Astra Ends Math, 1 Person $1M Companies, Hank Green Is Absolutely Right | John Gruber, Shaun Maguire & Isaiah Taylor, Justin Lopas, Beau Gaston, Ron Arel

26:54Those little sites are now getting sort of eaten by models and eaten by other people and you might be able to vibe code your own, but at the same time, like they might have a long, long term. Tyler, update. Okay, so they basically calculate the number of like solopreneurs based on how many people have, like, there's, like, special plugins or platforms that are directly for, like, the solopreneur. So they basically use that to, like, get a proxy of the general, like, percentage of solo people on Stripe. On Stripe? Oh, they have, like, a special flow for solopreneur? Interesting. Oh, cool. So they say that they're almost certainly underestimating the number.

27:31Hey, Julian Weiser, I know him, says the bar is getting, the bar for getting started has never been lower, said Julian Weiser, who runs a San Francisco-based accelerator for solo founders working in tech. The accelerator, which offers founders seed money and mentorship in exchange for an equity stake, attracted 4 ,500 applicants for 10 slots made available in its most recent cycle, nearly five times the number it drew when it launched last May. Now, he's been growing this a lot, but that is staggering. A lot of people want to be solopreneurs. Going it alone with AI can still be surprisingly expensive.

28:06Broca said he was losing money on many customers' accounts while paying to access Anthropics Cloud to run his clients' requests, that AI company, as well as others charged based on usage. He has since switched to free open source models from China. Broca says he has raised$30 million from investors and at the same time has saved millions in salary since he hasn't needed a team of software engineers. Another risk, if it's easy for one entrepreneur to launch an AI-assisted business, copying them can be easy too. This creates anxiety for founders like Troy Johnston, who runs an AI-assisted business alone in Orlando, Florida.

28:39Everybody has the sword, and we all have the ability to unsheathe Excalibur now. Johnston, what a great quote for the journal. I love it. He's 40. He used AI to code an app that helps people get the most out of credit card benefits. Huh. It's interesting. Pick which card you want to use because you might have multiple cards, one that's good for dining, and you build an app for that. There's been a few apps that do that. The Points Guy had a whole blog around it, a whole media company around it still does. But interesting to sort of like, yeah, go and actually Vibeco that. A lot of those things, it's like you could probably just use the models themselves for this.

29:22Just have a thread that says, hey, these are the cards I have. Go pull all of the data. When I'm about to buy something, let me know. But at the same time, there might be some value for something new with a deeper integration somewhere. The company makes around$3 ,000 a month in profit with no employees and continuing to grow. What a run for Troy Johnson. Great story. Who loves King Arthur-related metaphors for business. What one-person businesses will mean for the labor market remains to be seen. Polling has shown that Americans are worried that AI will replace jobs, and top economists are wrestling with that possibility too.

29:55But AI is also creating lots of new jobs. And the go-to-loan entrepreneurs show the technology can both open doors and limit employment opportunities. If everyone's hiring less, but you get four times more firms, what does that do to headcount? Said Rembrandt Koenig, an associate professor at Harvard Business School who studies entrepreneurship. He co-authored a recent study that found that among 50 ,000 startups the researchers examined, those focused on AI tended to operate with 25 % fewer employees. it's interesting because haven't we seen that the ramp data that said that AI adopting companies were hiring faster but maybe they still operate lower operational headcount but hiring faster because of hiring growth there's like three different factors that are going on here sort of mixing all together Koning the professor also believes in a soft believes a soft hiring environment that has left some people mired in log job searches has encouraged more to try their hand at launching businesses.

30:52That makes sense. Some founders cite different motives. It's a perfect storm of post-pandemic burnout and a re-evaluation of one's priorities and also booming AI and a sense of what's possible said Samir Ahmad, 39, who lives in Brenningsville, PA. Two years ago, Ahmad decided to leave the corporate job he'd worked at at Verizon for almost two decades to start a solo coaching and consulting business. He had been seeing social media posts touting the ease and virtues of AI, which he liked to chart, which he used to chart a business plan and help with marketing. It was like my chief of staff, second in command.

31:25The business ultimately petered out within months, though, and Ahmad is back to full-time corporate role with a utility company. For Claire Vo, 41, AI helped turn her passing impulse into a business. She was working full-time as a tech executive when she tapped AI in late 2023 to help code an app that would help manage documentation and design for new products, with customers ranging from financial services to healthcare firms. I was copying and pasting from ChatGPT, said Vo, who lives in San Francisco. She put her app online for$1 a month. Wow, that is cheap. And within weeks, people downloaded it.

32:01I thought we didn't know how to make apps that cheap anymore. Yeah. I mean, it is a subscription, at least. Not one time. But she put it online for$1 a month. And within weeks, people downloaded it thousands of times. Nearly three years later, Vo's company, which she ran solo for nine months before hiring an engineer, now has 100 ,000 users and is on track to make seven figures in profit this year. Wow. That's remarkable. At a dollar a month. That's crazy. AI handles the company's marketing, sales, and customer support. Well, AI is a shortcut. Vo said her network and credibility in the industry were key.

32:31I think people over-index on how easy AI is and under-index on how much I did to get to this point, she said. She's still. Yeah. I just want to see five companies that make more money from their business than they give Pulsia every month. Yes. So Pulsia has some public dashboards for how much people are spending or something like that? Yeah, they have a public dashboard. Let's see if I can find that again. While you're doing that, let me tell you about Cisco. Critical infrastructure for the AI era Unlocks seamless real-time experiences And new value with Cisco And if Jordy's continuing to look I'll also tell you about public Investing for those that take it seriously They got stocks, options, bonds, crypto, treasuries And more with great customer service Trying to find the dashboard I was looking at the dashboard that was showing There's some Okay, I think Tyler found it Thank you, Tyler Yeah, so right now you can see all the different things that the companies on Pulsia are doing, or at least some of them.

33:42Right now, so far today, companies on Pulsia have spent$373. Is that today? On ads. Today? Yeah. Well, it's still morning. It's still morning, so we're pacing. Okay. I don't know what time zone this is in. But yeah, the big question is like, is any of this stuff actually working or is it more like kind of a video game effectively? Yeah. That people just enjoy like watching the machine hum, but there's not really much happening. I mean, that was the thing for Mid Journey and Suno, I think in many ways. Like Mid Journey, when it launched, people were like. I mean, no. Okay. Hear me out. okay i'll hear you okay when mid-journey launched the steel man when mid-journey launched people were like this is going to take artists jobs and it was like okay so if that plays out then i'm going to go to the moma and there's going to be a show for someone that just prompted mid-journey and the highest auction at christie's is going to be some mid-journey artist and that's not really what happened like people aren't using mid-journey to make fine art but people love mid-journey like They love the activity of going on mid-journey and generating and prompting and getting an image back.

35:03And then maybe they send it to their friends. Maybe they use it a little bit. But it's not exactly the same of the process of becoming a fine artist. It's more like they're enjoying the process of just making. It's more like just having a guitar that you just like to practice and noodle on versus actually being a touring artist. and so like that's certainly my experience with suno is it's fun to try and make a song and then listen to it and then be like wrestling with the thing and and and it's possible that that could be the same activity for like okay i'm gonna go build an online business see if i can get something out but it's not really like a job it's more of like an entertainment product yeah i don't know what do you think yeah i mean you could easily see it turning into like ender's game scenario where it's like a game and then it's like oh that was actually a real business you were starting You offshored the last job.

35:55You sent all the labor overseas. You rolled up the entire HVAC industry. That wasn't a simulation. Now, based on my steal, man, do you agree or do you still disagree? No, I just think that you could use MidJourney to create a beautiful asset that you could get some enjoyment out of or you could use it for your business or whatever you're doing. Yeah. Or to just create AI art. uh yeah and you could use suno suno is just like deeply entertaining you go on there and in five seconds you make a real song that sounds like it was recorded in a real studio and there might be something to like okay it is fun to go and build a sass product like it is fun to go the the pro like the game it's a game right i don't know tyler uh okay just on your example earlier of like uh the moma artists like that's like the insane long tail of artists yeah that's not like the average you know center of the of the curve artist that like okay mid-journey like maybe is like doing a similar thing to what yeah maybe yeah maybe maybe i don't know i i i would just be surprised if uh if if the i don't know maybe maybe the way to put it is like i'd be surprised if like the like the majority of mid-journey users are using the product as a as a like a an artist career path or an artist career path.

37:21Like they're like, okay, I got my image. Now I need to get it printed. Now I need to go do a small gallery show. Now I need to go talk to an auction house. Now I need to go and pitch it to a bunch of collectors and like go through the process of being an artist. I think a lot of people are just like, cool. I got an image. Like this is nice. Like jobs done now back to whatever else I was doing. Oh, like I have some free time. I could go play a video game. I could go listen to music. I could go generate some images and have fun with that and see what those are like and then just enjoy them myself right i don't know i i think there's like a smaller tighter loop with some of these services that might be you know overridden and i'm wondering if there might be one in the like design a business like gamification of business world i don't know it does seem like ben has done a really good job positioning pulsia and like you know using some different methods to get attention i i have some something i feel like I have a little bit against the whole thing because I just get spammed DMs from him.

38:20He doesn't follow me on X, but he spams me with messages asking for different things, which I just think is somewhat entertaining if you're trying to run the anti-AI slop company, right? Yeah. Is it? I thought it was pro-AI slop. I thought that was the whole name. No, no. He's saying this is not slop. Because there is a world where you're like It's slop but it's good slop And it's fine There's a lot of programmers that say The answer for slop code is more slop And it'll be fine It's not a problem You might not like the way the code is written But it doesn't matter as long as it works You might not like those artifacts in the AI image But it's fine Because it illustrated the point Just like you wanted Anyway, Tyler, do you have something else?

39:09Yeah, I was going to say Yeah, slap is like a temporary term. At some point, the writing actually becomes good. The writing? Whatever, the output of the model. Yeah, writing does seem behind a little bit. It's like verifiable. It's not verifiable. No, people say it's good or bad. Not fast enough. The loop isn't tight enough. And there's too many people that say it's good when it's bad. You can make it. Maybe, maybe. It did get a lot better recently. Once a week, someone prominent posts fully AI-generated content. Yeah, what was happening with the cash? No, it just happens. It happens once a week.

39:48It's just an iron law of once a week someone really, really talented and smart posts something that is just entirely AI. Maybe we got to rip it. We got to try it just to feel something. Because maybe it's like a forbidden fruit. Like the full just like go to, go and prompt, like write me a blog post, thought leadership about business. That's the prompt. Copy, paste, rip it. Maybe it's also a strategy. You have something that's like you really want to get out, but it's a little bit boring. And so you know that if you use AI to write it, you'll get so much more ratio. Way more people see it. And as long as the first few sentences are kind of deliver the message you want.

40:27Yeah, this is good. This is really good. Yeah, I think we got to do it. We'll test it on Tyler's account first, though. Joe Weisenthal. Yeah, let's do it. I love all the AI people who are like, nobody's prepared for what's coming. It's like, maybe just speak for yourself. Joe's ready. Joe's ready. No, it is very funny that there's this whole meme of like, no one knows what's coming. It's not priced in. It's like all anyone talks about ever. It's on the front page of the Wall Street Journal every day. Lots of people are talking about this. And most people that say nobody's prepared for what's coming will not give you a really concrete.

41:07Yeah, like what is coming exactly. Like the last time we had this like nobody's prepared for what's coming moment was what was the guy who was comparing AI to COVID back in March? Was that Schumer, Matt Schumer? Yeah. So something big is coming? Yeah, something big is happening. And it's like, yeah, something big is happening. Like the models are getting better. like the math is is getting solved but uh like you can still go outside like yeah like at this point in 2020 unemployment had spiked to 10 and like it was very much like you're a bold patriot if you're going outside like it was a crazy crazy time and now it's like yeah there might be some softness in the in the job market like a hedge fund blew up there's like there's things that are happening some big things but but the hedge fund blew up from being a little too bullish like things maybe like they got the you know the basically directionally correct but got the timing wrong right uh mike isaac says yeah these mfs don't know how much i got stockpiled in my basement yeah and joe says these mfs don't know that if a man has a why he'll find his how

42:26well said yeah well said yeah buco capital was having fun with kevin roose over at formerly new york times now hard fork independent i don't know uh is it did they did they take the a did they take the ip are they still using hard fork outside of nyt i don't know but uh buco just this weird comment for a guy writing a book on agi he is exactly and literally wrong. Everyone is pricing it in. Why do you think OpenAI and Anthropic are priced at$1 trillion? What nobody is pricing in is that besides coding and math, very few domains have embedded verification. But Tyler over there thinks that you can formally verify whether or not the Odyssey is a good book in Lean or something.

