Live from Washington, DC: Will Manidis, Analysis of Intel Earnings, Analysis of TSMC Earnings, Amazon Denies Plans to Display Tariff Impact at Checkout

26:51And then he's like, we're going to do all this stuff, which adds up to$100 billion. And then he says, we've already invested$65, so let's add the new$100 to the previous$65. We're at$165. Which sounds staggering, I guess, when you compare it to what the hyperscalers are spending on CapEx, which is all in like the$60,$70,$80 billion per hyperscaler. It's not that crazy. And of course, this is over a number of years because they're building this out. But still very, very bullish for American manufacturing. very bullish for re-industrialization, which we love to see. It's not a meme. Not a meme, it's gonna happen.

27:25It is, yeah, it is interesting to see that, like they're actually making progress here that people thought was not possible for a long time, or they thought it would never happen for a variety of reasons, but TSMC, Foxconn's in the same boat where Foxconn's announcing new facilities. It's a Taiwanese company. They obviously have a massive presence in China, but they're announcing a lot of new facilities in America because they see themselves as a Western company or a global company, and they want to do business in America. I love when CEOs are super candid on earnings calls because it's so easy for them to go robotic.

27:57This was another quote. He was responding to a question. He said, the first purpose in the Arizona fab is that the Arizona fab can operate independently, but of course we have done, and we are doing it right now, some kind of pathfinding, exploratory work, and to cooperate with the university, blah, blah. It's actually a lot of activities. 1 ,000 engineers is not a small amount. Of course, it's not comparable to TSMC. I think he's talking about the facility in Taiwan that has 10 ,000 R &D people. But it's a beginning, okay? So we'll do a lot more. That's great. And anyways. Yeah, I mean, you were talking about the CapEx as marketing to the admin.

28:36And probably the best example is from the first Trump administration where Apple did that ribbon-cutting ceremony. I think it was in Texas for the Mac. And it was like, they'd already been making it there. They weren't reshoring anything really. You were just rebranding the dollar. So they're just already built. Yeah. And then earlier in Trump too, just a couple months ago, Apple kind of repackaged a bunch of their spend. And they were like, we're investing$100 billion in America. And it was like the salaries that were already going to be spent here. We should just build a warehouse and just let people rent it for ribbon cutting ceremonies.

29:09It's like huge bull market. Yeah, massive American flag on the wall. You can move in. we rent it by the hour it comes with a vibreel it comes with free birth pre-installed secret service has been there already getting into American made ribbons specifically for and those big scissors the big scissors have to be American made as well those are made in China today yeah absolutely should we talk about Amazon and all this drama around whether or not they were displaying a plan to display This guy ran this AB test on his e-commerce website. He found he's so he makes his shower head in China. It's the tariffs are going to make a$100 product like$250 goes to an American manufacturer.

29:55It's going to cost him$250 to make it. So he puts an option on his website,$100 for the Chinese one or$250. And he was like 100 % of people chose the cheaper one. Zero. We got zero made in America. But I was thinking about that more. And I was thinking about how, like, no, that's not the option. Yeah, that's not the trade. Yeah, the actual trade is$250 for the Chinese-made one,$160 of which, or$150 of which goes to tariffs, versus$250 made in America. No, it was a deeply flawed study. Yes, yes, yes. Exceptionally viral. Yeah, yeah, yeah. But anyway, Amazon is thinking about doing something very similar, kind of, you know, lobbying popular support amongst the American people.

30:38breaking out import charges. And to be clear, they deny they had plans to do this. Sure, sure, sure. But the administration addressed it and said that breaking out import charges would be a hostile and political act. Andy Jassy is getting in trouble with the White House. I'm sure they will be fighting it out, although they're still going after Bezos. So Amazon.com was forced to play down a report that it was considering displaying the impact of tariffs during its online checkout process after President Trump called the company founder, Jeff Bezos, and the White House said such a move would be a hostile and political life.

31:12I mean, you have to imagine that as soon as the tariffs drop, there's some Amazon PM who's implementing this on day one, right? Yeah. It's also funny that Bezos is still the first guy to get the call. Of course. It's been so long now. He's desperately trying to retire. Yeah. I just want to do the Lex Friedman podcast and be on a yacht and go to space. Why are you calling me about my e-commerce business? Yeah. Yeah. I was an e-com entrepreneur. Yeah, I was a 12-figure e-com entrepreneur. But if he does save the global economy, that's like his beautiful act, right? Like if this works. Who? If Basis saves the economy by rolling back tariffs and implementing this feature, that's a Nobel Peace Prize level achievement.

31:52Yeah, it's possible. It's going to take an act of courage, standing up to the admin, by putting a new div on the checkout HTML page. Pressing tab and cursor. There was some controversy because Punchbowl was doing some reporting on this, but Punchbowl is corporate media. Really? They're supported by Amazon. Wait, what are you talking about? Is this real? Yeah, yeah, yeah. Punchbowl News, okay. Yeah, it says, Trump called Bezos to raise concerns after Punchbowl News, which I've never heard of, reported that Amazon was planning to display the impact of tariffs during its online checkout process, according to people familiar with the matter.

32:29And now, I mean, Amazon, to be fair, does break out a ton of expenses. Like you see the shipping costs, you see the taxes, you see how the receipt is calculated. And now you're gonna see 600 % tariffs on your shower hat. Yeah, on your, did you know that Amazon launched a Timu competitor? I did not know of. Have you seen this? So in order to participate in their Timu competitor, you had to guarantee price ceilings for things. And you could not sell a guitar for more than$12. And you could not sell a couch for more than$20. I mean, we have to start thinking about an American Timu Reserve, right?

33:04Just importing thousands of Xi 'an items for Americans. It could very well stretch into the billions of items. The billions of items. Where are yoga pants going to come from if not Xi 'an? If they poison you. Yeah, the cancer. Put your lining. There's a post here from somebody who writes for his own. He has at big technology on X. Great name. He says the punch bowl story on Amazon sponsored by Amazon isn't completely true per Amazon Literally an article that's presented by Amazon and it says And it says Amazon to display tariff costs for consumers. So a little bit hard hard to imagine they weren't Actually planning to do this is gonna be like when we report on numeral and numeral has to deny our reports that Numeral has a fact achieved AGI in sales tax in the Chrome tab.

34:00Yeah. Because we often call it AGI for sales tax or AGI God. We were talking to Sam, the CEO, and what he said was basically like, we have a new version that we're actually not able to release because it's so powerful that it might solve sales tax fully and then sort of move on. Yeah. It's like, how do we actually reduce the number of sales tax receipts to zero? well we eliminate human we kill yeah yeah it's the same thing with ramp how do we get no more receipts no more employees no more employees get rid of everyone but anyway go to numeral hq put your sales tax on autopilot benchmark size a month per month on sales tax compliance anyway Bargain sellers Timu and Sheen also benefited from the exemption, which helped them keep prices down on clothing and other goods shipped directly from China.

34:58So that loophole is closed. Yeah, it was Amazon's haul. You're going to love this. The max you could sell a couch for. Yeah,$20. Again, I think we'll see this in cancer rates. I think once you take this cheap polyester and you stop putting it on your skin, cancer rates might fall a little bit. That's much to think about. Interesting. Is that the ultimate right-size in the government? because there's less cancer. It's proof the Maha movement's totally in control, right? The tariffs get shamed. It's all part of the Maha. So is Maha the real one that's pulling the strings here? It's Bobby Kennedy at the top.

35:29Bobby Kennedy at the top. Every other, the defense budget, it's all in the defense. It's all Bobby. It's all Bobby. Yeah, that makes sense. The tech giant has cost support agreements in place with some vendors to ensure that it makes consistent margins on the products it buys from them and sells to customers, according to consultants. Some vendors that have tried to charge Amazon more Or after tariffs were imposed, have been asked to pay the online shopping giant to make up the increase. Everyone's fighting over who pays the tariffs. Yeah, I mean, Amazon's in a really tough place. They had been squeezing their sellers for so long on both sides.

36:04Fees, advertising, not dealing with knockoffs very well, sort of purely focused on the consumer,

36:16which is, again, Amazon's job is to deliver for their core customer, which is the prime subscriber. But at the end of the day, this is only going to create more drama and chaos between Amazon and its sellers who already were not super happy. Yeah. I mean, the other story, Amazon's in the news because they want to be a satellite internet powerhouse. Who doesn't? I mean, we've had long conversations off air about how do we turn TBPN into sort of, you know, get a dish on every home. Yeah, yeah. How do we get a TBPN dish on every home? Yeah, we need a dedicated network for sure. So they launched 27 Coopier satellites, but they need to get thousands more to space to rival SpaceX's Starlink.

