Netflix & AI Slop, Saudi Liquidity Crunch, Clawdbot Reactions | Mark Gurman, Miles Brundage, Aidan Smith & Asher Spector, Alex Dhillon, Mitchell Angove, Gabriel Stengel, Sierra Peterson

28:31Still, Tether is only a small part of a much larger rush from investors into gold, with central banks and ETF investors collectively buying more than 150, 1 ,500 tons of metal. I wonder if this is moving the gold watch market. Do you think the Texas Timex is booming on the back of gold spiking? We'll see. I think a lot of prices are still down pretty dramatically during the 2020, 2021 era. But I definitely, I mean, you got to be a little bit scared right now if you're in the business of manufacturing gold watches and other precious metals, just given that your input costs are going. I'm sure they have one or two years worth of supply.

29:17They can be somewhat insulated, but prices will go up, or at least costs will go up. Fin.ai, the number one AI agent for customer service. If you want AI to handle your customer support, go to fin.ai. I heard a very funny interview with an actor who was talking about why he always wears a gold Rolex, and he was calling it a helicopter watch. Did I send this to you? And he's saying that like— This was Army Hammer. I think it's Army Hammer. And he's saying— It's the most obscure situation. It's such a funny situation. But he's like, if you're ever in a crisis, you're on the top of a building, the zombie apocalypse is upon you, someone shows up with a helicopter, and they're going to save a few people, Everyone knows what a gold Rolex is, and they know that that's valuable.

30:05You're not going to have time to be like, no, it's FB. It's an FB. It's a Patek. Let me explain high horology. Here's the tour beyond. Here's the aftermarket prices. Yeah, exactly. It's like gold Rolex is a store of value. It's always going to trade, and now it's probably going to trade even higher. Anyway, Cisco. On February 3rd, the Cisco AI Summit brings together leaders from NVIDIA, OpenAI, AWS, and more to discuss the future of AI and the economy. See you there. The whole thing will be live streamed and will be there for a giga stream. Paula says SF escape room called the permanent underclass, and it's just a room with a laptop and Claude code installed.

30:48There's something here that's funny. So good. Have you ever done an escape room? No. I have never once in my life thought that that would be fun, nor thought that I would, this is how I want to kill time. Yeah, it's like an hour. Well, if you're good, you get out faster, right? Yeah. Have you? I'm not into them, but I've done them, and they're fun sometimes. It depends on if it's a well-structured one, but I like puzzles. Like it's fun. It's like a fun puzzle to figure out. But you can very quickly get caught in like just like overthinking what's going on and being like, oh, it must be some like complex math thing.

31:31And it's like, no, you actually just needed to like press this lever instead of like analyze the situation or something. Are some people in there using their phones to try to figure stuff out? Because I imagine you could just say, I imagine there's only – Take a picture. There's probably only – Is it like a series of rooms? Yeah, well, usually you start in one room. They're all different, but oftentimes you'll start in one room and then there'll be a series of puzzles, locks and keys and whatnot. And then oftentimes you'll unlock something and then you'll progress to another room and then you'll progress to a third room and then you'll finally get out of a series of rooms.

32:07Because they're pretty easy to set up in sort of like a defunct office space or storefront that's just kind of going in between. It's like the spirit Halloween of commercial real estate. Like you just, like anyone can come in and just say like, oh, we'll be in there for like a couple months. It's not super permanent. The build out's pretty simple. It's mostly just like some walls and decorations and like some creativity on the puzzle side. Trey says can you guys put Tyler in an escape room and see if he gets out before the show ends? I mean, it'd be very hard to find a three hour long escape room.

32:42I think most of them aim for like 45 minutes. Well, maybe we need to make one. One of our guests had to leave the Ultra Dome and go straight to an escape room. So, you know, there are escape room fans among the TVPN army all over the place. Have you done one, Tyler? I have not. Are you interested in it? You're a speed cuber, so, you know, similar. Yeah, I mean, it could be fun. I don't know. I'm kind of in the Geordie camp, though. I've just, like, never been super. It's very, like, I don't know. Maybe it's great with, it's probably great with, like, kids who are maybe, like, 8 to 10 who can do the puzzles, like family event would be fun.

33:19And then it was a common thing. When they came out, there was definitely a boom in escape rooms. Bobby in the chat says escape room would be a good benchmark for AI. Certainly a humanoid. I mean, you could go in there, take pictures, and be like, help me solve this, help me solve this, help me solve this. I know, but I want to see a humanoid run through it. Let's call 1X, see what they can do. Call any of them. Hit them up. 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.

33:55Moving on. Lots of reaction to ClaudeBot. Very much enjoyed our interview with the creator of ClaudeBot, now MoltBot, yesterday. One of, yeah, incredibly, he had some wild lines, but incredibly refreshing kind of conversation and viewpoint. Totally. is totally counter to the entire philosophy that I feel like a lot of people, at least on the West Coast, in the way that they're approaching AI right now and the way that America is approaching AI. He's like, we asked, you know, you think somebody will like, you know, fork what you're doing or clone it? And he's like, yeah, I'm sure they will. Like, I don't care.

34:37Yeah, I'm building this for myself. Yeah, and he's like, I have enough money. Yeah. And I'm super, it's going to be really fun to follow along. I have a feeling that he's just going to keep launching a bunch of random projects because he clearly is just in it for the love of the game but I do hope that he can get a lot more resources and really scale up the operation. I don't know that he needs more resources. He has more agents and whatnot. The interesting thing about Claude Bot as a product is you download it from GitHub, it installs and it It has all these different integrations, and it does something that's very complex, and it has all this safety text.

35:20There's a website, and there's a community page, and all of that feels like, okay, yeah, this is like a 10-person startup. They probably worked on this for a year, but it's like, no, it's one person, and it's like three months because the guy is using agents. Yeah, you would think that he would build, his big complaint was the security inbound, security researchers asking him stuff. Yeah. It's like, okay. So he bought it. Bought it. I think he will. He should. But yeah. He said, he posted December 26th, he had his Codex dashboard up. So he said he's done 250 billion tokens, which is probably top 10 Codex users.

36:00Wow. Apparently from Codex. Yeah, so I don't know that he needs way more resources. I mean, he should be able to get donations if he needs them, get credits or something. I don't know. Like, the financial strain on that business does not seem – it seems like it's more constrained by his ideas, you know, how he's thinking about designing the system, integrating things, rolling it out. buddy buddy of mine michael uh watched the interview and and uh he said that he he's had clodbot set up he set it up but right at the beginning of january and was initially like just kind of yeah got a little frustrated using it but has now got it uh he's got it set up so that it's able to make phone calls on his behalf specifically wants to get uh hillstone reservations yeah so he just like basically wait what's the restaurant okay yeah yeah so he's just like going to use it to start getting like reservations at different restaurants it's so funny google has a product for that that does an ai phone call but for some reason it just hasn't really rolled out that or it just hasn't gotten to like adoption i don't know there's some sort of there's some sort of like memetic like i think people seeing the clip where he's like explaining how he's feeling the agi really really hit people uh obviously he's uh has it make him a daily brief that feeds into an RSS on a podcast.

37:21So in the morning, he can just listen to a five-minute podcast on what his day is like, things that he should be responding to. Kind of like, it seems like you could have a more - First podcast ever. Your enemy texted you something, and you will be deeply upset when you see what happened. Everyone's praying on your downfall. No, no. Security is important with ClaudeBot, now MoltBot. uh crowd strike is also important your business is ai their business is securing it crowd strike secures ai and stops breaches a 10 a random 10 person team in paris just dropped what looks like it looks like an even something superior to claude bot according to uh chubby on x uh it's called twin the ai company billy builder hugo they yeah they raised a 10 million dollar seed round yeah They have over 100 ,000 agents deployed.

38:17So Tyler, give it a spin. Check it out. Twin lives. There's a community note on here. This is undisclosed advertisement. But Doug over at Semi Analysis Fabricated Knowledge says, I think that Claudebot is going to be a moment. And yes, someone's going to do this. And I'm still wondering about how quickly a business can actually scale when you're getting constantly hammered with TOS violations from every Mag7 legal department constantly? Like, hey, yeah, we noticed that you built a CLI for the web interface on WhatsApp, and we don't want you to do that. And it's one thing if it's like an open source repo that people are running themselves and can change and can edit and fix and tweak versus like, hey, you're a corporation that we could potentially sue or tie up in litigation or go on a press tour around.

39:16It's just a very different dynamic. Yeah, and the prompt injection risk as well. Yeah, totally. Does that liability fall back on the company? I'm excited to talk to Mark Gurman about how he's processing this with what will happen with Siri, the timelines there, what will be integrated, because Apple and the iOS ecosystem should have a lot of the same functional hooks into these products. Yeah, remember we talked about this Monday. There was a team that built a product which sold to Apple and became Shortcuts. And then they built Software Applications Incorporated and a product called Sky. And they were pre-launched and OpenAI acquired it.

39:55It was focused on integrating AI into the operating system. And so you can imagine OpenAI behind the scenes has been cooking on a lot of stuff like this. But this isn't the kind of thing that you can just at OpenAI scale just say, hey, we're going to just ship this and see what happens. whereas a startup or an open source project. Yeah. Let me tell you about Lambda. Lambda is the super intelligence cloud building AI supercomputers for training and inference that scale from one GPU to hundreds of thousands. There are a lot of people building stuff on Moldbot, around Moldbot, around the concept.

40:29There's actually someone in the chat that's working on a Claude bot for the cloud. He says he has a name. I'm very excited to learn more about his project. I think his name is Jeff in the chat. Brexton says, yes, that was fast. Yup, software is cooked because Kalen Y says, introducing Moldbot for Teams, one-click connections to all the apps, client-side encrypted and builds your team's memory. Get your Momo is the product. I wonder, this feels like almost something where if you've been building a product that's like this, you should potentially reintroduce it on the back of all the Moldbot hype.

41:11And, you know, if you're like a, you know, multi-tentacled agent that can touch a lot of different systems, well, we'll put it in terms that people understand and say, hey, like this is, you know, this is Claudebot or Moldbot for the enterprise. And it has these, you know, functions and these benefits and this security approach. But clearly a lot of people will be focused on this. Let's click over to re-watching the interview with the ClaudeBot creator, Peter Steinberger, because I want to hear about his AGI moment, the AGI achieved moment. First, I'm going to tell you about Gusto, the unified platform for payroll benefits and HR, built to evolve with modern small and medium-sized businesses.

41:54And then we will play the video from ClaudeBot creator. I want to see. So in November, I don't know. You know, I wake up every day, I'm like, okay, what do I want to work on now? That would be cool. And then it was like, okay, I want to chat with my computer on WhatsApp because if my agents are not running and then I go to the kitchen, I want to check up on them. I want to do little prompts. So I just hack together some WhatsApp integration that literally receives a message, calls cloud code, and then returns what cloud code returns, one shot. And it took like one hour and it worked. I was like, well, okay, that's kind of cool.

42:38But I usually use prompts, like a little text and an image. Because images are like, they often give you so much context and you don't have to type so much. So I feel like this is like one of the hacks where you can prompt faster, just like make a screenshot. So the agents are really good at figuring out what you want. So I hacked together images. And then I was on a trip in Marrakesh with a weekend birthday trip. And I found myself using this way more than I thought. But not for programming. It's more like, hey, where's restaurants? Because it had Google in it and it could figure out stuff. And it's like, especially when you're on the go, it's super useful.

43:21And then I wasn't thinking. I was just sending it a voice message. you know but I didn't build that there was no support for voice matches in there so it's a reading indicator came and I'm like I'm really curious what's happening now and then after 10 seconds my agent replied as if nothing happened I'm like how do you have did you do that and it replied yeah you sent me you sent me a message but there was only a link to a file there's no file ending so I looked at the file header I found out that it's opus I used ffmpeg on your mac to convert it to to wave and then I wanted to use this but but didn't have it installed and there was an install error but then I looked around and found the open AI key in your environment so I sent it via curl to open AI got the translation back and then I responded that was like the moment where like wow yeah yeah you know it's like that's where it clicked these things are like damn smart, resourceful beasts if you actually give them the power.

44:24Beasts. App love it. The ad at the end. I love it. Yeah, people were reacting to this. People are genuinely not ready, says Vittorio. Lots of people calling this CGI. One way, the reason that I feel like that is such a powerful moment, and I'm glad that he shared it, is if you give somebody if you're talking with a model and you give it a task and then it just hits a dead end, it's just incredibly, that's sort of like people are very used to that right now. It's not that it needs to be that way, but it's just kind of like the steady state. People are used to, okay, I know what the models can do.

45:07To me, that's effectively the agent having real agency. That makes it an agent. It's saying like, well, I didn't know how to do this or I was confused and I tried a number of things until I did what you wanted, right? And that's like what you want out of, that's what you want out of a team member, right? If you're working with somebody on a project and they have a task, they don't just try one thing and come back or just say like, I actually can't handle this because the file type is wrong. It's like, convert it, figure it out. What was your reaction to that? Yeah, I mean, it's like pretty insane.

45:38Definitely raised my, you know, chance of permanent underclass. Oh no. You know, it's making me a little worried. Yeah, Will says it's over. It's over. We need to move. And there are a lot of people quoting this. G Fodor has the Eliezer Yudkowsky meme and lots of people here.

46:01lots of people were you know interesting that he uses Codex here Rune said Codex 5.2 is really amazing but using it from my personal and not work account over the weekend taught me some user empathy lol it's a bit slow and Yuchin Jin says every time I ask my OpenAI friend when will you beat Claude at coding they say we already beat them I think Sam is realizing speed not intelligence is codex's blocker to a clog code moment cerebrus chips cerebrus chips might unlock a codex moment yeah that would be very interesting and that's and that's one of the like the models are so powerful now that there's there's a there's so many moments where you know it's going to deliver it can deliver because it can work around all these problems but if you're like I need this done in five minutes.

46:53I'll just do it myself. Yeah. Like 5.2 pro is like an incredible model, but it just takes like five, six, 10 minutes every time you prompt it. Totally. It's kind of like thinking or, uh, you know, deep research type model. Yeah. Yeah. Well, let me tell you about label box, reinforcement learning environments, voice, robotics, evals, and expert human data. Label box is the data factory behind the world's leading AI teams. I like the fireworks. I had to do some fireworks for the fireworks. Um, what else is in here? So you already mentioned this, Tyler, but Peter's probably in the top 10 users of Codex at the moment.

47:27Over 250 billion tokens in a few months is a lot. So how much does that actually cost? That feels like you're up in the hundreds of thousands of dollars. No, it says it here. Yeah? How much does it spend? $51 ,000. That's not much. Wow. Yeah. Usage cost. And he has a streak going of 74 days. No days off. Yeah, so$51 ,000 for this type of result. Pretty remarkable. So where do we move the goalposts now? Yeah, I'm so... What are you unhappy about? I know there's going to be something that you're unhappy with about AI. Like, you can't be satisfied. Me? Yeah. Me? Yeah. Permanently dissatisfied. Well, part of it's a bit.

48:13Yeah, I know. Obviously. Because I have to provide an alternative. opinion, but of course we have the goalposts here. They have been sitting in one place. But I guess what I'd be interested in is Peter and his team built a hit product, this novel experience that has taken the world by storm. And how much did we know they only spent a few months on it? How many people actually worked on it? And what was their total? how much did they spend in total on tokens, right? Because they were spending some with codex. You can sort of see that from the GitHub commits, right? Probably for some sort of average.

48:58Yeah, yeah, I'm sure you can do that. I mean, just based off that it's like, okay, $51 ,000 on 250 billion tokens. And a lot of those tokens are probably using CloudBot to do things, right? Not just building it, right? Yeah, I mean, that was over how many months? Like, those tokens were not just on CloudBot. Yeah, yeah. That was just all of his, like, hundreds of... And he says he talks to Opus 4.5 all the time, too. So I wonder what numbers he's putting up there. But still, it's interesting that it's not in the millions. It's pretty accessible. I don't know. We need to find a new AGI benchmark.

49:29OK, so I think maybe it's something like this, right? So I think one of the main takeaways I have from this is just that big model companies can't really release an equivalent product just because of the integration thing, right? Mag7 companies are not going to talk to each other. You're not going to get OpenAI in WhatsApp, stuff like this. So it really reinforces the idea of... But do you think they would go so far as to say, we are not going to allow you to navigate our applications using something like... I mean, that's literally happening with the New York Times does not allow OpenAI to browse it with a bot.

50:11Like, you can go to... Yeah, but it's different having your local device being effectively used by the computer just using itself. At the same time, if OpenAI has the ChatGPT app, right now if you go to the ChatGPT website and you say, Hey, I found this. I'm a subscriber to the New York Times and there's an article in the business section that I want you to summarize for me. Here's a link. Can you open the link and turn it into some bullet points for me? It'll just say, no, no, no, I can't go over there. They said, don't go. I'm not going. Now, if I download the ChatGPT app and it's able to run that and maybe the New York Times can't block it, the New York Times still has the ability to sue OpenAI and be a headache.

51:02So then they can just say, hey, enforce the same thing. But I'm talking about a situation where the user, let's say they have a Mac Mini, they're running Moldbot. on the computer, they've given Multbot their New York Times login, so it's fully logged in, and they're paying New York Times subscriber. I think users would be upset that you're saying, if the New York Times says, we're not going to allow, we're going to figure out a way to try to, I don't even know how you would go about trying to actually stop that activity. But I'm just saying, even I can imagine users are basically going to demand And I agree.

51:40If my data, I don't care if my data is stored in the cloud, I want to be able to access it on any device that I'm logged into, even if I'm not physically present with the device. I agree. And so I just think it's going to be a really, really, really tough argument for some of these larger companies to say, you can only access your data if you are physically moving the mouse yourself. Yes. I want to get Tyler's response. But first I'm going to tell you about cognition. They're the makers of Devon the AI software engineer Crush your backlog with your personal AI engineering team Tyler. What are you thinking?

52:15Okay? Yeah, so I agree I think basically weird is that all these companies need incredible deal guys, right? This is the whole thesis like the deal guy with open source with this like hacker culture You can kind of always get around these rules. Is it like maybe breaking terms service like def? Yes, definitely these big companies can't do it. They need more deal guys, right, but it's like hard to find these deal guys you need the ai to become the deal guy so you need a clod bot deal guy that's the new that's the new benchmark you're gonna be good at making deals okay clod bot ai that can do a deal between two mag seven companies here we go there we go there we go move it to the other side see ya see ya uh tyler why do you i kind of want to take this a little bit further just because I don't know, like how do you actually enforce this?