43:11Well, I think we can formally verify that Tyler's goaded, right? A lot of people would say, Like it's kind of like gray area, right? You can't totally define what it means to be the greatest of all time. Yeah. But you've been working on that quantifying aura. As long as it's fully quantifiable, you'll be able to verify it. Yes. Andrew Curran says, shorten your timelines, friends. I've started this account to say this. And in many ways I've posted for the past four years has been saying the same thing. Some of you are increasingly feel it. We passed the threshold in November. We are already inside the singularity.

43:46Lots of people are picking up the, we're no longer in the foothills of the singularity. We're in the singularity now. Demis on stage at Google I.O. just a couple weeks ago, a couple months ago, saying we're in the foothills. Well, now we're on the mountain. If you follow this account for a long time, Andrew Curran says, I apologize for losing my mind a few times using GPT 3.5, and then Bing forced me to update all of this at once in one shot. And that was the correct time to update. talking to 3.5 and and the first chat gpt moment being sydney that it was if you if you updated on that you did very very well across everything both with the growth in the labs and the growth in the in the data center build out and everything like that was the key moment also there was a moment in 2020 this was coming up over the weekend yeah that somebody used gpt3 to generate like a fully functional react app you remember this yes i forget what it was called it was called like the something the yeah um but anyways that in in hindsight was like such a big moment yeah um but at the time it got like 2 000 likes and people were like wow this is really cool yeah but no one no one took it from there to be like businesses will be spending hundreds of billions of dollars on this in just a few years yeah like or i mean a lot of people did honestly yeah like tons of tons of people across venture and private market public markets a lot of people did but way way more people didn't update any of their behavior.

45:18Yeah, yeah, I guess. It's mind-blowing to me how few people realize their lives and everything they know will change drastically in the near future. At this point, it should be pretty clear, says Jerry Turek. Yeah, wild times. Elon Musk says 100%. He's completely agreeing. Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents, whether you're writing code, analyzing data, creating content, or automating business workflows. codex helps you move projects forward from start to finish um we gotta look at this simulator you're into racing simulators i gotta up you i gotta one up you with a train simulator look at this guy this uh this guy on z80.me has says uh has built a full scale train simulator controller for the three years i've been building a physical train simulator in my apartment modeled on the uk class That's ADX.

46:12Passenger train. I've been trying to replicate the instruments and controls in the real cab as closely as possible. Look at this. This is amazing. In many cases, managing to acquire real components and in others, building my own replicas. I've gone down rabbit holes for design from custom can bus transceiver board and a variety of daughter boards to tie the simulator together to the actual panels, all of which on the consoles and instruments are mounted. I usually wait until I complete a component or step before I write a blog post. But over the last half year, I've instead made forward progress on several disparate aspects of the project.

46:47Therefore, this blog post will update as a status update on many of those aspects. I'm feeling the acceleration, John. This is such a cool DIY project. Imagine just sitting there driving a train in the train simulator. So one thing I'm not seeing is any type of visuals. So is he? like you mean you mean the the software that's driving the uh yeah to me he only cares about the tactile experience of pushing the buttons right like he's optimizing for actually feeling like he's in the uh in the in the train but no there is a screen and that screen is running a game that simulates a train and the name of that game train simulator you can get a train-simulator.com It's also on Steam.

47:35Yes. Train Simulator Classic is, I believe, the one that he's playing. But, yeah, he's playing the full train simulator. AWS Sunflower? I don't know. Anyway, funny, funny story. Balaji is moving to Kazakhstan. This is huge for the Borat community. Absolutely huge for the Borat community. He says, this is Kazakhstan. Let's see. This is a beautiful video. I have no doubt it's a beautiful place. It's a little bit crazy because network school I thought of as sort of a startup incubator, a Y Combinator adjacent entity. He was in Singapore for a while, then Malaysia, and now maybe Kazakhstan. Did he actually move to Kazakhstan or is he just like touring and vacationing there?

48:25Because what is the actual story? Here's the news. He says, I'm pleased to announce that a memorandum of understanding has been signed between the Republic of Kazakhstan and network school. Our new campus will become a haven for global techno-optimism with expedited visas, streamlined re-dominization, and active recruitment of talent. Excited for Balaji, excited for Kazakhstan, excited for network school. Balaji is very smart and very entertaining, and I've enjoyed having him on the show. I do think it is very funny to be trying to recreate the incredible techno-optimism that many different sub-communities already have in the United States, where he effectively had all of his success at Coinbase and Andreessen Horowitz and I'm sure many other businesses.

49:28And I think this whole chapter is deeply entertaining to me, the Kazakhstan chapter. It's such a funny place. King of the Castle, I love it. So quick tip for anyone who's planning to do sort of like the bi-coastal thing, San Francisco, Kazakhstan. You're in for like a 30-hour trip because there are no nonstop flights, I think. You have to connect in Istanbul, Frankfurt, Seoul, Doha, or Dubai. You're looking at 25 to 35 hours depending on the layover. That is really, really far. The Miami thing was a tough pitch because, you know, so much activity is happening in New York, so much activity is happening in San Francisco.

50:17And it was still hard to get people to relocate, like great engineers. They come out with the beautiful— No, I mean, this is just truly— Full send. The toughest possible sell. I think he's trying to— I think he's— Weed out the week. I think, well, I think in some ways he's been so successful that he wants a challenge that to him feels almost impossible, which is to convince the best and brightest from all over the world to move to Kazakhstan. um it's it's i know so many i mean some major selection bias here but i know i know a bunch of bright people that are not u.s residents and they would do anything to be able to be in kazakson not quite maybe now maybe now but they would do anything to be in to be able to have free access to america to be able to set up shop here to build their business here um to be able to uh to be the king of the castle to be to be not even the king but a popper just someone just someone in the castle in the castle at all uh and so yeah and i haven't uh yeah just we gotta go So it's very clear that we've got to go or at least we have to send Tyler.

51:40I would go. It looks fun. Pack your bags, buddy. See you in 30 hours when you land. Absolutely wild. Let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. Mark Zuckerberg was answering questions about his AI strategy on the latest meta earnings call. and Ben Thompson wrote about what Meta's position is in AI, how they're grappling with a few things. There was a bunch of interesting points in this strategy update. One I wanted to call out was what Ben Thompson thought the best moment on the call was when an analyst asked him why the company can't just use other models.

52:23Like, why can't you just do the Apple thing? Do nothing, win. Like, partner with one of the labs, do some license agreement. If you need an image model, you get an image model. If you need a text model, you get a text model. If you need to speed up your programmers or your engineers, hire the best coding agent and negotiate with them, right? And here's how Mark Zuckerberg answered it. He said, I can take the open source question. Let's see. So basically the question is, do we think that because there are some open weight models that we can just rely on those? I mean, right now, the open source models are not as strong as the frontier models.

52:59Good point. So no is the basic answer. Meta needs to be on the frontier with their intelligence that they use. So they have to be there, according to him. He says, and then there's also just always the perpetual both policy debate and question around other companies' actions and whether that's a thing a company like Meta can rely on. So if you're using Chinese open source and there's some regulatory risk, it seems like that's sort of what he's getting at, is these things might not rely, they might not be available all the time. And then also some of these companies, they might be open source for a few years and then go close source and then start charging you an arm and a leg.

53:38So you don't want to be in a place where you become super dependent and then all of a sudden get hurt once you're super dependent on a particular product. So he says, and I think that's very tricky. So on both fronts, we believe we're going to be able to do better work and we think that there's some risk in that reliance. I don't believe that's the right thing to do. So that felt like not a great answer to me in the sense that the Apple approach seems to be working so well. We'll talk to John Gruber about that in a few minutes. But then he goes on to explain some of the history of Meta, and it's very, very interesting.

54:18He says, I think that we're a company that if you look at Meta from, take a step back on this. A lot of people view the surface layer of we build some social media apps and we have an ad business. We are really a full stack technology company. We build our own data centers, our own infrastructure, our own chips, our own low level software. When we got started, he says, my background in engineering, I wrote a lot of the systems code. A lot of the reason why Facebook worked was because it actually, it just worked, which is a crazy thing to say based on the history. But he makes a really good point.

54:53He said, like it literally worked when other social networks did not work fast and efficiently. And I think we just have the ability to build things that can be more personalized, more optimized, more efficient. Some qualitative experiences are just not even possible for others to build because we go all the way down the stack. And it just seems to me pretty clear that having kind of sovereignty over building your own models is going to be an important part of that stack going forward, which is why it's important for Meta. And so that was very interesting that that was a differentiator in the early days that certain other competitor sites would just be slower.

55:28They wouldn't be able to launch new features quickly. And by vertically integrating all up and down the tech stack, they were able to do things very aggressively. This is the Reels thing. They built two extra data centers to be able to do the Reels algorithm because if they didn't have that compute capacity, they could not have launched a competitor to TikTok on any normal time frame because it was actually a compute-intensive project, not just a design. People see, oh, they just put a new button there and there's some videos. But it's like behind those videos is a massive recommender system that is very computationally intensive and stores a lot of data.

56:08And you can't just spin that up for 3 billion users or however many billions of users they have on a dime if you don't have the infrastructure, if you're not actually vertically integrated. So there's a whole bunch of other things where in terms of personalization, understanding the user, like delivering a really first class experience. They sort of do need to bring it in house, but investors are very upset about this. They're not very happy because

56:37Ben Thompson calls it the financial tail wagging the dog and says that they have to sort of double spend right now. He makes a good point about this. They're double paying. The company right now is It's basically double paying for infrastructure without a clear path to monetization. And to make matters worse, it's improving monetization story lost a bit of its luster. So they're both renting AI compute, paying a bunch of money for new researchers, and then also spending all the capex for the next data center. So all of that needs to come together in this moment to actually deliver. And the investors are starting to ask all these questions about what the strategy is.

57:15But Zuck's sticking with it. He's not backing down, but we'll see. Yeah, tough position to be in when capital markets don't have a ton of faith. And you just see that in the stock price, the company broadly, right? There's a lot of infighting frustration around just how MSL is treated versus the rest of the company, which is paying for MSL. So Zuck is at war, fighting a war with multiple fronts. It is. It's a big war. I'll get through it, though. We'll dig into it more. I still think this Pierre Richelson tweet is so funny. Shower thought, why is no one doing outbound for pizza? Hey, this is Gigi from Gigi's Pizza calling.

58:00You ordered last week. We have a pepperoni pizza ready and could deliver it in 10 minutes. You hungry? Hilarious concept. Would be extremely annoying to have every possible low-tier business that you've bought anything from with. spamming you i mean they basically do this with uh you know email awareness like hey there's a super bowl coming like do you want to place an order what if we made a law that said that restaurants could only call between 5 30 and 6 p.m when you're hungry so you knew you'd be getting calls coming in and you could kind of play them off each other you get a pizza offer you're like look i'm kind of interested in pizza but i'm i have an open conversation with the taqueria yeah And I need to wait to understand what they can offer tonight.

58:46I'll let you know. Taqueria calls. They go four steak tacos, side of rice. You win. And then you get a price. You get a bid. You go back to the pizza. You kind of play them off of each other a little bit. And then you go with what you're really feeling at the end of the day. Do you know what it's called on Wall Street when you have multiple parties negotiating to sell a block of stock or debt or something like that? and you want to bring them all in really quickly. Let's say you're negotiating with Tyler and I have extra information. I might have a buyer and I want to jump in. Do you know what that's called?

59:23Barging your line. So like, oh, yeah, I'm going to barge his line. Jump in there and then we'll be on like a three-way call basically. And you can do that when your phone system is set up with multiple lines. So you could potentially have, okay, you got Domino's on line one and then you got Pizza Hut on line two and you could be like, okay, I'm just going to put you as all in on call. Conference all the time. Let's debate. Let's get to the bottom price because that's basically what you're doing. You're saying we're just going to hold an auction right here. Yeah. This might be the solution. I like this post.