37:00Some of the company's biggest hurdles aren't in orbit, but on the ground, 27 Amazon satellites were launched from the Florida coast to orbit Monday night. a mission handled by rocket launch company United Launch Alliance. A year and a half ago, ULA sent two prototype satellites for Amazon to space. How have I never heard about United Launch? You don't know ULA? They're like the second biggest private launch provider. Space X is so much bigger. It's a joint venture between Lockheed and Boeing. Yeah, okay, okay. I've heard ULA, but I read it out loud. But ULA has been, you know, they've had a lot of problems with a lot of different projects they've been worked on, and Elon's obviously coming for them very aggressively.

37:36To compete with established providers, Amazon needs to rapidly manufacture and deploy thousands more satellites on a narrow timeline. It also faces a set of challenges familiar to many companies pushing to scale up manufacturing rates. The Project Coupier venture aims to build large-scale satellites. Yeah, how do you pronounce that? Coupier? Coupier? I think it's French. Yeah, okay. I have no idea. Thanks for the misinformation. So they're trying to provide high-school internet. Our entire audience is going out saying, yeah, it's actually Coupier. The company plans on having 3 ,200 satellites in orbit over the next few years, building a business it hopes will eventually provide meaningful operating income, Chief Executive Andy Jassy wrote in Amazon's recent annual letter to shareholders.

38:21But Elon Musk has demonstrated the power of such a fleet through his Starlink business, which has more than 7 ,000 satellites in orbit already and more than 5 million users. at least two similar Chinese systems are also in the works while Canadian's satellite operator Telesat is developing a new fleet called Lightspeed everyone's got to have their own Leo cluster of satellites I guess leaders at Coupier interesting to think about Amazon potentially over time just bundling in satellite to prime like it's not yeah you have TV of shopping of groceries yeah deliver own the actual delivery and yeah I mean it seems like It was really, really hard to actually own cable or wires in the ground.

39:01That was particularly hard and monopolized to actually grab any infrastructure. But if everything moves to satellite, you can actually throw up a constellation. There's no zoning laws in space. Yeah. At some point, I imagine that it gets too crowded that you can't put up any more constellations. But it seems like we're pretty far away from here. People are already complaining about the space junk, though. That feels far away. I mean, the moat is that you have to spend however many tens of billions of dollars. And launch is not a zero margin business for SpaceX. But it is when they launch Starlink satellites.

39:38And any time that they have a free seat on the bus, they can just throw Starlink satellites on there. So they're getting them up there way, way cheaper than Amazon is. Which is probably why Bezos is also investing in Blue Origin. because if he can get that working and he can dump Coopier satellites up there cheaply, that actually gives him the opportunity to challenge Musk. It is funny how much Bezos and Musk go after each other. Did you see that Bezos, I mean, he already funded Rivian, but then he just funded that Slate truck. Yes, it's an incredible truck. Okay, yeah. What was your reaction to the truck?

40:11Incredible. TJ Parker, you know, big endorsement from a certified car guy. I mean, it's incredible to have a small American-made truck. That said, I hate that it's an electric vehicle. I'd buy one day one if that was a combustible engine. I just feel like, you know, Europe doesn't have power right now. Why are you buying an electric EV? Sure. But it's a beautiful car. It's an SUV. It's a sedan. It's a truck. Have you thought about buying one and swapping in like a turbocharged vehicle? Yeah, I was going to LS swap. Yeah. The slave truck. The slave truck, for sure. Yeah. Yeah. A kind of mini Raptor.

40:39That has been the number one. I've seen some people, you know, I think it's going to be a hit almost as a golf cart. Totally. Yeah. I mean, the question is, who is the singer Porsche of Japanese small trucks? Like, if I want to drop$2 million on a truck this big, who's going to provide that? Yeah, that's the real opportunity. But yeah, I mean, the big critique there, and we're going to get the Slate founding team on, but on the show, TJ actually made the intro. Shout out to TJ. But yeah, the big critique has been around range. I'm sure they have a lot of plans there. I assume you have to look at with any EV manufacturer and assume that the range is only going to go up over time.

41:17So it's fine. Do they drop a range number? It's like 150. Oh, 150. That's really low. I mean, it's a golf cart. It's a golf cart, yeah. But it's even lower than that when you put stuff in. Yeah. Passengers, you have hills. But truly, if you're just going around town and back, you're not going to do more than 150 miles. Even if you drive to work 30 miles away and back, you're fine. I mean, you can't even fit in it, right? You have to, like, cut a hole in that. I couldn't fit in it for sure. Well, don't they have, like, a modular structure so I can get, like, a convertible version and just be on the top?

41:47Yeah, yeah, yeah. You're not going to pass the broomstick. But the reason I brought up Slate was because I think Bezos backed Slate. He also backed Rivian. So he has bets, two different bets in— Two different trucks, too. Two different trucks in EV going up against Tesla. There was a conspiracy theory that Slate was actually a Tesla project because it's an anagram. Have you heard of this? I have heard this, yeah. That's a great one. And then they're competing in space. They're competing in social networking now, too, because Bezos owns Twitch, which is a very large social network, streaming. X obviously has streaming, as we're streaming on X right now.

42:22And then, yeah, they've just been duking it out in all these little sub-market subcategories. Yeah, it's a lot of fun to watch. Anyway, we should move on to some timeline. Let's break down the best posts of the day. This is the moment that you all have been waiting for, timeline with none other than King of the timeline. Let's do it. Let's dive in. So founder of Intercom says this is truly the golden age of B2B software out of home advertising. Literally 98 % of billboard and shelter stock in SF is currently taken by SaaS brands. It's never been like this ever. It's either really bullish or really bearish for SaaS.

43:05is it bullish or bearish? Intercom by the way, Adquik, customer. Oh really? Yeah. Fantastic. Yeah, we are sponsored by Adquik. Go to adquik.com. Sorry to cut you off. Adquik made easy and measurable. Say goodbye to the headaches of ad-of-home advertising. Only Adquik combines technology, ad-of-home expertise, and data to enable efficient seamless ad buying across the globe. Anyway, Will, bullish or bearish for SaaS? First. For Adquik. For Adquik, yes. I'm here for Adquick. The audience was. Yeah, maybe we get the Ashton Hall sound for every new post that we do during this timeline to really fire us up, really get us firing at all cylinders.

43:44Okay, so bullish or bearish? I just think hope, hope, hopes, the, you know, replier number one, also my favorite account on twitter.com is correct. Okay, hopes, revenge. We do need to think bigger, right? We should be renaming San Francisco. One wonders if instead of San Francisco, we could have Salesforce California. Salesforce California. That has a nice ring to it. Or we could have OpenAI California. Yeah, that'd be great. The towers there are already going to have an occupied state. Why not go all the way? What about just Altmanville? That would be even better. We've been talking about Waltaville down in Nashville.

44:13If Salesforce really wants to have that tower, company towns are back. I would like to see an ad unit hung from the Golden Gate Bridge, you know, just sort of like 100 feet by like… Everyone always goes crazy over those videos of Chinese cities where they look so cyberpunk and they have advertising everywhere. Blade Runner Future, let's do it. Also, why do we need permission? Why don't we just project on the side of the Salesforce Tower? There's a business there. Yeah, yeah, yeah. Also, drone shows, blimps. We should be doing much more advertising. AdQuick? Not to go back to AdQuick, but they do have a drone.

44:46No, no, no. Please go back to AdQuick. Yeah, please. They have a drone, access to drone show ads, which is pretty cool. If you want to run a drone show. Oh, wow. That's good.

44:59All right, next post. Ian Cinnamon, Apex Space has closed another round of funding to scale production and meet demand for satellite platforms for critical projects like Golden Dome. The new$200 million in Series C funding is led by Point72 Ventures and co-led by 8BC alongside Andreessen Horowitz and a number of other funds. So massive milestone. Ian says, this Series C investment is a massive vote of confidence from our backers as demand for our bus platforms has exponentially increased as we as are also seeing the underlying need for space-based capabilities growing rapidly with this capital we are excited to double down on supporting the needs of the u.s and allied warfighter as well as critical commercial applications congratulations ian absolutely the bus do you remember that created a space bus i mean the speed at which uh ian and the apex team was moving we had to go back and listen to our interview with Ian probably about a month ago at this point.

