Aaron Levie, Alexis Ohanian, R.J. Halperin, Cristóbal Valenzuela, David Zagaynov, Sam Lessin, xAI Merges with X, Isomorphic Labs Raises $600M

36:36But who knows? Either way, wild times. The most entertaining outcome is the most likely. Let's go through the Tesla-SolarCity merger. I just have some key bullet points that are kind of interesting to go down memory lane. SolarCity, founded in 2006 by Elon Musk's cousins, became number one in U.S. residential solar installation by 2013. So he kind of delivered on the initial pitch. And I heard this crazy story that like he thought about it. He thought of the idea of Burning Man or something like that. Have you heard this one? It's like a crazy like he was hanging out with a bunch of people and was like, maybe we should do solar panels on people's houses.

37:17And then they just like went for it. It's so funny because my entire life, as long as I've been truly conscious, solar panels on homes are just so normal yeah totally that it's funny to think it was maybe one or a handful of companies that actually normalized it but yeah it's usually how things work uh so at the time elon was like famous but really he was only known um uh by for paypal and spacex and 2006 like these weren't like paypal was a big deal spacex was still very early uh elon provided the initial$10 million in the startup. And Lyndon and Peter Reeve, his cousins, ran the company. He also served as SolarCity's chairman and was the largest individual shareholder, leveraging his vision of a sustainable energy future across his ventures.

38:07SolarCity's business model focused on solar panel leasing and what are called PPAs, which are power purchase agreements. The customer pays no upfront cost for rooftop solar installation, but then signs a long-term agreement to buy the power produced. So it's basically like a no money down solar leasing model. And it was wildly popular. And so there was a little bit of like hard tech here, but most of the panels, I don't think that they were manufacturing them at, you know, gigafactory scale. I think it was mostly, Hey, let's buy a commoditized product and then innovate on the financial side. And so this fuels rapid growth.

38:45But it was very capital intensive and debt fueled, Of course, because you're giving someone a, you know, I don't know. I don't exactly know how much these cost, but let's say like you're giving someone 10 ,000 bucks or$100 ,000 in value, like the solar panels and then the installation costs. And then you're paying that and you're amortizing it over 20 years, but you're not getting those payments from the customers for 10, 20 years. It takes a long time. And so every time you install new solar panels, you're accumulating debt on your balance sheet. And you're also accumulating like a revenue stream, which is great, but you have a major, major financing mismatch there.

39:22And so by 2015, warning signs emerged. SolarCity's stock price, which had peaked around$88 a share in 2014, began a steep decline. In October of 2025, the company lost a quarter of its market value in one day, falling below the coveted unicorn$1 billion valuation. The drop followed CEO Lyndon Rive's announcement that SolarCity would slow its expansion and cut costs as years of 80 to 90 % growth had led to huge investments that were not yielding cash flow yet. We talked about that. Internally, executives knew the company faced a looming liquidity crisis by late 2015. SolarCity's aggressive growth had left it with large operating losses and over$3 billion in debt by 2016.

40:06In fact, SolarCity was nearly insolvent prior to the Tesla merger. court filings later revealed that weeks after the acquisition, auditors warned that SolarCity lacked sufficient cash to meet its obligations on a standalone basis. And so there's always this question about like, you know, he kind of got a bailout. Should he have just let the company go bankrupt, buy it out of bankruptcy, do a proper bidding process? Like this was always like a hot button debated issue. So to sustain growth, SolarCity diversified within the solar value chain. in 2014, it acquired Solevo, a solar panel manufacturing startup, and embarked on construction of a massive one gigawatt solar panel factory in Buffalo, New York, with substantial subsidies from the state.

40:49The Buffalo Gigafactory 2 was part of a plan to produce advanced solar modules domestically and reduce costs. And we talked to Casey Hanmer about this a little bit, where he thinks that the Chinese manufacturers of solar panels are subsidizing them really, really heavily. And his argument was like, hey, when somebody's subsidizing something below their cost, you shouldn't tariff it. You should buy as much as possible to hurt them, which is kind of funny, kind of hot take. But there has been incredible international pressure with these commoditized products. We could have maybe pushed him a bit more on that and said, well, do you want to, you know, usually if somebody is selling something for less than what it actually costs, They have sort of some time, some sort of long-term strategy that maybe isn't in your best interest.

41:38It's the learning curve pricing issue. So yes, they are losing money on each solar panel right now. But if we all of a sudden provide to them 10x demand, they might be able to justify investments in 10x larger gigafactories. And they might actually be able to lower the price to the point where they are profitable at a price that we can never compete with because we just are not scaled up enough. And that's certainly what's happened with TSMC, where TSMC has seen so much demand from Google and NVIDIA and Apple and everyone else that they're getting to a place where no one else can compete. And they really have like essentially a monopoly on the three nanometer node.

42:20And they're running this with cars right now. Yeah, cars and phones and a lot of different stuff. It's all happening. uh so the cash squeeze got really severe the company was holding emergency weekly cash management meetings and its executives privately informed musk that it needed an infusion of 180 to 300 million dollars to avoid default i feel like elon nowadays could like raise that with one phone call for like literally anything he could text mark andreason and say ellison yeah ellison hey i need 300 mil don't ask what it's for it's just for something cool and be like yeah of course dude I got you.

42:57Attempts to raise equity or find outside investors in early 2016 fell through, which is unfathomable, forcing desperate measures. Musk's other company, SpaceX, loaned SolarCity money to prevent its cash balance from triggering a bond default. SolarCity had covenants requiring cash above$169 million. Hilarious number. Even his debt covenants include the number 69 classic elon like even in the debt world he's like memeing i don't know it's probably just a coincidence but you never know uh in february 2016 uh as solar city's board grappled with how to avert bankruptcy musk met with linton linden reeve to reeve to explore a bold situation tesla inc would acquire solar city uniting musk's electric vehicle and solar energy ventures and i want to go through a couple bullet points on how the deal came together so So by 2016, SolarCity was in trouble.

43:54They were nearly insolvent and they had$3.25 billion in debt. And again, this is the company that was trading in the billions and then the market cap goes down to less than 1 billion. So the equity value, debt to equity ratio is like really, really out of whack. And so in June of 2016, Elon proposes that Tesla acquire SolarCity to form a vertically integrated clean energy company. And it makes a ton of sense that a Tesla owner, again, it's like, it's crazy because it's one of these Elon deals where it's like him in one hand washing the other, him talking to himself and like he's on the board and he owns all this and stuff.

44:30But at the same time, if you think about it, it's like Tesla customers love clean energy. They have the Tesla battery pack on the wall. They have the Tesla charger. And of course, they want solar panels. So you imagine that it makes a ton of sense to just sell all of this to one person. There's tons of like sales channel stuff. And then also just in terms of manufacturing, like the gigafactory that produces the batteries and the cars can also produce the solar panels. Like it's not that crazy to me, but the deal is kind of crazy. And so there's an all stock deal that happens. SolarCity is valued at 2.6 billion and Tesla assumes 3 billion in debt.

45:05Shareholders got 0.11 Tesla shares per SolarCity share. And then Tesla went on like a generational run. And every SolarCity shareholder, if they held, is very happy about this. Of course, there was this conflict of interest because Elon was the chairman of SolarCity and the CEO of Tesla, and he owned 22 % of both. And his brother and cousins were board members and executives. Despite the controversy, 85 % of unaffiliated shareholders approved the merger in November 2016. And that makes sense. Like, if I'm an unaffiliated shareholder and I just hear about this and I've never even met Elon, I'm going to be like, yeah, of course.

45:41Like, this deal makes sense, even though it's a little wonky. Like, put these companies together. I'm happy to be a shareholder in the combined entity. And so, of course, there's a legal battle still because 85 % approved, but some of them are upset. And so there's a shareholder lawsuit. They sued Tesla's board of directors for breach of fiduciary duty, which is what you sue for whenever you're unhappy about something. All directors except Musk settled for$60 million in 2020 via insurance. And so they had some key man insurance there or some, I forget what it's called. It's the executive director, DNO, Directors and Officers Insurance.

46:19Musk alone went to trial in 2021, accused of pushing the deal to bail out his failing investment and family. Plaintiffs sought$13 billion in damages, which is crazy because that's like way more than the debt and more than the equity. It's like more than everything combined. But Musk won. So in April 2022, a Delaware judge ruled in Musk's favor, stating that Tesla paid a fair price. Of course they did. If anything, they overpaid. And so as a SolarCity shareholder, you've gotten a bunch of Tesla shares that basically like 10xed. You got a really high price at the time of the deal. Like you have nothing to complain about.

46:56Why are you doing that? It was a full zero, like literally three weeks later. Yeah, exactly. They were running out of cash. uh and so uh the delaware judge uh you know elon's been very rough on the delaware judges but in this case they say you know we got musk's back he's right uh and the deal this deal ultimately benefited benefited shareholders which i think we agree with um and then uh there was a post-merger decline too so even even as they combined and they had all the resources they took care of the cash management problem. Tesla's solar installs fell 60 % from 870 megawatts to 326 megawatts in just three years from 2015 to 2018.

47:34And Tesla lost its number one solar market position to Sunrun, which is another solar installer. Yes, Sunrun's been crushing it. They've been crushing it. And so today, Tesla has actually shifted away from highly commoditized solar panel installation to battery storage. So they sell those power walls. We looked at one, it was$192 million installation, right? Wasn't that? You could just go on their configurator and say, yeah, I actually need like a gigawatt of power. Can you send me it? And they'll be like, sure. But listen to this, in 2024, Tesla Energy hit 10 billion in annual revenue powered by power walls and mega packs.

48:16And so I think of the power wall as like, oh, it's like this cute little side project. It's like the AirPods of the Tesla world. Like that's a lot of money, 10 billion in revenue. And so even though like the solar angle didn't really pan out, like Musk was right over the very long term that like vertically integrated clean tech energy would be a good strategy, good company. And so just an interesting thing that like, you know, to think about with the X and XAI thing, like it's hard to predict where this goes in the next five, 10 years. Is it a good deal right now? Will there be a lawsuit? Like all of that kind of steps to the side of really trying to concretely visualize what does the social networking and AI slash LLM, consumer productization of AI look like in five years, 10 years?

49:06And can he figure out something that is good in that space in that time? Yeah. And I don't know. He has a lot of money and a lot of time. And bad time for Ben Thompson to be on vacation. I know. He did one unscripted podcast about it and he was pretty bullish on it. He said he thinks it makes a lot of sense. But we'll have to wait to get Ben's full take. Anyway, such a funny deal. Never a dull moment in tech. My number one advice for Elon is you're combining these two companies. It's more important to control costs more than ever. So go to ramp.com, Elon. Time is money. Save both. Save both. Easy to use corporate cards.

49:47bill payments, accounting and a whole lot more all in one place. I would love to see X and XAI both on ramp. Uh, really understand. I mean, big thing of ramp is like you're paying for two subscriptions and it identifies that automatically, right? It's like you're merging these two companies, you know? Well, actually I imagine a lot of their contracts are primarily seat based and maybe aren't saving much money at all. Uh, but, uh, you can always save money. You know, That man hates losing money. If anyone in our audience signs up for ramp and doesn't find a way to save at least$100 a month, I was going to say like$1 ,000.

50:27I don't know. And then somebody would - Depends on your scale. Yeah, it depends on your scale. But you sign up for ramp, you're going to save some money. And you're going to save some time, baby. And you're going to save some time. Well, let's stay with Musk and go on to this exclusive in the Wall Street Journal. That's right. Apple and Musk are clashing over satellite expansion plans. We talked about this a little bit. The iPhone is getting Starlink direct to cell, so you'll be able to communicate with the Starlink satellites from your phone. Very cool. It's already live in some places. You can send like a single text message on a satellite network.

51:04But obviously this is a really, really tough deal and a very big, there's a lot at stake obviously, because once you don't need the cell phone towers, you can use the satellite communication everywhere. That's going to be a big differentiator for any cell provider. So here's the key points. Apple and SpaceX are competing for spectrum rights to expand their satellite-based communication services. Apple's outer space ambitions have drawn the ire of SpaceX CEO Elon Musk, who has pushed regulators to stall expansion efforts from Apple partners. Despite their rivalry, Apple and SpaceX need each other to achieve their respective goals in the satellite market.