59:55Explain to my wife, explain to my ape wife that I have to spend nights and weekends learning the bone so we don't end up in the permanent underclass. Is this from 2001 Space Odyssey? Yeah. It's a good movie. Jordy, have you seen 2001 A Space Odyssey? Yes. You have? Yeah. No way. Yeah, I remember that scene. What? How did that happen? That's wild. I think I was forced to watch it. Hmm. Oh, yes. This old Leopold lore is coming back up. The author of this New York Times article definitely doesn't realize Leopold was being literal about the stars and galaxies. Asked in a 2004 podcast interview with Dwarkesh Patel what his goals were, he answered, Eventually, you are going to go to the stars.

1:00:41You are going to go to the galaxies, he added. Done right. There's a lot of money to be made. That was true for a while. At least while the fund was the center of the hottest trade on the planet, the fund made a return of more than 200 % after fees. As anything tied to AI shot higher, one investor said, yes, there was a funny line where Leopold's talking about buying galaxies. And some investors like, oh, like the particular brand of private jet that's referred to as a galaxy. Like the Galaxy 750 is the one that you'd want. And Leopold was like, no, no, no. I'm going to buy an actual Galaxy. Tay Kim says, who will play Leo and Ken in the movie?

1:01:17We know it's coming. The story is too juicy. Okay. We got to play this clip of Ken Griffin because this came up on my For You page. And it's a wild story of when Ken Griffin had his darkest moment, basically, and lost a whopping 4 % of the fund. something like that. Play this clip. You know, in 1994, I was in Switzerland. We'd had a rough year in 94. We lost about 4 % of our capital in 94. It was one of our only losing years in the history of the firm. And I'm in Switzerland. I mean, it was a rough day. My lunch, my lunch, I sat down at lunch. This person sits down. Oh, you're not John Griffin.

1:01:59No, I'm Ken Griffin. He goes, oh, I thought you were John Griffin from Fenchurch, another firm. He goes, I got to go. I'm like, great I flew all the way to Switzerland for my lunch date To get up and leave the table And then around 3 or 4 in the afternoon I was with another Swiss banker And we're in his office, his office It was like, almost the square footage of the stage Beautiful furniture He goes, do you mind if I smoke? He takes on a big cigar, he's smoking this cigar And we're talking for about 45 minutes

1:02:32Such a pity That such a bright young man so pick the wrong career. Like, well, that's the most graceful no I've gotten today. But you just have to tolerate, you're going to hear no a lot, but you need to become accustomed to have no more. That man, Leopold's dad. Yeah.

1:02:55Again, whether it's the people that you want to have work for you, or people who are trying to give you capital or customers, you need to get comfortable with the art of selling. Great clip. it's so funny that someone was like ken griffin like you're just in the wrong job you should be yeah i don't know this is this is not for you i wonder what that banker would have preferred ken griffin do like did he have a prescription was he like you should be course seller a hustler i don't know skier or something uh anyway fun little trip down memory lane we have john gruber from daring fireball in the waiting room let's bring him in to the tbp and ultra dome for the second time.

1:03:35Welcome back to the show, John. How are you doing? Good. How are you guys? Great to see you. How's your summer been? Yeah. Hot. Hot.

1:03:47Has the tech news been overwhelming or underwhelming this summer compared to previous summers? Oh, I'd say much busier. I wouldn't say overwhelming, but definitely busier. It feels like the, I don't know, I mean, at least there's like the traditional news, which is like a company does a thing, there's a process, something launches. is that felt that has felt very light recently but then there is like the meta drama of like situational awareness blowing up and that stuff has been really really crazy like the the hugging face hack and like mythos and all these different stories that are sort of like not planned in the same way of like there's a release cycle we all got to talk about the thing that's like goes through the pr turn you know yes i don't think that there was a marketing schedule for when we're when they're going to have a, hey, our AI broke out of the sandbox and attacked a major partner of ours during a testing run.

1:04:52Well, it just depends if you believe in like the Truman Show thesis. So the simulation theory was like, now have the hedge fund explode. That will be the entertaining thing. Well, the funny thing is there actually was people that were making the allegation that the hack was just marketing. Yeah, I think they don't mind, right? Like, I don't think it's – so I don't think there's a conspiracy in that it was fake or deliberate. But I do think there's a strong sense of, oh, no, you mean that the news cycle for the next 24 hours is going to be about how scarily effective and intelligent our model is?

1:05:34you know I do think that there is a laughing all the way to the bank aspect of it even though it wasn't deliberate yeah sure yeah or like a parent watching their kid and like the you know play sports and like the kid runs up like you know a bunch of points on the other team and the coach has to talk to the parent and say like hey look like your son is very good but like he's got to pass the ball a bit more and like it's not really that sportsman like to you know run up the score that crazy he should focus on you know teamwork yeah um i heard i while we're on conspiracy theories i heard a funny one that uh the person that's the most happy about uh situational awareness uh blowing up is potentially apple because they've been under all this pressure from memory situational awareness was of course very long memory there's a flywheel there where The stock goes up.

1:06:29The memory gets more expensive. Now, I feel like the real take here is that, in fact, pumping up memory stocks is the best way to lower the price of memory because they will all fund CapEx and that will ultimately make Apple devices cheaper. But how do you think about the pressure that Apple's been under around component pricing, supply chain pricing, getting caught flat-footed versus just reacting to what's happening in the world? It is extraordinary. And I really do think that to go the other way, I think, you know, Tim Cook's public remarks on it as being a once in 100 year flood situation and that he's been looking at this market his entire career.

1:07:14and there is no comparison point. And RAM in particular has always gone through these boom and bust cycles. And this is just a boom with a CapEx expenditure fueling it that is just such an extraordinary amount of money that it's like nothing else. And so I don't even know that Apple was caught flat-footed. It's like the sort of situation you just can't prepare for, right? If you live on a flood plain, you take precautions and you build levies and you pay insurance that's based on that. If you don't live anywhere where there's ever been a flood before, you don't pay exorbitant amounts of money for flood insurance.

1:08:06You don't build levies. And if something happens where you still experience a flood, well, then it's a catastrophe. And I wouldn't even – catastrophe is a strong word. But Apple having to go – I guess it was just a month ago, but it feels like – just speaking to how the summer is going, it feels like a while ago. But for them to come out and say, we're going to have to raise prices and then a week later raise prices mid-cycle, just Apple just does not do that. And for Apple, that's like a minor catastrophe. But I don't know what else they could do. I don't think there's any point where you could look back 18 months ago and say, oh, well, Apple should have really foreseen this and somehow done something different.

1:08:51What could they have done? How much do you see the leasing program as a direct response to higher prices? Do you think the leasing program is something that was rolled out because the prices were raised so abruptly? Or is this something that they've always been sort of moving towards? This program was probably in the works three years ago, and Apple works on a long time cycle. And so this is just the natural arc of things. Because they've had a refresh program where you could effectively subscribe for pre-memory price spike, right? Yeah. Yeah, they had the old iPhone upgrade program, which was just for iPhones.

1:09:34And I think they've seen it as a success. And so I think this was sort of a natural 2.0 way of – You know, and it's no, no, I mean, they're the ones who discontinued the old iPhone upgrade program the day that the new Apple upgrade program for all of their major products come out. I mean, you can even get like, I think you can even get like AirPods on Apple upgrade. It's not just for, I don't know, I think so. But certainly like all the Macs, all the iPads. So I think this was in the works. And I do think that they've been, I mean, the whole consumer market is moving towards, there's a lot of competition in paying over time.

1:10:23It's, you know, and for a while, I think that it was sort of at the consumer level, just locked into credit cards. You get a credit card, you buy something and you pay the credit card company over time. And I think a lot of people just looked at that and thought, we're leaving money on the table because they're buying our products. It's not just Apple. But everybody is looking at that and thinking, well, this is crazy. They're charging these consumers 13 % or 14 % APR. Well, how do we get involved in that? And, you know, Apple specifically, now they have the Apple Card, which is, I don't know, seven, eight, nine years old.

1:11:02I mean, it's been a while. And they've had a program through that where if you buy an iPhone, you get 24 months of zero interest. I think the big change, though, it's all clearly about the fact that for so many people, you know, I don't even want to pass judgment on them. But I think for some of them, they are mathematically disinclined to be able to extrapolate a monthly payment by the terms they're agreeing to and realize this is what you're going to pay over time. Sure, sure. They just look at the monthly payment and they say, I can pay that. And they look at the lump sum payment of$1 ,200 for the iPhone they want.

1:11:44They say, I can't pay that. But if I agree to this, I can get a new iPhone. I can walk out of the store with an iPhone 15 minutes from now. And so I think Apple just looked at that and thought, well, how can we offer something that's attractive? And the lease, I think the big difference with the lease is clearly that the monthly payment is even lower than the buy over time of the old iPhone upgrade program. But in either case, it is good for the consumer. And I know some really smart people who are like, I'm doing this because there's no interest on it. And if you do get a new iPhone every year, it is just an easy way to, you don't really lose anything.

1:12:25You're not paying any interest penalty and you just have like an easy system and they mail you the box to send the old one back in every year. So I think it's just a way for Apple to sort of kneecap the number of people who are buying their iPhones with credit cards and paying interest penalties over time. So I think it's better for everybody. It feels like I could bundle the leasing program, the MacBook Neo, into an Apple is going down market strategy. Do you think there are other plays that they will make to sort of calcify that strategy or really deliver on it? Or is this sort of like you think it's a pretty mature strategy that they have?

1:13:06I think it is incremental, let's say. These are not major moves. Even the MacBook Neo. I mean, I think the MacBook Neo is one of the, certainly the most interesting Mac product since the Apple Silicon in 2020. And it is that there were a couple of times where MacBook Airs dropped to like$899 or something like that. But that$999 starting point had been the starting point for a base MacBook Air for, I don't know, at least 15 years. And so when you consider inflation, the price has been coming down, but not down in the way that like what's the entry model for an HP laptop or something like that, which is like$400.

1:13:51So to drop the entry price, not just by like$100 or$200, but all the way down to$599. I guess it's up to$699 now with the price increases. But still, that is significantly lower. And I think they've been doing the same thing. I think on the iPhone side, the switch from selling SE models, iPhone SE that only get updated every three or four years to an annual schedule of having whatever the current number is, they stick an E at the end, 16E, 17E. There will be an 18E next year. Those are really low prices for a brand new iPhone that has the latest silicon. I mean, it doesn't have the greatest camera, but so many people do not care about the camera.

1:14:37And that camera on the e-phones, it is fine. It's one camera on the back. It's a 1X camera that you can get 2X optical with the fancy sensor system that they have. It takes great photos. I mean, but it's the same sort of thinking where they're not going to sell years-old products as they're, okay, fine, here's our affordable stuff. It's here's brand new stuff. It's just less technically advanced silicon wise, but it is brand new, right? The Neo is months old and it's at a record price. So they're incrementally moving down market. But I think only insofar as they think that there's people with money to spend.

1:15:20Yeah. Jordy? What percentage of the employees and the team over at Apple do you think are excited, truly excited about AI and think that it can make the Apple ecosystem significantly better over time? It feels like to me there's like very it feels like to me that the whole company is like, ah, like, why is this happening? to ask instead of like instead of saying like hey this is this like you know useful set of tools that can make our software better that can make our devices more magical to use uh i it's a good question and the last thing i would say is like there's i don't know anyone who's excited about ai that's saying like i gotta go work at apple because they have the best devices in the world they have billions of users biggest opportunity this is such an amazing opportunity like if I want to work on consumer AI, I have to go work there.

1:16:19Like that's just, I don't know anyone like that. Yeah, I think that's true. I don't either. But I don't know that it's a problem. And I do think clearly AI is where Apple was caught flat-footed. I don't see how else you could describe that, right? And I think as time goes on, looking back at the WWDC from two years ago, 2024, four, when they first announced Apple intelligence and then announced the stuff that eight months later was supposed to be rolling out. And they were like, you know what, we're going to have to postpone this by a year. And really, when they postponed it and said it'll be coming in the coming year, this is the coming year, right?