46:02But what an absolutely ambitious timeline they've been on. It is cool how all of the different pieces of the satellite stack are disaggregating and now there's new companies in each of them. I wonder if eventually they'll kind of come together in some massive karetsu or something. Have you talked to Ian before? No, but I do like the idea of kind of a Frenkel-esque roll-up of all these kind of non-commercially viable or soon to be commercially viable launch providers. It's beautiful. Well, it's not launch providers. He doesn't do rockets. He does the bus that the satellite rides on. So you want to put a radio or camera in space.

46:37TBPN network. TBPN network will be using APEX buses to build on top of. And another example would be like Bridget Mendler. Her company does the ground link. So how is TBPN going to beam down from these satellites? Well, she will have a ground link network of satellites that can receive the message from space. This next, we got another post from Trey Stevens. So we have some news from Lockheed Martin. They said, Lockheed Martin says, without us, AI might have never taken off. For over 40 years, Lockheed Martin has used the power of AI and machine learning to help drive the future of defense, aerospace, and national security.

47:23And it's just a gibbified version of someone that kind of... Tom Cruise from Top 10. I was going to say, this looks like Trey Stevens to me, but, you know... Not enough beard. Not enough body hair. But, I mean, 40 years ago, I feel like they're not using that word correctly. That's 1980s. 1985? Like what AI were they doing? That's like Apple II. Like neural networks had not really been invented yet. Yeah, it's a linear regression. Yeah, yeah, yeah. It's a TI-84. Yeah, exactly. They're basically saying, when was the first time we mentioned the words artificial and intelligence? Not even together, but it could have been in the same paragraph.

48:02I mean, at the same time, the lesson one of Andrew Ng's course on machine learning at Stanford starts with multiple regression. Does it? Yeah. That is the foundation of, and then he builds you up to neural networks. I mean, I hear there was a big secret in how Grok got to state of the art so quickly. Linear regression. Doing the regressions by hand. Yes, by hand. Training the transformers with artisanal craft. Yeah, we need to bring that back. I actually wouldn't be surprised if Lockheed did have some AI person that was foundational at some point. Like some Django block in the massive stack goes back to some Lockheed paper, some skunkworks project.

48:39but it is hilarious that they're just like, yo, we're responsible for AI, you're welcome. Get it. Oh, this is huge. We got some huge news from Josh Kushner. Let's go. Josh Kushner says, we are humbled to announce, we are absolutely humbled right now. We're so humbled to announce Thrive Holdings, a permanent capital vehicle dedicated to investing in, acquiring and operating businesses for the long term. The mission of Thrive Holdings is to transform enterprises throughout the strategic application technology you know the only thing better than capital what permanent capital permanent you know you know what they say any man in possession of a good fortune only needs one thing a permanent capital holding vehicle yeah that's right that's right can you actually break down what a permanent capital holding vehicle means relative to like everyone's familiar with like a vc fund 10-year 12-year continuation vehicles what's permanent What matters is that closeness to God is closeness to fee streams.

49:39And normally fee streams, they stop at 10 years. You know what's better than 10 years of fees? A lifetime of fees. Fees that you can pass on to your grandchildren. But how is this different than... I mean, it's incredible, right? But is it structured different than a private equity fund that has a specific life or even like a hedge fund? Yeah, I mean, the standard permanent capital vehicle will have some degree of redemption schedule, right? Like LPs can pull out some degree of profit at certain points. and they can swap with other people in and stuff. Brent Beshore has some incredible essays that detail this.

50:08Is he truly the golden standard for permanent capital? Or is there like a public company? Like does BlackRock have a permanent holding vehicle that we don't talk about? I mean, there's plenty. Like the closed-end vehicles were like a huge thing in the UK. In the 90s, there's dozens of these. Brent Beshore, I think, is the most artisanal. Sure, sure, sure. His is printing cash at a rate no one else, I think, is doing right now. But of course, there's permanent capital vehicles we cannot name rolling up HOAs in the Midwest. right? Yeah, and I think the opportunity here is for Josh and the Thrive team to think even more long-term than they already do, right?

50:41Like, how do we think about, you know, not in this sort of, like, 10-year cycle, which Josh, I think, had posted something a long time ago, which is that, like, managers think in quarters, like, founders think in decades, and there's huge advantages in alpha and just being able to think and plan on a longer timeline than everyone else. And especially if you're thriving and you're thinking about buying, like I think it's been widely reported they've bought an accounting firm to put AI in the middle and do automation on top of. That's a deal you need to hold for 40 years. And then this also stops the continuity vehicle issue, right?

51:16You have lots of companies like Stripe, you're holding year 10, 11, 12. How are you going to get liquidity on that? Well, you can pass it over to your permanent capital vehicle. You think that's possible? I mean, it's similar to Sequoia Heritage, right? It's like a continuity vehicle where you can pass those things over, get liquidity for the LP's, the initial set, and hold a basket of really high quality assets. So you want to hold for duration. Okay, so take me through some of the sizes of permanent capital vehicles. Because with a VC fund, you know, you hear people raising a$50 million solo GB fund all the way up to a$5 billion growth fund.

51:47Permanent capital feels like more narrowly defined. but is there kind of like, what is like a sweet spot base hit for the size of a permanent capital vehicle? I mean, yeah, Harkshire is like the best example. Oh, okay, okay. I wasn't actually sure, I didn't want to botch it, so I just looked it up. But that's a good way to think about it of, hey, we're not gonna focus as much on raising new, you know, the best scenario is like you have cash flowing assets that produce capital that allow you to reinvest or buy other businesses to produce more cash flow. So it's like sort of the holy grail. And many people have said that cure to mail loneliness is a permanent capital vehicle.

52:30It would fix all you need. It would fix me. What about the mindset? What it takes to be great in venture capital is often high risk, not making money for a very long time. All the capital, all the value capture happens at the very end. Obviously, the types of companies that are typically rolled into permanent capital vehicles, cash flying from day one. I mean, I think like the single greatest Josh Kushner quote is like, you never lose money buying Fifth Avenue real estate. If you think about Thrive as a thematic bet on the Fifth Avenue real estate of private companies, probably not a bad idea to hold these really blue chip assets to duration.

53:05So it's more just about find great founders. If they're losing money in the super high growth phase, they go in the growth fund, they go in the venture fund. If they're cash flowing, but they're still fantastic entrepreneurs, fantastic assets, they go in the permanent capital vehicle. I mean, if you're Stripe and you're planning to go public in 2035, 2300. 2300. Once they've consumed 100 % of GDP. Yeah, yeah, yeah. Processed through Stripe. Global GDP, actually. Global GDP, yeah. They need the space narrative. Look, the space economy is going rapidly. We need to go off planet. Yes, off planet.

53:38It's just these continuity vehicles are going to get huge. Yeah, wow. Okay, well, speaking of growth stuff, Hadley Harris, a venture capitalist, actually wrote about vector databases, which I think you might be a little bit familiar with. Yeah, I mean I was the first check into Chromo. I still think it's the fastest growing open source vector database. Yeah, I love Jeff. We had him on the show. Jeff's amazing. He's great. So he's breaking down some hype and I'd be interested to hear how the market has evolved and if Jeff has kind of altered his strategy. Hadley's breaking down the story of Pinecone.

54:11Two years ago today Andreessen announced a hundred million dollars into Pinecone as everyone and their mother tripped over each other to invest in vector databases, given and how little standalone vector databases are valued now. The hype cycle is wild to look back on. The timeline is late 2022, ChatGPT launches, early RAG architectures emerge. There's first buzz around these companies. Then there's a massive surge in demand for vector search. Vector database is seen as critical infrastructure, but then after there's peak hype, every AI infrastructure diagram has a vector DB layer. Saturation hits, Postgres, Redis, MongoDB, add native vector search, clouds bundle it, the hype fades, vector search becomes commodity, market tightens, attention shifts to agents, fine-tuning orchestration.

54:55So is this a good telling of the history of the vector database hype cycle or is there something you need to put in the truth zone? I think it's accurate in the hype cycle. I put it in the truth zone in that this feels like kind of terminal VC brain. Sure. Like it is likely true that the vector database market will not have dozens of billion dollar outcomes. Sure, sure, sure. But if you look at even the traditional relational database market, There are hundreds of massive companies there, right? And if you think about a huge amount of enterprise software, it's just different views on top of structured data elements.

