mean, there's so many. And companies are growing. the way the bar has been set with new e-commerce brands is insane. Best in class is like, okay, you're doing $200 million in your second year. I know exactly who you're talking about. And there's so many examples of this. Wow, you guys are hard graders.

3:00:29Doing$200 million and no gong for that. Of course we're hitting the gong for that. So let me say something. I brought on, my first time I was up on stage, I brought Richard from Mattel, Richard Dixon, who's now CEO of The Gap. Because basically what he'd figured out was that Mattel has this treasure trove of IP in their vault. And he was commercializing all of it. And so it was amazing that a company from 1945, I think, was basically dominating. The next year I brought Cat Cole from AG1. Oh, yeah. And obviously you guys know the story well, but one SKU,$600 million. Right? Like unbelievable. Another gong.

3:01:09Another gong. I mean, this could be a lot of gongs in this one. And then last year I brought someone you know, Sarah Foster from Favorite Daughter, who's been on the show, who effectively, I mean, Sarah and Aaron created a hit Netflix show to sell more apparel and succeeded with it. So this idea that, so I think one major thing, if you look at all of these different incredible brands, what they've done is they didn't just take a bigger piece of an existing pie. They effectively were TAM creators. I mean, shapewear existed. There were companies like Spanx out there, But the way that Kim and Emma built Skims is totally different than what Spanx is.

3:01:45It is almost like a cool version of an old school product. And then we announced four big things, which I'm very proud of. The first is that it's fairly clear. Sorry, John. Something went down the wrong pipe. Crazy. This show is so unhinged. Also, it's at the end of the show, I think, right? So get more unhinged as it goes. I'm like five Diet Cokes deep at this point, so anything can happen. Anyways, John was doing such a good job keeping it together. You got me. You got me. That's okay. That's no problem. Sorry. So the first thing we announced was actually Sundar announced it yesterday on stage, but with Google, we are making it possible for every agent to connect to every single merchant.

3:02:28we created this protocol called Universal Commerce Protocol, which effectively says universal language is open sourced so that all merchants can speak directly to every single one of the agents. And the best way to explain it is up until now, it was really just about like a single transaction. So I can buy something on ChatGBT or Gemini or Microsoft. It was like a single, you could only do like a single item at that time. It couldn't really be able to do that. But there's no concept of loyalty or subscription or bundling or if it's furniture, for example, please don't ship it to me on Thursday.

3:03:00I'm at home Thursday. Send it Friday. So this idea of creating this universal protocol that we co-developed with Google means that now merchants can actually tell these agents exactly how to show their products on these agentic tools and it should be as good as it is on the online store. So that was a really, really big one. The second thing we announced also with Google is that now we're actually expanding. You can sell everywhere commerce is happening from an agentic perspective. So we're going beyond the agentic storefronts of just chat GBT, which is what we said in Q3. Now it's also we're going to be working with Gemini with AI mode in Google search and also with Copilot.

3:03:39And maybe the last one, which I wouldn't bring up on any other show than TBPN is that we're actually bringing agentic commerce to every brand, whether or not they're on Shopify. So if you're not on Shopify, but you want to have your product syndicated and index, you can do so with our Argentic plan. And that seems to be the talk of the show, which is amazing. Oh, very, very interesting. Right, right. Because that's... A little hug in this ecosystem. I like that. I like it. Yeah. I like it. I mean, if you're going to... I think if Argentic is going to do what a lot of us think it's going to do from a commerce perspective, you have to give consumers all the brands.

3:04:14We obviously want them all on Shopify, but there's some brands that want to participate now, but it may take some time for them to migrate over. So this idea of opening up to anyone we think is a big opportunity. Very cool. Walk me through the customer experience journey that might shift over this year. We were following the Agenta Commerce thing so closely that we were cheering for ads in ChatGPT back in June when they were kind of like, maybe we'll do it. We were hoping that Black Friday would be the day that people were purchasing things in LLMs. Didn't quite happen yet. It's still in the experimental mode.

3:04:50I was using the shopping feature in Chachapiti to do shopping research. John's saying we want the commerce fast takeoff. Yeah, but I'm wondering, like, the early adopters, is this going to be a more techie community that's buying their next TV in an LLM? Do you think it'll jump straight to the Skims and the, you know, the Takovas category first? Like, how do you see this actually rolling out? Oh, look at you with all your direct-to-consumer brands. Way to go. I got it. A couple of things. I think it'll likely be something that most people use some of the time and some people use most of the time.