1:17:00And it still isn't really out. It's in the public betas this summer. It's coming out this fall. So they announced something that they were saying was going to come out to consumers, be in people's hands in the first half of 2025, and it's coming out at the end of 2026. And in this market, in the AI market, that is a big miss time-wise. Now, 10 years from now, will people look back at this and say, wow, that was a huge gap of time? Probably not. And when people look at the new Siri AI that's in the OS 27 betas right now, is anybody who really is juiced into the whole AI system saying, wow, they've really taken the lead here in any way?

1:17:47No, absolutely not. Right. This is very basic stuff. But it all does work. And it's going to be the intro to LLM-based generative AI for hundreds of millions of Apple customers. Hundreds of millions of them have never used any of the stuff. However popular ChatGPT and Claude remain in the App Store, there are hundreds of millions of Apple users who've never used this. And I've been using ChatGPT in particular for years. I'm definitely not amongst my peers in the tech media. I use it less than most, but I'm pretty familiar with it. But I have to say, testing it this summer, ever since WWDC, for all of the basic bitch questions, the Siri AI is great.

1:18:39And just being able to squeeze the side of the phone is a great way to do it. And I honestly think it's like what triggered OpenAI to have the fiasco of a launch of the new desktop app for ChatGPT? Clearly, the biggest driver of that is their fear of missing out with Claude and Anthropic taking the lead. If you just polled everybody who watches TBPN, who's in the lead right now? I think it's very clear. I mean, I think you'd be kind of nuts not to say Anthropic. Yeah, it's verifiable based on revenue run rate. So it shouldn't really be. And just the state of things, right? You just lick your finger, put it in the air, and who's got the momentum, right?

1:19:27And that's put OpenAI. They seem to have, and still, I think, are panicking over that. They really seem to have formed a sense of self, which was that they had already won, I think. you know, like somewhere around 18 months ago or so, that whole company had the, we've won this already. It's, we're just mopping up the chessboard at this point. And now they've found out that they're in a long-term race and they're not. So that's clearly the biggest driver is that they, they looked at the way Anthropic has bundled up cloud and cloud code and said, we need to do something more like that. But I do think, I absolutely think part of it is that the Siri AI app that Apple is rolling out, which visually, you just look at it, it looks like ChatGPT, right?

1:20:16It's black and gray. At a glance, if I showed it to you in May before WWDC and I just quick showed you a screenshot of it, you'd say, oh, yeah, that's ChatGPT. No, it's Siri AI. And I think that in the Apple world, we call it Sherlocking because there was the Sherlock thing 20 years ago where there was a Apple had an app called Sherlock and some third party developers made a much better version called Watson and then the new version of Sherlock came out. Sherlock was Apple's first. Then there was a third party thing called Watson. Yeah, it's funny. I've heard the story the other way. I always get it flipped.

1:20:54Yeah, thank you. This is helpful. No, and the third party one was called Watson and it was way cooler. And then the next version of Sherlock was like Watson. And it was sort of, you know, it's like tough luck to the outsider, but it's like everybody looked at Watson and was like, oh, this is the way Sherlock should be. This should, you know, and Apple looked at it and thought this is the way Sherlock should be. And, you know, Apple had this stance that was quite outspoken about it. Like Craig Federighi, I think it was with Joanna Stern last year, you know, said that Apple did not see chatbots as a good interface to AI.

1:21:32That's right. You know. Yeah. Yeah. And they're just like, you know what? It is a great interface to AI. So why don't we make one too? And I sort of think like ChatGPT sees that for non-advanced usage, nobody's going to go to these third-party ones when Siri can answer that. So no, is it exciting? And is it where the people who want to work at the frontier want to work? Do they want to work at Apple? No. I mean, why would you? They don't really have a product that's even aiming for that. But Siri AI isn't aiming for that. It's just aiming for the type of trivia questions that people just ask as like a one-off, two-off chat session.

1:22:12And it does a fantastic job. And it has these integrations with stuff like if you use Apple Mail, if you use iMessage, it finds stuff in your iMessage just like they promised. It really does work. All of the stuff that's in my Apple Notes, I ask questions about it, and it just finds the answer to it. It's really, really useful. Is this impressive to people who've been following AI for the last three, four years? Not at all. But this is the Apple way, is take something that is super exciting, super cutting edge, and boil it down to its basic bitch core and put it in a usable interface that normal people will be able to understand and give it to everybody.

1:22:57Okay, let's play it out a bit further because I'm curious where you imagine Siri AI goes. I've always felt similar to you in that there's so many questions that you don't need incredible advanced intelligence. You just need a simple answer to something. And I can see a lot of that flowing through Siri. But where does this product go over time? Is this something, because if you play it out far enough and the product gets enough traction, then at what point is Apple competing not just with other LLM providers, but competing with Google itself? And then they say, like, oh, we're for privacy and all these things.

1:23:42But then at what point does it say, hey, this is costing us a lot of money to run. We've got to start running ads. And then at that point, like, do we do targeted ads or non-targeted ads? Or do they try to take the moral high ground and say we're better than ads, even though they run ads in Maps and they run ads in the App Store and all these things? So play it out for me. How does this go? Let's assume that it's a success to the degree that they keep investing in it for many, many years to come.

1:24:18The ad question is interesting because obviously Apple never had to or never decided to make a search engine for Safari. And they just sat on, you know what, we'll just have Google as the default and Google will pay us for the traffic acquisition. And we all know from various court cases that however much they'd like to keep it under the radar, it's like$25 billion a year now. And it had been in that range. It's been a lot of money for a lot of time. And for a while in the early era of Tim Cook's services narrative, even though it was a lower number 10 years ago, it was a huge percentage of Apple's quote unquote services.

1:24:57So Apple's, you know, Tim Cook said to Wall Street, our growth area is services and their services number kept going up. But it was really for a long while just the Google traffic stuff. Why would Google keep paying this much money? Because when people would search, you see the Google results and what does Google show them? Ads, right? And there's a very simple, oh, this is kind of, you see how this works for everybody? There's a sort of flywheel. So Google gets all this traffic from the terrific audience, the demographically attractive Apple user audience, and is obviously very profitable at showing ads and search results.

1:25:37They pay a significant portion of that to Apple to have it as the default in Safari. And everybody just keeps searching in Google, and everybody can kind of bitch about the ads that have gone into the search results, but people still use Google. What happens with Siri where there are no ads? Why, you know, how does that financial relationship work with Apple and Google for the back end? Like, who's paying for it? That's a real mystery. And they've very, and I, it's one of those questions where when, like, Apple just reported results at the end of last week. every time i i i never listen because i find those calls in they're like 45 minutes long and they feel like 45 hours to me so i skim the the transcript and i'm like i would love if somebody could ask and i don't think that they would answer but if there's any kind of clue about how that it's so it's so interesting to me because google is help is like we're gonna help apple because we want to try to commoditize the llm space because we're behind and we're and our business is threatened for the first time because a billion people are using effectively another search engine an answer engine right they're using chat gbt they're using some gemini they're using some clod but like obviously by and large that usage is going uh going somewhere else besides google and so google makes the call to support apple and apple's effort to try to commoditize the sort of assistant category.

1:27:08But then through that, they, if you assume that partnership is going to be successful, then what happens to their existing Apple partnership? And then what happens to search over time? And then even with a demographic, even with a group of like customers, like iPhone customers, if you, there's a lot of people that are not going to sign up and say, oh, I'll spend another 20 bucks a month on, on AI with Apple. Right. Um, people already complain about the, like, you know, twice a month, you see a charge from Apple and it's like, what's that charge, you know? Um, and, and so there's the big question to me is like Apple right now would, to me is like positioning the entire company around being against ads and, and, you know, really pushing privacy.

1:28:00I see random out of home ads for like Safari and it's like, finally private browsing that is like already counter-positioned against google who is now their partner and then eventually i just think if you want to serve this product to the entire apple user base they're probably gonna it will probably make sense to do ads then at that point do they dance with google again and and use like the google ads network but it's so hard to see where this actually goes. They are, but Apple is dipping its toes into ads, right? The App Store is ground zero for this, where when you search in the App Store, they have a big ad in the top spot now.

1:28:43And sometimes if you're searching for, let's say, Signal, and Signal is the example I always go to because they don't pay. They don't seem to pay. But if Signal, the Signal group, or whatever the organization is, Bounce Signal, pays for an ad on the keyword Signal. You can buy your own name as the top result, and then you get an ad, and then you get the second spot too because that's where you naturally show up. Or you search for Roblox or something that kids play. Sometimes there's gambling apps that show up. It's crazy. And now they've added a second spot in the third spot. So for most things you search for in the app store, The top spot is a paid ad, often paid for by the app you're searching for, but they're still paying.

1:29:31Now they're like double dipping because Apple's taking the commission from the App Store transactions and making you pay to get the top spot. The second spot is the top natural search result. And the third spot is another paid ad now. And I think that's off brand for Apple. I really do. I think it is, you know, like I always go back to HBO, which it just was when I was a kid, it just seemed too good to be true that you could watch movies and there were no commercials. And it's like the only what's the catch where you're you got to get your parents to spend, you know, whatever, 10 bucks at the time in the 80s, like 10 bucks a month extra on the cable bill.

1:30:09But it seemed like such an amazingly good bill. Good deal. And that's, you know, always been sort of the Apple thing. It's like, oh, well, MacBooks cost more than other laptops and iPhones cost more than other iPhones. What do you get? Well, part of the experience is you don't get inundated with ads. And I go back to the Safari thing with Google search. Like it has been such a good deal for Apple. It is 20 to 25 billion dollars in cash that just comes in. It is almost no margin. You know, you can say that you could subtract. I mean, almost 100 percent margin. Almost, right? I mean, what's the salaries of the team that makes Safari and WebKit compared to$25 billion a year, right?

1:30:54It's as close to 100 % margin as anybody could reasonably get. It is year after year. And it doesn't look like Apple is serving you the ads. So you're just Joe Schmoe, Apple customer. You use Safari on your iPhone because it's the default browser. You search, you see ads. It looks like Google's Google is showing you the ads, but it's like Apple gets all this money and their hands are free. That that really can't happen with Siri, right? They could partner with Google to have Google sell the ads, but it's coming through the Siri. It would be in theory in the future. It would be coming through the Siri interface.

1:31:32I don't think they'll ever do it. I don't think they'll ever put ads in Siri. And I think if anything, they'll put like a cap on the number of queries you can do that need the server. And I think ultimately with Apple. The question is you just play it out, play it out, play it out. And how much of a threat, like is Google creating an even bigger monster than the one they were trying to, you know, keep down in ChatGPT, right? If you help make Siri a massive success and there's no ads, and I have gone a year ago, I didn't use ChatGPT for really any product-based searches. I just still thought that Google was just objectively better if I was looking up a car or a piece of clothing or anything like that.

1:32:21And more and more and more of those kind of searches, I've just ended up in Chat because they have better images now. and I can be like, hey, I'm looking at this exact type of product. Just go find it for me. Find me the cheapest version of it. And so you would expect that Apple is maybe like two years-ish behind or maybe 18 months behind in terms of AI capabilities. But I could imagine 18 months from now, Apple becomes pretty good at finding you a product that you want and you have Apple Pay built in. It's like they have your address. That is crazy. And so you potentially just fully cut Google out in a way that doesn't even necessarily, that becomes a cost center for Apple and sort of cannibalizes Google search revenue, but doesn't even necessarily get Apple that much out of it itself.

1:33:09And then like you were saying on the App Store, you're right in that searching for an app and then being flooded with a bunch of what feels like spam on an Apple surface area. The Apple, the company that's meant to be dedicated to just giving you the most magical product experiences, it doesn't feel Apple at all searching on the app store anymore because Apple knows exactly the app I'm looking for. And yet they're serving me junk. Right. And it's usually like. And and the Apple monetization for the app store should just be the commission they charge on all the transactions through the app store.