51:40And so Apple has been investing heavily in satellite-based communications that keep users connected in places where traditional wireless signals aren't available. Musk's SpaceX, meanwhile, has launched more than 550 satellites that provide cell phone connectivity via its Starlink service. To build capacity, the companies are competing for valuable spectrum rights, airways that carry the signals, which are in limited supply. It's very interesting like the the wireless spectrum when it goes when when there's a block up for sale there's this whole auction it's governed by the fcc brendan carr the new fcc commissioner is the guy who kind of oversees that now um but i don't know if you remember this but back in 2011 2012 tv stopped being over the air like do you remember rabbit ears do you remember those little antenna that you put on a tv that just stopped we had the digital transition tv you could still get tv but it wouldn't be over the air.

52:34And so that block of spectrum was then re-auctioned off for other things. And that's where like the 5G spectrum kind of comes in. And so Apple's outer space investments have drawn Musk's ire. SpaceX pushed federal regulators to stall an Apple-funded satellite expansion effort. The conflict intensified in recent months after SpaceX and its partner T-Mobile sought Apple's cooperation to offer Starlink on iPhones. The company's engaged in tense discussions, intense discussions, and eventually reached an agreement that allows SpaceX and T-Mobile satellite cell phone service, which will debut this summer.

53:09It appears to appear seamlessly on newer iPhones. So even though SpaceX has a deal with T-Mobile, they need Apple to opt in and say, yes, we're going to allow this. We're going to build this. This is going to be a function. It's not enough to just say, oh, SpaceX and T-Mobile agree. You Has Apple ever shared about what its long-term ambitions are? Because I imagine long-term, you'd want to just cut the carriers out completely and just own, you know, basically have another$100 a month sort of add-on. Yeah, it's kind of a match made in heaven with if Musk and Apple can partner up and just cut out the middleman entirely, it'd be amazing.

53:53But again, I think the satellite internet just does not work in buildings. It just does not penetrate. And so we're not anywhere near there. So it's very much like a nice to have add-on. Nothing penetrates my home because I use lead paint. You live in a Faraday gauge. Yeah, I've developed using lead paint to create a Faraday gauge in my home. I mean, a lot of people are doing that. They're worried about microplastics. So they're switching to leaded gasoline. Calling it now lead paint. Lead paint. Coming back in in a big way. Yeah, you don't want the microplastics. Let's just use the lead. for sure.

54:26Go LED. Full LED. Offering greater connectivity in hard to reach areas could spur more iPhone sales. Connectivity. Connectivity. What do you call it? Connectivity. Connectivity. Greater adoption of SpaceX. Sound off in the chat. SpaceX is a Starlink. In some ways, the companies need one another to pursue their respective strategies. So Apple's currently partnered with Global Star, which I think is one of those satellite internet companies that just puts up like a huge washing machine guy. It's like a massive, not in Leo, it's not a constellation like star starlink and global stars very high latency global star is a 2.6 billion dollar company yeah so um so there's a few of these companies astronis is one of them that puts up like a much larger satellite it's locked in geosynchronous orbit yeah there's fewer satellites but they're higher bandwidth and then elon because spacex has all this residual capability they can just throw up starlink satellites whenever they don't have a full ship basically throw it up there, create a whole constellation.

55:24It's a lot lower. They'll burn up a lot faster. But the benefit is that it's a lot lower latency. And so famously, you can game on a plane. You can take a FaceTime call on your private plane if you have Starlink. Whereas with Global Star, if you've used Apple's satellite service, you could definitely not have a conversation on it. It takes forever to upload even a small text message. But it's still very cool and it works. using apple service during the fires la fires because i was trying to help it was this you know it was a crazy time but i was trying to get my buddy's solar panel turned on because he wasn't around to help his mother like and i was using this sort of very it's a very dinky service you got to be around outside yep yeah pointing it around um this is interesting uh you said global star is worth what?

56:172 billion. Yeah. Not a lot. So last year, Apple invested more than 1 billion in global star. Yeah. And so they must be the biggest outside shareholder, unless they invested a really high valuation and just got wiped right now. But, uh, you know, 20 % this, okay. They own 20. Not bad. Uh, Apple service allows that$1 billion I think was like basically services revenue. Sure, sure, sure. Yeah. So invest it in the deal broadly. And of course, the service allows you to send text messages, call emergency assistance, seek roadside help, etc. The new spending is intended to help Globalstar develop a new global satellite fleet to improve space-based iPhone usage.

56:59This technology has already helped save lives, Apple said. These satellite features are designed to complement carrier offerings, giving users even more ways to stay connected. SpaceX recently asked the FCC to dismiss a Global Star application seeking permission to use certain spectrum for the new network of Apple finance satellites document show. It's called the airwaves it uses to carry Apple's users emergency signals an underused resource. SpaceX didn't respond to comment. so SpaceX and T-Mobile revealed their plan to offer always-on texting in remote areas during a rollout event in Texas inside Apple some staff believed SpaceX's announcement of the T-Mobile agreement was designed to front-run Apple's news disclosed soon after that it would provide emergency texting through satellites on some iPhones Musk's Tesla and Apple long ago competed over talent for developer for developing driverless cars the two sides also had spats related to X.

57:58There was that famously, Apple might be pulling out of advertising on X. Elon goes and meets with Tim Cook and then they kind of like figure it out. It was also the question of like, was there a way for Elon to make X a premium service that was based on subscription revenue without paying the 30 % to Apple, which would be obviously very cumbersome for him. The social media company relies on Apple for distribution on its app store and advertising dollars. And then there was also the question of like, if Musk goes too far with the free speech, will Apple kick them out of the app store? That didn't happen.

58:31But Musk at times has considered building a smartphone himself over frustration with how Apple controls distribution of third-party apps, but nothing's really, really happening there. And then he said, Elon said, the idea of making a phone makes me want to die, but if I have to make a phone, I will. I love it. Cracking a market dominated by Apple and Samsung, whose devices account for around 40 % of the global smartphone market would be tough, according to analysts. In orbit, SpaceX through its Starlink operation has far more satellites than Apple, and industry leaders believe Global Star alone won't be able to satisfy the iPhone maker's needs.

59:08Apple has held talks with other satellite providers over the years to secure more spectrum. People familiar with those discussions said it explored investments in the Colorado-based satellite operator EchoStar to provide more satellites and spectrum to support iPhone connectivity, the people said. Apple previously worked on a satellite plan with Boeing and everyone is trying to do deals with everyone in the great space phone race of 2025. It's fascinating because GlobalStar has contracted SpaceX to launch satellites. Of course, yeah. There's no vertically integrated satellite internet or satellite internet provider.

59:47They're all using either SpaceX's launch capacity or separately, and we'll talk about this today, their pure play on launch capacity. So SpaceX is definitely unique in the fact that they make the satellites and then also launch them. And so that's obviously much better for margins and all the different stuff. And again, part of the there's a world in the future where X is not only the everything app. It's your personal AI. It's your cell phone. It has sort of 24-7 connectivity via satellites, and you're hopelessly dependent on it in the way that you are the iPhone. And I'm sure. But can you imagine if your phone only worked on Starlink?

1:00:28So it's like, yeah, you got to go outside. You got to touch grass if you want to scroll the timeline. That would be bullish. That'd be bullish. I think it would be great. It'd be like Pokemon Go. I saw a bunch of people playing Pokemon Go at some meetup in Pasadena, and they were all on their phones. It was very dystopian, actually. But it was cool to see them all walking around a golf course. At least they were getting outside. in the middle of a driving range. Oh yeah. They shut down the whole golf course for an exclusive Pokemon Go event. Can you imagine being the golfer that's just showing up on a Saturday just like, ah, finally get a little time off.

1:00:58It is funny because they only shut down half the golf course. Like one set of the 18 holes was shut down for the event. So you could actually play and then you'd see like, you know, a thousand Pokemon Go players like heads down like, oh, I kind of Charizard. Like I'm very excited. It's great. Anyway, if you're traveling a lot and you have a Starlink and you want to hang out, why don't you find your happy place at wander.com. Find your happy place. Find your happy place. Book a wander with inspiring views, hotel-grade amenities, dreamy beds, top-tier cleaning, and 24-7 concierge services. It's a vacation home by better folks.

1:01:34I got to say, go follow Kyle Tibbetts on X. Oh, yeah, he's putting up a historical run. At Kyle, K-Y-L-E-T-I-B-B-I-T-T-S. He posts some of the best wanders every single day. He's their chief marketing officer. Yeah, he's their CMO. He's a friend of ours. And he just posted one a bit earlier called Malibu Ridge that just looks 10 out of 10. Iconic. We got to get over there. Shirts close to my house. And, yeah, we got some shoots coming up. We got some shoots in the pipeline. We're going to be leveraging some wanders. I love it. I'm excited about that. Well, should we move on to Isomorphic Labs?

1:02:17Yes. $600 million for AI drug development. They were backed by Google. Thank you for the size gong. Demis Hassabis, the CEO of Isomorphic Labs, of course, has been at Google for years. He founded DeepMind and famously beat Lisa Dahl in Go with his AlphaGo product and team. Fascinating documentary, if you haven't seen it, all about AlphaGo. Move 37 really shook Lisa Dahl to his core. One of the greatest moments in AI history, basically. But we'll take you through it. It's very interesting. It's very early. Everyone kind of expects AI will help in drug design. drug development could help us live forever.

1:03:01Like really, really amazing applications. But we haven't seen AI really move the needle or move the biotech markets yet. So still early, but Thrive Capital is getting in early as you should if you believe that this is the future. So they raised$600 million in the first external funding round to boost artificial intelligence drug design pipeline across multiple therapeutic areas. And so they're taking an ensemble approach, they're really looking at not just one drug, which is the kind of the normal biotech strategy. Even a company like Moderna, before they got into mRNA vaccines, they were focused on one specific biotech technology.

1:03:40And a lot of companies will go public as biotech companies saying like, we're an Alzheimer's company. We have a drug that's targeting Alzheimer's, not isomorphic labs. Isomorphic wants to work across multiple therapeutic areas. And so this is actually a subsidiary of Alphabet. Started four years ago. Thrive Capital led the funding round with participation from Google Ventures, but they're potentially spinning it out. And we've seen this with Waymo too, where Waymo is starting to take external funding and potentially spinning out as well. This company is based in London, and they'll receive follow-on capital from the existing investor.

1:04:14It's so fascinating that... that Google allows for this sort of massive external investment in companies that are, you know, They're effectively acting as some type of like venture studio, but you'd think with how much cash they produce, they would just want to own all of these. But sometimes having an external, you know, board members can can help. Yeah. I mean, I think that there's something about maybe this is a kind of an ad, like an admission by Google that says like, hey, like these companies, these ideas are great, but they can't really thrive when they're 100 percent owned by Google. Like we just need the, yeah, it's almost that external pressure to external pressure.

1:05:00Yeah. It's like, um, who are we talking to about this? Where, um, I think maybe Delian was, was talking about this, where it can be very tricky when you have, uh, when you're a company and you have like a billionaire patron who's just kind of paying payroll off their own balance sheet and they can kind of bail you out forever, but there's no pressure. and the venture model of like, hey, you raised enough money to get to the next milestone. You have 12 to 18 months. Maybe you stretch that to 24. Maybe it goes somewhere else. But like you have to deliver some sort of milestone and get on, you're off the research track.

1:05:36This is not go spend 20 years and figure out if you can create a grand unified theory of the universe. This is can you commercialize? Can you make really, really solid traction to justify the next round? And that treadmill has worked in technology development for decades. And so I think Google's maybe just admitting that. There's also the question of just, how do you incentivize someone who's at a subsidiary financially? Like if you go and take - And I wonder what these cap tables look like, right? Because you have to imagine Google - 100 % Google, it's at the start, right? Yeah, but post-investment, you've got the founders got to have some direct exposure, I would imagine a team that's now working on it would want exposure, you know, who knows?

1:06:25Yeah. I mean, I would imagine that Google outside of GV and Google Ventures probably owns 50 % of the company still or something like that. They probably gave a lot to employees and founders, but, uh, they, you know, it was definitely incubated there and built there. And so they, they deserve a large chunk of the equity, but that will probably get diluted down over time as the company does more investment rounds. So the company will use the investment to advance AI research and development and expand its team. Demis said, this funding will help further turbocharge the development of our next generation AI drug design engine.