3:05:29I don't think it's going to cross the threshold of most the way e-commerce does now. Yeah. It's just going to take time. It's going to take some time. I do think, though, that these are on very, very small numbers. But if you look at the velocity, I was asked as I was coming on stage today to do the keynote. The keynote after me was the CEO of PepsiCo. And he said, you know, last year you talked about, I asked you if, if, um, uh, if a gent was overhyped, you said it's underhyped. And I, and he said, well, you know, and I said, do you still believe it? I said, well, if you actually look January, 2025, the last NRF to this NRF, we've actually seen a 14 X increase in orders to Shopify stores coming from some sort of agent.

3:06:08Now that doesn't mean the transaction happened. the link wherever it was they may be going to the online store but 14x is is dramatic and it's on a very small base we processed 90 billion you know last quarter so it's a small base still but i think it's going to wrap up quickly what i think is really cool is i think you're going to see companies that are going to actually have these drops i think once you start seeing these like for example um last night i did an nrf event with tom sacks who is the designer behind the mars yard three shoes, which I think is the coolest sneaker in the world. He runs Nike Craft.

3:06:41Imagine he actually dropped something directly on one of the agentic applications. Now you obviously have people that go there specifically for it. You now may decide, okay, that experience is pretty good. I'm now going to book my entire ski trip, everything I need for my ski trip, directly in the application. So once you have a good experience, I think the actual friction reduces. You'll keep having it over and over again. But the thing that we felt was missing, and this is the reason why I think this UCP protocol is so important is it was very difficult to do merchandising inside of these applications.

3:07:15And this protocol allows you to do a lot more of that. You can say, like back to Cat Cole and AG1, it's obvious that anyone who studies the company, subscriptions play a huge role in the AG1 business model. Well, up until UCP happened, you couldn't actually do subscriptions. Now you can. Or this idea of bundling. For Gymshark, it's a huge part of their business is if you buy these, you'll also buy these as well. You can do that as well. So I think all of these things are sort of in line with creating a much more delightful experience in the chat. And I think ultimately what it leads to is like, this will be merit-based shopping, which will be different than I think some of the traditional retailers who were kind of leaning on their balance sheets to spend money on ads.

3:07:56You can't really game the system in that way. You actually have to be, from a context perspective, the right product for the right consumer. How do you think creative assets will change or maybe stay the same? I've noticed that a lot of LLMs, when you pull a link, it'll actually hydrate that link with an image of the product that you're looking at. I imagine that a lot of people were previously thinking, well, everything's going to collapse down to text. I just need the facts and the data. But now maybe the creative, the photography, the videos, that actually just gets pulled into the chat apps that folks are using and it stays important?

3:08:33How are you thinking about that evolving? I think you're probably, you're not going to, I mean, the idea of SEO won't exist in a Gentic because again, it's merit-based and it's mostly based on the context history you've had. But I do think though, you're going to have, these brands are going to have people at their companies who are thinking a lot about like consistent updates to UCP, consistent updates to the catalog. So they may pull something off the catalog and say, we don't want to sell it anymore this way. So I think there's going to be, I don't know if there going to be actual jobs, but there's going to be people inside of the company, potentially in the merchandising department, who say, actually, the way that we want to sell this, the way we want to describe this to these agents is a particular way.

3:09:09And then because of UCP and because of Shopify catalog, it gets easily disseminated across every single one of these agentic applications. So the experience just gets better and better. I think you have to be a little bit of a techno optimist. You guys obviously are certainly there as I am to believe that even if the experience is not incredible right now, it's likely just going to get better at this ridiculous pace. Yeah. Are the folks that you talk about, uh, the brands that you talk to, the leaders of those brands excited about, um, ads coming to, uh, chat surfaces. It feels like new unexplored territory.

3:09:42Whenever there's a new ad product, that's opportunity for fast moving brands, but, uh, what are they worried about? What are they excited about? What's the mood? most of the excitement is actually around this idea of like, is there a potential for this to level the playing field? Meaning, you know, if I've done a bunch of research historically on an agentic application about James Purse or about, you know, Tom Sachs or about the stuff that I love, the brands that I love and then I'm going on a hiking trip it probably should not show me a generic pair of boots. It should probably show me on running because it knows to some extent that I actually favor direct-to-consumer brands or more modern retail brands as well So the excitement actually is around like, is this going to introduce more brands that otherwise are unknown to more people or, you know, True Classic Tea, for example, which, you know, if you're looking for a black t-shirt, I suspect on a search engine, you're not going to see True Classic Tea come up that much, but it's an incredible product.

3:10:34And ultimately, it can be found on these agentic tools in a way that it probably couldn't historically. That seems to be a lot of the excitement. The other excitement is e-commerce as a percentage of total retail is still sub 20%. You've heard that literally, you hear that at the show all the time, that people are like, well, you know, we're inside the Shopify booth now. And people that are more traditional say, well, when's this going to eclipse 25 %? There is a chance, there is a school of thought that says that one of the things that a Gentic might do is increase that penetration rate at a faster clip, kind of a la, you know, COVID did.