1:33:43Right. And that's controversial enough and has been the source of antitrust and, you know, all sorts of complaints. But at least in terms of, well, how do you how do you fund how do you profit from the app store? The story was very simple. They take 30 to 15 percent of every transaction. And that should be good enough, right? Because you're paying. And at some level as a consumer, you have to assume that you're paying at least a little bit more for everything you buy through the App Store because Apple is taking that commission. and to double dip and show it'd be like hbo you pay extra money and in the middle of the movie they'd have one commercial interruption right and it's like with the world cup where oh they still don't have commercial breaks but they have hydration breaks and right and it's like that that's sort of what it feels like with the app store showing you these ads when you search and it's like i'm just trying to find a freaking thing how about you just fix search and their search still kind of sucks.

1:34:40So, Jordi, I think your big question is what's in it for Google, right? Why is Google helping Apple here? And I think it is from Google's perspective. And I don't think anybody knows. I really don't think anybody knows where any of this is going to be four years from now, right? It's so fast moving. But I think Google would rather dance with Apple who they know, and they know what Apple culturally wants to do. And I don't think there's any kind of I mean, Apple is doing AI research. They have AI teams. And I think Apple would obviously love to handle as much of this on their own as they could.

1:35:19But is Google seriously, is anybody seriously afraid that four years from now, Apple is going to be the producer of the leading edge models, the frontier models? Nobody. And I don't think anybody really thinks that they've got a path to that or that they're even trying for it. So I think Google is comfortable working with them where, well, in a way that they would just prefer to kneecap open AI and Anthropic. And if there is, if Apple was fishing around for a partner, like we need somebody who knows their shit to give us an LLM backend so that Siri is actually useful. And I think Google was willing to cut them a sweetheart deal to say, let's do this.

1:36:04And if it entrenches Siri for a decade to come as something that a billion iPhone users around the world are relying on, we can handle that. We know Apple. We're not worried. We'll still figure out a way that we'll make money. We'll still have everybody who's not using Apple products. but a world where like that two year ago deal with open AI, which now makes you laugh with, with the law, with the deterioration to say the least between Apple and open AI at a corporate level. But if that had continued and it was open AI who was partnering with Apple to do this in a way and and again the original deal two years ago had chat gpt branding yeah in the answers and it was like uh you know yeah the whole thing is so funny because you have google being like okay we need to help apple because like we need to help apple attack chat gpt because chat gpt is threatening us but if we if we are if apple is really successful then they'll end up threatening our core business and apple's like well we also want to take we also want to try to kneecap chat gpt because they're now building devices but then it google and apple's relationship is just like headed on a pretty interesting path and i and i can't imagine i i think they both companies will look back and be like man you remember those days where i could just give you tens of billions of dollars and we can just be friends and then you know you play it all out and and uh it doesn't it doesn't seem as as friendly yeah and i think basically it comes down to google is very comfortable playing in a world where the technology is commodity level and that they can, because of their scale, they can not just succeed, but thrive in a commodity world, right?

1:37:59Like, is the actual computer science behind Google search significantly better than what Microsoft has with Bing? No, not really. It's just that the scale is there on Google's side, and so it's perpetuating. And I think that if AI works out that way, there was somebody at Google years ago who, you know, Google invented all of this technology that we now consider AI, all the LLM stuff. But there was a paper that came out of Google where one of the scientists argued that there is no moat around this technology, right? No one company is really going to own this. And I think it's an open question whether that's true or not, right?

1:38:40The super intelligence hypothesis is that if somebody gets to the breakthrough of super intelligence first, then that super intelligence will accelerate their AI at a pace that nobody else can keep up with and no one will ever be able to catch. I don't think that's where it's going. And I don't really. And it just goes back to what we were talking about half an hour ago with the hugging face thing and chat GPT. And it's like, oh, coincident. And, you know, and this is what makes people roll their eyes and think it's a marketing stunt is then, you know, three or four days later, Anthropic came out with, well, we went back and looked at our logs and we found Claude broke into somebody four months ago.

1:39:21We didn't even know it, you know, and it's, you know, they're even Steven. And I think I just think ultimately Google looks at this as let's just shut those guys up. Let's let this bubble burst. We'll still be here. They won't. And Apple will certainly still be here. You know, Apple's the one big company with no exposure in this bubble. And I think Apple's bet is back to there is no moat here. And if they don't own this technology, it's fine. And I kind of think it's shaking out that way, at least for Apple's business. Well, last question. And you can answer in 30 seconds. Do you think Apple will make an acquisition of an AI company, could be an acqui-hire or product, but I'm thinking more acqui-hire, north of$10 billion?

1:40:16$10 billion? Whoa. I think you're going to say north of$100 million. Yeah, one. Okay, then let's bring it down. $1 billion. Because there's not that many great teams, to be honest, that you could get for less than$10 billion. I mean, they bought Siri. That was$100 billion. And I was kind of expecting them to try to pick up Poke, which Cognition just balked. Poke was like a really nicely designed assistant that worked in iMessage. And that felt like a no-brainer for them to just bring in some talent that is excited about AI that's already working in the Apple ecosystem. I would look at it on the Silicon side.

1:40:53And I think because I think that as this gets commodified, I think that, you know, something like the PA semi acquisition that led to Apple Silicon, that that's who Apple, I suspect, is hunting for is somebody who has like a breakthrough spitball idea. And again, at this point, one billion dollars probably isn't that much. So something in that range. But I would think hardware, something that Apple can do for Silicon. Yeah, if you have really optimized silicon serving lagging models that are a year or two old, the inference cost can actually be pretty low, and then that changes all the calculations about ads that you were talking about.

1:41:33So thank you so much for joining us. And ultimately, I think it's why Apple is staying out of the CapEx race and building out these data centers. It'll just wait. is the idea of why are we going to spend all this incredible sums of money, all of our free cash flow, to build data centers that are going to be completely technically outdated five years from now. Yeah. Well, I want to have you back to have a great debate about Electron versus native apps. We've got to go through all this. Get to the bottom of what's going on with these AI labs. They have such powerful coding agents, and yet they can't ship native code, apparently.

1:42:06We'll get to the bottom of that next time you're on the show. But thank you so much for taking the time to come chat with us. Have a great rest of your summer. Yeah, you do cool out there. Thanks so much. We'll talk to you soon. Goodbye. Let me tell you about Figma. Agents, meet the canvas. Your AI agents can now create and modify your Figma files with design system context. And up next, we have Sean McGuire and Isaiah Taylor. Dynamic duo. Two guests, third, fourth time on the show. How's it going? What's up, guys? Look at this guy. Look at you. We've got a nuclear reactor. We've got to get the eagle scream.

1:42:39every time this is like this is our intro this point exactly there we go what happened guys what happened one billion dollars series b

1:42:55with the wind up crazy crazy crazy moment um where should we start isaiah you want to kick it off yeah i mean i'm right here in the reactor hall right now with the hardware this is where i like to spend my time. And one thing that was just amazing about Sean's partnership in getting to know his company is he just wanted to know about the hardware. That's really my favorite type of investor. They're like, show me the stuff. Like, are you actually building things? I think that's what's special about this company is that we just build and we build fast. And yeah, it's been really great to meet the Sequoia team and get to know everybody.

1:43:29And this facility, you're no longer purely in the gundo. You've expanded. Where is this facility now? This is our Orangeville, Utah site. It's our first nuclear site. Okay. And what are the short, medium, long-term goals with this site and beyond? Is this still an R &D site or is this going to be generating power? When do you start building the factory, the machine that builds the machine? Yeah. So we've already made a tiny bit of nuclear power here. We became the first startup in history to make nuclear power about a month ago. But it's still a test unit. And what we're building toward is something that we call a gigasite.

1:44:07These are massive campuses of reactors, and we think they'll make the cheapest energy on Earth. And so that's what we're working toward here. And it starts right here in Utah for sure. This is a great place to serve data centers, AI factories, and eventually all of the industrial stack that has been falling behind in the United States that depends on energy. Aluminum, electrolysis, steel, all of the different input metals to making everything in the physical world starts on our gigasite. So that's what's coming next, And that's really what this raise is going to help support. Massive. Sean, I have to assume you've met every nuclear startup.

1:44:42There's a lot of them, but you've got enough time in the day to meet them all. I don't know. I doubt there's so many different applications of the technology. There's a lot of good ones. You know, it's not only Valor, but there's only one Isaiah. That's for sure. Yeah. Yeah. So talk about, yeah, talk about, yeah, what drew you to the company. And this is like, you know, this is not a it's not a series A check. This is a billion dollars. It's a big check. This is a big boy check here. This is a ultra high conviction investment. Look, I'm a former physicist. I have a PhD in physics. I've loved nuclear since I was a little kid.

1:45:16I was hoping that, you know, I read Richard Rhodes's Manhattan Project book when I was like 17. And I've kind of been waiting for there to be an opportunity in nuclear. Canly, I didn't think it was going to happen. I thought we were just going to go like solar battery or solar plus some other storage mechanism. Future, just given how regulated nuclear was, it just didn't seem like it'd be possible to get to scale. And then two things happened. One, like power became important in the West again, which is pretty amazing. Second, this administration. But I got to say this, I think a very bipartisan thing.

1:45:56Both the Democrats and Republicans, this power became a bottleneck. have been pushing for more favorable, you know, nuclear regulation, and it's starting to happen. So even three years ago, I just think the regulatory side was too scary for me. So we've had kind of regulatory breakthroughs with this next generation of founders that are, you know, really trying to bring in this nuclear future. But for me with Isaiah, there were a few things. One, just like when you get to know this guy, the level of intensity is psychopathic. And And I mean this in the best of sense, but like the day he went critical in this reactor he's in, first of all, like he thought he was going to go critical the day before it ended up being the next day.

1:46:42So he was up kind of all night, two nights in a row. Then that night that he went critical, they didn't go critical on like 9 p.m. or something. There was a candidate he wanted to close whose partner was in San Francisco. So the day he went critical after not sleeping for two days, he got on an airplane that night to fly to San Francisco to have drinks with, you know, a candidate and the guy's wife. I don't know what time he went to sleep. And then the next day he's back in, you know, I think it was Utah, maybe Los Angeles. But like getting to know Isaiah, the level of intensity is just absolutely incredible.

1:47:14And can't be the only person I just I don't know many people that have this level of intensity. And then just one more thing. something that I learned, like a lesson, a mistake I made in the space industry was, you know, Elon started off trying to build the Falcon 1, which was a pretty simple rocket relative to, you know, what NASA could do in the, you know, 80s, 90s. It was like a 1960s rocket, or maybe even earlier. And there were all these other companies that were telling these like advanced science stories. We're going to use carbon fiber, you know, vehicle frames to have, you know, less mass, or we're going to 3D print the rocket.

1:47:53So other people were like trying to do this very advanced technology to have better mass ratios, et cetera. And Elon was like, I'm going to do the simplest thing possible, put up one satellite, you know, get that revenue and then go from there to something that's, you know, Falcon nine, pretty damn hard, but still not state-of-the-art compared to what NASA had done in, say, the 90s. And then from there, go to what's truly state-of-the-art, reusability. And then from there, go to Starship, which is just completely pushing the limit. And Isaiah understands this. He's one of the only founders I've ever met ever in any industry that really understands the power of starting with the simplest unit where you can actually scale and win, and then climbing from there.

1:48:37So anyways, Isaiah should do the rest the talking. That was great. Yeah, I think that's one thing that Sean and I just connected on very early is like lots of investors trying to understand the nuclear space want to know what's special about this technology, what's really unique about this technology. They want to find this sort of like IP technical edge where you're doing some special sauce that nobody else is doing. And my approach is like, no, like we want this reactor to be as simple as we possibly can. If we could just staple this thing together from Ikea, then this would be a trillion-dollar company much faster.

1:49:11And so there's two philosophies that we use in building the reactors. We try to buy things that are completely off the shelf, like 100 % commodity, or we make it ourselves. And there's very few things in between, right? There's very few places where we have a supply chain that's dependent upon the existing nuclear industry. Because if you think about it, we're trying to go 100 times faster than the nuclear industry's ever gone before. And so if we're plugging too deeply into the existing network, it's not going to work that well. So we want to use standard off-the-shelf things and make things ourselves where we can't buy something off the shelf.