1:07:01And of course, while at DeepMind, Demis worked on the team that's created AlphaFold, solved the protein folding problem. That was an interesting development because everyone thought that protein folding was extremely resistant to computational efforts, that it was not going to be solved. They solved it very quickly. But also when they did solve it, it didn't really move the biotech markets in an interesting way. Like you would expect, it's like, hey, like the thing that's really hard just got solved. It's now trivial. Like you would expect all of the companies to kind of get a little bump, but nothing happened in the market because it turns out that protein folding, while it is important in drug development, It is not a major cost center.

1:07:44Yes, you need to use this. I think it's like an x-ray crystallography machine or something. You need to use some really expensive equipment. But at the end of the day, you kind of just pay a PhD to do it. And they're kind of a starving student. And it's not that big of a barrier to drug development, even though it is from like a philosophical and mathematical perspective, like a very, very difficult problem. And it's very impressive that they solved it. it didn't really unlock it wasn't like oh yeah protein folding alpha fold came out and then we got a new cancer drug the next day like that's that we were not that limited on it um and so uh this is the interesting quote demis says his mission is one day solving all disease with help of ai yeah he's coming for it all yeah not just one area and uh well open ai's gpus burn due to the Ghibli usage.

1:08:37Google's TPU is on fire trying to further Brian Johnson's goals of living forever. They already have partnerships with Eli Lilly and Novartis for drug design and development through AI. And they put out a press release. They talked a little bit more about this. They're bringing together top tier investor groups with deep AI and life sciences expertise. They want to transform the industry through an interdisciplinary approach. It feels like this press release was maybe written by Gemini 2.5. Who knows? Press releases just sound like boring all the time. I wish he'd just done a thread. He did post about it, but it was kind of like, okay, yeah, we get it.

1:09:23You're bullish. There was an interesting post I saw by Zachary Spiro here. He says, this means isomorphic has raised somewhere between 12 to 40 X the average UK funding round for companies at this stage. Even at that stage? Yeah. Well, how's he describing stage? Probably growth equity or like series B or something like that. I don't know. Because if this is the first, is this the first external round? It is the first external round, but yeah. 40 times the average pre-season. No, he's clearly saying it's more of a growth round. But this is the interesting benchmark. It's so large, it's equivalent to more than a quarter of all UK life science VC investment in 2024.

1:10:04So last year, all UK life science VC companies, VC-backed companies raised just four times what they just raised, like 2 billion. Thrive goes to London for one trip and they're like, we are venture capital in the UK now. Basically, but it makes sense. It's like Demis has had an incredible impact on the industry. He's built DeepMind, which became Google's like fantastic AI research organization. Google at one point had Google brain and then also deep mind and they wound up merging them because it was too much. And they went to research orgs that were kind of going back and forth. But he's a legend in the industry and he's, and he really has again and again and again, just created.

1:10:47He's an absolute dog. He's an absolute dog. He's an absolute dog. He really is like, I mean, solved go, mu zero, another legendary AI system. Same thing with, I mean, yeah, it's like Google's been doing an incredible amount of work on the research side. Just, I think that the productization has been a little bit slow because a lot of these organizations within Google, they feel like modern interpretations of the academic lab. They're very much just, hey, you have unlimited money, unlimited time. go work on science and research and don't even think about how this will actually make money. And that's produced some incredible results.

1:11:33But at the same time, it hasn't really produced incredible products or businesses or durable value. And so I'm in support of them spinning it out. But... Yeah, we don't want these moonshot projects just living within Google without any sort of urgency to actually deliver commercial results. Yeah, I agree. But until Demis solves all disease, I think the best thing that you can do for your health is just get some good sleep. Go to 8sleep.com. Nights that fuel your best days. And Tyler called out on X that you can use your TrueMed funds to buy your 8sleep. Yeah, shout out Justin. You can basically buy your 8sleep pre-tax.

1:12:23Yep. So if you have a FSA, HSA account, you got some money, burn in a hole in your pocket, head over to 8sleep.com and check out with TrueMed. Can't recommend it enough. Just do it. And a lot of these companies, they have FSA, HSA programs. You can just start accruing it, pay it down. There you go. It's all good. How'd you do last night? I imagine I did well. Let's check. i i i to be honest i was watching white lotus 98 and i don't recommend night lotus is night lotus night lotus never night lotus day lotus only uh white lotus before bed not a not a great combo i only got six hours and 42 minutes of sleep last night i got 7 24 so much of sleep is just getting and really getting the time in.

1:13:14Yeah, you got to put in the time. And once you're getting the time in, sleep on your pod. Yeah. Not in the pod. You need a good morning routine. You need a good nighttime routine. You got to go to bed. You got to start your bedtime routine at 5.30 p.m. I need a keg of Saratoga water to just do a quick keg stand. There you go. Keg stand in the evening, Saratoga water. Yeah, you're good. Sparkling. Well, you can do it with Aurora. Aurora is a keg. you should flip that bad boy over we should you're a huge keg stand guy i'm a huge keg you're bringing keg stands back yeah yeah we need to take back tech bro that was a great take back meathead that was a great post earlier it was fine maybe it was last night i think it was last but i missed it john said uh if you're trying to avoid microplastics don't drink beer out of those red plastic solo cuffs stick to keg stands stick to keg stands and that only got about 25 percent of the likes of your other posts which was the heat death of the universe is unacceptable we need to address entropy in a meaningful way within the next 10 to 100 years at gary tan so hopefully is that doing well it's uh it's over 400 okay we're good we're going with that one yeah the at gary tan one yeah if any any problems that you have right now yeah just tag gt this This, of course, comes from like a pretty serious post, Deep Prasad, who I've actually met years ago.

1:14:39Nice guy. He said last night around 3.30 to 4 a.m. Okay, that's your first mistake, Deep. You should have been sleeping. But this is crazy. He said he was shot at twice near his hotel in San Francisco. Very scary. He said they chased me to my hotel and shot the second time when I ran inside and got one of the staff to come out. We both heard a Glock and he saw the gun while I saw the bullet impacts. I'm still shaking. very sorry to hear that i hope you're okay that is very rough and it was no laughing matter until we found out that you were safe and then at the end of your post you just added gary tan like hey help me and uh so people had a lot of fun with this it turned into a meme glad that you're safe um but we're gonna meme i mean and she's safe new meme format says i'm sorry that happened but what is gary going to do and there is a little bit of like gary tan has just become like batman of san francisco you just like put up the bad signal anytime anything bad happens uh and so we had a lot of fun with this uh jordy put up a a similar post in the format talking about uh the difference between east coast and west coast time and how the east coast has a three-hour advantage on us and gary tan should fix it for us it's absolutely brutal it's absolutely perpetual three-hour head start on the day forever it should we should flip it we should flip it every six months it should flip yeah that wouldn't cause chaos at all in uh in sports you know you don't let one you know team just sort of like pick their side and china one time zone yeah face russia 13 time zones yeah so long and it's funny because it's like they're both like pretty big countries but it's clear that like russia was just like let's have as many times as possible and china was like let's have as few as possible um and so yeah i think if you're on like the i think if you're in the western side of china like the sun sets like really early yeah something something odd happens but you got to deal with it because they're not putting up an east-west rivalry it's one team well i got i got some breaking news okay hit me we have a new partner oh yeah uh if you look in the ticker right now, you'll see a lovely logo for Numeral.

1:16:56So shout out to Nate, Sam, and Matt over at Numeral. So break it down. You're about to know Numeral puts sales tax on autopilot for leading e-commerce and SaaS businesses. It is the platform for sales tax compliance, John. You already know that. They already work with a thousand plus different e-commerce and SaaS businesses. They power uh ridge wallet our friends sean and connor very cool they also power graza which is a company uh that is absolutely crushing uh it's in the hayes portfolio it's in the hayes family portfolio i think it's like a clean they got their own ticker on here uh they got their own ticker too so they're they're ticked up they're ticked up we love to see it you gotta scroll down fantastic our ticker doesn't block their ticker let's look at their benchmark series they got great teams.

1:17:45Uh, very cool. Uh, and, uh, I mean two things like, like sales tax is such a hassle, uh, because it's, there's just so many sales tax jurisdictions. It's like the perfect thing for automation and software. Um, I, I was definitely afraid cause I just yelled out benchmark series and I was like, Oh wait, did, did that not get announced yet? It did get it. it got uh it got announced uh congrats to the team um so we love a bunch of series and i'm very excited to be uh sam the ceo yeah splits is is spends time in la cool yeah we'll have lunch with them as soon as we're back on do you remember the golden age of e-commerce when you just didn't have to pay sales tax and amazon was just like yeah if you're not in seattle like we're a seattle company like we don't have nexus in california yeah and they just wouldn't and it it was this crazy arbitrage because you could buy things and you just wouldn't pay tax if you bought it online.

1:18:42And then eventually, Amazon was like, okay, yeah, we actually have Nexus everywhere and we got to pay tax everywhere. And it got a lot less fun. But now that it is a lot less fun, the fun's coming back because of sales tax automation. So what's the call to action? Go to numeralhq.com. numeralhq.com. Yeah, we're pumped to have them on the show. It's great. We are eventually going to sell some merch. And we will be running on numeral. That's great. Guillermo, friend of the show, former guest of the show, has a post here. There's no better investment in the world than America. It's people, it's culture, it's land, it's companies, it's freedom.

1:19:21Leaders in AI, there's no second place. I'd buy every stock and every square foot. It's only been in the blink of an eye since 1776. And the best is yet to come. Let's hear it for America. Let's hear it for America, folks. Thank you, Guillermo. John, what happened into the American flag that we used to have on the set. Did you take it for your bedroom? I don't know. I was thinking we need a massive American flag somewhere on the set. Yeah. Uh, well, we're working on a new studio folks. I'm sure we'll, maybe we should just paint the whole wall American flag. Uh, I'm also seriously looking at a cyber truck that's wrapped with the American flag and throwing a sticker on there.

1:19:57I bought this before Elon went crazy and see if people fall for it. Just the most aggressive screaming Eagle and be like, yeah, yeah. I bought it before or maybe you throw the i bought this after elon went crazy um i kind of want to do a copy pasta of this guillermo post that says there's no better investment in the world than versell it's culture it's land it's companies it's freedom leaders in ai there's no second place i'd buy every stock and every square foot it's only been a blink of an eye since 2014 when they incorporated, when Guillermo launched Next.js. Isn't that the one he launched?

1:20:39Very funny. Anyway, if you're looking to get in on America, on the action in America, buying every stock in America, go to public.com. To invest in America. Investing for those who take it seriously. Multi-asset investing, industry-leading yields, and they're trusted by millions, folks. Anyway, let's move on to Tom Bloomfield over at Y Combinator. He says, software engineers are highly paid farmers tending their crops by hand. We just invented the combine harvester. The world is going to have a lot more food and a lot fewer farmers in very short order. I think this is true of doctors and lawyers too, but maybe a year or two later, we're going to have an abundance of incredibly high quality knowledge work at very low prices.

1:21:19Instead of living off gruel, everyone is going to have bananas and kiwi fruit for pennies. the farmer but the farmers will need something to find something else to do with their time it's actually so wild that you can take a you can you can take any type of media whether it's text screenshot pdf image etc put it into any of these models and get and say pretend to be this and give me an analysis of this content. And then it just spits out a really robust analysis from the lens and you're effectively getting to hire a lawyer and intelligence operative, a doctor, et cetera, to do this sort of pretty in-depth analysis for effectively zero dollars.

1:22:11And that is still just dramatically underhyped. This is potentially a very stupid take, but I haven't seen it. And I wonder if there's something good here. Like we talked about Jevons paradox where deep seek drops the cost of inference and paradoxically the overall demand goes up, right? That's Jevons paradox. Now I've always wondered like, is Jevons paradox just price elasticity? Because if a product is very price elastic, like a price inelastic good is, is cigarettes. Like the price goes up, people buy the same amount, but with a price inelastic good, inspire myself to grind harder. course. But yeah, well, that's a Veblen good.