3:11:06because people that may not be big e-commerce consumers may be searching for recipes on ChatGBT and through that process may end up actually starting to buy stuff there as well. So that could be exciting too. The other thing is like, I think some of the time you go to a retail store, you're going because you want to have a conversation about the product in a way that you just don't get with a website. A website is like a beautiful catalog. And if you go around in the right order and you look at the FAQ and maybe you talk with support and hopefully they respond quickly and actually give you the insight, then you can get to the purchase decision.

3:11:43So a lot of people will just be like, I'll just drive by the store and get it. But I feel like LLMs can provide that. They can give people understanding on a subject or a category or a specific product that gives them the confidence to make a purchase. Actually, I'll go one step further. I think it's actually a better version of that because it's an unbiased discussion, an unbiased conversation. I mean, the bias is based on your context history and based on merit, as opposed to being based on, you know, what spiff is that particular salesperson getting today in a commission structure? So it does, I mean, you know, it's so cheesy, of course, but like when I describe it to like my mother, where this is going, I explain that this may very well be the greatest personal shopper in the history of retail.

3:12:23who knows everything about you, knows your sizes, knows in certain, you know, in European fashion, it's this size, but Americans like that, knows where you're located, knows where you travel to, and can provide this unbiased, objective recommendation. And then once you build in these applications, you build in checkout, which we're doing with our Argentic storefronts, right into the chat, it becomes this really delightful experience. Yeah. Well, congratulations to all the progress. I think it will be, yeah, the most transformative year for the way online commerce happens. How can we follow the numbers here?

3:12:58I can keep coming on and telling you the numbers. I will tell you that I did say on stage about 5 ,000 people, but an hour ago, that I do think this is a golden age of retail. And I think you're going to see brands. There are brands right now that none of us have heard of that in six months from now, we'll be like, I can't believe we didn't know that brand existed. And I think that type of velocity, there's going to be so many more of these skims and Gymsharks and Aloe Yogas and Viore's and Tecovis and AG1s and favorite. We know these brands because we kind of live in this world. It's what we're interested in.

3:13:30But for the most part, most consumers don't know. And I think they may end up knowing them a lot more readily through the agentic evolution. And Shopping wants to power the whole thing. Yeah, it's going to be so fascinating following this ramp relative to mobile in previous booms. Or actually social commerce would be another one. Because there was uncertainty around it. And then ultimately, actually, social is a wonderful place for discovery. Yeah, there were a lot of experiments where you'd be able to shop under YouTube videos or under Facebook marketplaces. Concurrently as well. It turned out it was a great place.

3:14:00Yeah, exactly. That's right. It was a great place for discovery. But the transaction still took place on the online store. This is different because a transaction happens embedded. Yeah, yeah, yeah. It's going to be fascinating to follow. I think we're in for a fast takeoff. It's going to be a good year. Love it. A great time to be in business and a great time to be at Shopify. So congratulations on all the success. Thank you very much. You guys are the best. Thank you for having me on. And I like to see this live broadcast from wherever. Fantastic. It's going to be a massive year. Have a great rest of your week.

3:14:24We'll talk to you soon, highly. See you. Goodbye. And you know their other partner, we've got to tell you about, Gemini 3 Pro. Google is powering all of that. It's Google's most intelligent model yet, Gemini 3 Pro. State-of-the-art reasoning, next-level vibe coding, and deep multimodal understanding. You can build them right there. We talked about this last week. California State Reservoir levels are now at 130 % of normal, holding 7 trillion gallons of water. It's remarkable. Allie Bell says, God wants us to feed Claude. And every model, every data center, there's plenty of water to go around.

3:15:01And Ashley had the same idea. Ashley Vance says, God is rewarding us for building AI. All the water people are super happy. Tyler, you have some breaking news? Yeah. So earlier on the show, I was reading into the Jack Clark blog. I was like, oh, maybe Jack Clark is pointing to something that - space data centers? Anthropics can build space data centers. I posted that, he responded, put me in the truth zone. He said, no, you should not be reading into this or any anthropic grant strategy. And he totally ratioed me. Oh, wow, got destroyed. Yeah, so pretty brutal. Wow. Brutal mogging. Brutal mogging.

3:15:36Oh, wow. So no, yeah, no anthropic data centers in space. Well, I mean, it's funny, I was talking to Andrew from Cerebrus and I was kind of giving him the layout. Like, I'd heard from other people that Cerebrus would be a unique beneficiary of space data centers if it happens. And he was like, nah, it's not really on my roadmap in the next three or five years. But we will see. You know, everything could be pulled forward if there's, you know, new advances there. Space, you know, payload to orbit comes way down. There's a bunch of ways that that could be exciting. But all over. Hadrian raised additional capital at a$1.6 billion valuation.