1:49:45And by necessity, that means the design is extremely, extremely simple. The more complexity you add to it, the harder it is to be one of those two options and the harder it is to scale. So I think especially in nuclear, this is a difficult thing because it's full of very smart people. It's full of physics people and PhDs and people who have spent their life doing complex analysis. And they actually want something that is a little bit complicated. It's like, it's an ego thing to design something that is like complicated and looks very sophisticated. And we have just really rooted that out of our minds at Valor.

1:50:19Like we work extremely hard to root that mentality out. Like it's our preference that this thing is so simple that somebody with a nuclear PhD looks at. It's like, that's like a toy. And it's like, great, because people make toys in like the millions, right? They just like stamp them out. And that's exactly what we want to do. And then the other aspect that's super unique here is just the safety of the overall architecture lends itself to this approach. If you build a really, really safe reactor, you're also by necessity building a really safe reactor because most of the engineering complexity in nuclear comes from safety engineering.

1:50:51If you look at a modern pressurized water reactor plant. They're very complicated, and 90 % of the complexity comes from trying to make it safe. So the approach that we've taken instead is design it to be safe from the physics, and you can actually just delete a huge amount of the bill of materials. Just completely remove it. It doesn't even exist in our bomb. So those are the philosophies we've taken here. Listen, I can't give enough credit to Sean in particular, and also the rest of the story. No, no, no. This is all Isaiah. This guy's building nuclear reactors. I'm wiring money. Get out of here.

1:51:25I got to give a couple of shout outs, though. I got to shout out Palmer Lucky, who gave me the hat tip that this is a special company. Thank you, Palmer. Hat tip to Liam Corrigan, newest investor at Sequoia, who was my wingman here, physics guy, road scholar, Olympic gold medalist, rower, 6 '5", Chad. Total Chad. Total Chad. Total Chad. and Max Ucroprina, who's on the team at Valor, who is someone I've known since I was a kid, who for years was trying to get me to come meet this company. I was like, oh, dude, nuclear is too hard. Regulatory is not favorable. And anyways, Max, I look stupid. Good job.

1:52:04Isaiah, have there been talent wars in nuclear? You're flying on such an important day to go meet someone, which many founders will have done, but that's felt like maybe urgent uh how intense has the competition for talent been this round i imagine will give you a lot of advantages just having um get more firepower to to continue compounding a great team but walk us through the last maybe two years in the category yeah look i i view my job essentially as trying to get the most talented people in the world to come and build this mission with me. There are no blockers in front of us. We have a regulatory environment that's ready to move.

1:52:47We have an enormous demand signal. We have customers that want to buy. We have an architecture that's very simple and very scalable. And now we have a lot of capital in the bank. And the blocker on us becoming the$10,$50,$100 trillion company that makes most of the world's energy is the smartest people in the world coming and joining us in this mission. So I spend an enormous amount of time of my time doing that. I actually would say that our primary talent competition is in other places where you can move the needle on a global scale that need incredibly talented engineers. I don't see it as sort of like, okay, other nuclear companies.

1:53:22It's more like, you know, what are the other companies that are genuinely going to change the force of humanity in the next 50 years? And like, those are the people that I'm fighting for. So yeah, it makes a lot of sense. Lessons from State. Oh, sorry. If I can say one thing on that, something I've seen from Isaiah that is very rare, he is looking forward to just ultra talented generalists and he tests people like crazy. And like real world tests, you know, like, hey, I'm going to be in, someone says like, hey, I want to interview. And he says, okay, I'm in Texas. Like meet me here tomorrow. And if they get to Texas, then they have a shot.

1:53:56And if they don't make it to Texas, then they're weeded out. It's just not having that level of like commitment. And anyways, there are not many people that understand. You have to kind of design the hiring process to find the people that select into a crazy mission. High agency. Yeah. And that is the lesson from SpaceX. You're revealing the secrets here, Sean. I know. Now the next three are going to show up to the meeting and I'll have to figure out some other way to weed them out. But yeah, that's true too. If they show up, it's still a good thing. It's true. It's still a great sign. Lessons from SpaceX.

1:54:29One of the interesting stories. Yeah, I can imagine. Yeah, sorry, I got a dentist appointment in Utah of all places. I got to get out there. It's like people being like, yeah, I got a dentist appointment. I have to take three commercial flights and drive three hours to get to my dentist appointment. One of the interesting SpaceX stories is residual capability. They build all this launch capacity. They have maybe too much launch capacity. You get Starlink, amazing business, becomes a telecom company. Probably not in the first pitch deck, Sean, you'd know, but is there a world where there's a residual capability from what you're building here where you're using the electrons that you're generating yourself?

1:55:10Or do you see there's just like so much demand that just that's something that's like is very, very unlikely to happen? Yeah, I mean, look, you know, you and I have talked about this before. Like, I think Valor actually started. Look, I grew up watching Elon, right? I read everything that I could about the Falcon 1 and the Falcon 9 and saw this very simple path. They call it flight rate. I think at SpaceX we call it tick rate. Tick rate is basically how quickly can you go from cores turning on each one. It's like a look-back average metric of the time between new cores turning on. And it is predictive of who's going to win, who's going to have the lowest cost and the highest capacity and all these different things.

1:55:50And we actually started with the idea of that excess capacity and being able to make the world's commodities, right? And I think that the AI thing happening as quickly as it did and power prices changing so dramatically where people will sign a$200 PPA is like, okay, obvious that we should sell electrons. But no question is our long-term vision to actually have the cheapest energy on earth and to use it for our own things. I mean, we see a vision of the world where steel is just way cheaper than it is today. Aluminum is way cheaper than it is today. The manipulation of matter is a lot cheaper because you have robotics hooked up to AI that's doing matter manipulation and vision and all these things just take energy.

1:56:29And so it'll be interesting to see where we decide to play in the stack. I think we want to be in the business of turning on thousands of reactors primarily. but there will be a couple of like massive, massive markets to be attached to that, that we can just make it a competitive price to no one else. And also like the nature of the AI boom is that you need a lot of energy in a single place, which is perfectly suited for you as opposed to if we were in some boom where we need everyone in their pocket needs twice as much energy, you would have to maybe transform into another source of energy or do something else in the supply chain.

1:57:03Sean. Did you have something? No, I agree. I concur. I'm just violent agreement. Fantastic. Well, thank you so much. Honestly, I was just going to say, the AI thing perfectly matches gigasites. The idea is nuclear is a thing that benefits from extreme scale. If you could build 1 ,000 nuclear reactors right next to each other, you should. You will get the cheapest energy anywhere in the universe if you do that. And so AI is the perfect thing to do with that first. but we will do many, many interesting things with it over the next century. Well, congratulations. Thanks for having us. Excited for you guys to partner up.

1:57:41We'll talk to you soon. These tag teams are fun. Peace. Goodbye. Cheers. Let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI. Own the data platform that powers it. We're going back-to-back energy rounds with another back-to-back billion-dollar round. We have Justin Lopez from Base Power Company. Justin, how are you doing? Good, how you doing? Good to see you. Give us the news. What happened today? Raised a bunch of money.

1:58:16Not just a capital raise, though. Break it down. But first, for anyone that's been living under a factory, reintroduce base power, what you're working on, why it's important, and then I want to talk about the news under the headline from today. Yeah, so welcome to Factory One. Good to see you guys. Thanks for having me on. Today we're announcing three different things. Number one, launch of the factory that I'm sitting in right now. It's behind me that builds batteries. Number two,$1 billion raise at$13 billion post-money valuation. And the third thing is BaseCore is our fully custom design, engineered, installed, manufactured here in Texas.

1:58:57What we do is we design batteries. We manufacture them here. we hold them on homes and then we own and operate them as a distributed fleet distributed power plant uh that's the business today uh incredible talk about the decision to not announce the factory or talk about the factory very much until it was actually producing uh products yeah you know it's like uh i i put this on x but you know a lot of factories get announced with a bunch of people in suits and hard hats shoveling like an ounce of dirt and that's cool it's groundbreaking that's exciting and all that but like factories are meant to make stuff and yeah we wanted to make stuff beforehand so uh we're we're doing that today we're just getting the line ramped up the the station behind me is starting to build some modules you'll see some come through uh as we talk here but um yeah look you want to have you want to have the real deal ready before you talk about it that's right and when did you when did you guys actually break ground on this site because i imagine so the site yeah the site was already here so this is actually the old printing press of the Austin American Statesman building.

2:00:00We started building the equipment behind me about five months ago. So it hasn't been a super long time. And we started warehousing here about eight months ago. Previously, we had a smaller facility just north of here in Austin. But yeah, we're live and running now. Very, very cool. Talk about the state of the business overall, how the market in Texas is evolving, what people can expect from base over the next couple of years in terms of new markets and things like that? Yeah. So as I mentioned, look, we install batteries on homes. Turns out there's a lot of homes, not just in Texas. So we recently launched our Chicago market, and we've also launched, I think it's six or seven utility partnerships now here throughout the state of Texas.

2:00:45You'll see us announce a few new states and a few new utility partners, hopefully by the end of this year as things get signed and under contract. And it allows us to go into more states. Here in Texas, Texas has got a partially deregulated market, which means that we can go to market without having a direct relationship with the utility. In other states like California, where I'm originally from, in Michigan, those are regulated states, so they require deals with utilities. So we have two go-to-market motions, regulated and deregulated. Both work in various different states in the country. and look, the goal is to have a battery on every home in the U.S.

2:01:21and eventually internationally. We've started here in Texas. We're in all the major markets now, also in Chicago, but we'll be launching new ones here pretty soon. All right. If somebody's a homeowner in Chicago or Austin or Texas broadly, give us the elevator pitch for why they should install. Yeah, look, so we make your power more affordable, more reliable. So if you're in a place like Houston or Dallas, where you can choose your power providers. You sign up with us. We sell you electricity every month at a very affordable rate. We also put a battery on your home that is only typically a few hundred dollars, depending on exactly where you live.

2:01:58That provides you backup protection if the grid goes out and also supports the grid when the grid's up and running. That's how we monetize and how we make money is that grid support function. And so it's affordable, reliable power. That's the simple pitch. Is there an element of price savings from drawing power from the grid at low rates, storing it, and then reusing it in the house when rates would be higher and just sort of load balancing at the house level? That's exactly right. That's basically how it works. So we charge the battery when energy is abundant and available, when there's not a lot of stress on the grid, and then we discharge it into the home.

2:02:39And sometimes if we decide to spin the meter backwards, push back onto the grid when there's some stress on the grid or when there's peak demand times. That typically happens in the dead of winter and in the height of summers. Okay. What about throughout the day, throughout the week? Like when are typical peak load times? Because during the day it's hot. People are running air conditioning. But at the same time, that might be when solar panels are collecting energy. So like what what is the actual reality of like a typical grid, the Austin grid? Does this vary grid to grid? Like what what is the the differences and the nuances of load balancing?

2:03:18Yeah, so you're exactly right. Basically, what you care about is the difference between available capacity or supply and the amount of demand. And so here in Texas and in many places, especially in the south where it's quite hot in the summer, you typically have these peak times in the summer, at least in the early evening. So kind of like, you know, 4 to 6 p.m., 5 p.m., 7 p.m., where the sun is setting. So supply is coming offline, but people are coming home, plugging in their EVs, turning on their air conditioners, et cetera. Got it. And so that's really where the supply and demand meets. Then in the winter here in Texas and in many places in the north, you have these early morning peaks.

2:03:56Basically, people are waking up using more electricity. Your heaters, your electric heaters are coming on. They're on throughout the night, but the sun hasn't risen yet. And so batteries help fill those gaps. That's at the macro level. And then at the more micro level, they're also able to shave peaks off of the lines on the grid. So you might have enough capacity in aggregate in bulk, but one part of Houston or Dallas or Chicago may need support. And so because we've got tens of thousands of these systems out there, we can say, look, in this neighborhood, we need some support. We're going to discharge just in that neighborhood.

2:04:28And that's the beauty of having a high volume of systems. I see a lot of robots of different types moving around in the background. Where are you guys getting the most leverage from robotics? I can imagine as you started the facility, it can make sense to figure out the sort of process using a lot of your human talent. But where are you guys getting the most leverage and how automated can this become over time? Can it become a lights out factory? Is that even something to aim for? But what's your view on all that? Yeah, you know, robotics is a great place, but it doesn't have a place for everything.