1:22:52And recently I said we should put 100 % tariffs on Veblen goods because Veblen goods are goods that the demand increases when the price goes up. So that's something like a Patek Philippe is a Veblen good because you're like, oh, it's a$200 ,000 one. I want to buy it more. I was joking about the auto tariffs being actually good for the economy because okay, if, you know, whatever, if a Daytona SP3 costs 25 % more, I'm going to work harder. I'm simply going to grind harder. And you're joking, but that's actually the economic analysis of what a Veblen good is. That's exactly how an economist would describe how a Veblen good works, which is why I think that post is funny.

1:23:34But what's interesting is like, if you have price inelasticity for AI inference, as the price goes down, if it's truly very price inelastic or price elastic, what will happen is your demand increases disproportionately to the price. And so you're actually selling more at the lower price. And so your overall revenue goes up. And that's what Jevons Paradox predicts about GPUs and new technologies. But is there a Jevons Paradox for software? And I think that's what we're hearing from a lot of folks in the industry who say, hey, yeah, it's going to be much easier to write code, but there's a lot more code to write.

1:24:15And so what's interesting is like, it's possible that the software engineering industry actually grows, even though the price of software engineering declines and you have a sort of Jevons paradox, but we'll have to figure it out. We certainly see that with Studio Ghibli's, like someone was chirping at Mike Tyson for posting a Studio Ghibli photo of him with a pigeon. And somebody was like, you're a millionaire. Like, why aren't you paying an artist to do this? Oh, I saw this. And the thing was like, look, like he was never going to pay an artist to do that because like he doesn't have the time to hire.

1:24:53He's not thinking about that. That's not something he was doing. Somebody quoted that and said something to the effect of, hey I'm looking for an artist to make me a picture of Mike Tyson you know with a pigeon and my budget is zero dollars yeah exactly let me know if anybody let me know if anybody's interested exactly and so and so you have this weird this weird paradox where where it's possible that like demand for for art and demand for engineering it all increases and it's kind of unclear what that does to jobs like you would imagine there's fewer software engineers like what tom bloomfield here is saying but uh we could see a lot more code a lot more doctoring and lawyering and we certainly can imagine that with with legal where all of a sudden like everyone's suing each other constantly because it's all ai powered and it's gonna get very weird um but yeah but uh yeah it's one of those things i think i do think more legal work will happen when you have these sort of chat GPT like I mean people are already using chat GPT to generate legal docs I think it's or dump in legal docs and usually those types of people are never going to hire a traditional lawyer yeah but then there's some intersection of people that would have hired a lawyer but now we'll just pay yeah you know or they're like I pay 200 bucks a month hey chat GPT uh draft a restraining order for a podcaster who stole my idea uh let's get using that 50 times a day aaron the ceo of box is coming on the show in just a few minutes i want to read his post because nope he's actually coming on at 12 45 we got yes at 15 oh sorry 15 minutes sorry um but i want to give some context about what i'm excited to talk to him about because he has been very rigorous in his evals.

1:26:47And I'm sure Avi Schiffman from friend.com is just kicking himself reading this post because Aaron's saying that Gemini 2.5 Pro is a big step towards AI being useful in the enterprise. So Box has an AI enterprise eval, an evaluation and gemini 2.5 pro scored three points better overall than gemini 2.0 pro for accurate single shot data extraction from enterprise docks this entails asking the model to pull out many fields at once from the dock and the jump is actually bigger specifically in areas where it must work through a complicated through complicated logic in a dock obviously this is extremely helpful for box you have a ton of documents in your enterprise uh kind of you know, document storage warehouse.

1:27:37And all of a sudden you can just query across everything. No, it doesn't matter if it's a PDF or Excel sheet, it's going to pull it all out, which is awesome. It performs much better when tasks require more time thinking and not just jumping to the first answer it assumes is right, such as deducing dates in a contract when you have to reason through multiple sections in the content to come up with the correct answer. Accuracy and data extraction is obviously insanely critical for enterprise AI use cases. You can't get data wrong when working with enterprise information for a mission critical use case.

1:28:06This is the difference between being able to deploy AI or not in financial services, healthcare, life sciences, and many other places. It's amazing to see the continued performance and capability breakthroughs happening in AI, but it only seems to be accelerating. So what I'm interested in is like Gemini 2.5 Pro, like it's a Google product. Google has Google Drive. Box obviously competes with Google Drive. He's going to be probably offering this as a product. How does that deal get done? What are the risks to partnering versus building your own foundation model? Is AI and data extraction a sustaining innovation or a disruptive innovation?

1:28:46Box is unique in that they're still a founder mode company. Aaron's still the CEO. He founded the company. And so you would imagine that he'd be able to move faster on product. And that could be a pretty big differentiator in the short term. So I'm excited to hear what he has to say about all that um and you know if he gets it really working he should run some more ads on ad quick out of home advertising made that's right measurable say goodbye to the headaches of out of home advertising only ad quick combines technology out of home expertise and data to enable efficient seamless ad buying across the globe um we have someone who is on the cap table of ad quick who alexis ohanian let's go the man himself and he's coming he's coming on with rj right now from uh see if he's here let's go and this is maybe our second time having two guests we're still figuring it out we might do a four up we might just see one person a little scroll back and forth depending on who's talking but we always love having alexis on the show uh he's got so much to talk about so much history and he's always cooking he's always cooking he's a chef and hey how you doing RJ hey guys I wasn't sure if you could see me or not not gonna lie how you doing I'm good guys how are you we're good can you start with just like a little bit intro yourself your company what you're working on yeah yeah really quickly I'm RJ I'm the CEO and co-founder of Playback I can go more detail in my background which is really high level I actually have a little bit of a non-traditional background for a startup founder.

1:30:21I was a hedge fund analyst at an activist hedge fund based here in San Francisco called Value Act Capital. So kind of investing in large public companies. In terms of playback, look, the idea is pretty simple. We're trying to create a better way to watch sports by integrating creators and communities directly into the content itself. In practice, what that looks like is it looks like a little bit of a combination of like a traditional NBA or MLB broadcast with a Twitch creator stream overlaid on top of it. So creators can basically add their own voice, own commentary, and most importantly, they can actually interact with an audience while they watch these games.

1:30:59And what we found is that that's just like a way better way to watch sports. And then I'm sure we can get into more detail on this stuff, but we had a couple of big announcements last week in terms of obviously adding Alexis and 776 to the team and and a new partnership with MLB. But yeah, that's the high level. Dude, I loved it right away for a few reasons. One, we basically are doing for technology and business news what you're doing for sports in a way where we don't do any journalism. We're basically sort of providing sort of this like commentary. Reactions. Reactions to the news. And I think what you tapped into is basically, you know, just an existing behavior, which is so many people are sort of consuming sports content in this sort of bifurcated way where they're getting their commentary from YouTube or Twitch and then they're getting the broadcasts, you know, through these sort of legacy providers and like sort of putting that in the same place just makes a lot of sense.

1:31:53Because I don't watch a ton of, I don't watch a ton of sports, but if I want to like get a recap on F1, I'm going to YouTube. I'm not turning on the sort of like post-race reaction. And a lot of those, yeah, I watch a lot of those like F1 reactions and they can barely even use photos. They have to go to Getty Images, license them. They can never use the real footage. And I'm just like, as a viewer, I want everyone to be able to use the footage in the right ways. And it just requires figuring out all the licensing and stuff. I remember seeing on Twitch, someone was watching a boxing match that was clearly pay-per-view.

1:32:28You've probably seen this. And they're making the controller. And they're like, don't worry. It's just a photo real video game I'm playing. I'm just playing the photo real video game. dodging and diving too. Yeah, yeah, exactly. They're like getting really into it. And it's like, clearly they're just, they're just like pirating the pay-per-view stream for their viewers. And the viewers are just like ignoring that guy and just watching the pay-per-view stream. Talk about how the licensing and the rights work for all this stuff. So just first off, totally agree with everything you guys are saying.

1:32:59Like it's one of those things where to me, like one of the internal maxims we use is like, you know, can the next Stephen A. Smith just be an independent creator on playback. Like they don't need to go work for ESPN. They don't need to go work for Fox or one of the big traditional media companies. And I think the only thing that's held back like independent sports creators is the lack of content and lack of rights. Like it's just been too hard to do it. So one of our goals was to kind of mesh those two things together. The way we've done it is actually pretty interesting. So essentially, well, this is true.

1:33:26And I think you guys know this, but we have partnerships now with NBA and NBA League Pass and MLB and MLB TV. And the way those partnerships actually work is MLB TV and NBA LeakPass are just apps on playback. So we kind of created this like bifurcated system where there's like an app layer where these like, you know, content companies and streaming services can put their content, still deliver it the exact same way in a lot of ways, still get all that user data, still have the same subscriber relationship. But then we can overlay on top of it, this creator experience, because sort of everyone agrees that this is a good idea.

1:33:56It's just really hard to actually fit it into the existing rights ecosystem. And the way we've been able to do this with this app layer just makes it a lot easier. It was still a lot of work to get the partnerships over the line. Like our MLB partnership went live on Thursday. We first agreed to that partnership in July of 2024 and had actually been negotiating and talking about it for like nine months prior to that. So it's still a long process, but I think the sort of structure we've set up here is pretty scalable and it can hopefully work with anyone basically. That's awesome. Amazing. Well, natural fit to partner with Alexis, given that he not only is a sports fan, but he's creating new leagues.

1:34:35Obviously, I wanted to give you an opportunity to talk about Athlos. But maybe before we get into that, I'd love to kind of get your take on it, Alexis, and what kind of got you over the line to partner up in a big way. dude i had a tweet from six years ago ranting about it only being a matter of time for someone to build the product because there is a lot of product that gets done so the syncing and everything works well and the partnerships and took a minute but thankfully rj and the team did it and and the only reason this one was an easy one to spot even six years ago is i think for anyone who spent time in gaming we watched the ascent of twitch you know shout out to justin and Emmett, who found that one of the most popular things people would want to watch people watch was video games, eSports, whether it was competitive or not.

1:35:27And what are games other than, you know, whether it's sporting events or League of Legends, it's abstracted away, it's the same damn thing. But what gets you in the way is the rights deal. And that was obviously not going to be solved easily. But I first got into sports. There's actually one of my first investments after leaving Reddit in 2020 and splitting initialized up was launching Angel City FC. Now the most viable women's soccer team in the world or sorry, women's professional team in the world. But what I found getting into sports was here was this world, especially in the emerging side where folks didn't really have any of the media dollars at stake.

1:36:02So they could move a lot faster with innovation because, you know, you weren't looking at billion dollar rights deals. You could just assess from first principles. And I think that's why we've had the success we've had in emerging sports. And then what it was going to take was really seeing the uptick of this new generation of sports fan. That's the reason Angel City has been so successful. We're talking to a very different sports fan than the traditional one. It's a fan that lives here, right, that watches on their terms. It's a younger fan. And so now, thanks to the success of emerging leagues, the big four can look and say, OK, like maybe we need to reach out to fans here and technology is the way to do it.

1:36:38It just took time, but the market's ready now and RJ delivered. Yeah. And talk about kind of with with playback, you have an opportunity to truly sort of productize sports culture in a way that I feel like Twitch early on, which is very much like video game culture. Like the entire thing was built around that. How are you approaching kind of like bringing sports culture into the kind of the product layer? Yeah, it's a good question. Honestly, the easiest way we've kind of done this is I think a hallmark of sports culture is that people like to get their takes off. Like they're very interactive and active in their fandom.

1:37:18And it's honestly like the biggest dividing line between what a 24-year-old fan looks like versus a 48-year-old fan is like, you know, a 48-year-old fan leans back, turns on the TV and just like relaxes and probably still yells at the TV. But a 24-year-old fan actually wants to yell and be heard while they're yelling. And so one of the first tweaks we made to the product is we actually built it originally in sort of the Twitch style of like a one to many. Like it's just a creator basically talking. And what we did is we switched that from like a one to many to actually a stage concept. So creators can actually bring people from the audience up on stage to chat with them, debate with them, do whatever.