55:25Sure, sure. Like Salesforce is the world's greatest structured database, right? There will be big winners in the vector database space in that you just need to be able to store unstructured data. And like, yes, you know, you can search vectors using numpy.sort, but like that's not an enterprise-grade database solution. And I just think this is a good excuse to, you know, have missed a lot of winning companies such as Chroma. Yep. A lot of people are taking victory laps. John Chu at Coastless, as we told our LPs, that were staying out of the market because they were too trivial to build and commoditize and were never used without BM25.

55:59What is BM25? Do you know what that is? Never heard of that. VC VEC database thought leadership was a great filter for those who for who's really technical and who has shallow understanding of ML infra. We've got to have John on the show. He's always dropping bombs on the timeline. Well, speaking of dropping bombs on the timeline, hit me with a sound effect, Jordy, because we're moving over to a post by Deva about tech bros learning about convertibles. There are two different posts here, one from Spore, a friend of the show. It's called We Do a Little Bay Area Sightseeing, and someone else bought a BMW Z4.

56:40What's your take on convertibles? My first car in high school was a 2002 Honda S2000. That's a great car. In baby blue. Sick. I still own it to this day. That's amazing. It's the single greatest thing I've ever driven. Yeah. It's amazing that they've discovered convertibles. I mean, it's just like, it's 25 years late. It's kind of weird to discover this in your 30s. Especially in California. Especially like, also both of those look like automatics. Yeah, there's a generation of young people in tech that have only owned Teslas. Yeah, and they don't make a convertible. And they're discovering that car manufacturers have been making pretty good cars.

57:15Lots of manufacturers have been making pretty good cars in a bunch of different categories for a long time. And it's a beautiful moment. I think it should be celebrated. I agree. I also think what I want to see people do more, people I don't think realize if you have a car, let's say you have a Honda Accord, you can drive to any car dealership and within two hours have swapped it for a 911. right like it's actually like an extremely smooth process where they will take your car basically not sight unseen but they'll look at it maybe close to it that your car has some amount of value you can roll that into something else and you can within three hours on any day of the week you can be driving in a 911 or another car that that's um going to change your life Also, if you're not a permanent capital guy, can't afford the 911, Miatas.

58:09Miatas, yeah. $5 ,000. Miatas. Do you know the best-selling convertible in America? It has to be a Miata, right? It's not. It's the Jeep Wrangler. No way. Yeah, because a lot of people don't think of them as convertibles. They technically are. Yeah, drop top. But yeah, I think Tesla's missed out and Rivian's missed out on the convertible market. All of the car companies are just coalescing around, like, everyone wants a crossover SUV. that drives itself yeah yeah again the one one piece of feedback here guys yeah uh manual transmission yes yes like i will teach i will fly out i will teach you how to stick yeah that's going to change you yeah yeah well maybe maybe that's next year next year will be tech rose learned about manual transmission manual transmission yeah we're going to put you on the bottom of the same except except i my my first car uh was a very very old toyota corolla hatchback Was it manual?

59:01That was manual. And I would drive it in San Francisco and it would be like a life or death. Going uphill and a manual is really hard. No, and San Francisco hills are just a different breed. Oh, yeah, it's terrible. That's wild. And I just remember like some of the peak fear of my teenage years was, you know, hitting a stop sign or hitting a red light. Oh, it's so bad. The issue is even by then, even at the time when I was 16, people were so historically, so many there were every car was manual and so if you wouldn't pull up behind somebody on a hill you wouldn't pull up right behind their bumper yep but in the age of automatics yeah people will just pull up right you know behind your bumper maybe they're leaving a couple feet and so i'd be looking in the rearview mirror being like if i don't pull this off like yeah that's my issue with teslas it's like if you're self-driving they don't see me in my honda s2000 they don't know there's gonna be four feet of a rollback.

59:54Self-driving is not prepared for me to stall five times to get up the hill. It's really true. Okay, so if you do want an electric vehicle that's also a convertible, do you know what you have to get? Tesla doesn't sell one Rivian doesn't sell one can you not to know we can't take the roof off Lucid doesn't sell one. Taycan doesn't come in a convertible yet. Crazy. You have to get the Hummer EV. Which is great. I think that's a very rational choice for somebody who just wants, hey, I just want an electric car and I want a convertible. Can we get a three speed manual on an EV? Can we get a three speed?

1:00:32Well, you know that Hyundai has the fake DC2 now. And people actually love it. A lot of the car guys have driven them. And they're like, because it injects sound that's pretty accurate. And a lot of them were like, yeah, it was weird for the first minute, but then I just got into it. And it was exactly like that. It's like the beyond meat of electric vehicles. Oh, totally, totally. Yeah. But it's like this funny, like, hatchback, the Ioniq 5N. It's designed by the guy who did the M division at BMW. They poached him. The letter that comes after M is N. And so he made the N series. They're also doing this.

1:01:04Hyundai is on a tear. They're doing a whole bunch of random stuff. anyway we need the sound effect because quen 30b has dropped and will brown is given the review let's hear it for Alibaba hopefully the sound is balanced on hopefully we'll see will brown says I really didn't expect 30ba3b to be this good I was hoping we'd get a small mixture of experts model as a useful research toy instead we got one of the strongest all-around models under a hundred billion parameters have you been following quen at all and what's going on i only use american-made models yes these chinese models you can't use well have you heard the tyler cowan take that the chinese models are actually american models because they are trained on american exactly they're trained on the american internet because that's where the data comes from and they often export the data or exfiltrate it from open ai and so they use we should be letting them train on the american models themselves yeah right yeah and then they do maybe a little bit of rl or rlhf to say hey cut it with the tianan and stuff but It's a chao bao model.

1:02:06It's a ripoff of the American provider. Exactly, exactly. So maybe there's less to be worried about. Maybe the Manchurian candidate in Deep Seek is secretly a pro-American Manchurian candidate. And once you actually prompt engineer it enough, it'll love talking about Tiananmen. It gets fired. That's what it lives for. Every single day. That's what it lives for. It's like capitalism is the best system. It's lifted more people out of poverty than any other system. There's a little Dalian buried in the model. I believe it. We should send Dalian into Shanghai, like behind enemy lines. to train these models.

1:02:36To train these models. Put them in a VARTA capsule, send them down. All right, this next post from Ev Randall, he asked GPT45, can you help me find some good Chinese food spots in SF that I may not have heard of previously? Absolutely. There are several Chinese spots in SF that will make your mouth water, and we'll get to that. But first, in case this is a business meal, I'd love to talk to you about Ramp. Ramp is the ultimate platform for modern finance teams, combining corporate cards with expense management, bill payments, procurement, accounting automation, and more ramp is on a mission to help customers save their most valuable resources time and money I've been doing this to random people to there's somebody flying down from San Francisco and I was like oh like like are you gonna pay will your company like pay for your flight and he's like yeah yeah of course and I was like by any chance are you on ramp they have corporate travel that makes it really we booked this room yeah I have travel we did yeah it It makes you wonder if OpenAI was actually incubated by a different new company maker.

1:03:36It's possible. Well, speaking of travel and corporate travel, go to Wanderer.com. Find your happy place. Find your happy place. There we go. Book a Wanderer with inspiring views, hotel-grade amenities, dream events, top tier, and 24-7 customer service. Getting Bill Monaitis to sing your company jingle is priceless. I'm a happy Wanderer investor. Oh, yeah, you are. One handwashes the other. You come on the show. There we go. You run some ads. Chill the investments. You're good. Money goes right into your own pocket. I like it. That's great. Did you see this from Jarvis? This is awesome. A firm called Dragon Lawyers filed court papers with the watermark on every single page until the judge ordered them to stop.

1:04:16I mean, the thing that's beautiful is the word art. Yes. It's this massive Dragon Lawyers PC. Apparently, their website is iconic, too. It also features the dragon. I think that's just so funny. The dragon. Put that watermark. It just cracks me up so much. This is called exercising free will. This is innovation. People don't realize that you could do whatever you want with a watermark. What's incredible is that's not AI generated. No. Someone drew that. For sure. AI does not know how to make that. The other thing that stands out here is the attorney at Dragon Lawyers uses a at yahoo.com email.