2:05:10There are certain tasks that do not make sense to automate or at least not as a starting point. And then there are other tasks that do. So like what you see behind me, you've got the robots that are driving through the tunnel. And that tunnel is placing these stacks of battery cells that are created also by robots outside of the frame here. That is a great thing to automate because it's relatively heavy, relatively repetitive, and requires a lot of fine precision. Then there are other tasks, loading in certain components, testing certain things that require a little bit more finesse, a little bit more dexterity that are pretty hard to, not impossible, but harder.

2:05:43And so we've said, look, we're not going to automate those as a starting point. We're going to focus on the things that are, you know, either dangerous or hard to do repeatably or are quality concerns, and then we'll automate more and more over time. So it's just, I'd say it's a phase-in approach. We're starting with what makes sense to automate and we'll likely trend towards more automation. But the fundamental thing is trying to not do tasks that you don't need to do in the first place. You delete the task, then you optimize it, and then you automate it. What's the state of blackouts, brownouts, blackout prevention?

2:06:16That feels like a huge selling point. You're selling a sense of security, a sense of comfort during a winter blackout or summer. But at the same time, how often are they actually happening? It seems like a known problem. Grids and energy providers have been working on this at a higher level than you. So is the problem still broad? How big is the problem of just losing power outright? And how big of a factor is that for you in the sales process? So it's huge in the sales process, right? People want to have more affordable, more reliable power. Yeah. The reliability is different from from place to place.

2:06:58So typically coastal regions. So we think of the Texas coast, Florida, the East Coast, et cetera, where you have a lot of hurricanes. You typically have more power outages. OK. Also in more rural areas. So if you're at the end of a line, basically any break in that line all the way up to the substation is is going to cause you to have an outage. And so if you have more line in front of you, basically, you have typically lower reliability. Yeah, it totally varies, though. So very, very, very rarely is there an outage because there's not enough power. Most outages occur because of the weather events, because, you know, a tree falls onto a onto a power line, et cetera, because of a storm.

2:07:33But but outages are rising generally across the country. There are places where they are getting better. OK, it's a factor in the sales process. But if I'm honest with you, the way we think about this is more about portability. First, reliability as a benefit of having this system on the grid. And it adds reliability to the whole system, right? It's not just your home. Obviously, it'll back up your home if and when the grid goes out. But it's more about adding reliability and capacity to the whole system so you can have more load and more EVs, more homes, more data centers, et cetera, on the grid.

2:08:05Got it. Got it. Absolutely crushing. My last question is, what does it actually take to expand to a new market? I mean, I want one of these in California. But it seems like you're growing the business and the factory. And there's immense scale, billion dollars raised,$2.5 billion raised. And yet, geographically, it feels like a little small. It feels tight. Sometimes you have companies that are, yeah, we're available in every market and all over the globe. We'll ship our thing everywhere. And we're a tiny company. You're sort of the opposite, very focused. What does it take to bring a new state, new city online?

2:08:39why the measured approach to actual go to market? Yeah. So I'll start with the latter point, which is the sort of measured approach. I'll remind you that Texas is larger than most. Yeah, that's true. Good point. We've got more homes and more electricity load than many large countries that you've heard of. Yep. Okay. Fair. So Texas is a big place, but regardless, I'd say the geographical tightness is a feature, not a buck. Okay. Yeah, of course. The reason for that is like not only do we have a factory behind me that produces these things, but more importantly, we have a factory in the field. We've got hundreds of people out in the field installing these things.

2:09:15And we've got trucks and crews and tooling and all this other stuff. And so having geographic density is very helpful from an efficiency standpoint. Now, to answer your question, though, on how do you expand to new markets, there are oversimplifying here, but there's basically two types of markets. There's ones like Texas where you can choose your power provider. Yep. Places in Texas, not all of Texas. And then there are markets like you have in California where, for the most part, especially in residences, you cannot choose your power provider. If you live in northern California, PG &E is the only game in town.

2:09:46If you live in southern California, it's South Carolina Center or LIDWB or other utilities. And so in the deregulated markets, we can essentially go there and start the business. And there's nothing really stopping us except for a bunch of regulatory hoop-jumping to do and setting up of a warehouse and hiring people and all that, which is its own challenge. In the regulated market, it's basically getting a deal with the utility. It's a B2B sales motion where we go to utility and say, hey, we offer megawatts as a service. We'll go install these batteries that we've built behind me. We'll put them on homes in your service territory.

2:10:17You can control them, operate them, and use them to add flexible capacity to the grid. And so that's a matter of B2B sales, long sales cycles, and working with both the regulators and the utilities in that state to go into those states. So I'll say we'll launch a few new regulated utility opportunities over the next few months here. And then the deregulated part will also launch a few new states. But as I said, it wouldn't be surprising to me if we're only in 10, 15, 20 states over the next few years just because, again, the geographical density is so helpful for us. Yeah, and of course, you only have so much manufacturing capacity.

2:10:51You have to load balance your own business across your sales people, your installers, your manufacturing capacity, your supply chain, all of this stuff. That makes a ton of sense. Well, congratulations on the progress, and thank you so much for coming on and sharing it with us. Love seeing you guys cook. Yeah, amazing work. Awesome. And thanks for the factory demo. Yeah, that is remarkable. I know. We'll come by next time we're in town. Yeah, that'd be awesome. We'll talk to you soon. Cheers, Justin. See ya. Goodbye. Let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange.

2:11:26Just do it, folks. I think we have to issue sort of like a warning to the viewer. The next guest is building something truly horrific. So avert your eyes if you are a scaredy cat, because our next guest has built something terrifying. Beau Gaston. Welcome to the show. How are you doing? Great. How about you guys? We're doing well. Talk to us about what you're building. Tell us about the journey, the goal, the mission. I want to get into this. everything, but also the applications? Yeah, sure. So the full timeline, a very high level. We'll start about six years ago. A paper comes out of MIT where a guy, Ben Katz, made this quasi-direct drive actuator quite cheap.

2:12:17I think it was about 700 bucks. These are kind of like the building block of these humanoid robots. And a large reason that you're starting to see a lot of these pop up, that he open sourced the design as well. So I'd seen back then, it kind of looked like it was going to become possible to build humanoids a few years in the future that are not only affordable to a huge lab, right? So around 2022, I start to take on this design that you see in the background. What I wanted to do with it was really make a robot first for myself. And what I would like to do with the robot is, you know, not do the dishes or fold my laundry.

2:13:01Do you want this thing to do the dishes? No, no, no. He's saying he doesn't care about a robot that will do the laundry. Oh, OK. Yeah, that'll be the next software update. I wouldn't try that quite yet. OK. My kind of thesis is it's going to be more useful to have something that's stronger than you at first. then maybe you could sacrifice some precision and intelligence, right? Sure, sure. And, yeah, running a chainsaw, I get that people have thought, like, wow, this is crazy to give a robot a chainsaw. It's, you know, a bit scary. But, I mean, what's really scary is operating a chainsaw as a human.

2:13:38So I kind of looked at that, you know, which is something I'm well familiar with living out here in the woods. So my kind of thought is this is a good tool to start with. I mean, running a saw is the most dangerous job in the U.S. It's 130 per 100 ,000 workers per year. It's 30 times more dangerous than the next most dangerous job. And yeah, I mean, you could imagine not only is a chainsaw dangerous as a tool, but what trees are you cutting down? the ones that are damaged near power lines, ones that are next to a fire break and a wildfire. So it's just a preposterous kind of situation. So looking at what robots can do.

2:14:21Yeah, the use case makes total sense. I'm curious to get into a couple questions. We should start with, I think, what everyone's wondering is why make it look like a demon from Hades? Yeah, yeah. Especially for some of these disaster use cases. I was thinking I've been in situations where I've never been really, really, really close to a wildfire. But I've been close enough where the sun gets a little bit blocked and it's dark and a little hazy. And if I saw one of these things kind of walking out of the smoke, I'd be a little freaked out. And I think a lot of people might feel that way. um but uh but yeah so so talk through the decision making on making it look uh demonic and then and then i really want to understand like the actual uh functionality of the form which we can get to yeah sure so um i will say yeah the the rescue stuff also was kind of propagated by the the virality a little bit that's definitely i was not uh aiming to pick you up out of a burning building, you know, like a hero in a movie, though I would point out if that did happen, like, you know, if someone's saving you, I'm not going to probably be too picky.

2:15:44And if it's this thing, I take it. Don't look at GIF Centaur in the mouth. Yeah, exactly. You just don't do it. Yeah, but, you know, it's interesting with the anthropomorphic stuff on humanoid robots. I think it's actually really disarming and people kind of aren't understanding what's really going on under the skin, which you can see a little bit here. But I'm like, have you guys ever used a drill press? Yeah. OK, so you kind of know they're dangerous and, you know, the workpiece could fly out of it and they got a lot of torque. So imagine if someone hooked 20 drill presses up series parallel all together and then used basically some arcane knowledge of machine learning to make it balance and walk.

2:16:26This is pretty much what a humanoid robot actually is, right? That's around the same power that you're seeing on leg actuators on most humanoids today. I mean, we see those videos where they go crazy. Yeah, where you're going is basically like a humanoid itself looks tame, but in reality, it's potentially incredibly dangerous to be around. So it's better to let people know. You're putting caution tape on it effectively. You're telling the user, hey, this thing is a chainsaw. It's going to be doing work in a dangerous environment. you should be triggered to say hey i gotta step back because this was serious business exactly like i don't know if you guys have ever been on a big work site or in a factory and first time you go in there maybe you're kind of putting your back to the wall and like you're a little overwhelmed and that's kind of the right attitude to take when you're around something that can knock your head off so from my perspective uh i made one in gray as well and you could kind of imagine it actually kind of looks like you know a little bit like a great go you know you might want to go bet it yeah um this is so do you do you think that you think you'll end up redesigning it because if you want to just show that uh at least like the head uh because if you want to just show that it's dangerous you could just have it uh you know maybe a speaker or a light that says like you know stay precaution taper you know a lot of way to do it besides besides the horns even though clearly clearly it sounds like you just wanted to make this sounds like you live in the woods sounds like you want a few of these patrolling around your house.

2:17:57Yeah, I mean, I think the other thing about humanoid robots, like we all kind of know that they're not purely like a perfect productivity maximizer device, right? Like they have to be emotive and cool and we've wanted them for 80 years or something and been dreaming of this stuff. And I'd say kind of like a car, like no one buys a car because it's the perfect like mobility blob with no personality that, you know, is exactly as safe as you want yeah maybe a few people do but you know humanoid robots like cars the ones that people really like you have to have their own personality a little bit so i think trying to redesign it to be like uh okay maybe the optimal safety would be to have a giant yellow siren on the top um but yeah it's just not as exciting so we've got to keep it interesting right yeah i get it and i'm sure there's i'm sure there's people out there that want to cut down some trees that genuinely would prefer this form factor than uh something more tame talk about uh for the types of environments that you're imagining uh the robot in why four legs is better than one i can imagine just wait one leg sorry two two one leg is really tough one leg but imagine the pirate robot is next yeah no four legs versus two legs i imagine if you want to be able to actually um manipulate like something like a saw you want to be able to stay able having the stability but also the terrain yeah you don't want to be on wheels and i think there's been enough videos now of the robot dogs kind of running around crazier terrain that show that But yeah, four legs is superior than two to two.

2:19:41I'm all in on centaurs. Let's hear it. All right. Yeah. I mean, bipeds are tough. You have to move the center of gravity up for the whole bot, right? Because you're putting the pack up in the human chest or I have it in the horse body. You're getting actuators closer to the ground because their ankles are actuated. So if you think about walking around in the forest, you're putting electronics really close to the ground. And, you know, if you have a linear actuator, then a moving shaft like right on the ground. Whereas this thing, you know, the closest actuator is nearly three feet off the ground.