1:37:53And like our best users are like very frequent stage requesters or even joiners. Like they're always trying to kind of get their takes off. and I think that's kind of a microcosm for how we think about why we're focused on sports is like if we can continue to tailor it to this user and this use case like we're gonna have a much better chance of like carving out our niche and kind of winning the use case uh but yeah there's lots of other little things but hopefully that gives you a good at least early single yeah do you see it you know maybe it's happening today already but do you see former pro athletes sort of like retiring in the same way they do now saying I'm going to start a podcast you know once once they're out of the league you know i'm going to get on playback and i'm going to just start being basically a super fan providing that commentary uh yeah is that part of the vision our biggest creator is gilbert arenas so very classic example of that and i've also been very successful as like an independent content creator he has a show called gills arena it's very similar on us to what you guys are doing live on youtube four times a week um been super successful in that venue uh and has transitioned into a playback creator too he's live you know four times a week with us at night and will be basically seven times a week for the playoffs on.

1:39:03And he's been super successful because I think he has, like, the right blend of, like, he has a content creator mindset. Like, he can actually think, like, what's going to go viral, what's interesting to people, what do people want to talk about, like, what's just basically fun. But he's also an X player, so people sort of naturally want to hear what he has to say. But, yeah, he's been super successful. I can send you guys a clip. He had a really fun clip that went very, very viral this past week when Giddy hit the game winner, basically, for Lakers versus Bulls. That was really fun. But yeah, I can follow up with that.

1:39:33Can you talk a little bit about monetization on your end? Obviously, I'm paying the MLB through their app. You're sitting on top of that. It makes a lot of sense that if I'm following a creator and I'm tipping them, that money just flows through. You take a cut of that. But on Twitch, they're actually running ads as well. I imagine that the MLB would be pretty upset if you just start running ads on top of their ads and kind of disintermediating them there? Is there some sort of partnership? Like you obviously as a digital platform have even more information, you can target better. There's a lot of different things.

1:40:04How are you thinking about monetization over the long-term? Yeah, so short-term, it's just focused on the creator monetization that you referenced. Like I think we have a long runway and we have great investors to kind of build to the long-term, but yeah, near-term, it's all about basically these independent sports creators. Can we actually help them build a business? So obviously tipping is one category, but even just room subscriptions is what we call them, like the ability to offer an actual subscription to their audience for either like limited interactive interactivity or even like premium streams.

1:40:31That's the primary focus right now. Long term, though, I think there's sort of two options. One is like there is already a proven model around like advertising sponsorships in partnership with the leagues. Right. So the leagues do have these partnerships with Amazon and YouTube and other like obviously fully scaled platforms to deliver ads for them. and then a healthy cut goes to the partner basically. So that's one category. The other category is like, I kind of think once we're big enough, there's going to be a really big opportunity for what we call like platform revenue, which is basically like some percent of our, of the subscribers we drive, obviously to these partners, we should earn a percent of that.

1:41:07But also there's going to be new services that we can kind of offer. So the two obvious ones that come to mind are e-commerce and sports betting, like being able to bet from within this experience, being able to buy a ticket from within this experience, Like all of these are going to be sort of like high quality opportunities. And we expect to both enable that and make money from that, if that makes sense. Yeah, for me, it's kind of silly right now that you have a creator who's providing reactions and commentary to sports on one platform and saying, use my code and go to PrizePix or go to, you know, Underdog or Joey and Jake's, you know, platform like better, better.

1:41:47uh like it's very silly that these are just happening in one place where a sports fan wants to like post up they're just like fully immersed in the game the commentary and whatever sort of activity they're doing on top of that so i mean maybe merging a social network and an llm foundation model company as well yeah just merge it all yeah yeah lbo a big quarter trillion i i see a quarter trillion dollar opportunity here yeah let's throw it all together um you got last uh we got a minute left alexis tell us i i wanted the the update on athlos uh like there was some stuff going on over the weekend yeah the greatest show in track uh we started this women's track league last year three million people tuned in for it i was the biggest person in the history of the sport fastest ladies in the world and it was a blast uh we're doing it again october 10th in new york and of course we're gonna be streaming it on playback yeah i was gonna say are you gonna have are you going to be providing commentary you don't want to hear my commentary but there is to your point though right this is why i love the idea of building in a new in this case building a brand new league but even in emerging sports is is you can re-underwrite everything from first principles right we can say because we don't have a billion dollar streaming deal media deal yet we can say okay what's the best thing for fans and what's wild is there is a long long long tale of creators who love this sport or even just mildly curious about the sport who are going to be able to bring their audience to watch it and i really believe if you haven't watched john boy like i'm not an mlb guy i'm not a baseball guy but i fell in love with this sport through this random creator happens to be a nick uh not a nicks fan a yankees fan nobody's perfect but he's a big yankees fan and does these voiceovers of baseball that delight even a non-fan like it's just so engaging and to me that that is the future of sports and it just seems kind of silly that we've had all this talent emerge from streamers talking about video games but we've never seen the new john madden or the next pat mcgafee and it's going to happen here imminently in these next years love it guys thank you for jumping on uh very excited i'm going to hit the size gong for you and uh we look forward to tuning in to some playbacks in the near future and uh hope you guys crush the rest of your monday yeah good luck thank you guys thank you cheers very cool it's crazy that that do people bet on running a lot yet horse betting's huge why is betting on you know people like these personalities track stars why is this not you should be able to do a whole track day just you head to the track get some get some i mean it's way i mean i respect horses.

1:44:27We love horses. I went to the Kentucky Derby. Why can't I just go to, you know, the human Derby? The Derb. The Derby. The Derb. How are we doing? Do we have our next guest? Not yet. Hopefully coming in soon. I think that's very cool. It's interesting. I mean, you have to wonder if Twitch is thinking about that. They're so big. Amazon has the rights to a bunch of stuff. It seems so logical, but man, I mean, it's been a decade and they haven't done it. I mean, part of it is it's great for it's great for business if your competitor is owned by am uh amazon yeah exactly and not that they can't innovate at all but i mean for a long time emma cheer was the ceo of twitch and was running essentially an independent organization from amazon but of course he stepped down and now twitch has been much more absorbed in the amazon ecosystem uh it is a fascinating model have you have you seen on twitch how uh the twitch prime model how if you have Amazon Prime, you get one free sub, subscription to a creator.

1:45:29And so most people who use Twitch, they really have one favorite person that they tune into every time. It's not really like YouTube where you're bouncing around watching 25 different creators. And so you go in and you get a free one and so they would give$5 a month and then the Twitch Prime I think delivered maybe$2. Only like the margin went through. But still, you would have these creators with like huge organizations uh what are you looking for water i was looking for a year but i thought you had two oh no we got aaron in the chat here bring him in there we go how you doing hey hey guys how's it going uh we're doing great we're doing great finally have you you should have been on like you know we've had like 50 plus guests yeah you should have been on the first one why wasn't i a launch partner for this thing i know seriously yeah oh well we're making it up for today okay great another podcast awesome good yes yeah yeah yeah okay that was the genesis right The world didn't need another podcast.

1:46:25We said technology needed a podcast. That was the idea for sure. I want to talk about the enterprise AI eval and how the different LLMs are interacting at Box. Is this a sustaining innovation? Is this a disruptive innovation? Is Google a rival? You're gassing them up. What's going on? There were like nine questions in there. Where do you want me to start? Wherever you want. Yeah, I mean, we're pretty pumped about AI as you can imagine. So we've been building this platform to help enterprises store their data for a couple of decades now. And we have hundreds of billions of files within Vox. And the one challenge with content, financial documents and resumes and marketing assets and contracts is you create it, you share it, and then you kind of never look at it again.

1:47:14And so what happens is you see this sort of curve of people, content remains hot for a day or two, and then it kind of goes into an archive state. And AI is the first thing that lets you actually tap into all of the value of all of your data, no matter how old it is and when it was created and who it was shared with and where it is. So for us, it's this massive breakthrough, which is you can turn all of this unstructured data into effectively usable knowledge for an organization. And so we're kind of betting the whole company on this idea of in the future, when you work with your content, you're going to have AI agents that go out and read your documents for you, create analysis, go through all your contracts, find exactly the data you're looking for, automate workflows across any business process.

1:48:00So that's effectively why we're all in. And so we're trying to think through what are the implications in the future of enterprise software. I'm very much on the side right now that it's a TAM expansion of software as opposed to kind of a TAM compression. I think this lets enterprise software, you know, startups or incumbents go after much bigger markets because you're now tapping into the, you know, just a completely different budget pool in an organization. But yeah, a lot's happening at the moment. What about training a foundation model? Did you consider it? It seems like, can you talk a little bit about value accrual versus foundation model layer versus application layer?

1:48:40We thought about it for like six and a half minutes. That's a long time in 2025. Exactly. So back in, you know, the moment we saw ChatGBT, like there was like one brainstorm of like, oh, do we need to like fine tune our own models or even train our own models on certain data sets? And we basically quickly concluded like, no, that doesn't make any sense. And the math is really clear. It's like, if you have Google and XAI, we didn't, you know, didn't have XAI, at that point, but you had OpenAI, Anthropic, Google, Meta at a minimum, you just don't want to be fighting that war from a capital standpoint or even just a deep research talent standpoint.

1:49:18So you always want to be as a startup or any size company, you kind of want to be on the side of the tailwinds of the market. And the tailwinds in this case are that there's like now five or six at-scale companies that are deploying hundreds of billions of dollars on computers and training data and talent. So you want to be riding that wave as much as possible. And so we've built an architecture that lets us tap into any of the AI models, you know, whoever makes them. Great. Jory? You had quoted somebody, I think it was a couple of days ago, that basically delivered a pretty good prompt about a paragraph of text describing a website that they wanted.

1:49:56And it just sort of pops out this beautiful, you know, landing page. And the designer said, we are so cooked. that was a reaction for a lot of people last week maybe that was reaction if you're a studio i'm kidding uh but but you said here's how this plays out ai makes producing amazing designs more affordable and faster designers get to offload the long tail work iterate faster and serve more clients or projects customers get better results faster or cheaper inducing more demand for great design it's basically calling jevin's paradox paradox on design uh you is there any like context that you feel like is worth adding there?

1:50:29Because I think they're generally, you see these sort of big, exciting releases or updates to models. And there's like a lot of sort of doom. But if I was in high school and I enjoyed design, I would be calling up every single SMB. And I wouldn't even, what I wouldn't do is say, hey, will you pay for this website? I would say, I designed this website for you already. And you can have it for whatever. a hundred bucks and you could just be printing in like seconds right so i had a positive takeaway i think you did too but i was curious if you had anything else i mean that that i took took it way way further than the initial thought on on you know how you kind of growth hack that uh which is which is i think a great idea the i think we we you know uh everybody got really excited about jevin's paradox uh because of of you know the the efficiency of ai models the theory that as they get more efficient, consumption goes up.

1:51:24But there's an equal amount of Jevons Paradox of all of the applied use cases of AI. And so that gets you closer to then the human labor side, which is if you just think about like most markets on the planet, I would argue are sort of smaller than they should be. If the cost of delivering the service was cheaper, you would have Jevons Paradox for almost every domain in the world. Like, and let's take a very easy one. If healthcare was cheaper, we'd probably consume it more. Like my biggest hurdle to seeing a doctor is just, it's a nightmare to just like even get, you know, a date on the calendar to see anybody.

1:52:06And so just think about all of the things, if you could increase the productivity of that particular category, which is an inverse or a corollary to lowering the cost of delivering that thing or making it more available or making it more accessible, would the consumption of that category go up? And I think there's not a lot of markets that are kind of at this perfect saturation level of what the potential demand is. And we saw this in the very, you know, I think in modern tech, we saw this, let's say, in Uber, as an example. There was this really funny article that some economists wrote like 15 years ago about Uber that was like, the best case scenario is this takes like half the taxi revenue, and it's like a$5 billion business or something.

1:52:47And what the guy like obviously totally missed the forest from the trees is if you just make Uber and liquidity of a taxi service 100 times more efficient, the market actually goes up by it, let's say, 100x. And he couldn't kind of quite process that. But we're doing the same thing over and over again in AI where we're sort of thinking like, oh, wow, we made this thing way more efficient. That means it's going to compress the size of that market. when actually what it will mean is that, to your exact point, all of the people on the planet that have really bad design right now, they'll have some entrepreneurial, some high-ingenuity person going and saying, okay, now I can actually do better design for the first time ever because it's affordable to do great design for this small bakery that would never have hired a designer.