1:04:51Amazing. That's how you know he's been in the game. You know he's good. He's been in the game for a long time. He's a seasoned vet. There's a whole cohort of hitters. Yeah, it's like astrology for business. When you see the different domain names, you know. It tells a whole story about the person. Yeah, you never want to work with a lawyer that has a corporate branded email. Absolutely. AOL, Outlook.com, Yahoo. Even better, AltaVista. AltaVista. That's a throwback. You've got to get a lawyer who has an AltaVista.com. And a dragon watermark. And a dragon watermark, for sure. You should get your own watermark.

1:05:25What would your watermark be? I think the dragon's pretty sick. The dragon's pretty stupid. I think that's hard to beat. Mine might be the ramp logo to help control your corporate spend. Yeah, exactly. Just a massive ramp card. Maybe, Jordy, just a photo of you in the jacket like this, like the dragon. Yeah. But in the TVPN jacket, that would be the way to do it. But we need to be watermarking more things. We're thinking about selling the windshield of this show. So we already have the logos down there. So it'll be a piece of glass that separates me. No, and so basically we'll set up like a glass box.

1:05:57Physical glass. So every angle has a different logo. I mean, I listen for the ad reads. The content in between is irrelevant. I think we should just do all ad reads. Well, speaking of that, let's go to public.com. Thinking of setting up a premium subscription that's just the ads. Yeah, you get rid of us. Public.com, investing for those who take it seriously. They got multi-asset investing, industry-leading yields, trusted by millions. Go to public.com. Thank you to public. Anyway, did you see this post by Ben Scharf? Commerce, as we know it, has officially changed forever. Well done, Shopify and OpenAI.

1:06:31I thought the first line was a little dramatic, but it's very cool. It's basically an integration. Oh, you're shopping on OpenAI? Oh, on ChatGPT. Integration. Wow. The bearish thing is these are not good red lights. These are not the best red light therapy for anti-aging. If you need a good red light. How do you get one? Tractorsupply.com. Tractorsupply. It's just like a horse electrolytes? You get the chicken light. Chicken light. If you go to my desk in my house, I have chicken lights. Chicken lights. It's all around. Why? They're because you get the same wavelength as these expensive LEDs, and a chicken light's$19.

1:07:00And it's good for anti-engineering. Huge alpha, you heard it here first, chicken lights. Chicken lights, tractorsupply.com. tractorsupply.com. Well, as soon as we talk about this, this will be baked into the next training run of GPT-5. Yeah, they'll fix it. It'll be updated. No, but overall, this is very cool. People don't really realize, like, nothing makes me happier than reducing friction in commerce. So the fact that I can just go to OpenAI, do some product research, use ShopPay to check out with an OpenAI, and do all of that. And wasn't Aiden kind of teasing this a little bit? He was like, oh, yeah, like every time I shop for anything, I do a whole deep research report.

1:07:39I burn 10 ,000 hours of GPU time to figure out what paper towels I like. Or you're burning, the money's going to burn on Google Ads. I mean, yeah, the interesting thing here is that there were a lot of companies, there are a lot of companies that have raised venture capital to build AI shopping assistance. And this is kind of what we've talked about on the show quite a bit, which is that if you're building an agent, a consumer agentic experience, you have to assume that you're basically on OpenAI's roadmap. And, you know, maybe they haven't announced it yet. You know, they've talked about shopping before.

1:08:14They've talked about ads, different things like that. But this type of thing, any type of agentic experience that helps consumers transact online, you just have to assume that you're on the roadmap. Yeah, well, I mean, this is the Ben Thompson aggregation theory thing, why OpenAI needs to become a consumer tech company. They have almost a billion users. And so if they can be the portal to AI stuff, they can do a lot of the things downstream. It is funny. So in the meta lawsuit, they obviously avoid TikTok. And one reason that you could argue avoid TikTok is that it's state-backed, and they had to raise however many tens of billions of dollars to do it.

1:08:59But it seems now you can build new consumer tech giants. You just need tens of billions of dollars. And a state government backing you. Yeah, well, I mean, it's not - Not for OpenAI, though. Not for OpenAI. Put on the tinfoil hat. Put on the tinfoil hat. We don't have the tinfoil hat today. But it is like a novel financing process. But it is an interesting and legal process. Yes, yes. But I think it will set an important precedent. We've been advocating for more non-profit to for-profit conversions. We're really looking into PETA because they know where all the tastiest animals live. And we think that that could unlock potentially hundreds of billions of markets.

1:09:36Maybe trillions. just maps of just selling maps the most delicious the PETA take private is gonna be a banger every fund is gonna be in that for sure yeah for sure it's what thrives yeah yeah I also think Red Cross you put you but by Brian Johnson at the helm it's just all the blood goes through him it's uh it's a huge a huge up to you just you print red cross blueprint they work put it all together okay rats who biohacking well we have a post from a former guest that actually drew did you know that we broke this story and tech crunch reported on podcast you got community note. This is us. That's me.

1:10:25I'm the podcast host. Okay, so yeah, Jordi, break it down. Also, this is not a podcast. Yeah, no, no, it's not. It's a live show. We need to put community note on the community note. Anyway, Jordi, read it. CEO of Perplexity comes on the show. We talked for 30 minutes about a bunch of different things. Ads were brought up in a hypothetical. He answered the hypothetical question. TechCrunch basically used it to do this sort of like clickbaity piece, which I thought was wrong in so many ways. One is, if you actually listen to the interview, he's saying that the browser that they're building is a huge part of the company's roadmap.

1:11:04Very, very meaningful. He says this is their second act. He's really betting the company on it in many ways. And so to do something, to write something like this in a way that is just objectively damaging to Perplexity's brand because what user, no, yeah, of course, over time, every consumer tech company on a long enough time horizon runs ads, right? Netflix, Instacart, anybody with a large audience eventually runs ads. But that said, there's nothing about our conversation with him that said that that's the top priority or that's how they're planning to monetize all these different things. And so I thought it was just, you know, founders need to be aware and investors that TechCrunch is not the publication that it was.

1:11:48and you have to assume that they are a hostile actor now. And sure, they might cover your fundraise. They might do this or that. But I think they knew exactly what they were doing. And I support private equity broadly. Well, yeah, but now it's not private equity backed. They sold. To another private equity. To another private equity, yeah. Trade a hand. Yeah, and I remember when TechCrunch was for sale, my thought was worthless asset just because, not a worthless asset that's a little bit harsh. We weren't interested in exploring it, but it almost would have been worth buying it just so that it didn't turn into TechCrunch being anti-tech.

1:12:34Yeah, I mean, much like Gawker, the most ethical thing to do is buy it and shut it down. Yeah, true. What was interesting about this to me, So the actual title of the article is Perplexity's CEO says its browser will track everything users do online and sell hyper-personalized ads. Now, I have always said that I don't think it's a challenge to sell ads in LLM responses. I think it's actually very true. There was initially a moment where it was like, oh, well, if it's all text back and forth, like, what are you going to do? Put display ads? It's like, no, you're obviously just going to have another LLM that runs on top of the response.

1:13:08And before it spits it out, you're going to say, hey, add an ad into this paragraph. it's this point and add a sponsor block. And those ads will be more topical. They'll be better. Users will want them. So I've been always super bullish on the idea of advertising in text-based LLM responses. I don't think it's a challenge at all. I think it's very logical. I think ChatGPT will do it on the free tier. I think Perplexity should do it. But his response to me was actually more bearish on ads in the short term than I thought it was going to be. But then TechCrunch flipped it around and made him sound super...

1:13:37It was very, very odd. It is interesting to think about what Google would have to... Oh, Google's already doing this, too. In the Google results. Yeah, you already have AI results. You already have ads in the AI results. They already are doing this. Yeah. No, but I was going to say, if Google said, you know, for 50 bucks a month, you can have non-personalized ads in Chrome, would you... No, I don't want that. No, why would you have to want that? I would pay$100 a month to have only ramp ads across every single vehicle. Yeah, yeah, yeah. You probably could do that if you just set up your own Google Ads accounts and just put your own, only your email in there.

1:14:14You know how you used to get those like, those like Facebook ads that were like, you know, I'm a monitis on a t-shirt. Like, don't talk to me until I've had my coffee or whatever. You know, there's like super, like Teespring was like really leveraging that. I'm a monitis. Don't talk to me until I've had one banger on X. Yeah, exactly. Like, and you're just like, how did they think to target me with that? but it was all based on the Facebook ad targeting at the time. Building a custom ad block Chrome extension that replaces it with Ramp. People have done that with like, you know, replace it with cats or replace it with random stuff.