2:20:14The way I'm driving the knee, it's easier. You know, I'm not a this is not a huge project. I think another kind of misconception from the viral explosions for some huge stealth startup that's popping up. And these things are going to go knocking door to door. but uh no this is more of a passion project so doing a biped today is is really tough um you're not going to see all but you know the best people doing uh and even then i'll point out there's of course a lot of curation for what you see from anyone i think everyone kind of knows that uh they're pretty tough to to get stable and i think another question is why not do tracks or wheels that's pretty tough in robotics too because if you're trying to keep all four wheels on the ground like not like these uh the ones you see out of china that are wheels and a dog but just wheels you know then you're gonna have to have some suspension uh which is kind of the enemy of robotics is having these unknown spring forces and stuff like that and tracks are just really heavy and really actually not all that stable for the size that they are if you think about a little tracked square driving around the forest if you're on hard pack ground it's actually could still be a little tippy so uh yeah four legs works pretty well yeah have you gotten any death threats since the viral moment he's got a robot army at his disposal i imagine some people might might think now is the right moment before you have you know a hundred of these on your property yeah um i guess though the few people that were offended by the designer are christian so that you know peaceful people so uh you know unfortunately i've had you know people uh say i should stop or uh you know that it's uh someone sent me an email today that said my company is now owned by god uh who will save me and stuff like that but no nothing too crazy um the more surprising reach outs have been police departments.

2:22:18That is what I wouldn't have thought would have happened. To partner with you or arrest you? Yeah, to ask if it could be used for public safety. Okay, yeah, yeah. If you've ever been at an event where a bunch of police cavalry have been and then there's just like people just dissipate. There's something about being around a large horse where it just doesn't I was going to say the theses that gets uh i hadn't even thought about that yeah i i just think for for riot control the the horse is just like it won't quite it's not like you're gonna get hit by a car but there's a natural human reaction to sort of just moving out of the way the chat wants to know about the pogo humanoid uh there is one pogo stick there is one there is yeah there's at least one um look yeah if you search that is potentially even a scarier format where the robot can jump 30 feet in the air and And it's just logo stick.

2:23:17It's a psycho clown on it. Talk to me about controlling this thing. I imagine you're not doing full autonomous control, heavy teleoperation. But what am I because you have multiple cameras. So are you looking at a screen with multiple camera feeds and then controlling with an Xbox controller? What is the process? and then I imagine that as you go forward towards controlling a chainsaw, that's a little bit harder to control with a Xbox controller. So will you have gloves, VR? How are you thinking about that? I got a solution for that. Okay. Let's see it. Let's see if we could see. No one's seen this chainsaw, so I guess you guys could do it on your show.

2:24:01But there we go. So this chainsaw is to a degree of freedom, kind of like a – let's see if I can line it up a little better. you can kind of see what's going on yeah yeah i gotta go this way my thing's mirrored yeah um but yeah it's something i thought about so uh yes it's think about like this like you have you can split robotics into moving around and doing something with the end effector and the good thing about a quadruped is you know you could kind of park it and then the hips have enough mobility where you could kind of control it almost like it's an excavator if you've ever been in one of those so it's a little bit more intuitive to do than trying to do something like dynamically like you see uh you know people with vr goggles and gloves to control a humanoid that has hands or a biped and all that that's again like way outside what i'm able to accomplish so you can think about it more like you're driving a big car that keeps his hips square to the ground And then the top half kind of acts more like it's an excavator where you're kind of controlling it on two plane.

2:25:11Do you think there's more of a consumer market for robotic horses? Because a lot of people out there that love horses, but horses are quite expensive to maintain. And I imagine if you made a robotic horse that you could control with an Xbox controller and sit on it, there would be some some consumer market for it like a riding one there was a demo that came out of japan i think maybe we shouldn't be surprised that they're on the cutting edge of making that sort of thing uh but yeah i don't know would you buy one i think i would i think i would potentially get one for uh for my daughter yeah oh not to like commute in uh the the robo horse commuter galloping to work if you could get it up to like 40 miles an hour would be would be quite appealing to me but um maybe maybe more just for fun use case early on so something to consider you need to make the head a lot a lot prettier though i think like an angler fish head you're thinking i'm thinking more like a normal horse oh okay i thought about that oh well thank you so much for coming on the show and breaking it down for us uh good luck with wherever this project goes we'd love to stay yeah keep us updated as things keep going i will say when when john showed me the website i said this is clearly somebody just messing around that vibe coded a site i love that it's a real thing but it's amazing to see uh yeah yeah it's a it's gonna be amazing to see where this goes congratulations and thank you so much we'll talk to you soon Have a good one.

2:26:49Cheers, pal. Up next, we have Ron Irel from Intology. He's the co-founder. And we're talking about RSI, recursive self-improvement. Is it here? It might be. We'll get his take. Ron, how are you doing? I'm doing well. Thanks for having me on, guys. Thanks for hopping on. First time on the show, why don't you introduce yourself and the company a little bit? Yeah, of course. So my name is Ron. I'm the co-founder of Intology. At Intology, our mission is to automate scientific discovery. and we're starting with AI R &D. So today we have some pretty exciting results to talk about regarding automated post-tranning of other language models and lots more to discuss, so glad to be on.

2:27:27Yeah, how would you characterize the progress, the announcement? Because people can get sort of lost in benchmarks. Are you 20 % on this thing, 99 % on that thing? Qualitatively, where do you see the technology today? Yeah, absolutely. So fundamentally, we believe that automating discovery is a domain agnostic problem. What that means is that the structure of discovery problems is pretty similar across domains in the sense of, no matter what problem you're looking at, like drug discovery or materials discovery, even like improving language models, there's always going to be some sort of process of proposing an experiment, getting feedback from that experiment, learning from that experiment, and then using that information to propose the next set and continuing until you make the discovery.

2:28:11So it's a great question because we think about the problem on that access of how do we build and scale systems to these problems in which the evaluations and experiments become more expensive, more difficult, harder to access. And post-training, for example, represents a problem space where experiments are pretty expensive. They can be quite noisy and they take a long time. So if you're trying to build an automated research system that develops the next state-of-the-art 70 trillion parameter model or whatever, you can't really imagine a system training 15 ,000 models until it discovers the best one because training is expensive.

2:28:52So you have to think about how do you run efficient experiments? How do you gain information from those experiments better? And I think that today we're showing kind of a closer step in that direction because in the past we were doing things like kernel optimization, like MLE bench style problems, and now post-training where obviously it's more expensive and takes longer. On the cost side, this does sound expensive. If you want to hammer a bunch of different experiments, what has been your approach? Just sort of suck it up and use venture capital dollars or partner with companies and labs that have big compute allocations, they have the resources, have them sort of front the cost?

2:29:32Or is there another way to solve it? Because scale seems very important here, and yet your whole job feels like burning compute? Yeah, so definitely a little bit of both. At the beginning, it was a lot of burning venture capital money. And now we have partners and systems in place to run experiments that, you know, don't just burn money for no reason. You know, we have our own cluster. We have our own infrastructure that efficiently utilizes those resources. So we're at a place where we're not just like throwing money at the wall for no reason. So we're, I mean, obviously it's an unsolved problem.

2:30:03We will always continue to work on making our system better at this. But I would say it's a mixture of having some great compute partners, you know, spending a lot of time on our infrastructure, even before we start running experiments. And now we're at a place where, you know, I feel comfortable throwing hundreds of thousands of dollars at the wall if we feel like the system can actually make progress. Gary Marcus, he says that it doesn't count if it's not a pure LLM, that AI is only making progress in verifiable domains. You need a lean output that's fully verifiable. How optimistic are you?

2:30:35And I'm somewhat sympathetic to it because it does actually seem like we are moving much faster in verifiable domains like math than unverifiable domains like, I don't know, coming up with a script for the next great movie or joke writing, a comedy writing, or even some of the bio stuff that's maybe verifiable but over a multi-year process of going through FDA applications and testing in vitro, testing in mice and monkeys. and humans. There's just things that the verification loop isn't just run some lean really quickly and see if it checks out, right? So what do you see the future of transferring all the amazing learnings and ability for AI to make discoveries in ML and computer science and math to anything that's a little bit less verifiable?

2:31:29Yeah, for sure. So I think that it really just comes down to, I mean, so we focus on verifiable domains, but I think it really just comes down to, you know, how much you can actually query the evaluator, right? So back to that example of, you know, post-training the next large state-of-the-art model. You could argue that, you know, that system could, in theory, train the whole model, get the ground truth feedback, see how it did, but that's not fully realistic for every run, right? So it would have to do some sort of experimentation with either its own rewards or, you know, not full ground truth rewards before making that progress.

2:32:03I think it's my take is I really think it's just about building systems that almost like wean off the requirement of the evaluator. So for example, you know, if you're building these kinds of systems for training small models, no problem, you train an infinite amount of models and you find the best one. But yeah, like it's almost like as you scale on this axis of difficulty, it's almost like as a forcing function of compute availability or cost, you start having to think like, okay, well, now I can't actually have my system train the whole model every time, or I can't run an entire clinical trial every time I want to test a drug.

2:32:35It's almost like a nature of the research direction. And so I guess that I, I don't really think it's that different of a problem. I think it's more that if we continue on this path, like for example, with our system, if we continue on this path where it doesn't need to, you know, query the full evaluate every time, eventually it'll get to the point where it might not even need the evaluator or it'll only need it at the end when sending a drug to clinical trial or, or, you know, a material and a fabricator. So I think it's really just, I think, I guess we, and collectively the AI for science community, I think we're heading in the right direction.

2:33:05I don't think I view it as like black and white as we're only doing verifiable and then we're going to go to unfair viable domains and like figure it out. Interesting. I have one more question. Jordy, do you have anything? Yeah. I was just going to ask, what do you, what do you think your business looks like two years from now? I won't say five or 10, but like, where is this work going? Yeah, for sure. So, you know, we, I guess we really believe in not building co-pilots. You know, what we are really trying to do here is build fully autonomous systems that are deployed in R &D environments and just run the entire loop, you know, autonomously, perpetually.

2:33:40Obviously, we're quite far away from that right now. You know, we have to deploy the system, we have to monitor it. We have to see it work, make sure it's succeeding. But I think at the end of the day, definitely two years from now, we want to be at the point where our system can basically be deployed as infrastructure in any computational R &D problem. And then it gets access to the data on the problem. It gets access to the ability to run experiments on the problem. And then it gets deployed in an environment in which it's hard to hack and you're getting good signal out of the experiments. And then boom, it runs autonomously.

2:34:09It's cranking out discoveries. It's shipping them. And humans can be there to take a look and make sure it's not messing up. But eventually we want it to be running end to end. Last question. Tell us a little bit about the company. Where are you based? How big is the team? Who are you hiring? What you were doing before this? Yeah, for sure. So we're based in San Francisco. We just moved into our new office. That's why my background is pretty boring right now. We're growing pretty quickly. I've been really proud of the team we've been putting together. I mean, we have researchers coming from DeepMind, Anthropic, Factory AI, both in, I guess, industry and in academia.

2:34:42I think this is like a really fundamental problem that needs to be worked on. And it's not just a research problem. It's not just an infrastructure problem. It's the whole stack. And we've been putting a pretty incredible team together to do so. And I guess before this, my co-founder and I ran a pretty large nonprofit research group. We were primarily funded by the National Science Foundation, obviously very different from running a company nowadays. But we did a lot of fundamental research in coding language capabilities, published some of the first work in test time scaling with language models.

2:35:11And it kind of felt like a natural progression because, you know, we were really curious about how do we model agentic behavior in this kind of search process. And it kind of just made sense that, you know, this is the time, this is the place let's make it happen. And that's what we're doing in ontology. Yeah. Well, thank you so much for coming on the show. I'm sure you'll be back on soon. Have a great rest. Very cool update. Great to meet you. Goodbye. Let's go to this Paul Graham post. he had such a wild experience as he bought a book it was awful didn't want it on my shelves but i couldn't throw it away so it sat on a table near the door i know someone that will rip it apart feed it to a machine and burn it i know someone i don't know them personally but i know they they would love they would love to to take this off your hands rushing to an appointment this morning.

2:36:07I grabbed it to read it. First mistake. Then went to breakfast and had nothing else. So I spent the morning reading the worst book I have known. It's such a funny, such a funny, like, just like, I don't know. It feels like a Curb Your Enthusiasm episode or something like that. Anyway, thank you so much for tuning in to TBPN today. We'll see you tomorrow at 11 a.m. Pacific. We love you. Leave us five hours on Apple Podcasts and Spotify. It's been an honor. Sign up for our newsletter. or tbca.com. Goodbye.