1:53:32They never would have thought even to make their website look good, but now it's actually affordable to do so. And so I think that's actually going to play out, you know, many, many times over in AI. There'll be some categories where we are at sort of equilibrium of sort of supply and demand. But I'd argue most categories are this is not the case. You know, even as an example, taking like another knowledge workspace, let's say, you know, there's a lot of AI startups doing legal work, you know, AI for lawyers. And, you know, one theory is, OK, wow, is it a really bad time to be a lawyer? On the other hand, if you make it much more accessible to now ask a question that's a legal question, I'm still going to follow up and say, OK, can you paper this now as a contract or can you just give me a little bit extra advice?

1:54:19But we've lowered the barrier to now getting access to even thinking about what is the legal implication for a particular thing that then grows the demand of that space. And even in the category of, let's say, internal legal work at a company, if I could make it so we could review contracts faster with customers that we're going back and forth on, I'm not going to have fewer lawyers in the company. We're going to actually just have higher throughput of our deal review process. So there's a lot of these areas where you're just going to actually see an increase in demand of the overall category as AI comes and drives productivity in these spaces.

1:54:54Talk about just hiring and headcount expansion. Generally, you have thousands of employees already at Box. When you're talking with the team about doing headcount planning, are you pushing people to say, do we actually need this person? Or if you just figured out how to use these tools better, could we get away with without adding that incremental person? because there's been a lot of, you know, the Klarna CEO came out and was basically like, we're not, we're never hiring anyone again. And like, obviously that's really good marketing, but like, what, what's the reality and how are you approaching it?

1:55:30Yeah. So, so we, we are, uh, we're a hundred percent committed to being, uh, an AI first company just across the entire business. So we want every employee using AI to be as productive as possible, you know, with all the right asterisks of, of, you know, if you're doing production code, you're still responsible for what the AI produces. It's not the AI agent. So So all the normal kind of T's and C's on that. But for the most part, we look at it through the lens of if we can cause efficiency in a particular area of the business, we want to reinvest those gains back into another area in the business, maybe even the same category or same function.

1:56:07But we want to use those dollars that we're freeing up to actually reinvest into the areas that were previously constrained. And this is sort of, you know, even in the case of things like customer success, where you're getting kind of like the first line of defense, you know, hey, I need to reset a password, those types of things that we can now begin to automate with AI. The dollars, at least in our case, that we save on that kind of human labor, that will actually go back into the same exact organization. But now for more proactive customer success managers that we're always constrained by, we can't hire enough people to go out and do a proactive outreach to our customers and help them strategically because we have to have people that are responding to tickets.

1:56:52And what's interesting in this case, and I think it's an optimistic story about AI, in most cases, it will be the same person that moves from one type of work to the other. Our classic customer success manager type roles oftentimes started in support at either our company or some other one. So really what it lets you do is recalibrate your talent to increasingly more strategic ways of using it. You know, the sort of, you know, go in and translate this marketing asset into 10 languages is a much less strategic use of time than help go work on the next marketing campaign that we want to go deliver.

1:57:30So it really just lets you readjust into more and more strategic things. And that I think is an analog for what we can expect for most AI impact in an enterprise context. Can you talk about going public? Is it as bad as people say? You went public young, You're a young guy. Your hair is gray. Is it responsible of the is because the SEC? It's actually not because of the SEC. It's because of the hedge funds. Okay. My, you know, there's always this conversation of, you know, if we magically change the listing requirements or something of a company, you know, would that be better? I actually don't think I think that's a red herring.

1:58:06I think that the scrutiny that the SEC requires, the governance that you have to have, I actually think is all net positive on companies. I think that that like just lets you run a more stable, mature, you know, thoughtful business. And, you know, any reduction in that probably, if anything, just causes adverse selection for the kind of companies that eventually go public. I think I think the the the and this, you know, Brad Gerstner kind of kicked this conversation off a couple of days ago or yesterday online of like, you know, about the supply of IPOs. I think this is firmly actually in Brad's hands.

1:58:42This is a private capital market dynamic, which is as long as you have him and the KOTUs and the Sequoias and everybody else doing very large late stage deals, as long as you have capital in the private market to stay private longer, that is the real thing that will reduce IPOs. And I don't necessarily know that there's any kind of grand economic loss for that. But this is now, SEC has nothing to do with this. This is firmly in the private sector. We could decide do we want more IPOs or not. Last question. How did you react to the XAIX merger? You have more followers on this platform than probably anyone I follow, but Elon himself.

1:59:29Yeah, I'm curious what your take is. you've been here since day one? I was early on the Twitter wave. I mean, Elon clearly likes to keep buying X, so he will find another way. He's buying twice now. Yeah, I don't have a particularly strong opinion other than I do wonder if Tesla is the ultimate acquirer of XAI, but I don't know how he's going to kind of, you know, make, make all these worlds come together, but you know, we'll be interesting to see and, and, but no, no impact on my life. So yeah. Yeah. As long as we can post, as long as we can post, we're good. Just don't turn off the posting. Don't turn off the posting.

2:00:14And what I don't want is I don't really want like a write this tweet button. I hope that's the one thing that they won't do. I think, I think that like a lot of these platforms, like LinkedIn has this, you know, like help AI write your post. And like, I just think you get slop. It would be nice if this could be like a slop-free environment. Slop on X? No way. Unfathomable. No, we want a handcrafted, artisanal, organic farm-to-table post. Yes, exactly. We need artisanal content. We need artisanal content. I agree with that. It's great to have you. We'd love to have you back on when you have Box News.

2:00:52We'll talk to you soon. Bye. Bye. Bye. uh who got coming in we got sam lesson that was a lot of waiting room oh we can keep it on on on x and xai Let's bring him in. Get a hot take. Get him out of here. The take Smith himself. I want to hear it. I want to hear it. You know it's going to be good. Is that a yes or a no? He's here. He's here. There he is. Welcome to TBPN. You going to get me in here and get me out of here? Want some hot takes? Spice it up? Yeah. I mean, you know what we're talking about. X, XAI. What's the take? Sustaining. Disruptive. We technically have like 12 minutes, so we should probably just let you just go on a giga rant.

2:01:28Just rant. But first, I want to say you're the only person I know that's been getting more tan throughout winter. And I think it's because of all the tennis. But you also were skiing recently. It's because of the skiing and I'm not done. I'm not done. He's not done. He's putting up 60 days on the Epic Pass. No, there's no passes. What do you think I am, a basic bitch? There's no passes where he skis. There's an app that you track your days in. I forget what it's called. but it's people people track garmin maybe it's a garmin okay garmin yeah garmin uh there's no epic where i go okay yeah there's no chairlips it's all helicopters and cats right oh it is listen i'm just trying to live the vc dream you know you're doing it keep it up yeah uh speaking of vc dream pine on something yeah you were probably you know uh midway down some peak when the news broke uh but talk to us what was your reaction we we had sort of a uh you know sort of i remain sort of long-term bullish elon having a sort of an ai bet and how could he not yeah yeah yeah and and on you know this platform that we've built uh our you know show on our show on but uh to me it's like you know it's generally like complicated there's a lot of ways to kind of like you know maybe poke some some holes in it but i was curious to get your initial reaction as somebody who's been bearish on just a lot of stuff in AI.

2:02:56On everything except meta. Because the reaction was like people were really... Yeah, big tech. Yeah, people were just like immediately like, yeah, this makes total sense, et cetera. But there's more to it. So look, I actually think to Elon's credit, it makes total sense of your Elon, right? Like Elon is the king of cheap money, right? No one played Zerp in history better than Elon, right? You basically come up with a long-term vision that can require infinite capital. You're the world's greatest marketer. You sell the shit out of it. You raise infinite free money and you build stuff, right? He did that great.

2:03:29Now, it turns out the Zerp free money is over, but you know where there's free money? AI narratives, right? AI narratives is where the free money is. So obviously, Elon's got to have an AI narrative because he's the king of free money. And so I think when you look at the Twitter X thing, I mean, it seemed, I mean, he literally named them. They're like, how could they not end up combining? Like literally one was distributing the other. They're layered on top of each other. They basically have the same name. And to Elon's credit, in terms of pulling this off, I mean, what is this really? This is AOL Time Warner if Time Warner had like a ton of debt, right?

2:04:03Like, and so you use an inflated, you use the zero cost capital inflated valuation thing, which is the AI thing, to buy the thing you also happen to own that's an actual asset. Twitter's great. I love Twitter, right? But it also has tens of billions of dollar debt noose around it, right? It's 12. It's not tens. It's 12, exactly. Oh, 12. Sorry. A 12 of debt around it. And what's not to like about it, right? If you're Elon, it's a great deal all around, is what I would say. But what does it actually add up to? I mean, mid-market AI players actually worth, right? And that's the, I mean, I, again, as you know, I've been very consistent on this for years.

2:04:50If you're the biggest in the world and your cost of capital is zero, um, sure. Like build out a huge AI infrastructure, ideally one where it's non-cannibalistic like meta, ideally one where you know how to make money off of it like meta. If you're a tiny baby player, take all the leverage, fire your customer service team, like be more efficient hooray but i think these mid-market players i don't understand the numbers i mean you said it makes sense for elon but i feel like when we were looking through the deal it's like the solar city tesla acquisition 100 and it looks and it looks good for the tesla shareholders too because they they effectively get out at the same price they got in and at they're in this new thing if you're a tesla shareholder are you happy or not well wait which one's Tesla now?

2:05:36Are we saying that X.ai is Tesla? No, no, no. So if you went in on the Twitter deal and you're sitting at X and the valuation probably went down and then it goes back up and you get out at like 44, it's about where you got in, but now you're in this combined entity. Well, let's put it this way. I think the question is, there's probably, and I don't know this, you guys probably do, there's gotta be some market for X.ai shares floating around and markets, things that are worth what people are willing to pay for them. So if you bought into Twitter at 44 and it's not worth 44, despite the fact I love it, despite the fact you guys love it, just not, right?

2:06:12And all of a sudden there's a market for XAI plus Twitter at whatever the number is and you can get out and you want out. Look, who cares, right? So this seems like a win-win. It's not just Elon. The original Twitter slash X shareholders are also happy. But I guess... But if you're, I don't know, if you're an X.AI shareholder, it's just kind of, I mean, this kind of fits in the world of, you know, we live in a world of spheres of influence right now this is not my line by the way it's a smarter hedge fund friends of mine line which is we've gone from kind of a legalistic globally open market to a world a multi-polar worlds of fears spheres of influence you know if you bought into x.ai you bought into elon sphere if you bought it in twitter you bought it the elon sphere what are things actually worth yeah you know like who knows but like you're kind of just betting on elon holistically and this is the nice consolidation which by the way gets him out of a very large debt bill right that he was personally liable for in some way shape or form but i mean it just feels like tesla solar city all over again but if you play out tesla solar city that deal worked out for everyone well it's not clear it actually worked out for tesla shareholder it's fine because it's a rounding error they're making 10 billion now they're making tesla energy on batteries and stuff yeah they pivoted away from that.

2:07:27Did they need SolarCity for that? Maybe not. If you think about it, the SolarCity story was pretty weak when it was acquired. It was we're going to stick solar panels on the roofs of Teslas and that's why we should buy this thing. It didn't really make any sense. You take it to stick it on the house of the Tesla owner, right? You're already there to install the battery pack, install the charger. I get it. I'm saying we all know, you look back on the history of that deal, this thing is about to die, bail out i'll come up with a one-page letter that makes that justifies it because we're going to put some solar panels on the roofs or some shit right it was like completely made up now when number go up all sins are forgiven right and so what i'd say is number go up so everyone's fine and no one's the master of number go up he's the master of number go up right so everyone should be happy so is i guess that's kind of that's the kind of story in general i think there's a real elon narrative which is master of zero cost money in master of narrative number go up right and the reality is it's kind of like very roman is like as long as everyone's getting paid everyone's on your side even doge right like the whole thing with the dogecoin is brilliant like the best way to build more followers is to get them paid right and elon has been the king of that now the question is it works in both directions and we'll see how that plays out now um but you know i think what I'm what I'm interested in is like I want to hear Elon speak for an hour about the combined entity and what that really means versus data and distribution right like the two narratives have been like you know xai gets distribution x uh and and sort of data so so it makes sense but for me it was it was more easy to visualize how x was worth 44 billion dollars when it's sort of an elon pure play sort of social media bet in a you know strategic uh influential asset and if he is able to continue to grow earnings and grow the importance of the platform it's easy to see how that's like you know maybe it's a hundred billion dollar company someday now you have xai and it's you know you're you're sort of looking at this collective entity being like it's a founder it's a pure play like founder bet at this price i mean look i in the end of the day the problem with all its AI stuff is it's either worth infinity or zero.