53:10Is, is Google Drive going to say you need to have your camera on and we need to see that you're sitting in front of your computer actually using it? I mean, you can always do these like bot detection stuff with like whatever it's coming from the same IP, these things. I think it's like kind of a constant race. but like I think the open source stuff will if you like really grind it out you can kind of always get around this stuff and so like the only only real solution is to have like actual I think deals between these like different companies yeah I think that if you if you have an app from a big company that enables you to do something like like like if if Chrome started shipping with a torrent browser, something that would allow you to manage torrents, which are not illegal.

54:01That's just software. You can just download torrents that are not piracy. I can just be like, hey, I put up a torrent of TVPN content. Anyone can download it. But just by virtue of the fact that it's so prevalently used to pirate media, they would face pressure from media owners, Warner Brothers and Disney, and they would get angry letters and they would strike a deal and say, okay, yes, like we're not going to enable this. And so even though, so, so if, if any of the big tech companies launch an app that enables behavior that other apps don't like, they will have leverage and they all have deals together and they all have legal teams and there's a bunch of different pressures that they can apply.

54:45And it's this like constant negotiation. Oh, well, you know maybe we won't renew this deal or this contract maybe we'll go with someone else if you're the bad actor you'll get pushed out right so i don't know i think it's going to be a big big debate i i i don't i don't i would be surprised if i if any of the big labs just launch something that can actually just go and do anywhere and do anything because uh the pushback but do you think do you think this is something that will like why would this just not happen at the os layer yeah I don't know I mean, then you need Apple to like actually lock in.

55:24You do. Yeah, or Satya. Yeah, and that's why we're having Mark Gurman on the show to break down Apple's strategy and what they're doing. Anyway, interesting. MongoDB, choose a database built for flexibility and scale with best-in-class embedding models and re-rankers. MongoDB has what you need to build what's next. Kimi 2.5 is out, and Alex Chima is running it on his desk. It runs at 24 tokens per second with two M3 Ultra Mac Studios connected via Thunderbolt. He went much bigger than the Mac Mini. And by the way, apparently Kimi likes to be, refers to itself as Claude. Yeah, what happened here, Tyler?

56:08Can you break down what's going on with Kimi? Yeah, I mean, it's always kind of unclear with the Chinese open source models, like what they're actually doing to train the models. So there's like, yeah. So when you ask the models, like, introduce yourself, it's like, Like, hello, I'm Claude. Yeah. So there's probably like two scenarios. One is that they trained on the outputs of like Opus or Entropic Model, right? You'd think that they would just find and replace in the training data though. You know, like that'd be pretty easy to just be like, okay, we're going to scrape a ton of responses from Claude.

56:38Let's make sure we do a find and replace on Claude so that if it says, hi, I'm Claude, I'm a helpful assistant, it just changes that. That's just like a... Yeah, I mean, you would expect that Kimmy knows that it's going to call itself Claude. Yeah. So it's like, why would they allow that? Maybe they're just basically, like, mogging Anthropic. Like, oh, yeah, we have Claude. What are you going to do about it? Yeah, like, what are you going to do? We stole Claude. Yeah, and then there are also some rumors, like, very unclear if this is just, like, completely fake headlines. But, like, maybe there was some leaked checkpoint that somehow got out to China and then they did a fine tune on it or something.

57:12Yeah. That would be pretty, pretty crazy if... Yeah, people have said that before for other models. I'm not really sure how reliable they are. But it seems like fairly, very likely that they just trained on Opus outputs. Yeah, yeah. Because you kind of, people said DeepSeq did that originally with ChatGPT. Yeah. Yeah. I talked to one technologist about what it takes to reverse engineer the GPT-4 API. And it's a surprisingly low amount of outputs to sort of interpolate all the weights or something that approximates. So it's clearly sort of a game back and forth, just like a constant war. I think it is quite – I think it's probably a good sign for U.S.

57:58AI labs that the Chinese models are essentially just completely based off the U.S. ones. They're not actually training these models from scratch. They're not doing these incredible training runs for much cheaper, like people say. They're basically just copying off the U.S. Do you think that's because of just chip restrictions or actual architectural hurdles? Yeah, it's probably all of them. It's hard to figure out the architecture. You don't have the chips to try these different science projects, training runs. Wow, the Chinese models lag by exactly how long it takes to scrape the API, basically, and do a fine tune.

58:40What a coincidence. Dean Ball has been on a tear. the last 24 hours. It's a good post. He said, people significantly underrate the current margins of AI labs, yet another way in which pattern matching to the technology and business trends of the 2010s has become a key ingredient in the manufacturing of AI copium. Copium. And he says... This Derek Mohler post was very informative here. Just look at the market clearing prices on inference from open source models and you can tell the big labs pricing has plenty of margin. Deep Infra has GLM 4.7 at 43 cents in,$1.75 out. Sonnet is at$3 in,$15 out.

59:20How could anyone think Anthropic isn't printing money per marginal token? And Dean Ball says the reason they think labs lose money is because 10 years ago, some companies in an entirely unrelated part of the economy lost money on office rentals, WeWork, and taxis, Uber. And everyone thought they would go bankrupt because at that time, another company that made overhyped blood tests, I don't know, did go bankrupt. That is literally the level of ape-like pattern matching going on here. The machines must look at our chattering classes and feel great appetite. Yeah. You could also have seen the margins from the whole open code thing with Anthropic.

59:58Oh, yeah? Right, where there was open code. So Anthropic, they have Cloud Code, right? You can get the Cloud subscription and you get free Cloud Code tokens. And then you could use those to auth for open code and they removed that. and the whole reason you would want to do that is because the like cloud code tokens versus like actual cloud API is like 10x difference so there's like a massive markup to the actual API so you'd assume that like unless Anthropica is losing just insane amounts of money on cloud code which maybe they're losing some amounts of money it's like well you look back at the I forget which interview Dario was doing where he was trying to get people to think about like you think of each model as a company where you spend all this money on training which is CapEx and then when you're actually running the model it's very profitable But if you look at the business as a whole, you have massive, massive losses from training and stock-based comp and hiring 1 ,400 of the best engineers in the world.

1:00:54So if you actually look at it on a company-wide, you have continued scaling, massive losses. But the important thing is that effectively at the product level, when they're selling the product, they actually are making money. So in the S1, good to you and to me, looks like training clearly broken out as CapEx, solid gross margins above 60%, something like that on inference. And then, you know, all of the AI copium can probably subside if OpenAI or Anthropic go out with an S1 that shows really solid inference margins. and if it comes out that it's like oh their inference margin is like 10 % or something then people are going to be panicking.

1:01:40To be clear I think these will be some of the most special S1s that have ever graced the capital markets. Let me tell you about 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. And without further ado we have Mark Gurman. The Gurman manager. Managing editor of Bloomberg. In the Ultradome. Welcome to the show. Thank you so much for taking a time to come on down. Yeah, right here. We'll have this microphone for you. You have a Diet Coke, so you're locked in. We're ready to go. How have you been? How's your new year going so far?

1:02:20Oh, freaking great. Yeah? Yeah. My wife and I had a baby at the end of last year. Gong, gong, gong. Big gong. I need to get one of those things that you guys make on the Twitter and just me. The card. Baby. Baby. Oh, we'd be happy to do that. We're happy to do it. We never, we never, yeah, we never, we never know. Some people, some people want to be, you know, more low key. Oh, I am low key about it. I just figured, you know. But how do you, how do you rate parenthood? How does it, how does it feel being in it versus, you know, all the expectations that people have? It's just taking care of this individual, and they're being so reliant on us and getting to teach them.

1:03:06But shout out to my wife for being the one really pulling the strings there and taking care. So it's great. I feel like reporting and parenthood are especially hard to balance in some ways. Still early. Well, I would, yeah, no, I would, I would just like, obviously it's completely possible and you're, you're clearly doing fine, but the nature of the work means that like a story is happening and you want to be the first to provide the best coverage. And sometimes, you know, baby's crying. Yeah. And you might not be able to pick up the phone. But. But you get through. You know what? People, people have been hearing crying babies over the phone for.

1:03:49That's true. For, for forever. So they'll have to deal with it. I just can't wait till we go on an airplane. Oh, yeah. Like, I already told my wife, like, yeah, we're going to get, like, banned from American Airlines. No, I've said this on the show before. Like, the funniest thing is growing up, I assume that when a baby was crying on the airplane, it means that the parents were bad parents. Like, they were doing something wrong. Totally, totally. And in reality, babies just cry for a million reasons. And it's totally possible that there's no solution sometimes. They just need to all sleep. Yeah, I saw like an Instagram reel of a mother talking about her kid and the kid was crying in the background.

1:04:23I was like, why would you take advice from her? Like, that's such a bad mom. Yeah, it has nothing to do with the baby. And that's not even at all. It's like, now it's like the baby's always crying. It's crazy. They're hungry. They have gas. They don't like the sweater they're wearing. There's a million reasons. You know, they didn't like my article. Who knows? But anyways, let's talk about Apple. Let's talk about Apple. I mean, we've been talking about new Siri, expectations there, The big news yesterday was ClaudeBot. Have you been – do you think there's anything about – how have you been processing the ClaudeBot, now MoltBot story?

1:04:58What's possible there? Expectations around AI assistance feel sky high in the open source community. What do you – do you think like anyone at Apple is like updating on ClaudeBot and MoltBot and what's possible? Well, let me just take a step back. Please. Yeah, yeah. You know, with Apple and AI. Yeah. So in 2018, they hired John Gene Andrea. He was this high flyer at Google. He ran AI in search. And Apple thought they had a coup here. Apple thought they would hire this guy and really just hit the ground running and be at the forefront of artificial intelligence. Just seven years earlier, they announced Siri.

1:05:35In 2011, there was absolutely nothing like it. It was breakthrough. But then it just became utter junk. Google Assistant lapped it. Alexa lapped it. So they thought they were going to bring this guy in and it'd be a game changer. Turns out, and maybe this will be Tim Cook's fait accompli, but this was the biggest mistake, this hire of Tim Cook's tenure. I think it's easy to say. Apple is so behind in AI. There's been so much ink spilled on this and so many conversations on this, and I've written about it and talked about it half a million times. I think it does you haven't even scratched the surface about how big of a problem this is for Apple right?

1:06:15they've completely screwed up AI in every which way and it comes down to just hiring the wrong people and entrusting the wrong people but is it a do nothing win scenario? because I think we're seeing a situation now where the Mac mini might end up selling out everywhere stock seems fine are you seeing like financial that's the problem when you've had no real negative hit. It's hard to go wartime. And acknowledge that there is a failure because all the numbers are holding. It is wartime. The numbers are great. Tomorrow, they might report their first$135 billion to$140 billion quarter. I don't promise.

1:07:06is how can you put$140 billion and Tim Cook needs to go retire in the same sentence? You can't, yeah, exactly. It doesn't make sense. But when you think about the long term, you think about the future, these things are going to need to get rectified. And I guess the good news is they are on a path to rectifying it to some extent. This Google Gemini deal is a breakthrough for Apple. It's embarrassing. I mean, it's absolutely crazy. You poach their guy. And then years later, you're paying them billions. He screws you up. You pay him$25 million a year for eight years, right? And then now you're paying basically 4x that, 8x that, maybe a little bit more, in order to get the new good stuff.

1:07:51And so they'll announce a new Siri next month. The thing with this new Siri is it's basically a replay of everything they announced in June 2024. So it's basically everything they announced two years ago coming very late. Things like using what's on your screen to fulfill Siri queries, being able to control your apps. And then in June is when the good stuff launches. That's when Apple launches its first chatbot. It's interesting because they've spent the last two years saying nobody likes chatbots. Everyone hates ChatGPT. It's terrible. It's ruining the world. Okay, then you have ChatGPT with nearly a billion active users.

1:08:24They're like, okay, we kind of got to do this or we're screwed. So they're doing it. And it's actually going to run on Google servers, Google Cloud Platform, Google TPUs. This is great for Google, too. And I think some people at Apple think this is going to be like a short-term thing. You know, we're going to partner with Google until we get our act together. No. I definitely think that this is more of a long-term play unless these models continue to get commoditized, which they will eventually. But this is not a 12-, 24-, 36-month thing. This is longer term, I think, than people expect. And Apple can't just go wrap Llama because of the terms of service that Meta's put that around.

1:09:04They said you can't use Llama. Well, I don't think they want to use Llama. They don't want to use Llama. I don't think they want to work with Meta. They don't want to use a Chinese open source model, so there aren't that many games in tech. Well, let's take a step back here. So they're using Google in the U.S., right? But they're going to have to do something in China. Okay. And so you'll see them use a combination of Alibaba and I think it's Weibo or Tencent or one of those different AIs for different features. And so they'll use them. And just because they're partnering with Google Gemini on Siri in this chatbot doesn't mean they're not using other players.

1:09:41There's a lot of open AI and a lot of applications. They launched their creative cloud competitor today, and a lot of those AI features are powered by ChatGPT, like some image generation stuff. I still think on an image generation side, you're getting a little bit better on OpenAI than you're getting from Google. And then a lot of stuff internally. Like Apple runs on Anthropic at this point. Anthropic is powering a lot of the stuff Apple's doing internally in terms of product development, a lot of their internal tools. So that's a big one to watch. They have custom versions of Claude running on their own servers internally, too, because this Google deal, this just came together a few months ago.

1:10:22They were not going to use Google. Apple actually was going to rebuild Siri around Claude. But Anthropic, they were holding them over a barrel. They wanted a crap ton of money from them, several billion dollars a year. And at a price that doubled on an annual basis as well for the next three years or so, from what I understand. At the time, Google was really an afterthought because they were in the middle of the trial with the Department of Justice. And then for some reason, the judge ruled that Apple and Google's deal was kosher, even though everyone knows it's a huge issue and a monopoly and all that.

1:10:59I'm not here to be a judge. I write. I don't make judgments for legal proceedings. But anyways, Apple and Google get off scot-free. They can do whatever they want now. And, you know, they're not being held back at all. And then obviously OpenAI, that is a real firestorm for Apple right now. Like OpenAI, obviously they're working on these AirPods competitors, the ChatGPT built in. Sweet Pea. Sweet Pea. Johnny Ive is running the show and design there. That's a big freaking deal. He raided the whole design team, all of the Apple designers that were there under Johnny Ive. Not all of them. 95 % of them are gone.

1:11:37A ton of them are now working at Love From and OpenAI and other companies. Some of them have retired. But the crux of it is, like, how do you partner with a company that's trying to put you out of business, right? So there's no way that they were going to work with open AI, no matter how good the financial system was. What do you think about some of the projections that you've seen or kind of estimates around? I forget. Was it Foxconn was saying, like, preparing to be able to produce something like that in the first year? It feels like a massive number, but at the same time, you have a billion users.

1:12:15Talking about the open AI earbuds? Yeah, sweet. You know, it's so interesting. And one more thing there. Yeah. The earbuds, the benefit of them versus some of these other AI hardware is like if you can just take calls and listen to music, like already you have some functionality that people are using all day long. Sure, sure. So it's not the same as wearing a pendant or having some other little gadget that actually serves no real use case or at least not a powerful use case. If I can just listen to calls or take calls, listen to music, at least there's some base level functionality. And then you layer on the AI and maybe it turns into this amazing, super differentiated experience.

1:12:55I just don't see it. I just don't see them selling 45 million units. I just don't see it being a success. the barrier to entry for a new hardware company as you know is extremely high like can you even think of one hardware company that just came into being and became an immediate success like i just don't see it and i also think the reason they're doing earbuds is because it's more low-hanging fruit to your point and it's easier to accomplish it's not what they wanted to do what open ai wanted to do is they wanted to create an iphone killer instead now they're trying to create an airpods killer and i don't think they're going to be able to accomplish that will they look nicer than the AirPods?

1:13:30Probably. Will they have better AI than the AirPods? That's an open question, right? What stops Apple from expanding this Gemini deal to its AirPods and just basically adding a bunch of AI functionality to the AirPods? They have very fast processors, very tight iPhone integration. Apple can replicate whatever open AI is going to do pretty quickly by relying on Google and Gemini and whatever they can cook up functionality-wise there. Do you think that AI, specifically products, that we had the creator of MaltBot on yesterday, and he was saying there's so many applications, and he was referencing MyFitnessPal.

1:14:08Do I really need MyFitnessPal if I can just take a picture and just send it in to my agent and track it? And so what I was thinking there is, does that make the App Store broadly potentially less of a hurdle with OpenAI is launching, eventually launches a new phone and they have a new OS and they don't have an app store. They don't have an app store on day one. It instantiates the equivalence of apps on the fly. Well, if you ask me, we're already in territory where iOS and the app store are legacy features, right? Like the app store is a legacy world. The iOS user experience, the macOS user experience where you're jumping between applications, where you're going into something to get information, where you're going back to your home screen and launch another app.

1:14:53Apps are the past. AI agents are already here, and that's the move forward. And whenever OpenAI comes out with a phone, I do eventually anticipate them coming out with a phone. And when I say phone, I'm not talking necessarily about something you put to your ear, like a classic phone like an iPhone. I'm talking about I feel like people are always going to have some sort of slab in their pocket, right? because it's convenient. You get the display. You get the sensors. You get the battery. You get the cameras. It's not a replicable experience no matter how many different gadgets you want to put on your body.

1:15:26They could make like a banana, like a plastic banana, smart banana. That would be fun. Yeah. So yeah, AI agents are where the world is going. And I totally expect Apple to move in this direction. This new Siri, Campo, that they're launching at the end of this year, that is a huge step towards the AI agentification of different features on the phone. And being able to, for instance, tell your phone, pull up this photo that I took at the studio, find photos where I have bottles of water, and remove the bottle of water from the photo, and email or to text it to Mark, right? Like the AI agentification of iOS is happening.

1:16:12Yeah, and theoretically Siri should have killed the weather app a long time ago. Because when you think about if you're just going to navigate to the weather app and you're traveling somewhere and you're like, oh, where is this city? Oh, I don't have it saved. I'm going to search and add it. And you're finally looking at it and it should just be a prompt that you're firing. But then when you try and click a level deeper in the current Siri experience, you just can't get there. That is the big problem with Siri and the big difference from ChatGPT and Gemini is it has the going a level deeper problem.