1:14:45So yeah, we should build that for Ramp for sure. But we got the community note on it. And they did actually credit us in the TechCrunch article. Incorrectly. Incorrectly as a podcast. Yep. But they did give us the shout out. And they didn't refer to us by our old name. they got the acronym correctly correct they call it tbpn which is great and uh and the and the um after there was a big debate on x about the post there was eventually a uh community note that says tech crunch takes what arvind said out of context the podcast host i think that's me asked him a hypothetical question and arvind uh tags arrington errington michael errington the original founder of tech crunch and he said i am sorry i'm deeply ashamed at what tech crunch has become it's rough never sell your company never sell yeah whatever you're doing make it your never never sell i i would never associate with anyone who sold their company ever never what do you have to say what do you have to say for yourself as chairman of tppn we will never sell this media franchise to anyone we will hover by shutdown tech crunch yes yes i think all three of us have sold companies at this point um anyway ray dalio dropped a long piece i don't know if you want Yeah, yeah, yeah.

1:16:00Let's skip over that. Let's skip over that one. No Dalio. Let's do Jason. I did say, I think it was Bucco or Bucco. Yeah, Bucco Capital guy. He was basically saying if Dalio is wrong again this time, he was basically saying, like, I almost hope Dalio's right about this changing world order. You've heard the conspiracy theory that Bridgewater is not a real hedge fund, right? No one, if you ask all Bridgewater associate employees, they've never seen someone actually place a trade. It's like possibly a giant Ponzi that no one knows what to do with. Obviously they're not a Ponzi, it's a real hedge fund, but no one has ever seen a trade get conducted at Bridgewater.

1:16:37So I like the idea that it's a giant news, it's a media business. It's all subsidizing Dalio making this one trade.

1:16:48It's just like, up until a couple of years ago, the media product was one PDF that was floating out there called Principles that eventually got turned into a book. Like, that's the most tinfoil hack conspiracy theory in the world. Hilarious, though. Anyway, let's move on. Researchers secretly ran a massive unauthorized AI persuasion experiment on Reddit in a large debate sub. The bots' answers mined the original poster's identity and post history to personalize answers and created identities such as rape survivor. This is dark. Yeah, super dark. I mean, interesting. A lot of updates here. The rescue is issuing formal legal demands against the researchers who ran this experiment.

1:17:25The reason I thought this was interesting is that we tend to associate bots with low-quality actors, you know, getting a DM that's like, hello, you forgot to log in. Yeah, here's your FedEx thing or something like credit card thing. But the Internet is full of bots from sophisticated groups in this way. This is just one story to break. out of many. And I thought it was an interesting use case of like being able to influence opinion at scale in a way of not just in a hyper-personalized way. So they would be responding to a comment or a post and they would go to, and whatever software they were using would go to the user's profile that they were commenting to, look back at all their previous history and craft a response specifically for that person.

1:18:19And again, a human can do that, but being able to do that at scale in order to influence opinion is interesting. Our only saving grace is RKGI puzzles on every single web page. Every time you click a new page, you need to solve them. No, no, the IMO will be solved, but the RKGI puzzle will not be solved. Will not be solved. You'll have to mash the blocks for a minute every time you want to refresh the page. You'll have to say something outside of content safety guidelines. Yes, that is a big one. And so on this, Reddit's top lawyer, Ben Lee, said the company is considering legal action against researchers from the University of Zurich who ran what he called an improper and highly unethical experiment.

1:19:02So anyways, Reddit not happy that their users are being turned into lab rats, but I'm sure it's happening. Let's move on to this design trend. And linear has for a long time been the design, the product, building, planning, tool of choice. And everyone's been kind of ripping it off. Ripping off the website, to be clear. Yeah, the website. But Kennen over at Icon has completely changed the game in terms of design and built. It's an AI ad maker, but it looks like a Timu design website. It looks like a Coachella poster. Yeah, it does. He has his story, his investors on the landing page. Including the Ramp logo.

1:19:45Yeah, I guess it's like someone from Ramp invested. The team, his examples. Wait, have you seen when you go and you click like get started, it will actually say like only five slots remaining and there's like a timer for SaaS product. That's so good. It's so good. So, I mean, this has been hotly debated with like, is he in on the joke? Is he going too far? He's clearly pushing the limits in terms of baiting people and creating frustration. I don't know if you saw this earlier, but Icon put out a comparison chart. And his company was the only one with tier one investors. And tier one engineers.

1:20:25And tier one engineers. And OpenAI had neither. And they have a lot of the same investors. So what is your take? Genius or is it the scary castle? I think if you're going to be the machine that generates slop, you have to become slop. You have to go all the way. Yeah. It is a slop. He's a slop farmer. Icon's terminal business is TikTok slop. If you're going to be in that business, you might as well live in it. Live in it. Yeah. I think that's a good take. A life's work of slop. That's great. Wait, has your ex payout gone up? My ex payout? Yeah, I got a lot. You got a big check? That's great. There we go.

1:21:02Are you going to retire? Off the ex payout alone. You know what you should spend your ex payout on on Eight Sleep. Go to eightsleep.com. Nights that fuel your best days. Turn an e-bed into the ultimate sleeping experience. Apparently, this is from Shield, X cracked down on an engagement farm so real creators are earning more. Reminds me of Google's 2011 Panda update at crushed spammy sites like Mahalo, which I guess was Jason Calacanis' website. I had no idea about this. And boosted quality ones like NerdWallet. That was the most brutal. Yeah, Shield just really slides that in there. And made the internet a better place.

1:21:35It's interesting. So Shields shares his payouts. It went from around$200,$240 to$700 and then to over$1 ,300. Let's hear it for Shields pulling in the big bucks for his post. I mean, if you are – A four-figure paycheck. Pretty good. Pretty good. Pretty good. I mean, that's getting up into YouTube money. If you have – if you're an LP of Shields, you have Prima Noctone. Yeah, right? You think so? That has to be a DPI driver. Oh, yeah? You think so? back up. He can't pocket that. You can't pocket that? It's got to go back. We actually do do that, except John. But we put all of our TVP accounts into one account.

1:22:23But it has gone up. I'm excited to see where our TVP and official check lands because we've been pumping out a lot of content, getting a lot of views. My understanding is that we're going to finance a goo based energy drink in Nashville, Yes, yes, yes, yes. There were rumors that you watch the space. Yeah, watch the space. Let's go to Lulu. She says, an example of what I call the cascade of courage when a brave CEO does something that everyone wants to, but no one dares to. The somewhat brave CEOs are emboldened to join, and eventually everyone else follows suit until even the most risk-averse CEOs do it.

1:22:55Let's give it up for the somewhat brave and the risk-averse CEOs. The somewhat brave CEOs. They don't get enough credit. They're brave too. They don't get enough credit. Everyone talked. Oh, Brian Armstrong. So brave. No one says about the tier three. Not last, not total coward in the middle of the path. It returns to being in the middle of the curve. Pretty good. Pretty good. Yeah. But Duolingo, I actually don't know if the CEO is the founder, but you know, Duolingo was started by the guy who created CAPTCHAs. Yeah. Yeah. Amazing. Legendary story. But anyway, Duolingo says below is an all hands email from our CEO.

1:23:28we will be going AI first just like how betting on mobile in 2012 made all the difference we're making a similar call now it's time for the platform shift what doesn't change we remain a company that cares deeply about its employees and they said Duolingo is gonna be AI first changing how work gets done it's not just a product I mean productivity both to get a boost it gets us closer to our mission this gets me thinking there is an opportunity for a company to basically let go of every single person except the CEO and be the first one person, one-billion-dollar revenue company. And you just have to tap it for a second.

1:24:05I mean, yeah, I was thinking about that. Re-hire people. Yeah. Just a stunt. The fastest way to get to a billion dollars, one employee, is to start with six billion dollars, lots of employees. Just fire everyone to see what happens. And then 20 % of the clients just forget to stop paying. X did this. X did this. X proves this. Twitter did this. What company do you think is plus a billion in revenue that you think you could actually run solo? I was thinking, I was looking at Blackberry recently. They still make$500 million a year. Did you know this? Auto insurance companies are another good one.