2:09:46And no rational person can do a discounted cash flow on an infinity or zero, right? And I think that's kind of where everyone's stuck, which is why it's so narrative dominated. Personally, I have to say, I actually quite like Grok. I use it. Like I have on my home screen, Grok and OpenAI and all of them next to each other. And when my finger goes, which button do I press? I press Grok. You know why? Because it's kind of marginally faster and fun, whatever. I don't know what that's worth. Right. Um, but we'll see, right. And the, and the, and the bull case for that behavior is that Grok could replace Google.

2:10:21Exactly. I say Google's a multi-trillion dollar company. You slice it up and all of a sudden you're getting a couple hundred billion dollar bets. Maybe, maybe this is a mistake that people make, I think, consistently when thinking about AI, which is the story of will you use LLMs to look things up? Yeah, of course. Yeah, you're definitely going to use them. Like, you know, but the idea, it's kind of the same stories when people say, hey, I pay accountant $120 ,000 a year, AI accountant$10 a year, therefore AI accountant worth 120 times per, that's not how it works. It's just, it's worth 10. And like the mark, it's all about the marginal advantage.

2:11:01And so it's not clear to me. I mean, it is very clear to me that a lot of search volume that currently goes through Google and gets resolved by, you know, shitty web pages people have set up to farm those searches goes, that behavior goes away. Where commerce go, where the ads dollars actually go, and whether just searches like it's just not the same market configuration anymore and who gets that money. That's all like really TBD is my view. And it's really unclear clear. I mean, again, it's a very binary game. So that's the one that's why I think it's very hard to value right now. But I think the big picture is like so long as private investors are willing to foot the Hail Marys, the binary outcomes and believe that, you know, again, Elon is the king of cheap money.

2:11:48How do you feel about market concentration? Google has like 99 % of the search market. Meta has 99 % of the social networking market. It feels like LLMs could be a little bit more oligopolistic on the consumer side. But if history is any guide, like these monopolies kind of run away with it at some point, what are you feeling? You know, actually, I think the question is, if you think really what I'd say what Google and meta are, is they're not technologies, they're economies, right? And this is the really, I think the mistake people make is, if you think that Google is a technology company, and Google searches technology purely, then and you're like, wow, technology tends to compound and run away, then And you're like, yeah, like this would end up that way.

2:12:29Same thing with social networking. There's no actual technology. I mean, yes, there was like some stuff Facebook invented on the way to scale, but like it's not fundamentally a technology company. It's an economy. And I think that the problem with LLMs is people are equating them to an economy, but they're not. They're just a technical level. Now, if someone actually nailed the everyone's pouring every single thought they have into it and all the knowledge of the world into the LLM, and then that single element aside is like the smartest fucking thing in the world. But that's what they're doing with this deal.

2:12:59Yeah, but they're not. You go on X and you get paid a creator payout for bringing information to X and then that gets siphoned into XAI and Gronk gets better answers. Agreed. But here's the thing that you have to understand and this is really important is one of the big economic failures of the internet is that the economy of the internet, both social and search, is actually driven around entertainment, right? And engagement, not around information or truth, right? And so the reality is what you're basically at best, that's what the economies work. That's why we're so screwed on so many things is because there's actually no economy of truth.

2:13:31There's an economy of entertainment. And I think that's getting to get quashed in a bunch of ways because, again, if nothing else, and LLMs do generate infinite free personalized entertainment, right? And that's going to crush that economy. It's not clear to me. I mean, I can tell you the vision where, yes, for some reason, X becomes this machine that pumps all of the world's knowledge. and then somehow Grok has some fundamental advantage on knowledge. I can tell you the story, but I don't see any indication of that actually being what's played out. And I would argue that, again, that was the same story Google had.

2:14:01It didn't play out that way. It ended up being a commerce engine and a bunch of dead internet theory stuff. Unfortunately, it's not the way that social played out either, because social is just like a dopamine feed, unfortunately. It's not an information machine. So that's really the question of how you get there. And we'll see. I am very skeptical that the mid-market players who are saying LLMs are a compounding economy actually have anything right now. I think they have technology that's pretty fun and it's all pretty undifferentiated. But we'll see. Where do you think XAI would trade if it were to IPO tomorrow?

2:14:38Well, so this is a really interesting question, right? Because basically you have two dynamics at play. Either it's traded as a meme coin, right? If it's traded as a meme coin, it might be huge, right? It's the same thing with like, you know, Mr. Beast wants to go public. We talked about him the last time I was on. It's like if Mr. Beast's, you know, public company trades like a meme coin, it trains like GME, then it might be a very, very large number. If it trades on a discounted cash flow basis or as like a traditional accepted business, like not very well, right? And so I think that's really just the question, you know, as the retail versus, you know, enterprise versus like effectively real serious business world compete, how these things trade is all about the marginal buyer right and what that marginal buyers belief system is as much as anything else yeah got it well we got to move on to the next interview talk to you soon he said bye friends great to have you sam thanks for coming on um we got chris uh is it chris oh wait no no we got david is he in the waiting room that's right from poseidon let's do it i'm gonna let you take this intro himself can you grab me a yubamate thank you uh david what's going on big day for you and a big flag thank you yeah no it's uh it's been a great day here we finally launched in public we've been in sell for a bit over a year now so it's really exciting to show the world what we've been working on uh it's amazing and And if I remember correctly, some of the very first posts we featured on this show at the beginning of last year, weren't they yours?

2:16:13I feel like you came back. I think there was one. Yeah. Way back. So thank you for contributing to us getting this off the ground. But talk about the Poseidon Seagull, kind of the genesis of the company. Would love to hear kind of like your backstory. And then the first, you know, has it been a year and a half of the business? How long have you even been in market? Yeah, so we I can go way back to the beginning. We started right around like December 23, January of 24, raised our initial pre-seed. And the way that the company got together was it's me and two of my co-founders, Isaac, Zeke Dup, and Parker.

2:16:52And so we were, I think, ideating about logistics and aerospace. And I think logistics especially has been a market that's been largely unchanged for the past 100 years. The types of planes that we're flying are largely the same, same 737s, same Cessnas. And on the ship cargo side, it's been pretty much containerized cargo with very much spoken hub model. And so we were going through a bunch of ideas of like what we could build to disrupt this, thought about airships for a while. And those are very interesting, but have their own set of problems. A couple other things. But then I think one night, a random party in San Francisco, a girl walked up to Isaac and was like, hey, you should look into like ekranoplans.

2:17:37And we're like, oh, like what are ekranoplans? So we spent the next night just like binging YouTube videos and like documentary series. So ekranoplans were these like types of vehicles that were built in the Soviet Union in the 60s and 70s. And at the time, they were like the highest cargo capacity, like planes ever built, like they're like flying ships, essentially. And so they're constrained to fly in ground effect, which is a regime when you're flying extremely close to the surface of the ground, or in this case, water, because water is extremely flat, where you have an aerodynamic boost, it does two things.

2:18:08A, largely reduces your drag, and it increases your lift. So you can carry a lot more and spend a lot less fuel. So it's like increase of your lift drag ratio by 30 to 50%. It was a huge deal. And so we started looking at those and the really crazy thing about ground effect vehicles in the modern age is like a lot of the problems that the Soviets had are solved for the most part. So it's really down to a controls issue and a material science problem. And composites are incredible material science. So like carbon fiber and controls is like much like, I think we've solved those farms pretty well now.

2:18:44And so, but the really cool thing is like the ground-fake vehicles are not regulated by the FAA. They're regulated by the Coast Guard. And so Coast Guard certification is significantly easier than FAA. And so you have like really great companies that have built cool like EV tool stuff and like they've IPO'd or SPAC'd or whatever for billions of dollars, but they've never actually had a commercial flight because the FAA regulatory process is so long. uh so it's like a decade plus to get certified um but if you're going through the coast guard and so it's if you're flying sub 150 meters and it's the same thing internationally europe under the imo you get certified as like what is technically a boat uh so what we're building are flying boats yeah we talked to billy thalamar at regent he's doing yes personnel carriers but uh that that regulatory arbitrage is just like amazing uh john even off air after we had billy you were just like, I'm so excited about this.

2:19:38I love that. It's just like, it's previously the domain of crypto stuff, but now in hard tech, we're taking advantage of the video. So we have the video. I want to pull up the video and I want you to just kind of like talk over it. Uh, Ben, can you pull it up? And, and, uh, while we're waiting for that, I'd love to know a little bit more. Oh, here we go. You can mute it, Ben. I don't see it. Okay. Okay. Uh, maybe try to watch. i mean we'll just uh seagull okay

2:20:15pretty cool so this is uh starlink on it my heart um so this is it taking off um yeah like flying at about eight to ten feet is uh peak and so how fast is this uh it cruises 65 miles an hour we We hit 115 miles an hour. 115, let's go. Yeah, some crop changes, I think we'll hit 160 on our next set of tests. You got to get the sports car numbers. I want to see 200, Bugatti Veyron goes, what, 250? Yeah, we got to race one. You got to race one. Yeah, do the Nürburgring on this thing. I love, I mean, the design looks fantastic. So when you say ground effect, does it have to fly over water? Or, I mean, this looks just like a plane.

2:20:59Can you just fly it over land too? Yeah, so if you're flying in ground effect, you generally want to fly over water because water is extremely flat. So you'll have waves and even in waves you'll fly slightly higher and it'll average out. But you can fly ground effect in the salt flats or something. But ours is what's considered to be a class B, class C ground effect vehicle. And so we're not restricted to ground effect. We've just optimized aerodynamically to take advantage of ground effect. got it but we could fly out of it um up to 150 meters can you talk about where were you yeah can you talk a little bit more about the tech tree that enables this it sounded like there's some material science stuff like carbon fiber is cheaper then there's also like the miniaturization of electronics like you have starlink on that thing that's not possible a couple years ago the question's really just like why now but i want like more of a breakdown of like of like all the different things from the regulatory to the tech tree that's enabled this to happen now and not earlier because at the same time like this feels like something that could have happened decades ago but why are we just why are we just breaking through with this now yeah so i think like starting on like the material science side i think is really interesting like composites are really coming to their own in the past like 10 15 years um in some terms of like being having like actually like very like good many like there's very good manufacturing here so like atv composites is where we get all of ours and that's in labor more it's like an hour drive oh cool it's kind of Awesome.

2:22:27And that's both like the science of those composites, but also the cost, or is it just the cost that's come down? Both and like the amount to which they're being used. And so like, sure. You're having your like large planes, like your air buses are being built with like composite wings. And so like aerospace as a whole has like really moved into composites as opposed to just like aluminum. And so then like on the electronic side, I think like there's in In the same way that GPUs for gamers enabled machine learning, the hobbyist grade RC stuff has gotten extremely good. And so just really move forward, the quality of motors available, the quality of electronic systems, sensors, the types of sensors that are available right now have gone small enough that we can fit them on here.

2:23:15And really, the SQL has started as a quarter scale prototype of a full scale vehicle, which is going to be a 50 foot two ton capacity vehicle um but that's what i was going to ask i was going to talk about you know sort of long-term capacity scale and then like the applications of this because at that size to me it looks like something you would use for like potentially defense applications is a big one but i thought you're going to say pleasure craft and like going out and just getting on with your boys with your boys on the ground effect vehicle with your boys having a glass of champagne but yeah so is is your guys's first focus like logistics and and sort of like long sort of longer term uh longer distance transport that kind of thing yeah so um i think like as i was saying this like started as just like a four-scale like prototype demonstrator figure out the aerobody design validate a bunch of stuff uh but as we were building it um it became where they like it was like pretty capable of vehicle in and of itself and like we designed it for manufacturing.

2:24:16So we reached out to the DOD, specifically like the Naval Surface Warfare Center people and SOCOM, and they all had their own specific mission profiles that they were interested in with the SEGAL. And so it's like pretty large pilot capacity relative to size and being able to operate in like marine domains. And so doing stuff like anti-submarine warfare, like UV detection, very like urgent and like high priority logistics and goods. is on the on the defense side for the seagull and then we're starting operations commercially with it uh for doing like very like mostly like medical deliveries to remote islands um and then at the the full scale carrying two tons the idea is to be able to compete um with air freight for coastal and island uh logistics and cargo um and through a bunch of things we are going to be able to get to roughly a quarter or a third of the price per kilogram per mile.