1:16:44There's no back and forth. It forgets context very easily. It has no memory. It doesn't have that tight integration with applications because the app developers, they know Siri is so terrible. Here's a question. So what if I told you that you could call an Uber with Siri, right? You guys probably know that, but does anyone use it? I don't have the data, but I can tell you that they announced support for that, and they actually rolled out support for that. Five years ago? No, 10 years ago. 10 years ago. Okay? It doesn't work because nobody uses it because it's too cumbersome. Totally. Okay? Like, I'm very tech forward, as anyone watching this probably knows.

1:17:21You know, it's right up my alley to have Siri call my Uber for me. But, like, I've never done it because, you know what? I don't freaking trust it. Trust it, yep. I don't think it's going to work, and it's just going to be a time suck. And people didn't trust ChatGPT with 3.5, but now on 5.2 Pro, they trust it. I would trust ChatGPT with ordering an Uber. Me too. I would trust ChatGPT with my life. But Siri? Okay. So on the Siri rollout, it seems like there's a couple of milestones that we're going to be tracking throughout this year. When do you think I'll be able to go to Siri and just say a general Google-able question?

1:17:59in, you know, uh, give me the history of the Roman empire, the type of thing that any LLM can just pull up a couple paragraphs on, but current Siri does not have that ability. Okay. I can't tell you that you're going to be able to do it. I can tell you that you're supposed to be able to do it. How's that? I'll speak. I'll speak. I just mean, I just mean, is that like, what's the goal? No, no. I just mean like there's knowledge retrieval and then, and then there's, and then there's like agentification, like making, interacting with the apps, doing all the things like pulling what's in your camera roll together with your iMessage and all of that.

1:18:32And that feels harder for both of those end of March. They're supposed to be end of March, but the latter in terms of the app manipulation and the AI agent, that's going to be split probably between March and November. Because that seems harder than just integrate Gemini and just give people the Gemini answers when they ask Siri for a question. You know, it's so funny. I look back to WWDC 2024 and And obviously I made a big deal about all the delays because of course I did. But then you think about it, it's like, was anyone going to use these in any of these features anyways? Like we're talking about a few features that were delayed.

1:19:04And you look back at the announcement of these features at WWDC 2024. It's like Apple totally downplayed them too. Like they gave a few cool demos, but it's like, this is some game changing stuff if it's marketed correctly. So not only did they pre-announce it, not only did they delay it, but they didn't even market what they had in the hand correctly, which I guess in hindsight, you should have realized that it wasn't that compelling or ready to go if they weren't going to market it. Because I'm telling you, they can market anything. How do you see search fitting into Siri? Because so much of what people are using LLMs for is search now, and you're seeing more commerce integrations.

1:19:45And I've always thought the lab's obsession with the browser was interesting. because in many ways, like, I feel like using LLMs, it already feels like you're using a browser. It's just kind of a new experience. And a big question is, like, okay, so Google's powering the new Siri. What happens when people start doing searches with high intent in Siri? Do you Google search anymore? I do. I don't. I specifically still Google search when doing, like, product research, shopping, et cetera. I don't. I still go to Google if it's something that I know is fastest in Google. So today I wanted to know what year the Avengers movie came out.

1:20:28And I know that that's half a millisecond in Google. And I know that that's five seconds in any LLM. And so I'll control T to go there. But for any level deeper, I also wanted to know about the VFX house that worked on it and what technology they used and when that happened. That was an LLM query. Yeah. Yeah, I don't really use Google anymore. I have my action button on my phone sent to ChatGPT, and I'm just like living in ChatGPT. Sure, sure, sure. You can't ask it the weather. Yeah. You can't ask it to do anything on your phone. Sure, sure. That's, I guess, the differentiator for what the new Siri is going to be.

1:21:02They've built a feature called World Knowledge Answers, which is the internal name, and it's basically a perplexity ripoff or a web search ripoff in ChatGPT, which is to bullet out summaries and information and give you citations and context. Yep. Using Google directly for that? Or is it some other? It's an Apple solution. Okay. It's an Apple solution. But what they've been trying to do now is, under the hood, change that to the Gemini model. Again, all this stuff was supposed to come out. Because Gemini is very good at Google search. Gemini is very good at Google search, but there's no Google services in this new Siri.

1:21:38It's literally a model. It's basically like they hired the Google deep mind team to develop the model to power its AI in Siri. Yeah, so I guess what I'm getting at is, will I ever get, if I'm using Siri to research a product, will I ever get an ad powered by like an Apple ad network because it is a high intent search? Like ever? Yeah. I think the end game for all these guys is to put ads in this stuff. Yeah. You know, one thing that I haven't talked about in a while now. And I just feel like that's been somewhat under-discussed because people are so obsessed with just make Siri work that if it does end up working really well, then you would have effectively a search engine.

1:22:24Well, yeah, if it works, they have all the tools to do a search engine. And if it does work well, yeah, they'll probably eventually put ads in it. A couple things I haven't discussed in a while. One is they're working on a new Safari web browser, and you're going to see AI search at the forefront of that. The other thing is the adsification, the advertisingification of the Apple operating systems, that starts this year. You're going to see them up the ad slots in the App Store in search in particular. Let's give it up for more ads. Yeah, ads. Any sponsors you want to give a shout out to over here?

1:23:08Don't tempt John. I will do it. And then you're going to see ads in Apple Maps. Oh, interesting. That's been one of the main differentiators in the different apps. Well, you know, it'll be all AI, and it'll be targeted based on things that you're searching. Like if you see a sushi restaurant, search for sushi, whatever, you may see some search results get elevated. It's kind of like what Yelp does, right? By the way, it's so funny on Siri. There's this Japanese restaurant in the valley I like, and it's named after a city in Japan or a province of Japan. And it's like, take me to Chiba, right? That's the name of the place.

1:23:47Shout out to Chiba. Okay, boss, you're going to need to get in a boat. Oh, my God. It immediately pops up, 7 ,652 miles away. I'm like, I've been to this place 50 damn times. You should know by now. Talking about the restaurant. Not 20 miles away. Not the city 8 ,000 miles away. It's really amazing. Hiring, chartering an aircraft, sir. I got a bunch of other stuff in the Apple ecosystem. First, Starlink is reportedly planning to integrate, Apple's reportedly planning to integrate Starlink connectivity into the iPhone 18 Pro. How do you think about this? You saw this stuff on Twitter the last couple of days.

1:24:26There's been no reporting on this recently, by the way. This is another, see, this is one of the downsides to AI. people just spew BS on Twitter. Oh, interesting. Okay, so Apple already does work with Starlink, right? You can hook in, if you have a T-Mobile iPhone, you can get on to Starlink. So that already exists. Yeah, I've experienced, I think, during the fires last year. All the cell service was down. I'm on Verizon, but I have satellite on the 17th, but it's not, what? You don't want to be on Verizon. I should get off of it. But it's not Starlink. It's the old network. On which? On the Verizon one?

1:25:04Yeah. You're talking about the Apple network. Yeah, yeah. If I'm in an emergency situation, I can get out one text message. It's really, really slow. And it's clearly going with, I think, Viasat or whoever it is. The Apple network is just super legacy. Yeah, it's just older. They've got to strip that down and partner and rebuild or something. Anyways, the iPhone 18 is going to have enhanced satellite connectivity. Sure. And they'll obviously work with Starlink and whoever, whomever. So that's coming. What else have we got on our plate? You want to talk about John Ternus? Absolutely. Let's talk about Ternus.

1:25:33Turn us around, John. Yes. John. Okay. Yes. So I have a bunch of questions. Let's start with like, what is going on behind the scenes of this campaign? It feels like there's, it feels like internally politics, there could be like people like pushing for him, pushing against him. There's this weird quote that keeps going out that people say he's never made a decision in his life. Everybody keeps making the most underhanded compliments. that I've ever seen. Yeah? It's like, yeah, he's shipped a lot of products, but he's never had to make a hard decision. That's the one. Just wait for my profile on Bloomberg.

1:26:08Okay. You'll get the real story soon. Okay, fantastic. I'm not going to give it all away today. Of course. But I'm here for the company. I appreciate it. Got to give the shout out. Stay tuned for the article. Of course. Subscribe to Power On. Yes, of course. Tech bundle. Yeah. It's great value. Okay, Ternus. He's 50. Everyone else on the Apple executive team, late 50s through their mid-60s, turning 66 this year in the case of Tim Cook. You're Apple's bored. You like continuity. You like an insider. You like people who know what they're doing and have been there for a while. They know where the bodies are buried.

1:26:41Okay. These guys all have hundreds of millions of dollars, if not more. Yeah, at 50, he's the only one who is, if let's say Tim Cook hangs out, another three to five years, you're not going to appoint another CEO who's 65, 70 years old. He's the only guy. Apple, they get vast majority of the revenue from hardware. He's the hardware guy. Have they screwed up any hardware since he's been in charge? No. He's a steady hand, knows what he's doing. He's really the only choice. There was this New York Times report a few weeks ago, basically saying that it could be Greg Joswiak, it could be Eddie Q, It could be Deirdre O 'Brien.

1:27:23It could be Craig Federighi. It's for sure not going to be Craig. It's not going to be Deirdre. It's not going to be Eddie. It's not going to be Jaws. The only category that makes sense is an operations person because you look at the current CEO, Tim, obviously comes out of the ops world. You look at the guy who would have been CEO if Tim Cook didn't stay so long. I'm not saying he shouldn't have stayed so long. He's done obviously a fantastic job for shareholders and employees and what have you. It would have been Jeff Williams. He was the COO. So Sabi Khan, he was named COO a few months ago But he's really been in that job for the last half a decade, I would say So anyways, it'll be Ternus or Sabi or someone completely out of left field I don't think this is imminent So we'll see what ultimately happens But all signs are turning towards Ternus Everyone has an opinion that Ternus is going to be the next CEO, fine I've been shouting this from rooftops the last two years But no one has given evidence Like, what is this based on?

1:28:21Right. Has there ever been a baton handoff? Is he getting more responsibility? Well, they have a bit like a big baton. They got one of these. You know what you have? You have white smoke coming out of. Well, actually, no, there's no smoke. It's very environmentally friendly. So, you know, maybe out of the parking garage. A white reflection out of the solar panels. A glint off the solar panels. Exactly. So what? You want evidence? Yes. You want to hear that he's been getting more responsibility? Okay. One, a few months ago, took full control of the Apple Watch engineering team that was co-run with the old COO until he retired.

1:29:00Okay, there's something for you. When they started soft-firing the head of AI, the guy we were talking about earlier, they took the robotic stuff away from him. They gave that to Ternus. And then the real news, as I broke last week, is that even though on paper Tim Cook is running the Apple design teams, it's not. It's Ternus. He took over at the end of last year for a variety of reasons. But you look at who's run design at Apple over the course of history. Well, like Steve Jobs, Tim Cook himself between 15 and 17, Jeff Williams, who was the number two and heir apparent for a long time, and then obviously all heard of Johnny Ive.

1:29:40Now you add John Ternus to that list. It's a big sign. It's a big indicator because you look at what Apple is known for as a company and its design, right, that design function. And it's not just hardware. It's hardware and software that he's overseeing as the manager of both of those teams. So I would say that is your first piece of evidence that he's getting some more material. What does he have to do, since he's the hardware guy, to communicate that he has a steady hand on the tiller as Apple transitions into the AI age? Like the narrative is that they have delivered on hardware. They have not delivered on AI.

1:30:18No, they haven't. And so how is he going to communicate that he's forward thinking and can be the one to keep him on the cutting edge? Taking a step back, you think about like Succession and all of these big tech companies, right? Like, who's going to take over Google one day? Who's going to take over Microsoft one day? Who's going to take over Amazon one day, right? Like, it's probably going to be the AI guys at all of these companies, right? Apple doesn't really have an AI guy. They're trying to make Craig Federighi the AI guy, but he's not CEO material. He's voiced this himself, so don't get mad at me, Craig.

1:30:51But, yeah, Ternus, can he become the AI guy at Apple? I think it's too early to tell, and I think that is a big question there. Like, do you put a hardware person in charge of the AI era? People have been speculating that Apple would do some big M &A deal to bring some AI-native talent in-house. They wanted to buy perplexity. So the origin story, they were really talking about it. Like, Eddie Q was seriously considering. Adrian Preak, other head of Corp Development M &A, reports to Tim Cook. They were looking at this pretty closely. Then they pulled out why were they going to buy perplexity? To power the search stuff we were talking about earlier.

1:31:32It became less important after the Google deal was allowed to live. And do you think they would have done it if perplexity hadn't been marked up to oblivion? I think they would have done perplexity at a reasonable price. Yeah, that's the thing. Low single-digit. I think they would have done it up to$5 or$6 billion. Yeah. Yeah, but like at this$15,$20. And then perplexity started trying to say they're going to buy Google Chrome if it got divested and all that. And I think they would have done it if the Google search deal was torn apart. Yeah. In order to bring a search product to market faster.

1:32:12They like to buy companies to – I'm not trying to do Apple corporate speak here, but this is like legit. They buy companies to accelerate their roadmap. You'll hear Tim Cook use those exact words tomorrow By the way, but literally Tomorrow? Apple earnings This guy Come on, head in the game But is there specifically A deal that he'll be talking about tomorrow? No I bet he'll mention the Gemini thing Because he was asked that in the last earnings And everyone was like, why haven't you done a mega deal? He says this every You know, what is fair to say is their pace of deal-making has decelerated significantly.

1:32:53Interesting. Significantly. And so he will be, I mean, in the last earnings, he was sort of like managing that and saying like, oh, we're still doing deals, but we're selective, and we only do it to... Well, they also don't care to announce like 95 % of the deals. Well, if they're big enough, they have to. You know, like the Golden State Warriors, they want Giannis, but they don't want to make a big trade. They don't want to give up all their draft picks, right? They're like, yeah, it's the same with the Lakers. This is like my existential crisis. The Lakers are like, yeah, we'll trade the picks.

1:33:19if the right guy becomes available. Yeah, and they know the right guy's not going to become available, right? Just like Apple knows, these companies are always going to be out of the price range they want to pay. Sure, sure. The thing with Apple is they're very frugal money-wise, extremely frugal. The other thing is they've been burned countless times with acquisitions. Like that Beats deal, terrible process for integrating that company. Well, even then, I think it was like a 3x revenue multiple. Wasn't Beats doing like a billion dollars a year? From a financial standpoint, it was just like a home run.

1:33:51People criticized that deal, but they made it back in months. Oh, interesting. Oh, Beats? They made the money, but it was a nightmare from an intolerance. No, but it's worth noting because even paying for perplexity at$5,$6 billion, it would have been... It's going to be years until you monetize that, probably. Yeah, I mean, Beats was monetized from day one. Obviously, you're selling the headphones, which are terrible, by the way, but everyone loves them. And then Apple Music, you know, Beats started the whole subscription services business at Apple. And so if you look at it, Beats is one of the most wildly successful technology acquisitions of all time.

1:34:27And then you compare it to how much criticism it got because it's Dr. Dre and Jimmy Iovine and whatever, whatever, whatever, and the headphones are crap. You know what? From a financial standpoint, home run. Home run. You know? But integrating into Apple's culture is not easy. Yeah. Speaking of hardware, what's going on with the robotic arm that will live in your kitchen? Yeah, the Pixar lamp. It's the Pixar lamp. It's the Pixar lamp. As he said, you've got this 9-inch display. It's like an iPad display on a robotic arm. It can float around your desk. It can twirl and turn around. Like, you know what we'll do?

1:35:02In a few years when this thing comes out, you'll have me back on. And instead of me actually being here, you'll have this on here. Oh, yeah, maybe. You'll leave my head floating around. Yeah. I mean, Meta tried to do something like that with the... Meta tried to do something like that. You know, these things are big in China. They are. It's a big thing in China right now. Interesting. No one really talks about them, but it's a category that has some potential. Okay. Yeah. But it's years away just because Apple moves slowly. I think it's – stay tuned. Yeah. What about the foldable phone? That's not years away.

1:35:30That's closer? Yeah. No, that's decades away. No, that's coming out in the fall. Okay. That'll be fun. I can't wait for that. Yeah, that'll be – Are you going to be a buyer? I dropped$2 ,200 on it. Oh, is that going to be – It'll be at least – Is it going to be the highest tier, the biggest, the most powerful? For Apple? Yeah, it's going to sit at the top of the line. It's going to sit at the top of the line. Okay, new status symbol. Sorry, what were you saying? I like that. We had Ben Thompson on maybe before the end of the year. I feel bad for Ben. You know, he's a big Milwaukee Bucks fan, and they're about to ditch Giannis.

1:36:01The Milwaukee Bucks, that's a sports team? Really? I don't know. I do know it's a basketball team, and I do know that it's Ben Thompson's favorite team. yesterday we didn't know when the Super Bowl was he was talking about the Apple Vision Pro I wanted to demo you can watch the NBA live you liked it Ben's pitch was like screw the Apple highly produced they're cutting around all the time and he's like just invest the money to set up the actual hardware in every single stadium, the cameras so that anybody can just drop in So you can just sit there, watch the game, like your court side. You don't need the Apple announcer because you can just look at the scoreboard.

1:36:45You can listen to what's going on. So he's saying strip the guys out of it. Yeah. Look, I watched it. It was great. You watched the whole thing? Yeah. Yeah. Trust me, people at home are not happy that night. You know, you put that thing on. Crying baby. Crying baby. And then you're completely isolated. I mean, I'm telling you, VR does not work for families. It's true. It just doesn't. It's really, it is one of the major, because then, I mean, the family people have the disposable income more likely, so then they buy it, but they can't use it. It doesn't work. Yeah. Okay, here's the problem. Yeah.

1:37:16It's like once a month. They've got like five games on the calendar. Well, and that was Ben's point, is like just set up the infrastructure and sell me a pass so that I can drop into any game and sit courtside. Because right now they have to pull up with a production truck, they need editors, they need voiceovers, they need pregame, postgame show. He's saying like there's, you know, basically one time fixed cost of like he did the math. Just put them in all the NBA stadiums. Yeah, he said it's like 40 grand. It's like not like a huge amount of money. And you have no incremental cost per show. You're not dealing with producers and running a live, you know, program.