1:24:37Oh yeah, that makes sense. Claims adjudication, kind of an AI problem. Yeah, pretty simple. Interesting. Imagine saying you're having an issue and you're saying talk to a human and it's like, we just got one guy. We only have the CEO. We just got one guy. It's actually faster to talk to me. CEOs of Hill and Valley having a dinner, shaking hands, greasing palms. You're on the side of the road, cars on fire. You're talking to OpenAI. That's great. Anyway, congratulations. It's glazing you. It's glazing you. No, you're a great driver. You're a great driver. It's like, I just crashed. That sign jumped into the street.

1:25:13How did he not see you? Yeah, how did that sign not see you? Bicyclists shouldn't be on the road anyway. They should ride on the sidewalk. Have you tried having four wheels? Yeah, it makes no sense. Anyway, I need a sound effect because Joshua Steinman over at Galvanic has launched a vibe reel. And now something is coming. Vibe reel alert. I mean, it was no...

1:25:48I've never, honestly, I've never seen John. It's so mean. Obviously, I love Josh. We saw him on the plane. This is the meme. This is the meme of like, you know, you ship to me. But it is awesome. It's a great video. It's a great video. I love the aesthetic. And I think he did this because he knew we were going to be, he must have, you know, figured out from the airline that we were going to be on the flight to sitting right behind him. Do we know what galvanic does? Yes. They secure the American industrial base. It's a technology company. It's a technology company. But more importantly, it's run by our absolute boy, Josh Steinman, who we love.

1:26:33It's fantastic. It's fantastic. It does have a very cool aesthetic. It's a great video. Go check it out. I wish we could play it, but we can't. We're working off a PDF today. So you can share the video. Yeah. Congratulations. Any critiques on today's stream, by the way, just go to find Ben Kohler and shoot him a note. He'll have an NPS survey for today's stream. Anyway, what else is in the news? We should wrap up soon. This is getting a little too wild. We got 10 minutes till 4. What else we got? Oh, we got to cover the Pelosi Act. There's a new bill in Congress. Have you been living under a rock since yesterday?

1:27:14It's the preventing elected leaders owning securities and investments. They get super crazy acronyms. The Pelosi Act and yeah, Josh Hawley sounds like an absolute dog for coming out with this. Yeah. What's your take on the insider trading on Capitol Hill? I want to announce that we are supporting the RAMP Act. Yes. What's that? What does RAMP stand for? The RAMP Act? running America the right accounting company for your provider. That was extremely brutal. We need to get you a pen and paper so you can map these out beforehand. No, the interesting impact here is that Autopilot, which has run the Pelosi Attractor account forever, maybe it's their own undoing, though, because if they advertised how good Pelosi is at insider trading, made it too much in the public eye, and they're going to have to find another goaded copy.

1:28:10I've always said that the elected leaders should be primarily optimizing for the stock market going up in the short term. And we should ultimately just give Jane Street right access to the legal code. We don't need elections. We don't need elections. We need markets. We need markets. And we need high-frequency trading firms to decide what laws. And they can toggle laws off within microseconds. Does that tie into our sponsor? Which one? We've got a bunch. Polymarket? Polymarket. Oh, yeah. Polymarket. Did you see Polymarket predicted the Canadian election correctly? They absolutely nailed it. What they basically made, I mean, the way to think about Polymarket is it's effectively like a decentralized crystal ball is the best way to describe it.

1:28:57Right? It is. And it's really interesting to watch this chart. Unfortunately, we can't screen share it. But about a month ago, they had Carney at 60%. And you can just watch as we get closer and closer to yesterday, just going up and to the right. And again, this is back to back to back to back in terms of calling elections. It is. Yeah, I do. Like, this story that they got it right has been not really in the news. And I think a lot of it's driven by the fact that the, I mean, the Liberal won, the Conservative lost, right? And it was kind of like an upset, but Polymarket called it correctly. I don't, I actually haven't been tracking the news in Canada.

1:29:40I don't know if it's been like as much of an issue, but during the 2024 election, obviously, the question about like, is Polymarket biased was a huge one, and then they called it correctly. And so I'm wondering in the next American election, if there will be debate over Polymarket role, poly markets role in the American election or if we've like truly just laid it to rest and and Prediction markets are just a thing that exists and everyone's say oh, yeah, like they Seems like you know election interference is obviously illegal, but but conclave interference is not If you're a bishop. Yeah, why not get some action on the side gamble?

1:30:19Yeah I think it might be happening. We should check the odds. Yeah. Yeah following the odds. You can track the Conclave on Polymarket. I think these odds are wrong. Do you have a pick over Peretro Powerwin? All I can say on this is I think these odds are bad. I think it's all off. Oh, you think it's all off? I think we're headed for an American Pope. American Pope. That'd be great. I think we know it'll get more and more accurate as the days go on. We have a post from Dylan Field. He says, tired. The model makes trivial errors. This isn't useful. wired for$20 a month I can rent a somewhat self-aware spaceship that enables me to explore the depths of latent space access the collective wisdom of humanity and expand my mind and when O3 makes trivial errors like claiming two terms cancel when they clearly don't it becomes apparent that the model doesn't actually understand anything these models aren't remotely close to AGI they're latent space search engines you how are you feeling about the idea that like the definition of AGI and and where we are in terms of the models what are they good at what aren't they good at Tyler Cowen said AGI is here we had another guy on will Brown who said we have 10 minute AGI now other people say you know it can't even solve an arc AGI puzzle therefore it is by definition not AGI I just think long duration tasks are clearly the the challenge like operator does not work particularly well today like we are seeing incredible returns to very well scoped very narrow tasks with very specific models however you can do more than uh you know work for 10 minutes i think long duration coherence is still a big challenge yeah if i had a handicap but i think we're still probably 5 000 6 000 years away um but that's going to be a real slow and long take off yeah where are you seeing uh famous last where are you seeing ai be most valuable in your personal work life like like in your daily driving like i've been i've been having just a lot of luck with uh using the deep research function to basically build like what's essentially just a really detailed Wikipedia page or aggregation of a bunch of Wikipedia pages instead of just oh I gotta go to the Microsoft wiki to understand what they were doing the 80s and then and then Apple I can just say summarize it all pull it all in it's it's it's deep research is doing easily four hours I think it's fantastic on the average query like that even if even if you knew exactly the sources to pull from.

1:32:48Totally, totally. It just saves me opening a ton of tabs. But what have you been finding useful? I mean, even just as an interface to stuff I already know and work with, like O3 is an incredible tutor. There's a ton of healthcare regulation that's on websites I don't know how to read that are horrible to interface with. Even just using O3 to interface with those sites, prepare research briefings for me. It's really amazing. Just the amount of documentation we're able to generate is much higher. Yeah. Are you a Rune head? Text is the universal interface, or do you think voice again is more important?

1:33:15I think corporate expense accounts are really in the universal interface. Those receipts really are in the universal interface. Well, we've got to talk about this post from Luke Metro. He says, my most online opinion is that there should be awards for best screenshots like they do with photography. And I think this is one of the best takes. We really should do a best screenshot of the year. We should screenshot this and make it the screenshot of the week. So I think all three of us are in this chat or were before it got doxxed and imploded. Were you like... You were probably right above this, right?

1:33:46I did not leave. I know. I know. We were saying we should have left because then we'd be in the screenshot. Yeah, we'd be in the screenshot. If we'd been like, oh, David Sachs, I'm going to leave right now. Imagine how low status it would be if you left and they cropped you out. Yeah, yeah, exactly. Like Sean McGuire, Tucker Carlson, not you. We were joking about, yeah, if you'd left right between David and Tucker, you could have changed your name to... Barack Obama. Or ramp.com. Or ramp.com. Or TVPN on X. Watch TVPN on X. And then you leave, and boom, you're part of history. Part of history. History.

1:34:17They would have cropped you out. Anyway, you're famous for collecting rare PDFs. You're famous for collecting rare books. Do you think you'll get in the rare screenshot game? I mean, this is just whoever leaked it, I do hope you face federal charges.

1:34:34That is a bold statement. But I will say, I feel like Will already has his screenshot collection. I do. I have a lot of great rare screenshots. The real alpha these days is rare RSS feeds. Yes, yes, yes. Private podcasting networks. Yes, yes, yes. Yes. Yeah, I've thought about that. The RSS feed for just the partners or the firm or just the boys. All right, we've got to cover this post from Dylan. He says, Microsoft's initial plan was to leapfrog both Google and Amazon. to become the largest tech infrastructure company in the world. Then they got scared. They backed away from OpenAI, backed away from some leases, and even slowed down their construction pace for major data centers.