3:16:14valuation their series c2 this was

3:16:19love to see it they got t-row price altimeter d1 stepstone we got to get 1789 founders fund lux capital a16z 137 deli and company all all the absolute boys yeah back in a run from chris power uh what else we got uh this post here from horse hater horse hater went very viral yesterday i didn't uh i didn't like this post the whole post is horses don't like boom we love horses here horse hater was asked why and horse hater replies they are evil creatures i don't know that they're i i know that there is only malice in their hearts completely false Fake news. Just exposing yourself for not spending time.

3:17:08Why is there no community note on this? We should community note this. I'm going to write a community note right now. I'm going to get kicked out of the program if I do this. Horse hater hates horses. Huge surprise. Alexis Ohanian is sharing an article, luxury watch prices hit a two-year high in the secondary market, partially due to and maybe almost fully due to tariffs. Oh, yeah. Tariffs were a big part of the story. So if you like spending more money on watches, you'll love tariffs. Yes. But Alexis has a great anti-AI bet I've been making is in human craftsmanship, human excellence, collectibles will keep thriving.

3:17:48So if you needed a reason to buy another watch, this is it. This is it. But it's interesting because he frames it as an anti-AI bet. It's sort of an anti-AI bet in that it's not a beneficiary of AI, but it's still, it's a very AGI-pilled bet in the sense that in a moment when AGI-AI is accelerating, maybe human's craftsmanship is more valuable than ever. And that it's sort of an anti-slop, but it's still aligned with a bullish view on AI broadly. So he's not saying, I don't think AI is going to work, so I'm buying watches. He's saying, hey, I will work, and so I'm buying watches. Yeah, I mean, humanoids are going to have to be stunting.

3:18:31They will. They will need something on the wrist of your humanoid, and why not an FP Journe or a Patek Philippe? Well, yeah, so it's interesting. Watches, Babylon goods, right, more desirable as they get more expensive. That means that if a robot can make a Swiss watch, like right now, Rolex might cost$2 ,000 to make, right? Some parts are automated. Some is, you know, handcrafted. And suddenly Rolex can make the watch for, you know, 50 bucks just due to automation. They're not automatically going to say like, okay, let's bring the prices down, right? Because the price is part of the product. So I do think it's a good piece.

3:19:13There is a little bit of a potential scoop here. Bobby Goodlatte replies to Alexis Ohanian here and says, yep, I love that both Zuck, we all knew Zuck is into watches, and Sam Altman are both into watches, shows that they still value handcrafted objects. I didn't know that Sam Altman was into watches. It makes sense. He's into cars. We know that. But we'll have to figure out what's on the wrist. Does he do like a Jacob and Co on each wrist? Truthfully, I've not seen a photo of Sam with a hitter on the wrist. I've seen him in the Koenigsegg. I've seen him in the McLaren F1. I know that there's a car collection there.

3:19:50but I haven't seen anything hit the timeline of a wrist check. But maybe 2026 is the year of the Sam Altman wrist check. I'm sure his PR team would love that going out on the internet. But I would support it, and I will fight for you. I will be your strongest soldier, Sam, if you want to drop a little photo on the Instagram of whatever's on your wrist. Anyway, fantastic show. Great show, folks. Great show, folks. Thank you for hanging out with us today. Well, there's one last scoop that we should talk about is from Kylie Robison over at Core Memory with Ashley Vance. She's got another scoop. This was just from the weekend.

3:20:29We didn't talk about this. But XAI staff has been using Anthropics models internally through Cursor until Anthropics cut off the startup's access this week. Everyone's using Anthropics, apparently. So this went viral, got 7 ,000 likes. and yeah I think at this point if you are a lab and you haven't had your access to anthropic cut off yeah polish up your resume because clearly they're just going to do this to any company that that is a meaningful threat yeah there's another dynamic where you can be using the model because you think it's good but you can also be trying to distill the model because if there's good data that can be pulled out you you might just say okay we want this capability that's uniquely available at this model endpoint.

3:21:11So let's grab that. There's a whole bunch of reasons why this could be happening, but it is dramatic. Tyler. Yeah, I was just going to say that was like the original deep deep seek controversy. People thought they were distilling from chat GPT. Yeah, GPT. Before we go, I got to give a shout out to Cooper in the chat. Every single day when the market closes, he comes in and he says, time to like the stream. I love it. Thank you so much. It's a movement. It's a movement. Thank you everyone for watching. Thank you for leaving us five stars on Apple Podcasts and Spotify. We can't thank you for subscribing to our newsletter at tbpn.com.

3:21:42We will see you tomorrow. Goodbye. Just one more sleep. Goodbye.