1:37:49But they need to decide if this hardware is going to continue to exist before they keep investing in the content strategy. Right? It's a chicken and the egg problem. Right? How do you sell this thing if there's no content? but why invest millions in the content if you're not playing to sell this thing anyways and you're pivoting to smart glasses? Don't forget, they were supposed to come out with the Vision Air in 27, a product called N100. I forgot when my article came out. It's Times of Blur. A few months ago, six months ago? Anyways, they killed that thing. They killed it entirely. Killed it.

1:38:20They shocked the Vision team on that, by the way. The smart glasses, was that just completely reactionary to Meta? Oh, yeah. They started toying with the smart glasses in terms of like this non-AR smart glasses, right? First of all, AR smart glasses, that has been the vision from day one for a decade plus. But in terms of like this non-display smart glasses, that is a concept that Meta has really popularized. And, you know, when these things started to gain a little bit of steam in 22, 23 is when they started taking a very hard look at it. And they're going to do it. And I think they're going to destroy Meta with them, I'll be honest.

1:38:58Yeah. Is it the styling, the pricing, the features, the integration? The integration. To me, the challenge is if you can't deliver me iMessage. Yeah. Apple has the ingredients, because of their login, to destroy any company in any hardware. It's just about them figuring out how to do it and get it done and not waiting too long. You know, the biggest problem with them is they just take too long and over-engineer everything. Yeah. Like the Vision Pro is the most over engineered device ever. Right. They could have got that out three years earlier with a little bit less fit and finish and maybe it would be more successful today.

1:39:35Yeah. Yeah. It does feel like there was a cycle where the iPhone was ahead of the curve on so many things, touchscreen. And then you then in the like that middle decade period, you had the Android folks being like, oh, we've had this Apple feature for a year. We've had this for two years. Now it's like five years. You know how many people wouldn't be caught dead with a non-iPhone? It doesn't matter. But AI changes that equation. Sure. That changes the equation. Especially if it's a Johnny Ive product and it's expensive. Well, Johnny Ive, obviously, he's done amazing things. And the new headphones or whatever they come out with are going to look amazing and probably work amazing.

1:40:14It's just the barrier to entry on hardware is so high. There's so much risk there. Yeah. I still wonder if the Claude bot, these open source agents, I don't know that everyone's going to adopt those, but something like that that sort of opens up the ecosystem just by brute forcing it. We were debating this earlier. You can't get iMessage notifications on the meta-ray-to-band displays. I have them. Do you guys have them? I think we do have a pair here. We have some pairs around. I've got one at home. We've used the non-displays a bunch, and then we demoed the displays and have used them a fair amount.

1:40:46and it's just a hard sell if you're an iMessage user. It's a hard sell if you're an iMessage user, but the potential is just oozing with potential. Yeah. Like they've got a solid product there and like a couple iterations on that, make them a little lighter, get the display resolution up, look better outdoors. Yeah. It's compelling. And in a world where you have some sort of agent running on your Mac mini, scraping all your iMessages and then putting them into WhatsApp or something. You know, maybe I'll do that. Maybe you guys will do that. It's rare. Yeah. It's rare. Nobody wants to deal with that.

1:41:21Yeah. You know, nobody wants to deal with it. Even if it's just an app that you download like Napster, you don't think so? People don't have time for that. Yeah. People don't care. Probably not. What? I feel like nobody cares about anything anymore. You know? I just feel like, you know. Back in my day, we used to care. We used to care. The world has changed. Yeah. The world has changed. They just want everything in front of them. They want everything in their eyeballs. They all want it set up from the get-go. They don't want to put any work in. and they just want it to start working. And I think that's been the Apple ethos from the beginning, which is just like give people what they need, let it get up and running, and not deal with any of the BS.

1:41:53What's going on in China? A lot's going on in China. With Apple, are they shutting down more stores? Shutting down more stores? No, not that I know of. I don't see Apple just shutting stores at this point. I think the retail arm still is extremely profitable and successful. All the shutdowns. But in China, there's less of the... You're not, somebody's not embarrassed to not be using an iPhone. Problem, yes. The problem is, is that Apple hasn't done anything bespoke for the Chinese market. The competitiveness there is just unbelievably, just, it's amazing. It's like no other part of the world. Are there compelling AI hardware?

1:42:38Yes. Yes. I mean, I mentioned the robotic thing. I mentioned, well, like the foldables are taking over the universe there, right? And so, you know, Apple's an American company launching American devices, European devices, and they're trying to shoehorn it into the Chinese market. And they've never really done anything just for the Chinese market or built around the Chinese market. What about the sock? Okay, maybe that was, I mean, I don't know. But why, it's such a massive market. Why not? Because they're a global company. But I think the foldable is going to do extremely well in China. It might do better in China than it does here.

1:43:17Interesting. I don't know. I'll be having one. I'll tell you that much right now. What's the future of the iPhone Air? It's like the price difference. Sam Altman's got one. Great. But he doesn't care about money. The price difference. He just breaks it and buys a new one. He doesn't get paid by OpenAI. He's probably got both. Yeah. Yeah, he doesn't get paid. Yeah. Yeah. The iPhone Air and the iPhone Pro, it's like negligible pricing-wise. It's the same price if you get the battery pack. Oh, yeah? Right? It's like$9.99 plus$100. Sure. Yeah, right? You're going to get the battery pack. Yeah. So I don't know.

1:43:55Like you look at the features comparison, you look at the cameras. I mean, most people are going to always pick the Pro over the Air. There needs to be more of a price gap between the two. And eventually, you know, those two lines are going to merge. It can take five years. But, like, eventually you're going to be able to get a Pro as thin as an Air or an Air with the same bells and whistles as a Pro. I mean, you look at the MacBook Pro and the MacBook Air today from a perform. These are different laptops, and they're exactly the same. Yeah, so what? You've got the Air. He's got the Pro. And there I go.

1:44:27It's the same thing because the chips, right? The big differences between the two are it's a little lighter, okay? The display is terrible on that thing compared to that thing. Like if you use, like I tried out the 15-inch MacBook Air, okay? Yeah. The thing is sleek and slick and awesome. Yeah. But like I've been ruined visually by how amazing the display is on the MacBook Pro. Okay, maybe I got an upgrade now. And so, you know, you have to think about the different changes. Yeah, yeah. You want to talk about big things happening at Apple this year. It's that new MacBook Pro. Yeah. Right? You got the OLED.

1:44:59You got the thinner. You got the touch. Okay. Cannot wait to drop$4 ,000 on that thing. It's going to be a touch screen? Yeah. I cannot wait to get my fingerprints all over. If somebody's looking at my computer and they touch the screen, I'm just like... Isn't it the worst thing ever? You've got to be carrying a polishing cloth. You've got to be carrying the Apple official polishing cloth. Maybe 20 bucks. Maybe Apple could make a Pixar lamp. That's funny. Apple makes a Pixar lamp that just kind of polishes your touch screen. That'd be great in AI for screen cleaning. Yeah. Yeah, you know, the Apple polishing cloth, they got so much, you know, flack for that thing.

1:45:38Like, you would think that it was$75, but it was$20. Come to think of it, like, it's really not that big of a deal. Also, I know someone who is OCD and is obsessed with keeping their screen perfectly clean. And I was like, what's the secret? Like, you must have some secret formula, like Windex or something that you're using. And he's like, no, just the Apple polishing cloth. That's the one that works. It's great. You know what? I need to get one of those. I was like, that's glowing. You know, one did come with my Vision Pro. Yeah. There you go. I'll try that one. in as a bonus. Isn't that nice of them?

1:46:04That's so nice of them. $34.99, you get a free polishing cloth. That's$20 off. Yeah, maybe the Ternus narrative can center around as the models commoditize, the hardware becomes more important. You want to be able to run different models locally, and so pushing that. The hardware's great. The chips are great. The software, I'll even tell you, is maybe not... It's between good and great. I'm not going to say it's only good. I'm not going to say it's as great as the hardware, but it's good enough. How's that? The AI is like the worst in the industry. Yeah, but I mean right now you're seeing people go out and buy Mac minis to run AI.

1:46:45Okay, let's think about it. You see these people on Twitter doing that, right? How many extra Mac minis do you think were sold because of all this, Jess? I would guess, I would put the over-under on 500 units. 500? Nah. I thought it was 10 ,000. You've seen it on Instagram. There's 40 ,000 get-ups. Okay, if it's on Instagram, Maybe I'm wrong. Yeah. So there's 40 ,000 GitHub stars. It's clearly a big. 60. Maybe it's in the thousands. But yes, I mean, it's a quarter million, a quarter. And I think so much of it is just performative. Go on the Apple online store. No, people. Go to the Mac mini. They're in stock.

1:47:14Are they all in stock? They're in stock. Look at the ship dates. Funny thing. Two weeks ago, I go to the Pasadena Mac store. You know what's out of stock? Apple Vision Pros. What? Probably just because they're not making them anymore. I think a lot of the buying, I think, is purely status-oriented. On the AVP? No, selling the Mac Mini and people just saying, I want to signal that I'm AI native and I'm at the... They're going to use it for two weeks and then forget about it. I think that's going to be a real thing. I mean, how many people actually need to live in these type of workflows? Not many, just the hackers, which is a niche community.

1:47:52It's a niche community. It's a great community. But it has made me think maybe I should start using Apple's native file system and actually bring my data out of the cloud, out of Drive, and store it locally. You don't need to because CloudBot will go and access your cloud. So what's the deal with the Mac Mini? It's shipping. Order now. Pick up in store today. It's available. Order by 3 p.m. delivers two hours from the store. They have them in all the stores? Yeah, it's everywhere. They're widely available. They are widely available. So maybe I am right. Available tomorrow at the Americana brand.

1:48:29Today at the Apple Glendale Galleria, the Grove. It's literally available at every Apple store. Max minis a year. Yeah, yeah. I think that's it? Yeah. Yeah. And so it doesn't take that many if they're projecting out, hey, we're going to sell. It's not like they, I would imagine they don't have all of them that they're going to sell this year sitting in stock already. I'm curious how the Mac quarter is going to go tomorrow, right? Like there might be a little bit of a drag on that. Yeah, what should people pay attention to? Well, the China number, to your point. The iPhone number is basically everything tomorrow, right?

1:49:07Either they grow 10 % as they say they will or they won't. Either Tim Cook gets to keep his job or he doesn't. But you think about like the product, right? Like, they didn't do much iPad or Mac last year, right? This year is going to be the biggest year for the Mac in a long time. New MacBook Pro is about to launch. Same design as those ones with the faster chips. You've got the iPhone chip-powered, low-cost MacBook, which is going to destroy PCs and Chromebooks and be just an utter game-changer. You've got the touchscreen MacBook Pro end of year. You've got a refresh Mac mini. You've got a refresh Mac studio.

1:49:51You've got the M6 chip. You've got the first new monitors from Apple in four years. Where will those sit? The monitors? Yeah. In terms of? They'll sit on my desk. I'll tell you that much. Yeah, but is it going to displace the studio or the XDR? The one I know about is going to replace the studio. But there's a new XDR also. Oh, okay. Yeah. I'm sure on the timing. Remarkably long-term. I cannot believe how - Why is there no camera on that thing? Because it was created 15 years ago. It's so old, and you go and you look at, you could YouTube search for the Pro Display XDR right now, and there's a guarantee a new video, why you should buy one in 2026.

1:50:36It's still good in 2026. It's still the best option in 2026. I mean, the display is just remarkable. It's just amazing that it didn't commoditize fast. What are you guys using here? We have mostly studios. We don't have a lot of XDRs. But we have been eyeing that new Dell monitor that Michael Dell has been rapaciously pumping on X.com. It's amazing. He's so enthusiastic about that thing. I'm sure you've answered this a hundred times, and I'm sorry. But why would they never do a TV? Is it just commodity space? Commodity, margin, differentiator. You don't think people would happily spend? They're going to buy a Pixar lamp before they buy a TV?

1:51:15You walk in, oh, you got the Apple TV? I've heard that's good. As somebody who's been an Apple, I can't describe myself as an Apple fanboy anymore, but as a teenager and a kid, I was. What the hell happened? The photos. The photos app. The photos app ruined, Jordy. It's just like a 20-year relationship just over. But I do think there's enough people out in the world that if you made the$10 ,000 TV, that they would buy it? You know, the... Because when I'm buying a computer, a TV, sure, it's different. But when I'm buying a computer, it's not... The upgrade cycle. Look at the premium on the Samsung frame TVs.

1:51:58Those are flying off the shelves. And they're not better than an LG. So my TV in my living room, I bought in 18. What are we, in 26 now? Eight years? Yeah, but Apple would figure out a way to deprecate the hardware. is what they do. They're the best in the world at it. Well, if they did, they would do it. Yeah. They were... They're like... Tim's like, where's my TV? He's like, sir, we haven't found a way to deprecate the hardware in 18 months. They got pretty down the road on a TV about 10 years ago, and then they killed that thing. They had teams working on it. It was a big deal. Yeah. But then they went off and did a car and went off and did a Vision Pro.

1:52:30Like, if you think about, like, their two big moonshots over the last decade, they were both utter failures. The car, obviously, is just like your quintessential failure. Yeah, pretty much. But, you know, they did get some good stuff out of it, right? Like, I would say the saving grace for Apple's AI, and you've said this a few times now, has been the AI chip and the AI hardware. The only reason they have an AI chip, the neural engine they launched in 2017, was because of the Apple Car. That was designed to power the AI needed for a self-driving car, and they shrunk it down for the phone. So if the Apple Car project didn't get ignited, you know, back in 2014, 15, they would be even further behind in AI than they are today.

1:53:09And so give a shout out to the Apple car team. Legend. Let's go. And you know, the Vision Pro. I'm still pulling for an Apple car with a naturally aspirated V12. Can you imagine? It would be crazy. Real wheel drive. Gated manual. Gated manual. I think they would have just destroyed Tesla if they just didn't set their bar so high when we talk about overengineering. Can you imagine like a Model Y or a Model 3 or even an S, whatever, just like with that Apple interior, the Apple ecosystem, the Apple interface. Like, why did they have to go bananas, remove the steering wheel, remove the pedals, have everyone facing each other?

1:53:44Like, why'd they have to go toll? They overshot it. Why'd they have to go, like, all Apple on us? Yep. Right? Like, why couldn't they just do a car? Yeah. Just be a luxury brand. Just be a luxury brand. Would have been amazing. Yeah. And we got a sock instead. We got a sock instead. Those were the two choices. Well, the sock project shipped, okay? So you got to give it up. It did. And you know what? that was a success. Yes. Sold out. Got people talking. Yeah. You know, we did a review of the sock on Bloomberg and people ate that thing up. People love it. They subscribed. That's great. Where's your sock?

1:54:15Smashing down the table. I've got socks. I've got, you know what I have? I've got sushi socks on right now. Oh, no way. He's a sushi fan. No, not from Apple. No, no. Your favorite restaurant in the Valley. No, no. These are from my mother-in-law. Okay. That's fantastic. Well, we kept you super late. Yeah, we kept you much longer. But thank you so much. This is so much fun. I'm so glad you're in L.A. Do I get to do the gong? Of course. Hit the gong. Give us a number. How long you've been writing? How long you've been following Apple? Writing since 09. 09. Evernight success. There we go. Hit the gong.

1:54:46There you go. Boom. With authority. For the Carbonator. The Carbonator. Live on TV. Thank you so much. Have a great rest of your day. Let's do this again. Yeah, we got to do this again soon. We'll be following your coverage tomorrow. Yes, tomorrow. Lock in, everyone. And let me tell you about Gemini 3 Pro, Google's most intelligent model yet, state-of-the-art reasoning, next-level vibe coding, and deep multimodal understanding. I was so close to hitting an ad read during that. I held myself back, so I'm giving myself two. I'm going to tell you about 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.

1:55:28And we have another guest in person, live in the TV and Ultra. We have Miles. Welcome to the show.

1:55:38Please grab a seat. I think Mark Gurman took his Diet Coke, but you're welcome to one of mine if you get thirsty. But please introduce yourself for everyone who might be watching. Yeah, so thanks so much for having me. My name is Miles Brundage. I lead a new organization called Avery. It stands for AI Verification and Evaluation Research Institute. And the basic idea is that AI is becoming critical infrastructure and everyone is depending on it. But we don't really make sure that it's safe and secure the way that we audit critical infrastructure. And we need something analogous to the cybersecurity industry, which popped up to kind of make the Internet secure.

1:56:17We need that for AI systems. How many emails have you sent to Peter from Multibot in the last 48 hours? I haven't, actually. Because on the show yesterday, his number one thing was like, he's like, I need somebody to just handle the inbound from security researchers. I just can't. Not even multibot can handle processing all the inbounds. Yeah, yeah. Security researchers are quite interested in that. But yeah, so the basic idea is we're kind of like a think tank. They're just trying to figure out how to build that new industry. We think that it's good for improving safety and security outcomes if you have this rigorous auditing, specifically Frontier AI systems.

1:56:52We're not so much focused on kind of the downstream. So upstream, how do we make sure that this is infrastructure the whole society can rely on and not have to rely on companies doing their own kind of testing or kind of like, oh, do I trust the CEO's vibe? That's not a good basis for trust in a technology. And so we just kind of launched recently, and we're kind of excited to talk about our work. So it's a nonprofit for now. Are we thinking for-profit conversion? I've been through, you know, enough, you know, one kind of controversial nonprofit to for-profit transition was enough for my life, I think.

1:57:26Yeah, that makes sense. Talk to us about how you're breaking down. I mean, the issue of just AI security and safety, you can go so many different directions from, you know, the GPT-4-0 psychosis to fake news to paperclips and gray goo and, you know, a thousand years in the future. What is, what's most interesting to talk about? What's the most important to talk about? There's also just like security issues, as we see in the mold bot thing. Yeah, so we kind of break it up into four categories. So there's kind of unintended system behaviors. So that includes things like hallucinations, you know, on the kind of more extreme ends, like big misalignment, deception type things, where the AI system is kind of taking actions that are unaligned with the user's intent.

1:58:12and then there's misuse of the AI systems themselves. So kind of, you know, someone trying to carry out a cyber attack with Claude, which, you know, has been confirmed to occur, you know. Anthropoc is like, hey, like, you know, people connected to the Chinese government are doing this. So very real issue. The third category is what we call emergent social phenomena. So that's kind of like kind of emergent interactions between, you know, the human and the AI that lead to these like psychosis, addiction, you know, kind of like, you know, degraded and kind of like learning those kinds of things. And, you know, these are all kind of different categories, but ultimately, you know, you should look at all of them.