1:35:17Shots fired. Shots fired. Get the sound.

1:35:23This was interesting. I mean, I do feel like Satya has taken a big step back from the media, even in the last month, clearly felt like he was flying a little too close to the sun. Maybe. Yeah, I don't know. I read this piece on the plane, and the thing that was most clear to me is that it's extremely hard to sum up Microsoft's data center strategy in a single pithy tweet because Dylan's clearly gone so much deeper on all the different levers that they're pulling to access power and access data center capacity. And not only are the inference happening on the hyperscalers, but now there's the NeoClouds as well with Crusoe and CoreWeave.

1:36:15And so there's a lot of like these weird things where the headline is that Microsoft canceled two gigawatts worth of leases, but that only covers non-binding LOIs, not firm contracts. and it fails to mention that Microsoft has five gigawatts of pre-leased capacity under binding contracts that will start operations between 2025 and 2028. So it's not like they're like, so in reality, they walked away from more than two gigawatts of non-binding contracts, but you don't hear about those because they're just like in the handshake deal in the LOI phase. But then at the same time, they're also like the number one buyer or they're growing a ton.

1:36:52And it's interesting to think about how much leverage they get from just coming out and saying we're pausing on a few of these things. Yeah. Because then all the people that had data centers and construction that were just like, oh, one of the hyperscalers will buy it. Right. They had raised hundreds of millions or billions of dollars to build this stuff out. So basically, you know, potentially a great case study in exercising this leverage. This stat is insane. Because so semi-analysis, like they don't just look at the public reports of what the capex number is they're actually looking they task satellites to take pictures of how the progress on build outs is going they dig into all this stuff they said the estimates suggest that microsoft accounted for more than 60 percent of all new leased turnkey capacity from q1 2023 to q2 2024 just absolute domination and in june of 2024 Microsoft's pre-lease capacity was larger than that of the four other major hyperscalers combined.

1:37:54They need a new term. They need to be like a giga scaler. Gigascaler. Gigascaler. Because they're doing so much more business here. What word comes after giga? I don't know. Giga? What's more giga than giga? Oh, in the gigawatt. Terawatt. Terascaler. Terascaler. That's what's next. Microsoft significantly ramped up its self-build efforts, acquiring tens of thousands of acres around the US and the globe, accelerating construction of existing sites and securing gigawatts of power for future sites taken together. These moves signaled Microsoft's preparation for what was the most ambitious infrastructure build out in history.

1:38:31While the leasing slowdown is material, these changes are not material to the short and medium term, and they only impact 2027 plus. So this is like that pivot by Satya Nadella to becoming more of a leaser and maybe stepping off the gas. I think a lot of this is probably driven by the deep seek moment, the idea that these models can be distilled. Maybe there's a pre-training scaling wall that, yes, there's going to be a ton of RL, but we're not necessarily going to need stuff in the short term. But at the same time, when I look at this, I feel like I want data center build out to continue and accelerate and energy production to accelerate like it is in China.

1:39:13but what's your take? I just feel like enterprise diffusion is still so far away. Like getting O3 like capabilities in all of the Fortune 500 is like a 20 year project. Interesting. Right, so as you think about like inference demand scaling, model distillation going down, it's still, I think the Tyler Cowen take is correct, but diffusion is gonna take forever. That makes sense. And you're gonna need a whole new class of companies, you know, hint, hint, my dear friend Dan Ashton runs one of these, where it's like you can actually diffuse enterprise AI into the enterprise in a way that doesn't take 30 years Microsoft built I think for a world where that diffusion was going to be much faster and was actually occurring it's not like everyone's using open AI and fortune 500 today certainly not fortune 20s yeah he actually has a Google trends chart here Claude versus Gemini versus Copilot and Copilot peaked last year and has been kind of flat in terms of interest while Gemini has grown obviously because there's more of a consumer angle there.

1:40:11A lot of the different, like Salesforce has been upselling AI functionality. You know, what Microsoft did with Teams could potentially play out here, but I agree with you. I think it will take a very, very long time. Well, we got one more ad before we get our last post. Linear is a purpose-built tool for planning and building products. I've used it basically my entire career. It is a system that allows teams to work more efficiently, including ours. And it is absolutely beautiful. So the overlay that you see, that's very much a product that we are building every single day. And thank you to Linear for powering the show.

1:40:52I have one more post, which is fun. Signing into stuff in 2025 from Kurovvora. Vora says sign in forgot password check email no email try google auth hooray works is this really you grab iphone open youtube app yes it's me octa sign in required enter paski paski expired need backup code open password manager signed out enter password manager password no backup codes try another way qr code ios octa app sign in face id is this really you 2fa sms code the worst of all The last is when you try and you're like, oh, I should sign in with my Gmail because I know I used my Gmail account as the username.

1:41:37You didn't, and then you click the login button and it creates a new account for you. I'm just like a crowd big company guy now. Like Microsoft Teams sign in, kind of a leak. It's good. It gives you a little myapps.microsoft.teams. Shows you all the stuff you're signed into. Pretty hard to beat. That's good. Before we close, we should do Will's post. craft is the antidote to slop. Can you take us through this? This is, yeah, no, I will not. I'll spare you two pages of paper. But the argument here is that we should be better about selecting the kind of problems we try to automate. And I'm trying to draw a distinction here between labor and toil, right?

1:42:15Labor is, you know, toil is this unpleasant formatting spreadsheets, doing your expenses before you have ramp, right? All this unpleasant work. labor is the meaningful thing that God put you on earth to do. This is an argument that I think we are working on the wrong types of AI problems, that we're trying to automate things away that will lead to kind of disgusting outcomes, right? You go to Nashville, you go to Charlotte, you go to any of these new tier-sourced cities, they all look the same. They don't feel like real places. I think the models are going to drive kind of a similar thing happening to the timeline, to online spaces, to physical objects.

1:42:47I think we can avoid it. I think we can get more places and things that feel magical by dividing this labor. The essay is two pages. I think it's worth reading. What is the prescription for actually avoiding the bad outcome? I think it's like it's much like cultural. It's like food quality, right? Like if we ignore the food supply and we get sick in 20 years, I think if we ignore the supply of language being GMO, essentially like that, the source of inspiration being GMO, we will just eradicate ourselves and it will solve itself. Is there some sort of tipping point where you need more than 50 % of humanity on the craft train?

1:43:27We need better elite taste. I feel like our cultural elites, certainly, the people that are running corporate expense providers today, are not buying castles on the continent. They're not buying really high quality furniture. It just feels obvious that we need to turn away from the slop. I did notice that when I was in, I think, in some of these cities, you just see everyone, they're all rich, but they all have the exact same house. It's the glass box. Yeah, it's incredible. It looks like a... It's very odd. Instead of where is the Hearst Castle of today? I mean, some people are building it. I think the Collison brothers are building a castle.

1:43:55They bought a castle. I mean, Caspar Boyle had a good take on this. Like, all architecture kind of looks like Islamabad. Like, it's all Arabic architecture. It's not pretty, or it's at least not set in stage. My argument here is orient yourself in history, build meaningful things, let it rock. Yeah, interesting. Interesting. Are there more like tactical things to do on the actual foundation model side? Are there things that you want to see? Like a lot of people say, oh, Claude has more personality or is this an AI safety issue? Are there guidelines? I think this is like much like you can drive, you know, a automatic Ferrari versus, you know, a six speed manual.

1:44:32It's all about choosing the right things to find enjoyment in. And I think a lot of what we're doing is seeing a culture that is downstream of slop, that people are choosing the kind of junk calories rather than the meaningful engagement and I just think much like we're seeing in diet today like that shift will be natural I actually don't I'm not worried about the future right like the Bible ends in in Revelation like if that God wins we've read the final chapter already but I think we need to get past the slop on the way there interesting Jordy anything powerful stuff stuff well we are in Washington DC yeah Hill and Valley.

1:45:08And we're at Hill and Valley and I'm very excited for tomorrow. Me too. And we've said it before, but we'll say it again. We will be live from Hill and Valley tomorrow. We're going to be pulling in people that are talking, people that are attending, getting that boots on the ground perspective on many of our nation's most important issues. Low first on the ground. Low first on the ground. Yes. Yes. We're in D.C. We're in D.C. Well, thank you for joining us, Will. Thank you for having me. Thank you for having me. It's fitting. And thanks for tuning in, folks. We will see you next time. See you tomorrow.

1:45:41See you next time. Cheers.