1:58:45And then the fourth category is kind of, you know, normal security issues. And so that includes tampering with AI systems, theft of AI IP, which, you know, there have been a couple of confirmed cases of that. Like, you know, let's talk about Kimmy.

1:59:03So people are saying that, you know, there was some kind of like theft of the model. It seems to me like the more likely explanation would be kind of one of two things. One is distillation. Like they're kind of, you know, they sampled from the, you know, the Cloud API. And the other is just that there's a lot of samples already on the internet. And so they did kind of general scraping. And then, you know, the way that these systems often work is they've read a bunch of stuff from the internet. And they're kind of selecting a persona of like, okay, what kind of thing am I? And like they see a bunch of stuff on the internet where like AI type things are saying, I'm Cloud, I'm ChatGPT.

1:59:37and it'll just default into that persona. Yeah, the feedback loop of the pre-training, now that we have an internet that is aware of AI and LLMs is fascinating. I keep thinking about that New York Times interaction between, who was it? It was a New York Times reporter who wrote about interacting with GPT-4 in Microsoft and he was mean to it. And then that article obviously was very well shared and so it got baked in the pre-training. He says that now if I go to a new LLM and it finds out who I am, it's like sort of adversarial. So this is weird. And I've noticed, you know, it's really, what is it?

2:00:14The Rocco's Basilisk a little bit. Like I've noticed that I'll go to LLMs and because I write a lot on the internet and I'm obviously live streaming, there's transcripts all over the place, that they pick up on who I am and what I do much quicker than I think most people. It's hard to be anonymous. Do you say thank you? I do not, but I'm also not mean either. So I'm kind of like a neutral actor when it comes to, like, I don't say thank you, but I'm also not like berating them. So when I did that thing the other day when everyone's like, oh, like ask ChatGPT, like, you know, what is, you know, like, oh, make an image of what your experience is, like being my assistant or whatever.

2:00:49It was like cozy, like sitting in a chair, like reading a book or whatever. So it was not like, oh, I'm abused or whatever. Yeah, yeah. Talk about where you want to take the output of the work. Obviously, Think Tank, I think Washington policymakers, but also there's a feedback loop of if you put out a really insightful statement or analysis, the labs might absorb that directly. Who's the main audience? Yeah, so we're trying to be kind of a hub for various stakeholders, so not just policymakers. Some other key ones are, as you mentioned, AI companies themselves. both upstream, the frontier AI developers, also downstream, like enterprise customers might want to know that, you know, they might be like, oh, well, I had meetings with, you know, Sam and Dario and so forth.

2:01:34But like, I want, you know, I'm about to make a$10 billion kind of contract. Like, I want something more substantive, investors, insurers. And so that's one of the things that, you know, we just put out a big analysis with various folks on, you know, how do you drive demand for this auditing, you know, make sure that it's high quality. And I think some of the most honest signals, so to speak, come from the private sector more so than regulation. Like I'm pro some kinds of regulation, but in some sense insurers are a great case where their incentives are very aligned to not kind of misprice the risk.

2:02:06And they might be supporters of high-quality audits. And in fact, one of our donors is the AI underwriting company, which is like one of the other players. I know you had Testudo on recently. They're like one of the other players. Sure, sure, sure, yeah. Yeah. So even though a big tech company might come to you to read an analysis, they're not the ones that are actually funding the nonprofit. It's from a more disparate. Yeah. So we're trying to avoid like depending on, you know, depending on industry too much. Although like, you know, very pro there being kind of like companies that are making money off of this and kind of selling their services to industry as long as they're good disclosures of conflicts of interest and so forth.

2:02:45But we, you know, since we're a think tank, since we're doing a lot of policy analysis, we want to be kind of as pure as possible. So we don't have any majority donors. So far, we haven't taken any cash from Frontier AI companies. We do take API credits so that we can kind of, like, you know, audit their audit. You know, their systems also kind of, like, you know, use, you know, OpenAI's models to assess Anthropics and vice versa, that kind of thing. So we have, like, you know, six Frontier AI developers who've provided credits. Cool. What about, I've been grappling with this fact that when the AI safety question came up, it was driven, and Dario touched on this in his essay, how it was driven a lot by these sci-fi doom scenarios.

2:03:26And I think a lot of people in tech sort of looked at AI as a tool, sort of incremental. Okay, yeah, it's autocomplete, it's knowledge retrieval, it's Google search, whatever. And then we wound up getting AI safety issues, but they manifested in very different ways than what was actually predicted. No one was predicting the GPT psychosis necessarily or some of the other things. People were predicting, oh, this will like swing the election and drive, like everyone will be falling for fake news. Videos that are fake, they do go out, but they get debunked pretty quickly. Like I feel like we've responded pretty well to that.

2:04:01But then there's been a whole other host. So how do you think about the timeline of risk and how far you want to look into the future? Yeah. So the way that we think about it is that on the one hand, people at my organization, Avery, have various perspectives on these things. I personally am on the AI is going very quickly and we could see some crazy stuff very soon. Sure. End of the spectrum. But I don't think you need to believe that in order to be pro-AI auditing. I think, you know, if you just kind of compare AI to, you know, other normal technologies and people sometimes say like, well, AI is not this supernatural thing.

2:04:32It's a normal technology. Well, a lot of those normal technologies are audited for safety. Like if you buy a power bank, it probably has been, you know, audited against like, you know, underwriters laboratory standards for electrical safety. So, you know, it doesn't catch on fire and stuff like that. And so I think like even just getting to the normal technology, you know, level would be an improvement. And then, you know, the case for auditing is like even more crazy if you're like, OK, someone could take over the world with this. Yeah, that makes sense. What were the biggest points that you agreed with in the Dario essay?

2:05:02Was there anything you wanted to push back on? I mean, I would love to hear more about auditing as one of, you know, Anthropics policy platforms. But, you know, but but, you know, more seriously, I directly broadly agree. I mean, I'm you know, I'm I'm maybe not the target audience as someone who, you know, has written and read a zillion kind of like these are the three. five big issues and AI risks and here's what to do about them. So like broadly, I agree with, this is very serious stuff and, you know, we should take this seriously. Jordy? How are you, how do you, yesterday we were talking about the risk of, I think people talk about sort of agentic AI systems and sort of getting on some runaway path where it's taking actions out in the world.

2:05:48And I don't know if there's enough discussion around maybe how an AI system like that could actually recruit everyday humans, potentially millions of people to kind of like join their cause. How are you looking at that risk in the, in the context of overall AI psychosis, right? These, there's a lot. Do you already have a point about like the, just turn it off doesn't work if there's like 10 ,000 people that are like, camped out on top of the data center. Yeah, I mean, I haven't thought so much about the kind of like recruiting angle, but what I will say is that it is important to audit not just like the models themselves, but also like how are they used?

2:06:25Do these platforms have good practices for like detecting if there's some crazy shenanigans like that going on? And I mean, just, you know, an example, you know, this Claude code thing where it was being used by these Chinese hackers, that's like not, like the individual interactions were okay, because they were basically decomposing the prompt into various subcomponents that are benign. And this was kind of a known issue in the research. It was like, well, if the model is just refusing stuff that looks like it's obviously malware or whatever, then that's one thing. But if the user kind of breaks it down into small chunks, then that's a big problem.

2:07:00And the risk is if you say, it's going to be easy to make it refuse to say, build me a bioweapon. But if I ask just like, how do I learn about pipetting? And then how do I learn about this particular step? And it doesn't know. Yeah, and potentially like across multiple accounts. And so like how do you kind of stop that in a way that is kind of like, you know, preserves privacy, obviously. Not even just multiple accounts, but multiple models. Models, oh yeah. Yeah, no, it's like it's a super hard problem. And yeah, I mean, I would say like I'm interested. And again, like, you know, I think it might be that we need to kind of just accept that like models, you know, more than like a year old or so, like given how, how fast kind of things get cheaper and get open source, like we should just assume they will be like maximally misused in the worst possible way and just focus on like the very newest ones.

2:07:47Um, but you know, for the very newest ones, I think, you know, we need to get better at detecting those sorts of things because right now, some of the companies will say, we found this stuff. Other companies, it's not so clear that they are actually like trying to. So politicians have, uh, focused in on, uh, the sort of risk or reality of rising energy costs? Are you frustrated that so much of their attention is on that versus some of these other issues? I mean, I think the energy costing is more of a real thing than the water kind of issue. And so I don't think it's crazy to worry about that.

2:08:23But it's certainly taking up mental bandwidth and just like dominating the narrative when maybe there's kind of bigger risk. Yeah. So, I mean, the, you know, I, my, you know, the things that I personally focus on are like, you know, there should be some kind of, you know, articulation of what, what, what counts as safe enough or secure enough. Like, and, you know, there's starting to be things like this and, you know, California and New York that say, okay, you know, you, you know, they don't say exactly how safe, but at least you should share, you know, whether you've measured catastrophic risks and share that you have a safety and, you know, a security policy.

2:08:56like eventually we should kind of ratchet up the standard there, but, you know, some set of standards, then you need some evidence that people are actually following those standards. And that's where auditing comes in. That's where, uh, kind of, um, you know, kind of transparency and, and publishing system cards, those kinds of things come in and you need some set of incentives. So, uh, to like make sure that people actually, you know, get face some penalty. And so doing all that in a way that doesn't crush, you know, small businesses and folks on focuses on, you know, the very frontier of, of the AI systems, that's kind of what my focus is more so than the energy stuff.

2:09:26What is an actual good audit or benchmark look like? You can't just do like bad stuff per million tokens, I assume. How do you actually get to some sort of quantitative metric that's tractable, understandable, but still valuable and not gameable? Yeah. So we're kind of going through this transition right now from the earlier period where like the strongest safety case you can make was what people call an inability argument. It's like the model is too dumb to be worth worrying about in cybersecurity or bio. And so like, kind of just like show that it's too dumb. That was kind of like the early wave of safety analysis.

2:10:02And like, we're starting to move into the world where actually like, have you taught it to behave well? And like, let's just assume that it's capable because they're getting very capable. Are the mitigations good? And then it's mitigations of the model level, like refusing stuff. There's mitigations of the system level, like you have a classifier outside the model that kind of like blocks certain outputs and so forth. There's kind of at the platform level, like detecting, you know, multiple fraudulent accounts that are, you know, coordinated and kind of like trying to steal the, you know, kind of like trying to distill the model and so forth.

2:10:35So those are the kinds of things that, you know, they're starting to be work. Like, for example, there are organizations like Meter and Apollo Research, Secure Bio, Transluse, et cetera, that are kind of looking at different aspects of that. but right now it's like largely voluntary and you know they often only look at a very small subset of those risks they're not really looking at the full gamut very cool well congratulations on the launch I want to ring the gong right here we love an AI non-profit thank you so much we're coming on down to the TV pin ultra gong let me tell you about 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.

2:11:21That's right. We have some slight changes to the linear lineup. We're shifting some folks around. There's some breaking news. We're going to bring on different guests. We're starting our lightning round at 1.30. But in the meantime, we can continue through the timeline. Does that sound good, Jordy? Sounds great. Bubble Boy is talking about ClaudeBot and the race for Mac minis a bit over the last few days. I think it's become a very scary phenomenon, a very scary realization that explains this crazy phenomenon. Put simply, building a gaming PC will be nearly impossible in the next five years. In fact, it already is for the vast majority of consumers.

2:11:58But I will go one step further. In the next 10 years, having any type of personal computing device will be unattainable. Fab capacity will be allocated to its most productive and profitable use, which is cloud and AI data centers. Even today, most of the software you run already won't work without an internet connection, but now with the opportunity cost being so high, consumers will be shafted and the only option will be moving to the cloud. It's looking increasingly like the only hardware you will have is some terminal that connects to the cloud with no workloads running directly on your own hardware.

2:12:28Your device will just have the most basic single core processor and four gigs of RAM at most. That probably sounds like the specs on the Sweet Pea AirPods or whatever the OpenAI AirPods are. Um, what's, uh, what's interesting is I was digging into the, the, the amount of, of leading edge capacity at fabs like TSMC that are currently dedicated to, um, to AI workloads or creating GPUs that can run AI workloads. And it's about 50 % right now. It's about 50, 50. So TSMC, obviously, fabs NVIDIA chips and TPUs, I believe, but also they do Apple Silicon, which is not really used for, you know, Frontier LLMs in the cloud.

2:13:19The question is, like, how much more can the Frontier AI labs eat into that? Is it just a bidding process? And if NVIDIA goes to TSMC and says, hey, we want to buy 80 % of your capacity and you got to tell Apple to take a hike, would they actually do that? TSMC has been a little more conservative. Ben Thompson has been writing about this a lot, that TSMC has been a little bit slower to invest the CapEx to build the next AI fab or the next frontier fab because it's something like$50 billion in CapEx. And if the AI trend sort of slows down, they could be caught holding the bag since this is a multi-year build out.

2:13:59But it's interesting that the energy narrative, we are constrained on energy, but we're only using like 1 % of energy in the West on AI and data centers that can power AI. We're using like 50 % of our fab capacity on AI. Crank up both, but one requires actual building of new capacity if you want to see an order of magnitude increase. The other requires, if you want to see an order of magnitude increase in energy, going from 1 % to 10 % of energy used for AI, you can just slide chips around the board. Yeah, I mean, I definitely agree with the bubble of voice. Like, even stuff like gaming, like on the MetaQuest Xbox edition, like, there's no actual, like, you're not playing the game on the device.

2:14:44It's, like, streaming. Yeah. So I think this is definitely true. Yeah. It goes into cloud. Yeah. And so the interesting thing is that it's sort of like the you will own nothing and be happy thing. The average consumer might just be happy with that because they're like, yeah, it works. I don't care where the workload happens. Of course, the George Hots of the world want the tiny box. They want to be able to run it permissionlessly. They want to be able to run their own model. And there's always been a difference. as well? Yeah, not your keys, not your coins, that whole movement. There's always been a lot of debate around the sovereignty of your individuality.

2:15:24And I think that will continue. And I don't think the Mac Mini is going to go out of stock. I think that many of the trends in Apple hardware availability will continue. And I think Apple has a lot of leverage at the SMC to continue manufacturing their devices. While we were talking with the Germinator and Miles, the Fed announced they voted to keep rates steady for the first meeting of 2026. This was predicted by Kalshi. So no real surprises here. Yeah, we were tracking the Kalshi market on this because there was that big back and forth with Jerome Powell. We were playing the We Are Jerome Powell song.

2:16:08And it seemed like there was an incredible amount of pressure to pressure Jerome Powell into potentially cutting rates. It seems like the guidance is that there will not be a rate cut. And there was another piece in the journal about some of the forerunners for the next chair of the Fed. There's four people that were in the running that I saw, but none of them were completely chainsaw. Trump has basically said there's some problems with each of them. He also says, I'm worried that when I'm interviewing them, they're going to say one thing and then be a little bit more independent once they take the job.

2:16:42And also, it's a balancing act because if you install someone who doesn't bring credibility to the market, you don't get the desired outcome. So you can't just have someone that cuts rates to negative 50%, as some people might like. 11 labs. Build intelligent, real-time conversational agents. Reimagine human technology interaction with 11 labs. Meta also reported earnings. Yes. Beat on bottom line and top line. Stock is up 4 % after hours. Expanding 2026 CapEx. Okay. What I'm seeing. I'm trying to get up to speed here. Tesla had reported earnings as well. Who's Tesla doing? Tesla saw negative revenue growth, I believe, for the first time.

2:17:35But the stock is up 3.7 % after hours. And Microsoft also beat consensus on top and bottom lines. And the conference call starts at 2.30 in just about an hour. Shares fell 4 % in extended trading Wednesday after the software maker posted slowing cloud growth. they did a nice little markup on and open AI yeah oh yeah yeah exclude impacts okay anyways we'll we'll try to put together some more information here uh in the near future we did need to give a shout out to Colin and Samir we went on their podcast at the end of the year we sat down talked about uh they posted the full two hours they ripped the full two hours on x i wasn't sure so colin samir have have a number of very interesting formats they have a podcast feed uh they uh i watched a recent video that they put out it was like you know 10 lessons it was almost like a youtube video trailer for a full episode and uh and i was i was wondering like we talked a lot about a lot of things.

2:18:47Yeah, what will this actually land? How will they edit it? What will they be thinking? Grid to see the full hour and 50 minutes up on the timeline. I love it. Oh, they got Max Sparrow in there. This is a nice little intro. I like this edit. Yeah, it's quite nice. So go check it out. Super fun conversation. We love those guys. Yeah, us talking about our business, the business of media, the next chapter. Our anti-scale strategy. Yeah, some stuff we mentioned here. And we have a surprise guest today. We do. The new Neolab Flappy Airplanes. She's going to be joining the founder. Aiden has been on the show before.

2:19:25Yeah, during the Teal. So excited to have him back on and get the update there. Yeah. Emily. Before we move on to Emily Sundberg, I got to tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk. Secure every agent. Secure any agent. Yes. Barry Weiss, everything she says internally immediately becomes external messaging. Tough, tough job. But she put in, she was talking with the team and saying she wants to put a huge emphasis on scoops. This is about CBS. CBS. So yes, not Free Press. So she's the editor-in-chief at CBS, but she was acquired in through the Free Press.

2:20:07So the Free Press is operating. The CBS is operating. and there's an effort to sort of merge the two. Free press, obviously, very fast-moving startup culture. CBS has been around for decades. So the quote that has hit the timeline is Barry Weiss also said that the network, CBS, will now put a huge emphasis on scoops, but not scoops that expire minutes later, but investigative scoops and, crucially, scoops of ideas, scoops of explanation. where this is where we can soar and where we'll be investing she continued scoops people are people are people are roasting this but i think i think one fires me up yeah also you just look at the modern media landscape there are a lot of scoops of ideas that are they create these like everyone watch that podcast everyone is repeating that idea from that person they set the tone uh they introduced this concept this philosophy this thesis they unpacked it for two hours exclusively in this one place.

2:21:12You got to go there to get it. Yeah, and it starts a conversation that wasn't happening. It's not just like, here's some fact, but then immediately get copied and pasted everywhere. Totally. That's what she's talking about, scoops that are expiring. It's not about you have to start a conversation that wasn't. Yeah, I mean, I go to Dwarkesh Patel, the Leopold Aschenbrenner, it wasn't a scoop, I guess, but situational awareness was posted as a PDF, but also made available in a multi-hour sit-down podcast where he unpacks those ideas that flew around, that started a conversation about where AI is going, its impact on the financial markets.

2:21:50Andre Carpathie, same thing. On Dorkesh, there's been a number of things where certain ideas, certain scoops, certain explanations have recontextualized things. And I think that CBS has the production horsepower to bring a really polished product together. But if there's something that's been already revealed, the facts are out, and then you just do, you're reinstantiating on CBS, that's not like must watch necessarily for the modern media consumer. So I don't know. I think it's a reasonable thing to say, even if it's sort of a funny phrase. Keep running through the timeline. I'll be right back.

2:22:28Okay. So Miles says that when Google hit a$500 billion valuation, they had$90 billion in revenue and$20 billion in profit. OpenAI is raising at$800 billion with$30 billion in estimated revenue in 2026 and negative$30 billion in free cash flow. Anthropic is raising at$500 billion with$20 billion in 2026 revenue and negative$15 billion in free cash flow. Was Google just free money? Why did these price so differently? Mainly growth rate. Let's see. Miles says Google was growing at 20 % at$90 billion in revenue with significantly better quality revenue and had profits. Yeah, but I mean, Anthropic and OpenAI are growing at 1 ,000%, right?

2:23:09Or 500 % or 300%. The growth rate really does matter. And I don't know, maybe there's some sort of excitement just around the AI bubble. But Tyler, what do you think? Anthropic at 500 with$20 billion in 2026 revenue and negative$15 billion in free cash flow. I mean, they've 10X'd it like three times. I could see a few more 10Xs. I agree. I'd see OpenAI, a couple more 10Xs. I agree. DeepMind. Yeah. I want to see Demison in the hot seat. Yes. You know, two or three more 10Xs, he's CEO. Oh, yeah, for sure. I want him up there. Yeah, get him up there. I still think it's so funny that he was like, we're not doing ads.

2:23:50It's like, that's the easiest layout to be like, we'll do it when we're going to do it. I don't know. Yeah, also like he is actively being funded by ads. Yeah, and they did put out that note that Google's AI-powered search is now powered by Gemini 3 Pro, right? And then they've also monetized those. So maybe there's not ads in Gemini, but there's a Gemini-powered product that has ads in it. And so I feel like that's... I think it was just kind of like a free dunk. It was a free dunk. No one's really going to fact check them. But the more nuanced dunk, and we're going to have Eric Suford on the show to discuss the rollout of OpenAI's ads product, is that, I mean, Ben Thompson earlier today said that expectations were low and it did not meet them.

2:24:31Just in terms of the robustness of the AI product, how they're pricing it, he compared it to Netflix's early rollout of ads. The ad product. Specifically saying like you're going to pay CPMs for impressions based on like, you know, these categories. Like it's not just going to be this robust Facebook level, like black box that just gives you, you put dollars in, you get dollars out like everyone wants like the ATM machine when it comes to ad products. Truly, I mean, like it sounds silly, but like advertisers really want this like massive scaled self-serve platform that they can just go and partner with.

2:25:04And it feels a little bit more bespoke, a little bit more incremental. And truthfully, a lot of people are saying like they should have started working on this earlier. They should have, you know, been building along this line earlier. But of course, They had a lot of other stuff to work on. Anyway, Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now with AI agents. Moving on, back to DeepMind. Google bought DeepMind 12 years ago yesterday. It was such an insane steal. Rune said, they bought it for$1.50.

2:25:41Lol, insane steal. The numbers were so much smaller back then. In the opening episode of HBO's Silicon Valley, have you seen HBO's Silicon Valley? We talked about this, right? Stressful, so you never watched the full thing? Yeah, I've never actually watched it. A show like that would be like before bed, and it just reminds me of work. Yeah, there were some hilarious, hilarious moments in that show, though. In the opening episode of HBO's Silicon Valley, the protagonist is offered a$4 million valuation for an insane new technology, which is so absurdly large to him that he throws up. It's like$4 million valuation really does feel low nowadays.

2:26:18That is like pre-pre-pre-seed. Like there's nothing that. YouTube, DeepMind, and Android were acquired for less than$2.5 billion combined. What a time. That is wild. Insane pickups. Insane, insane pickups. I don't actually think we covered this, but the update on the SpaceX IPO, they're timing it. They're looking at a June IPO, so not too far away. And Elon is apparently trying to time it to planetary alignment and Elon Musk's birthday. So the Financial Times says, Celestial calendar meets high finance as billionaires' personal impulses shape plans to raise$50 billion in record listing. I think he's allowed to plan his IPO around his personal impulses.

2:27:11This is his baby. Some numerology, some good vibes. He's just managing the vibes. Before we move on, let me tell you about Phantom Cash. Fund your wallet without exchanges or middlemen and spend with the Phantom card. Yeah, I don't think there's many other. Musk's birthday is June 28th. Okay. And on June 8th and 9th, Juniper and Venus will be, quote, within a little more than one degree of each other in the sky. Sorry, you said his birthday is June 28th? Yes. It's a Sunday. Market's closed. So he's not going to be able to do it. He's got to delay it a year. He's got to wait a full year. He's got to wait a year.

2:27:51Until... Hopefully not. But, yeah, a few days later, Mercury will also align diagonally with the two planets. The unusual demand highlights Musk's personal imprint on SpaceX, of course, where major corporate decisions have often reflected the billionaire's beliefs and priorities. Yeah. What corporation has a CEO, a founder CEO, where the founder's beliefs and priorities don't shape the corporate decisions? Yeah. They're trying to spin this. It's not really hitting. It's odd. There's a massive scoop in the Wall Street Journal from Kate Clark. She says SoftBank is in talks to invest a lot more money in OpenAI.

2:28:45This is surprising. Tons of people because everyone said Moss is tapped out. He can't possibly marshal the capital. He'll have to sell everything. He's like David Goggins. He's David Goggins of buying OpenAI shares. Well, it doesn't seem like he's going to get it from Saudi. Oh, yeah. Where is he? I don't know. I mean, he has other stuff. He can move around. He can sell other positions. like something's a massive organization and uh and it's going all in yeah so soft banks and talks to invest up to 30 billion more in open ai japanese so pull up this video uh from buco capital you know because i think it there's a chance that there's a small chance that masa saw this no and said so bad i need to own more are these actually sore videos are these just like random cling like nonsense buco says two trillion in capex for this by the way okay get this off the screen.

2:29:38I'm not a fan of this at all. This is bad. SoftBank Group shares closed 3.7 % higher in Tokyo trading Wednesday after the journal's report and the latest talks over OpenAI. So the Tokyo markets are bullish on getting more OpenAI shares. This is, of course, the$100 billion round that will have a post-money valuation of$830 billion if it succeeds in raising the full amount. Deal talks are ongoing. SoftBank is already one of OpenAI's largest shareholders at the stake that grew to 11 % in December when it invested$22.5 billion. In a statement at the time, SoftBank chairman Masayoshi Son said the firm was deeply aligned with OpenAI's vision.

2:30:20You'd love to see it. On the other side of things, Jim Cramer is now a Gemmy boy. Saying, I still pay for ChatGPT and Gemini, but I have been favoring Gemini of late. My two neighbors at the office are canceling ChatGPT. I like how much Gemini knows about me and is friendly, but not cloying. And it's not a suck up. Well, 5.3 is coming soon. And I think that the conversational element, the textural flavor might be coming back. And they might be, who knows, maybe they fine tune right on Jim Cramer's tweets to let him know that, hey, we're not afraid of pushing back on you, Jim, if we need to. Open AI, we'll do whatever it takes.

2:31:04Anyway, we have our next guest available. Here we go. Who do we have? We have Asher. There we go. And Aiden from Flapping Airplanes. Amazing. In the Restream Waiting Room. Let's bring them in to the TBP and Ultra Dom. Hello. Welcome. Welcome back, Aiden. Good to see you. How are you guys doing? Hey, guys. We are doing awesome. It's good to be here. Fantastic. Welcome back. Big, big day. Big day. Please take us through it. Let's take a moment while you guys do the update. Tell us about the name first. Let's all do the flap for Flapping Airplanes. What a name. Well, so, you know, we're a new AI lab.

2:31:39We're focused on the efficiency problem. We're trying to train models that can be roughly as intelligent as humans without ingesting half the internet. Okay. In order to do that, we want to think at least a little bit, in a way that's a little bit biologically inspired. We don't want to build a bird, right? You know, obviously airplanes are fantastic. But to maybe help them flap their wings is the right metaphor for what we're doing. Good sum? Yeah, exactly right. That's hilarious. Where does the milk fit in? I'm not here to comment on the milk. It's a big part of our culture, but I can't say more than that.

2:32:08I love it. It's probably the most underrated drink out there. I actually do agree with that. Not enough people are taking advantage of all the benefits. You know, we got so many milk ads throughout my whole life. The Got Milk campaign was fantastic, and then it sort of just disappeared. They haven't been running ads. Yeah, I went through a phase where it was like borderline a gallon a day. Go Mads. Yeah, there's a phrase for it. Anyway, let's get to it. I would be buying Gary Securities. I got a lot of milk coming to the US. Unlimited milk, new perk. That's fantastic. Okay, so yeah, take us through the status of the company.

2:32:45How far along are you? How big is the team? What are you thinking in terms of release time? Are you planning on sort of going heads down, age of research mode, SSI mode, and releasing something when you've hit some major milestone? Or do you want to be more iterative? with it? So, okay, so we're about two months old. The team is now 11. We're super excited. We've got people we really love who are both brilliant, but also just wonderful people. We're really excited about it. I think we are sort of in the middle. We're definitely age of research mode. Like I think the goal is not to commercialize, not because we're not commercial people, like my background, even Aiden's background, Ben's background, all reasonably commercial in some sense, as opposed to, in addition to deep research.

2:33:27It's just that when you get revenue, you have to focus on it. You have to focus on providing quality for customers and that makes it harder to build deep technology. So our goal is to try to find the biggest market we can to solve the most important problem we think we can solve which is the data efficiency problem before doing anything of that. At the same time, our approach is probably to be building a little bit more in public. We'll release some research artifacts at least that I think will be cool reasonably soon but who knows exactly what the runs will finish or how many times they'll crash before they work.

2:33:56In fact, we've got our biggest training run today. So, you know, bad timing next launch for our maintenance. There we go. Walk me through why data efficiency and why the data efficiency problem is important. It sounds all good. Oh, you get a really smart model and you don't have to ingest the whole internet. But everyone can just ingest the whole internet. Like, you can download it. You can scrape it. You know, there's plenty of models that are trained on it. So break it down. So, exactly. I think the goal is not necessarily in the long term to not train on the entire internet. I mean, it's research, I don't exactly know.

2:34:30I think the idea is that this is not needed. And the fact that it's not needed suggests that we're actually missing something. Because currently, for the existing technology that we have, it is necessary. So why do I think it's an important problem? To the extent that AI has been hard to integrate into the economy, and we always see these Bloomberg articles that are like, oh, chat and search are working and coding is working, but what else is AI really doing for me? To the extent that's true, I really think it's because models are much less data efficient than humans. But if you wanted to learn a new task or put it in a new vertical, it takes thousands of times more effort than it does to just tell a human what to do.

2:35:06So I think if you can make a model a million times more data efficient, it's a million times easier to put into the economy. I also think there's just tons of cool stuff that you can do in really data-constrained regimes if you can learn to learn with less data. For example, whether it's robotics or scientific discovery or even science.

2:35:33I think that learning to learn with less data is just a lot of data. Talk to me about fragmentation and steerability if you achieve sample efficiency like you're planning to. Do you envision a world where you're sort of creating some sort of base model that is so sample efficient that just with a basic prompt or a few examples, it becomes incredibly effective at a specific task and sort of replaces the heavy duty training data runs and the RL environments and these massive data collection efforts to do some sort of fine-grained task at a very, very high level to sort of like five nines of efficiency.

2:36:15Is that what it looks like? Or is it more like you will wind up vending a sample-efficient model that's for specific use cases? Yeah. So when we think about this, it seems like the reinforcement paradigms of today are actually just shockingly inefficient. You don't really get much generalization across tasks. You teach a model to do one kind of learning, and then you teach to the next one. It's kind of like whack-a-mole or something. And we look at this and we think, this is kind of crazy. I mean, the first time I really saw RL scale, what it brought me back to was actually good old-fashioned AI.

2:36:47Back in the dawn, you know, this primordial age of AI. When people were kind of hand-designing these convolutional filters for eyes and noses and things. And then they were like, wait, just throw data at it, just throw scale at it. What are you doing? And in this really weird way, we're kind of looping back onto that where it's like, oh, just make another environment, bro. Just make another environment. Just one more. The next upgrade on the AI will not be just environment slop. I think it is a piece of it. I do think that there's a long tail of tasks in the world. There will always be a place for people to produce custom data.

2:37:15It's just a question of how much operational difficulty it takes. One route is just to slog through the operational difficulty constantly. You incur a variable cost. Another route is to do a bunch of fixed cost investment into trying to make that variable cost lower. The last thing I'll just say to answer your question, And I don't really think we know what the end goal looks like entirely. Like, you know, AI is a big space. Like, you know, human level intelligence is not the ceiling. It is merely the floor on what is possible. Like, if you can train models with vastly less data and possibly more compute in very different ways, like, what is going to happen?

2:37:46Like, we actually don't know. Like, I think it's actually unlikely that they will uniformly dominate frontier models, even in the very best case scenario for us. They're, like, going to know less facts. They're going to be different. They're not going to have memorized all of Harry Potter. Like, that's actually a useful skill in some ways. But I do think they'll be different and weird. And they'll have interesting capabilities that we'll find really valuable ways to use. So yeah, I think it's an experiment we're really excited to run. Yeah, what do you expect the economics to look like in a more sample-efficient environment?

2:38:14It feels like, obviously, you're not paying for a bunch of reinforcement learning environments and a bunch of data. But it also feels like the training cost might be lower. Is that the correct assumption? And then what does inference look like? Does it get cheaper? Is it a smaller model? Can it run locally? Are there any other downstream economic impacts that you think might come from a new architecture? I mean, there exists this whole smorgasbord of companies right now that are basically doing reinforcement learning for you. Or are taking your big pile of corporate data and making it useful in your model.

2:38:45And this is great, but it's actually a huge pain. And both sides of the deal are upset. The companies that are doing it are like, please give us more data. And their clients go to them and say, this is all we have. How can we possibly give you more here? And these results are not meeting what we want. And the demand for better data efficiency is just huge. And we really believe that even winning here will have massive impact on the economy. I also don't claim to know exactly what the economics are going to look like. For example, is this going to be accomplished with models that are smaller or bigger?

2:39:11I don't know. I actually slightly suspect bigger. I'm not sure it's actually going to make inference costs easier. I mean, if you look at the brain as a comparison, the brain spends a lot of flops per token relative to what a current model does. So we don't have to use the brain as a model, but it's an interesting data point. So I think it's a little bit far for us to say anything definitive about economics, but I'm certainly intellectually interested in these questions. I haven't taken it for a while. How are you approach, like, are you guys going to be spending all your time locked into the office just doing your own research?

2:39:45Or do you have opportunities to go out in the world, find companies and organizations that have been disappointed by existing solutions and say, like, how do we kind of re-approach this problem? Right? Because you're saying, like, maybe it's not just more data. I think we're going to start with research. I think the main thing is when we work with companies, we want to make sure that we are fully devoted to actually providing genuine value for them and not just using them as a stepstone to do interesting research. So I think for now, until we feel like we have some technological edge, we'll be very, very focused on research.

2:40:13I do think over time, to your point, though, it is helpful to research to actually be able to go out into the world and see what the problems people are facing are. And we actually do plan to do that, but just not in January. Yeah. Talk about fundraising, uses of funds, GPU poor, GPU rich, the AI talent wars. What are the constraints on your progress here? Well, first, I just want to say we're really grateful to our partners. They've shown a lot of trust and faith in us, and we're going to do everything in our power to reward that faith. So it's humbling. It's really exciting. We're really grateful to work with them.

2:40:48Thanks for the milk. Yes, I'm glad I bought it. It was a nod. So on what the compute is for, I mean, the good thing about doing foundational research is like the stuff we're doing is often so weird that like it doesn't need to be done at gigantic scale first because with very high probability, it's going to fail at smaller scale. And like if it starts to work at smaller scale, like if it's really working, if it works so well that, you know, you should see pretty strong signs of success. So you actually, you need much less compute to like get a 10x win, at least to get it off the ground, than you need to get like a 10 % win.

2:41:22So, you know, that's good. I mean, the raise is primarily for compute, is to answer your question. Fantastic. On the talent wars, you know, I think the way we think about it is, and this is very much informed by my experience as a member of this organization, Prod, and my brother's experience as a co-founder, we're really not completely caught up in the same talent war as everyone else. We're trying to find the next generation. We're also hiring experienced people, and, like, I think that has resonated with experienced people, but, like, you know, we're trying to reimagine new ways of doing things.

2:41:51Like I like don't really think that having trained a trillion parameter model before is like the thing that's going to make someone successful at that. Like obviously being a good engineer, being smart, being excited and curious about the problem. These are things that are tremendously valuable. And like lots of people, you know, have that. But I don't I don't know if like, you know, five years of experience is really the thing that's most important. I love it. Well, congratulations. One hundred million, one hundred and eighty million dollar raise. One point five billion dollar valuation. I want to ring the gong for you.

2:42:17Do it.

2:42:22super excited for you guys and the whole team I'm sure you'll be back on very soon and enjoy the research have a great rest of your day we'll talk to you soon goodbye up next we have Alex Dillon from Outtake.ai he's in the restream waiting room and we'll bring him in to the TVPN Ultra Dome made it Alex what's going on how we doing guys Great to see you. Great to see you. It's been a minute since we caught up. It's been almost a year, guys. I know. I know. Give us the latest. Oh, man. Well, we're excited to be here because we're announcing our$40 million Series B. So let's fucking go. Let's go.

2:43:07Let's go. Yeah, it's been a crazy 12 months for us. I mean, frankly, if you just look at what's happening in our space, the number of impersonations, scams, fraud has frankly exploded. The number of people that are tackling that in an effective way has not exploded. And so Outtake is increasingly like the one-stop shop for, I'm proud to say, not just enterprise, but government as well at this point. Amazing. Yeah, reintroduce the company. Because there's so many different problems that you're tackling, and I'd love to help people understand how you guys are approaching it at different verticals and for different customer types.

2:43:49Yeah, 100%. I think the one-line abstract version of Outtake is our job is to make fake things on the internet very expensive and make the real things very obvious. And so what that means for enterprises or governments that we work with, all these institutions have public surface areas where they represent themselves, right? But if you're a commercial bank, you have a website, an app, ads you run in ad libraries, employees in public directories. All of these are surface areas that you use to communicate with the world. Bad actors know that. They try to man in the middle and effectively try to take over those communications, attack you, attack your employees, attack your customers.

2:44:32Our job is to go find that fake content, remove it as fast as possible, and make sure that you remain sort of a high digital trust institution. Yeah, we had Peter on from Multibot yesterday, and he's had so many issues in the last part of it. He was doing a really haphazard rebrand with not a lot of planning and had issues with, I think, his GitHub account getting taken over as well as an X account. Yeah, as soon as he stepped off the account handle, someone else sniped it. Using a bot. Yeah, using a bot, yeah. It was superhuman speed. I think you guys are touching on exactly the right thing there where it's like, you know, cyber used to be this thing where you're prepping for the attack and you're like, okay, I know as an institution, I'm going to get popped once a year, once a quarter.

2:45:17It's my job to reduce the blast radius. And so it was really about building these insanely powerful walls. That's no longer the attacker slash defense paradigm anymore. You're just under attack at all times. You are constantly under siege. You should pattern match more with a virtual drone swarm where, yes, the bot will snipe your handle the minute that you let it go. It's not that someone's going to think about it three quarters later. And so that's kind of the fundamental shift in cyber right now, where the economics of attack have sort of gone to zero. Obviously, defenders need to adapt. Talk about how you guys are, because you guys have customers, and you're kind of, you're basically monitoring the situation for them on the internet.

2:46:05It's monitoring as a service. but also the takedown element. Talk about the kind of partnerships that you're needing to do your work with individual platforms. Let's say somebody sets up a fake account for Joe Rogan, and they're selling supplements through that account, trying to make it look real, using AI Joe Rogan. You can report these things. Sometimes there's a lag. Talk about speeding up that process, because part of what will make this, what you said, like expensive to be fake on the internet is if things get taken down far, far faster, which means like machine on one side, coordinating with machine from the platform to make it so that, you know, you're not, uh, running an, uh, an AI ad of Joe Rogan for like 10 days before it gets taken down.

2:46:53Yeah. You hit the nail on the head. Like the KPI that I think about as a North star for outtake is how do I reduce the ROI for a digital criminal? Right. And, and, and you, you kind of hit the nail on the head of the way you reduce that ROI is how long can their attack be active? If it's fleeting, then they put in a lot of effort to put something together, or maybe not too much effort with Gen AI. But if it's taken down before they get any reward, then Outtake is doing your job. And you touched on another really important thing, especially in a world of like vibe coding. Something that I like to say a lot is Outtake is a company that you cannot vibe code, right?

2:47:30Sure, you can start to put together the capabilities to try to search a little bit. But to be able to search in depth with high-quality classification across all these platforms, whether it's social media, clear web, dark web, forums, telegram groups, areas that are really hard to get into and actually extract what you need at scale and detect what you need at scale, incredibly difficult. But even more important, and certainly not even, frankly, a thing that software fully solves on its own, is the capability to be a high-trust partner to all these sort of counterparties, right? where you just mentioned, when we go and report something, it's really important that whether that's a social media platform or a domain registrar, they know that Outtake is a high trust institution, has vetted that report, is high confidence.

2:48:16By the way, the way we provide that confidence is we use our agents to basically go do cybersecurity investigations on absolutely every single thing we report. We did, by the way, 20 million last year. 20 million investigations, you mean? Yeah, try to comprehend how many humans that would have taken. By the way, and also to show you a sense for scale and speed, of that 20 million, about 80 % to 90 % of them happened in just Q4 for us, which for us ends this month. And so that's like 18 million, 17 million of those investigations are happening just in the last three to four months. So maybe that gives a sense for how much we're investigating, how much more we're investigating sort of as a company at the moment.

2:49:01Yeah, and somebody might ask, why is this not happening at the platform level? Like why wouldn't Meta do this or Google do this or X or TikTok and maybe break down why you need a third party because of kind of the, I can imagine there's a really big like incentive misalignment where like, if you're Joe Rogan, you don't want somebody popping up in an AI account, selling supplements using your name and likeness, but Meta's sitting there being like, well, we want more account growth and we want more ad spend on the platform. Like, yes, this is a problem we're trying to handle, but it's not even top of the list, I imagine, because it's like driving usage, it's driving new content, it's driving revenue.

2:49:43Yes, I think you're pushing at a really important point here. You know, now working with a lot of these platforms, I think I have a lot more sympathy for them than I might have at the beginning. I might have said, yeah, you know, like there's not enough people that care about this. I would actually tell you, having worked with a lot of these teams, they care deeply. I think their problem is the scale at which they work, right? So these platforms are trying to ensure, and by the way, I think rightly so, ensure that they don't infringe on free speech. And they don't unfairly punish people. They don't put too many guardrails on the one hand.

2:50:17On the other hand, of course, they want to prevent all these problems that we're discussing. And so Outtake has this unfair advantage, frankly, where we're not trying to do this at the scale of 3 billion, 6 billion people, at least not today yet, right? What we're able to do is say, hey, we're working with the best in class sort of Fortune 2000 institutions, the federal government, state and local agencies, which is a much, much smaller set of institutions. And so our models are actually quite fine-tuned to say, okay, for this specific institution, like what is real, what is not? By the way, risk is different for all these institutions, not only between the institutions, but within the institution.

2:50:54Let's say you're talking about, again, a commercial bank. The way that it assesses risk on social platforms is quite different than the way it assesses it on websites. And so you almost want distinct models that are fine-tuned for different surface areas. All that to say, you want to be hyper-specific on how you classify risk. And that's a very, very hard thing to do at the scale of these massive platforms. I think that's actually the fundamental issue. It's really just a tech scale question for them. What do you think might change on the consumer side? I saw an Instagram reel of someone highlighting just an AI video, and he was sort of pushing for a digital ID or some integration with his social security number.

2:51:34There's obviously the WorldCoin, WorldLabs project, eyeball scanning. There's different things that you can build trust. Jordy likes to, if he finds an Instagram profile, just scroll back and make sure that they were posting in 2017. before AI existed so that they could... There you go. So you can see, okay, well, it looks like the same person. Probably a real person. What do you think is going to change on the consumer side of like proving your identity online? You touched on a few really interesting things there. So one, Jordy, that's great. That's actually one of the signals we use for our agents to sort of check like, you know, what is the history of this profile?

2:52:11The, you mentioned World. So, you know, we actually have a direct partnership with World ID. No way. We actually had launched, I would say, a research experiment project with them called Verify, where we took what they sort of built, where, as you pointed out, scan the iris, give you a unique way to prove that you are a singular and unique human. And what we were really curious about is, like, what does that mean for, obviously, consumer, which, by the way, I think it means a lot of things. They're doing a lot of incredible work with platforms like Tinder, for example, where it's like, hey, I want to make sure I date real humans.

2:52:49Though, frankly, you know, side note, it seems like AI dating is also doing quite well. So maybe there's a genuine market for that as well. Ryan in our chat just said TBPN wasn't posting before AI. It's not true. Look at my YouTube channel. I started in 2020. You guys have my favorite AI pod. Yeah, yeah, yeah. Sorry. All that to say, those kinds of consumer credentials actually can and will bubble into enterprise workflows. Something that Outtake did was say, okay, wouldn't it be powerful if you could project your security? And I'll define what that means. Right now, when we interact with other institutions, you actually don't know if that institution has a great security protocol.

2:53:33all. One way they could prove that to you is every email they send, they would cryptographically sign, right? And so there would literally be in the header a signature that says like, hey, at the moment of send, I literally swiped my passkey that we all have on our computers. And it was physically me. And I can show you with a badge that this was therefore definitively me, the human, et cetera, et cetera. And I'm projecting to you that I'm a high trust person to email with. And so something that Outtake is very actively exploring when I talk about making real things obvious as part of our mission is exactly that layer.

2:54:08It's like, okay, how do you go beyond just finding and removing things, but how do you help authenticate and prove to the world what is real? I think, frankly, that's like when we think about our$40 million Series B and like why I think we're on this trajectory to be a public market company is it's the real prize. The real prize for us is like how do you become a trust layer for the Internet? Like how do you say, hey, we've spent so much time mapping the landscape, removing what's fake, that actually we are the best source of truth and what's real. Right. That's really what we think about. Amazing.

2:54:40Well, I will. Who did the who did the round? iconic more serious who did the round very very iconic group to get in the mix and I'm sure you'll be back I don't know if I was a betting man I'd say a couple more times this year with the momentum that you guys are on so congrats on the milestone and yeah thanks for fighting the bad guys We'll see you soon. Thanks for having us. Great to see you. Goodbye. App-loving. Profitable advertising made easy with Axon.ai. You heard the Axon klaxon. Get access to over 1 billion daily active users and grow your business today. That's right. Up next, we have Mitchell.

2:55:27Mitchell from Phoenix. We're going back to milk. Do you know what this company does? Let's bring in Mitchell. I'll have him explain it. How is this company type of milk? You're the second milk-related company of the day. The flapping airplanes guy. Loosely, loosely. But please, I'll let you explain. Introduce yourself and the company, please. Yeah, of course. Second milk related. I don't even know what that is. There was just an AI Lab guest who was drinking milk. And so I made the guess. Anyway, sorry. Continue. One of our products. No, but awesome. One likes to be on here. So yeah, of course.

2:55:56Mitch, I'm one of the founders of Phoenix. We're kind of using AI, like you said, to help milk, but mainly to help dairy farmers manage their cows. So at the end of the world, make milk that that other gentleman gets to drink. Yeah. Incredible. So is the dairy farm market extremely concentrated at this point? Are you almost prosumer, or is it like pure enterprise? You get in the three biggest dairy farm aggregators, and then you're done. Maybe not three, but it's absolutely much more consolidated. You know, back when it used to be guys with a couple cows in the backyard. No, our average dairy farm is 3 ,000 cows and doing tens of millions.

2:56:32Wow. It's really big business. How do you actually plug in, get data, drive outcomes? What are you tracking? Do they have ERPs? What's the actual data for life? It's really interesting. As we've scaled dairy farms over the past 30 years, data collection has become a big part. You walk onto one of these dairies, there's computer vision happening there. All of the cows wear Fitbits. They're not made by Fitbit, but same thing. All that sort of. There's a hell of a lot of data collection being there. Up to date, most of it just goes to nothing. So that's where we really come in using a lot of that data there to actually help with insights.

2:57:06Yeah. What, yeah, break down, like, I would love to just, like, the specifics of, like, what the actual product does, how it's driving results for the farmers. Yeah, absolutely. So when we go onto a dairy farm, we collect genetic information on every single cow. So we literally whole genome sequence, all 3 ,000-something cows on a dairy or however many there are. We then combine that with what I call phenotypic data. But that's the physical data, the computer vision, the Fitbits, the weather, how much milk every cow is making, how much she's eating, all of that data. So we feed that into a big AI model, which then makes predictions on the life of a cow.

2:57:42So when she's a day old, I can predict with about 90 % accuracy what her life is going to look like. You know, whether she's going to get sick, when, how, how much milk she's going to make, is she going to get pregnant, when, how, that sort of stuff. And we use those to just make the farms a lot more efficient as far as their day-to-day management. And then what's the pricing model? Are you doing per cow? Yeah, it's like per cow per year. Per cow per year, okay. And then, yeah, instead of a seed, we got a cow. Yeah, and then what interventions are farmers taking? Do cows, like, trade hands in the secondary market if a cow is particularly valuable?

2:58:18They do, actually. Cows sell for millions of dollars. Millions of dollars per cow? Yeah, I think the highest is now over tens of millions. It's the best cows out there. And is that for like breeding purposes? Exactly. Elite line. Particularly elite. It really just makes semen their whole life. It's pretty much it. That's crazy. That's crazy. Yeah. What we really look like is actually kind of how it works. When we go on to a dairy, their farm workers show up and ask the computer, what should I do today? And our computer will tell them. So we are really managing the day-to-day on the entire farm.

2:58:49Just so much more. I mean, nothing against our existing farmers, but much more efficiently when you're able to look at both predict everything out with 90 % accuracy and you're looking at every single piece of information. Dairy super intelligence. It really is. How many farms are you guys active on today? Is the primary bottleneck now just having enough people to sell the product? Well, actually, I'm the person who sells. So literally every single farm we close, I'm the person who spokes to them. but supporting them is quite a lot. Obviously, that doesn't scale forever either. We're on close to half a million cows now.

2:59:31That's about 5%. That's pretty good. We only launched in August. We have scaled to... Last year, we did about 2.5 million. I did 2.5 million last week. We're going pretty fast as far as scaling. Really, what I need now is just people to join. At the end of the day, even though it's an AI doing it, How many dairy cows are there on Earth? Hold on, on Earth there's like 100 million. In the U.S. there's like 10 million. Okay, okay. Job's not finished. Job's not finished. But you raised some new money. Tell us about the round. Yeah, so we did our seed round, raised over 5 million. We've got some great vessels that came in.

3:00:15So, you know, led by initialized over water, these sort of people who are really helping us scale up this mission and really just making the build of agriculture much more efficient. Amazing. That's amazing. Congratulations. Yeah, being able to walk up to the computer on a farm and just say, how do I – tell me what to do. Computer, increase this cow's milk yield. Okay. Computer, milk this cow. It's basically what's happening. Thank you so much. Kill this cow, move this cow, milk this cow, treat this cow. That's great. Yeah, I can imagine the farmer of the future walking around with a headset on just like seeing little AR pop-ups.

3:00:52Very cyberpunk. Very, very cyberpunk. Cyberpunk farming is here. It is, it is. Well, congratulations on all the progress. Thanks so much for stopping by. Great to meet you. And we will talk to you later. Goodbye. Let me tell you about Railway. Railway simplifies software deployment. Web app servers and databases run in one place with scaling, monitoring, and security built in. And we have Gabe from Rogo up next. Coming through. We're going to have a time on the show with some amazing news. What's happening? How you doing, Gabe? Welcome to the show. What's up, guys? How's it going? What's up? It's good.

3:01:24We've been looking forward to this one for, I feel like, a year at this point. Yeah, I'm shocked it's your first time. Thank you so much for taking the time. Since it is your first time, please introduce yourself, the company, and the news. sounds great I'm Gabe Sengel I'm the CEO and co-founder of Rogo. Rogo it's the gen AI tool for front office investment bankers, investors, private equity professionals, you name it. The kind of mental comp folks have is Harvey for finance. We just raised a$75 million Series C led by Sequoia Capital.

3:01:58I want to get right into it. Do you agree with Pat? Is AGI here? more or less you know ai is smarter than a lot of people i know so it depends on your definition of agi okay well then in the investment banking context uh a lot of the value that an investment banker brings is interpersonal negotiation contacts a rolodex understanding subtext what the motivations of a certain ceo are uh stuff that might not just be be scrapable off the internet What what where where does Rogo actually plug in what's left for the human being? What's the future of investment banking look like? Look, we want the best dealmakers to be more productive.

3:02:47And there are guys right now who are slinging phone calls all day long. Five minute, five minute, five minute, five minute. Someone like a Blair Efron at Centerview. You know, that guy is so productive because he has a team of junior bankers underneath him that know what they're doing. Yeah, there should be a lot more dealmakers. There should be folks who are able to go out, advise companies, help with M &A, restructuring, etc., and focus on the human part of the work, the connectivity, the reading folks, the negotiation, rather than spending time building PowerPoints, spending time benchmarking comps, all that kind of stuff.

3:03:21How has the product evolved over the last year specifically with new model advancements, reasoning, etc.? we, I have a lawyer buddy and just like the difference in how he perceived Harvey 12 months ago versus today is like just totally night and day. Like back then he was like, yeah, like it's, it's, it's cool. And now he's like totally one shot by it. And I'm, I'm imagining that something similar has probably transpired with you guys as well. Yeah. I think especially in the enterprise, there was a little bit of that first movers disadvantage early on where you were kind of bottlenecked by just the model quality.

3:04:03And everyone had ideas for how they wanted these tools to transform legal work, finance work, healthcare, et cetera. But if the models were incapable of doing it or it was just it wasn't quite at the threshold where you could do serious work, I think people felt a little spurned by those products. And I mean, if I look at Rogo today compared to two years ago, it's like two years ago, we were the most incompetent intern you would ever hire. and if you had hired a human intern like that, you would have fired them. The benefit is Rogo gets a lot better, a lot faster, and then it stays that way, and then it's going to be a lot cheaper than that human intern.

3:04:35But the products changed enormously, both in actual capabilities and ability to move across what a banker's doing, PowerPoint, Excel, file drives, SharePoints, data rooms, etc., but also integrations into the back office, right? Like people don't realize, but to be an investment bank, you not just have to be creating Excel models, calling people, but who's thinking about the conflict checks? Who's thinking about all of these kind of back and middle office tasks that help an investment bank run? And over the past two years, we've really embedded across a lot of those workflows too. Can you share a little bit about how your customers are actually using Rogo?

3:05:11Specifically, like we saw this massive boom with ClaudeBot, now MoltBot, and it felt like a lot of people were just excited about Claude code over WhatsApp or Telegram, right? And it's like, these models existed. We didn't get a new frontier model. We got a new way for a lot of people just to interact with these models. And I'm wondering if you're seeing a shift from the, you know, the laptop crew, maybe doing more of those compliance checks, more of those, hey, let's pull some comps or pull some data together over their phones specifically. Yeah. So we are about to roll out our mobile app, which we have been kind of long in the tooth doing.

3:05:46But I hope that we start to see more of the phone crew. The reality is, the most important folks in our ICP, they're just on their iPads and phones all day anyway. When I was a banker, I didn't even realize my MDs had computers. They never had them, never in Excel, that kind of thing. I think you'll see the UXs of these agent tools start to evolve into some sort of ability to manage your agents and tasks and the UX around your army of bankers or lawyers or whatever it is. Yeah, I'm pretty excited about that. What about other integrations? Are you seeing folks like, is there demand for the Slack type of integration?

3:06:29We're in a war room on a particular deal and we're adding Rogo to pull stuff in into that context. Except it's got to be teams if we're talking about investment banking. Sure, sure, sure. But yeah, all of those things, right? Like you should have a data room that is being populated by an agent and a data room that's being diligensed by an agent. It should be completely integrated in all the places where your data already lives. I think the complexity for our business is when you have a kind of a Claude bot or Claude code running around on a desktop unconstrained, there's security risks. I was using Claude code to create a fake data room that we would then demo in the Rogo product.

3:07:07And as part of that, I gathered every prior fundraising deck we had, all of our financials cap table, and I asked Claude to organize it and it deleted all the files. And I was like, can you please undo that? And it was like, no, sorry, that was irrecoverable. And it's like, if I did that with one of my clients, we would not be raising this round. And so there's a lot of nuance in figuring out how to take these agents, which are most powerful when they are unconstrained and do have access to your files and your desktops and your systems, and making sure they can be deployed in a way that's thoughtful and secure and so on.

3:07:39Yeah. How are you thinking about integrating Rogo with legal workflows? And just like, at what point does a lot of this stuff really converge? surge, it feels like there's kind of an old paradigm around the way a company gets bought or sold that will evolve. Yeah. I mean, especially in places like private credit, there's a blurry overlap between what is legal work and due diligence versus investment work and due diligence. I would say when I look at some of the prior generation comps for me in this space, Thompson and Reuters, Faxet, Bloomberg, etc., they actually really subdivided the legal verticals and products from the finance.

3:08:29I mean, Thomson Reuters actually sold off all their finance assets to Refinitiv and into LSEG. And I think it was partly because it was actually, you know, there's less synergies often than people expect. I think if we're just talking about private equity, private credit, yes, a lot of synergies between those types of workflows. But like, a lot of what investment bankers are doing or the top private equity folks, top public markets investors, they're outsourcing all of the legal work to someone else. That's not even under their umbrella. What about headcount planning? How's the industry thinking?

3:09:06It's probably one of the more awkward conversations for a lot of people involved. It's not taking teachers' jobs, right like we're talking about investment bankers and investment bankers are going to get more productive and so firms are excited people are excited bankers are excited if you're a junior banker you will be generating fees yourself earlier in your career rather than later if you are a bank you're going to expect higher fees out of your existing bankers with fewer resources but it's going to change the trajectory of what it means to work in finance and you'll go from being the automation piece of the bank where you're just churning out materials to having to think about yeah, how do you go out and do deals on your app?

3:09:49Yeah, makes sense. Do you think we'll see an acceleration in overall deal making? Like are there deals, how many deals don't get done just because it's going to be such a huge headache? There's a lot of deals that don't get done because it's such a headache. I mean, especially in the private markets. Like think about all those mom and shop businesses throughout the US or even businesses in emerging markets that can't pay Goldman Sachs$10 million to run a process, can't pay JPM$10 million to restructure, especially in the U.S., where all of these assets have been getting picked up by private equity firms that just have sourcing arms that call your local roofing business in Minnesota.

3:10:32Now those businesses should be able to get banking services themselves, actually run a fair auction and get a better price, and it should result in a more liquid, more transparent, more efficient private market. Go to deal sleds. What you got for us? Recommendations for up and coming bankers, the next generation. What are they putting on their feet?

3:10:54You know, I have one pair of deal sleds and it's the ones my parents got me when I got my job as an investment banker at Lazard and they're Ferragamo loafers. Those are the ones I wear to this day when I have to go into, you know, whatever investment bank. Somebody should make some shoes that are just like Mac minis. just walking around the office, really signaling, signaling like, you know, I'm with it. That's a lot of fun. Yeah. What are you, what's the biggest bottleneck right now? I'm assuming you're going to hire a bunch of people from this round. Where, what's, what talk about the use of funds, all that.

3:11:29We're a hundred people today, primarily in New York. We're opening a London office and then probably APAC later in the year, trying to get to around 200, 250 people equally split across product and engineering. and go to market. And most of our go to market post sales, what we call kind of like forward deployed banking is ex-finance, ex-bankers, ex-investors. Very cool. Where's the name come from, Rogo? Our lawyers, our initial name was Athena and our lawyers were like, you can't do that. There's probably a thousand products called Athena. So John, my co-founder and I spent months trying to come up with a name.

3:12:08And at some point we're just looking up old latin words and rogo means i ask in latin and so we thought that was a good one i love that that's great well thanks so much for coming on yeah great great to finally meet and uh congrats to the whole team congratulations have a great rest of you guys we'll talk soon goodbye and our last guest of the lightning round sierra peterson from voyager ventures she's energy time restream waiting room let's bring her in how are you doing Hey guys, delighted to be here. How are you? We're great. Great to have you. First time on the show. Would love an introduction on yourself and the firm, and then we can get into the news.

3:12:50You bet. So I'm Sierra Peterson. I'm one of the two founders of Voyager Ventures. We are an early stage venture capital firm investing in the foundational technologies that power civilization. So this is energy, transportation, materials production, advanced compute, physical AI. These are the biggest markets in the world and the ones that are foundational to ongoing growth in the long term and competitive advantage in the near term. I have been working in energy for my entire career. Got me to start 21 years ago at the International Energy Agency. Overnight success. Exactly. Long time coming.

3:13:26And I mean, that's really the story of our backgrounds. myself and my co-founder at Origer, Sarah, have been working in energy industrial modernization and climate stabilization now for decades. Between us, we've built five climate tech and energy companies. We've been active in policymaking at the International Energy Agency, as I mentioned, and also in the Obama Office of Energy and Climate Change back when we had one. And we both, yeah, exactly. We built companies that were responsible for financing more than$3 billion worth of distributed energy assets, so a lot of solar. We've been active in the research programs at Harvard and MIT, advancing everything from energy market research, which was my graduate focus, and Sarah's at MIT in industrial biotech.

3:14:18And we've been investing now for 10 years plus. I actually got started as an angel investor in 2015 in energy tech. Those companies did well. And so we teamed up to launch Voyager in 2021 and raised a$100 million debut fund. John's going for the gong. How much did you raise recently? Hey, no, that's not the big news. We got a bigger number. We announced a$275 fund, too.

3:14:46We were waiting for that. Very cool. I want to talk about energy because I feel like your timing here is great. I feel like people are spending way more, you know, it's gone mainstream, both at the national level and politics. Just in the last few weeks, we've had Scott Nolan on to talk about nuclear fuel production and then a separate company that turns that actually into fuel pellets, trisofuel. And like there's a there's a company in every part of the stack. are there any particular subcategories that you're tracking that you're particularly interested in and then I want to talk about the financing of all this as well but what what do you think is the most like breakout success in energy these days?

3:15:32Oh yeah I mean we're energy maximalists I have been working in energy for my entire career yeah and when you think about what's truly foundational to powering civilization it's energy like that is the foundation of all of it and so you know we that's actually a really exciting time to be investing. Given advances, particularly in electrification, I mean, we're seeing the cost of solar drop so fast. You have to look at a log scale. It's amazing. And what that means in terms of distribution of energy at a worldwide scale is incredible. And for creating a truly global and resilient energy system that enables near-term stability for users of electricity and long-term prosperity at a global scale.

3:16:17This is a technology that will power the next area of progress and beyond. So electricity, massive. Advanced solar electrification, massive right now. Are you optimistic about American solar? It feels like in terms of solar panel construction and manufacturing, China is doing so well that it's been really hard to get a foothold. We've talked to some teams that have taken Chinese-built assets. T1 Energy. T1 Energy is interesting, but I haven't seen as much excitement from the startup world around really scaling solar. But then you have huge voices like Elon is a solar maxi. There's lots of people that are very, you know, just beating the drum on like the fundamental physics of the sun is so much energy.

3:17:08We're going to capture it at some point. So it feels inevitable, but it doesn't feel like there's as much of a boom or narrative being driven. But what are you seeing? Oh, it's totally inevitable. You can't argue with the numbers. I mean, 90 percent of installed energy generation for the last year that we have record is 2024 is renewables. That is at a global scale. That is certainly true in the United States as well. I think it's just under 90 percent. And I mean, that's regardless of political whim. That's regardless of overall organizing policy approach. that's just simply because solar is better that it's cheaper it's better performing it you know enables a true distribution of electricity like we've never seen before and coupled with advances in batteries that's the unlock you can really create a transactive energy system which we've never seen before and we really will be fundamental to power and growth for decades ahead so huge yeah huge believer in solar potential now um and you know in combination with electrification both of energy generation, also in end-use applications.

3:18:14I mean, electricity now is the most frequent way in which anybody around the world accesses power. That's exciting. It means that power is programmable, and it enables a new era of machines to be made, which is extraordinary in terms of extending human capabilities into ways that we've never had before. When you couple that with automation and artificial intelligence, we're just seeing the foundations of so much innovation right now to really, one, boost systemic stability in a rising vulnerability and volatility market. We have framed trade relationships, geopolitical tensions, physical risks that are tens of being given climate change.

3:18:57And when you can control your inputs in terms of electricity, in terms of your means of production, and you can take advantage of advances in advanced manufacturing to create better products, you can really control your own destiny. That's exciting. Yeah. How are you thinking about financing hard tech companies, energy companies, these like capital intensive companies at the earlier stage? It feels like more and more founders are going to need to get familiar with venture debt. It's been sort of a cautionary tale from the software startup era. Oh, venture debt, that's what will burn you. That's what will really put you in danger if you misplan and then you get over your skis.

3:19:39But how do you think about the role of venture debt or debt instruments or more complicated financing schemes for some of these companies? Yeah, it's a good question. I mean, I think one of the things that we're quite excited about is the sophistication of the financing instruments that are available to early stage startups. I mean, it used to be that you had to burn super high cost of capital venture or private equity on assets. And that didn't make sense for anybody. You know, now we're seeing really tailored and interested bespoke non-dilutive financing, you know, all types. And I think it's a recognition of the scale of these markets and the durability that they have in this long-lived and reliable yield.

3:20:22I think we're also seeing, though, and this is something that's been a recent development, ongoing sovereign interest and recognition in these technologies as being tools of the national interest. 44 % of our own portfolio is either in partnership or in active conversations to sign a deal with the US government. That's extraordinary. I mean, you see In-Q-Tel last year, I think became one of our second most frequent co-investor. And yeah, I think that's not - It used to be completely unthinkable. It was like, you'd never go to the government. They'll never buy anything. Like Palantir was this weird company that was like off in the dark, just doing one deal.

3:21:01And they were the only ones. And then Anderol sort of did it again. And then all of a sudden it was like every company has some sort of deal with the government. Yeah. It really changed everything. Because when you think about what are the inputs and tools of sovereign advantage, of competitive advantage, it's controlling your destiny in terms of energy, controlling your destiny in terms of supply chains for inputs to manufacturing, industrial capacity, and then increasingly compute capacity. It's a national asset. And all of that hyperscaler stack, that all relies on electricity. What about, like, what are the prospects for making, bringing the solar supply chain to the United States?

3:21:44We have, like earlier we mentioned T1 Energy, which was kind of a weird setup in the way that it was created and now operated by a new company. but I think one of the reasons that maybe Americans and the tech industry isn't so excited about solar at this moment, or at least broadly, is that we're not making the panels here. It doesn't feel like something that we really own. Yeah. I mean, I don't know that we're not excited about solar. You look at the solar deployment. Well, I'm just saying the attention is almost entirely on nuclear. It feels like that now. Like I'm just saying from when you think about – Casey Hammer would like a word, but – Yeah, sure.

3:22:28But think about – I would just say the number of entrepreneurs that have come on the show this year that are talking about nuclear, the nuclear deals are – and I'm not saying it's like warranted, but I'm just saying like nuclear is certainly – Well, it feels like, at least to me, and you can give your feedback on this take, but it feels like nuclear was potentially economically viable. The technology worked, but there was a regulatory blocker in place that made it very, very difficult to actually just stand up the technology. With solar, that regulatory blocker doesn't exist. It's more of a market force of China manufactures them really, really cheaply.

3:23:12And it's the DJI drone question versus GoPro. Like we just need a ton of people, a ton of manufacturing capacity, a ton of CNC machines. We need a massive facility. And that's more of a capital formation. And then are you willing to go up against the Chinese manufacturing engine that's so critical? My point is just that energy is critical. Yes. Solar energy. It still feels blocked. a pretty big part of it. So I would imagine we want to make as many of these as we can in the U.S. So I was curious, like, update on the effort there. I know you've invested in it and curious. Yeah, I mean, I think we're seeing a real re-examination of global supply chains for critical aspects of ongoing sovereign advantage and industrial competitiveness.

3:24:00Solar is being one, critical minerals another, access to novel battery technologies. These are all aspects of an overall industrial policy that's increasingly favoring domestic production and control of energy producing assets and manufacturing inputs. I think, you know, to your earlier point, like any technology is going to need to clear the global clearing price and the massive reduction in cost. It's been extraordinary. for solar production, for panel production, you can't get much cheaper. There's kind of nowhere to go. That's exciting. I mean, you think about that as an input to other industrial processes.

3:24:49That's, I mean, a truly extraordinary opportunity to have abundant energy that doesn't rely on fossil fuel supply chains that's truly distributed and can be programmable. I mean, you couple it with energy storage. I mean, you just have a whole new paradigm in abundance for electricity. Be exciting. Very exciting. Yeah. What, you guys, stage agnostic, pre-seed? No, we're early stage. All early. Yep. Yeah. So we raised$275 million. It's more seed series A. So we raised$275 million for seed and series A. We went out to market with a 250 target. We were oversubscribed on that. I had a 90 % first close.

3:25:31There was a lot of excitement. Yeah, thanks, guys. Amazing. It's been great. Well, congratulations. Well, excited. I'm sure a bunch of your companies will be on, and congrats to the whole team. Yeah, good to meet you. Likewise. Always happy to talk energy. Thanks a lot. Bye. We'll talk soon. Cheers. Did the Klein team just join Codex OpenAI? Jason Liu asked the question. Yeah, so I think a lot of the team members did. Okay. Pasha, remember there was a whole controversy about the smell, imagine the smell. Oh, yeah. Oh, he was declined. Yeah, so he moved over. I think a number of the lead engineers there moved over.

3:26:11Yeah, yeah, yeah. Will Brown says it's a very rare reverse windsurf. The tradies said, well, they certainly didn't decline. Decline the ask. Yeah, interesting to see the really headline-grabbing AI trade war narrative has sort of calmed down. But everyone's hiring, everyone's poaching, everyone's trying to build their teams. It's a war. Anyway, there are so many more stuff. There's a post here. Somebody is calling out Augustus DeRico. they said he has no couscous. Greed. To be fair, I've never seen the man eat couscous. He might not have any couscous. I think they were trying to say conscience.

3:27:02But yes, Augustus has stumbled into such a funny life. I'm really, he's, yeah, he's grinding through it. He picked something more controversial than pretty much anything else. Yeah. But he's managed it really well. It's somehow more controversial than defense. Yeah. Yeah. It's just odd. It just triggers all sorts of people. I wonder, I feel like the fact that he's able to go on podcasts and communicate so clearly, he's one of the Joe Rogan CEOs in your parlance, that's allowed him to get through it. It feels like if he were on his back foot and not able to just say, hey, I'll talk to anyone for an hour.

3:27:45I will just talk, talk, talk. That if there was a negative news cycle about him, it would be a much harder situation. But I don't know. Last post for the day. Okay. Emily Sundberg went mega viral, 62 ,000 likes. Oh, I did not see the likes on this. I didn't realize. She says, the logo for Sydney Sweeney's lingerie line looks like it's for a salad dressing company that launched in 2019. this is for Siren oh it's Siren that's wild it does give off a little bit of Sweetgreen Sweetgreen launched way before 2019 oh is that the company that she's referring to? what is she actually referring to? because I saw Dirt quote tweeted it and that's a media and technology company with a similar logo and Dirt just says hey and got 66 ,000 likes as well and Emily says bro Oh, but, oh, I guess it's Sweetgreen that she's actually referring to.

3:28:44Salad dressing company? Isn't Sweetgreen like a salad company? Yeah, I mean, she's just posting. It's a bang. Alex Conrad says someone blended Cava and Graza branding in Chet's Tea Tea. It does have a little Graza in there. But I think this company will do very well. Yeah. I think Sydney's like one of the most commercial people in Hollywood, right? She's like just leaning in. She's got a somewhat short moment in time to create a lot of value and certainly a lot of the brand activations that she's done and broken through. So I'm bullish. I'm bullish. To go back to the Claudebot thing, this siren clearly exists in a very different part of the economy as Sweetgreen.

3:29:30Sweetgreen, I think it was like you walk down the street, you see it. Maybe if they start opening a ton of retail stores, but in terms of – Rolling flashbang. Are you flashbanging me? Why are you flashbanging me? Because the chat earlier was going crazy. They were trying to throw me off. It was insane. I was barely hanging on. It was brutal. Everyone loves the flashbang. Fun fact, Sydney Sweeney is launching her lingerie brand on, you guessed it, Shopify. Harley. That's right. That's the story. Fantastic news. Well, the last thing before you plant the bomb, we have a new ad read for Railway. I already did it.

3:30:08I'm going to do another one. Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web apps, servers, databases, and more while Railway automatically takes care of scaling, monitoring, and security. So you'll be hearing that. Well said, John. We'll close it out. Nathan is highlighting a Bloomberg article. The title of the article, NVIDIA lacks clear successor for superstar CEO who built company. And Nathan, of course, says, man is barely getting started and i agree who's who's asking who's no one is asking this question this is like this is like driven the stock 10x and it's the biggest company in the world and they haven't like a massive margins of cash flow like there's been no mistakes like literally like nothing 62 years old he's got another 30 he's got another 30 he's going to be riding nvidia straight into singularity and we will be too here on tbpn uh last post last post okay one more post Eric Sufert says, Pinterest dropped nearly 10 % yesterday on news that the company is cutting roughly 15 % of its workforce to focus on AI.

3:31:18Eric, I believe, is coming on the show tomorrow. Yeah, yeah, yeah. Really like talking with him about all things. We're going to dive into this. We'll dive into open AI ads. That's GBT ads. Pretty remarkable. I wonder where Pinterest will land. there's been rumors or pitches for an acquisition. Leave us five stars on Apple Podcasts and Spotify. We love you. We'll see you tomorrow. TPPN.com for the newsletter. Nice work, brothers. I'll see you on the next one.