2:25:12It seems like Zipline's done pretty well with the medical deliveries stuff. I was surprised to see how long that company's been around and how much they've grown. So clearly there's like these niche markets that are really important in logistics and transportation. Jordan, you got another question? No, I mean, that was it. I just wanted to bring you on to say thank you for the early, I was saying you missed it earlier. Oh, yeah. uh i just remember i don't i don't remember much of what you posted but i do remember covering your in terms of military applications it sounds like uh logistics first then maybe isr but is there gonna you're gonna put a bomb on this thing at some point i don't know what they stick in it but there are discussions about a wide range of different options but i mean just just like stepping away from like the business applications like is that something even like we even need like we hear about the hypersonic missiles being able to kind of bob and weave and they don't fly on a traditional like parabolic icbm trajectory so they're harder to shoot down this thing seems obviously maneuverable uh is there military value on a weapons platform for this thing or is it just like too slow and too small to to really like matter yeah the it's definitely faster than like torpedo per se but the unique advantage of it is when you're flying at such a low elevation you're flying below radar yeah it's a like very stealthy platform like a ship um so you can get pretty close to the enemy lines without being spotted that's cool uh back on the tech tree um are you are you doing any ai on board even like open cv to do collision detection like you'd get on like a dji drone or is everything kind of remote controlled right now tele-operated how are you thinking about the evolution of that uh tech yes so the the way that we built it out is it starts with being teleoperated because you need to like train the systems and so running it's like the on a hardware level it's like pixawk with ardupilot that we've like modified it and ardupilot it's like an open source platform cool i know guys at andrew use it as well and so uh there's a few sensors that we have on board uh we have sonar we have a radar and just like a visual camera sensor so yeah um taking stuff in terms of like collision detection like elevation detection maintaining stable flight so yeah we're building some of our own models so when you're flying it are you doing fpv or is it just at a computer screen like what what does that actually look like for the pilot yeah for for the pilot when you're like flying in just like remotely um we have the the computer screen with all the um different like signals on it um you get all your stats flight simulator yeah basically and then you have the camera on board so you can actually see where you're going that makes sense cool super cool that's amazing well good luck really impressive what you've built uh in so little time with in a very efficient way so congrats again congrats to the whole team appreciate it looking forward to the next update yeah we'll talk to you soon thank you guys cheers bye and we're going deeper on ai image generation we got the ceo of runway ml coming in the temple of technology, Cristobal Valenzuela from Runway.

2:28:21I'm a huge fan of Runway. I used it years ago. They had a fantastic feature that would allow you to basically create a green screen, even if you didn't have, even if you didn't record the footage on a green screen. And so you could do all these funny things. Remember Zuck using the flag when he was like surfing on Lake Tahoe? I was able to use Runway to select just the flag, and then I could put a different flag there. And it would have been so hard to do in After Effects, which is a software I kind of know pretty well, but it takes a ton of time to do rotoscoping. And that was the moment when I was like, oh, this company is doing really cool stuff.

2:29:01Now they've gotten into image generation, video generation, and a whole lot more. And today they launched a new Gen 4 video model. And I want to hear all about it. So welcome to the stream, Cristobal. Hey, yeah. Thank you for having me here. Thanks so much for joining. I have a million questions, but let's just start with, can you introduce yourself, the company, and what you launched today? Yeah, I mean, you did kind of a pretty great job. OG, you've been using Runway since a couple of years. Yeah. That's great. Yeah, so we just introduced Runway just a couple of hours ago. Sorry, Gen 4, just a couple of hours ago.

2:29:36It's our latest video model. It's basically a new frontier, I would say, of not only video generation, but it's storytelling. It's one of the most pressing things of AI video these days, which is consistency. So you can create really consistent worlds. And the way we wanted to emphasize this idea of consistency is that it's actually releasing a model. And I think it's kind of the model part is like the first stage. We put together a collection of different shorts and films that if you watch them, if you guys haven't watched them, definitely try watching them. It's just like a few minutes. There's short films.

2:30:06It's crazy. It's such a trip because you're looking at it and you know it's AI and it does not look anything like. We're passing the uncanny valley for sure. And faster than the VFX stuff did. I mean, that took 20 years and it feels like AI video generation, it's been like a few years. There was an uncanny valley for a while, but we're getting the other side really fast. Yeah. I mean, we were working on runway for almost seven years. So it's definitely been like more than like a couple, like two or three years. But yeah, I would say the last two or three years have been like really fast. And I don't think we're even close to like the final stage of our media will come.

2:30:41Yeah, I think there's still a long way to go. But the point is really to point that once you start like seeing how good the stuff that you can do at Runway is and you layer in a really good creative mind, you can make stuff that you just like forget how you're watching something that's completely generated. Yeah, we need a new eval. We had the Will Smith eating spaghetti. We blew past that. That's photo real. The next one I think should be Born Identity style shaky cam. You know that? Because I think AI could not do that today. I think the evil should be just like how good the movie is, how good the shirt is.

2:31:14You know, like how many awards, how many people are watching it. That's evil. I think that's our internal goal is basically to make sure that you judge by the quality of the story more than anything else. uh i'm sure last wednesday whenever the ghibli meme started ripping i'm sure some of your investors big and small reached out hey did you see this i'm curious what uh if you had to answer that question like you know what what was your immediate reaction to it uh it's amazing that you guys so quickly followed it up with something that is you know uh yeah so you come on the show we ask you the stupid questions and then you don't have to answer the stupid questions from your investors yeah yeah you just have to answer one time you share the clip so so i have to go honest uh no one reached out to be honest like oh that's great like yeah it wasn't like a like a thing i think it was it i think there's a separation between like memes and like the cash flow like fun stuff that i think it's great you can just kind of like have fun it's like a trend but i think if you're making like good professional content if you make a living out of making content like making a meme or like one image it's not gonna get you far it's like You need a suit of different tools.

2:32:21And so I don't think we've, I saw the trend and I saw the images and the things are great, but I don't think they're, no one really thought about us to be honest. Yeah. The thing that I think was part of what made that so viral is that it didn't require sophisticated prompt engineering. You image, one line, Studio Ghibli, you can misspell it. I actually have like a text replacement on my iPhone, a shortcut now. I just uploaded it. SGS and it automatically expands to Studio Ghibli style and then it just does it. Yeah, but so more importantly, as it comes to using runway and video, how important do you think prompt engineering is in the context of runway long-term?

2:33:05I'm assuming it's just not going to be a thing. You just sort of describe exactly what you're kind of feeling off the cuff and it just creates something beautiful. Exactly. No, so prompting is a way of like sampling something out of the model, like basically getting what you need. But there's many other ways of getting what you need with much more control because language doesn't really like have all the nuances of like describing a composition or the style, the mood or the feel of something. So the idea, for example, with Gen 4 is that I saw like the demo, but you can combine images of real things.

2:33:36You can take a photo of something you've actually like have. You can take a color grade and a kind of like a moody scene and you can take a location and you can, there's no prompting. Just like put them in and like create like the composition that you like. And you can get a much more like a specific detail thing rather than just like trying to use the words correctly to like get it to where you need to go. So it really depends on like what you need. But I think prompting is overrated. Like just language as a unique interface will not get you too far because like storytelling and films, particularly in video, it's just so hard to just describe words, you know?

2:34:11yeah how how uh how many years away do you think we are from being able to one shot a a movie that could you know win uh you know an award-winning movie right it feels like so many movies in hollywoods are sort of derivative hero's journey it's set of sort of characters but it's somewhat predictable you could imagine that like from a storytelling standpoint a lot of models could already deliver something that's like pretty good from a story standpoint and so to catch up you need uh the the i mean from what we're hearing like the gpu cost alone it would be like hundreds of thousands of dollars right now but sure sure but but you know if you're a movie theater today i mean we we're so we're here in la we've been actually touring studio spaces and almost every single studio space is completely empty because like it's just too expensive to film in LA so I I can see a world in the future where a big studio is just like run does these expensive you know sort of prompts yeah then they get no no totally remix it or whatever but you know I'm curious what your take is yeah I mean if you look at the the lonely little flame the shirt I was just mentioning before that's production value before June 4 we're talking about hundreds of thousands of dollars that's a very expensive thing to make it took us like two to three days to make uh like the first rough sketch of that and then like another few days to just like get the final comp so you can by a version of one the whole thing would sound by one person wow so um i think you you are there you can technically just take that same process and it's something we're doing already and do it like for longer future films it's more of like you're gonna sustain the attention of your viewer not by how fancy or how expensive the thing is like but just how good the store is and just iterating on the shots takes you a lot of time and so i think the issue with the way we've traditionally making like very high-end production things is that um the budget type like skyrocket to the point it's unsustainable like i mean to your point like you're spending hundreds of millions of dollars trying to make something and then if you need to make something you're gonna bet for the thing that the safest the safest which is like superheroes and and franchises that you know will return which i think from a from like uh filmmaking perspective has kind of like prevented like new independent stories from actually like being made because it's more much more risky actually um and i think part of uh part of this new wave of models is that it will kind of like flip that a little bit and allow this like independent set of people actually don't don't use those studios just like use runway like make the films to yourself there's no excuses like you can just do it yourself you know do you think uh the big hollywood studios are reacting quickly enough to the advancements of runway and other you know are that do you feel like they're sort of properly planning around this sort of the biggest change to filmmaking in our lifetimes yeah i mean i i think so many are i mean bob eiger said it was the most uh important technology of their company uh just a couple of days ago i think many more that we work with are pretty much in that same vein i think some are more advanced than others i think for a long time it wasn't clear to many what this meant and to be clear i think that's fair because if you look at like video generation four years ago no one took it seriously like no one cared it was so i mean maybe you were like using runway for like rotoscoping and green screen but like that was a small amount of people right and i think now i can show you a film that makes you feel something and it was done like in two to three days all generated and that's when you wake up and i think Some have woken up before, but some are still like Dermot.

2:37:50But I think, yeah, everyone's really paying attention to Dermot. Can you talk a little bit about just like style transfer, animatics, and the importance of like generating the first image in something like images in chat GPT? I could imagine a world where it's a lot easier and maybe even more efficient to block out a story in Minecraft or in Roblox and really direct it and be very, very solid. I kind of program the camera move. I know the shots. I know the layout. But then I just want to take it to photo real. And that style transfer is what you guys are great at. Versus I'm going to be purely prompting and really hands off.

2:38:33how do these things evolve? What's most important? Is there an important like step function where we want to get to the next? No, I think you're right. There's many people who've been using Runway to do exactly what you're saying. You block the scenes by using either like 3D software or just like props. We have a way of rendering content using existing videos. So I don't know if you guys have seen it, but you use our performance of yourself. It's like advanced motion tracking. And then that drives the performance of the generation that you're creating. And that's great because it allows you basically to like, you become your storyboard.

2:39:06You are defining the shots in the way you want it. And I think the essence also is that once you create like a mood, there's a few, for example, another short film that we released called The Hurt. It's like the story of like two guys like hiding from each other in like this dark moody scene. We start with one image and then you can use Gen 4 to kind of position the camera in different parts of the scene. And you kind of like, it's interesting because the model becomes like a cinematographer. It's kind of helping you understand or provoking, provoking you with ideas of how you want to do it, which is different from how you do filmmaking traditionally, because you can't afford to do that in traditional filmmaking.

2:39:42You have a kind of like a set of shots that you need to make, and it's very expensive to like not do those here. You can just like wander around and kind of like find your angle. And I think that's also interesting. It's changes the nature of how you make things in the first place. It's a much more experimental, like you're kind of finding ways and you're getting to stories that you never thought you could. that that to me is the most exciting thing because historically if you're a director and you're like i'm gonna bring you know these you know superstars out to this like desert and we have one day to get this shot and then we got to go this other location yeah this other shot storyboarding you're basically like you could have a better idea midway through the shoot and not actually be able to do it necessarily without calling the producer and being like hey i need another 20 million bucks because yeah exactly we miss it it's like one person sort of deciding in the moment it's that and if you think about how different startups have been historically where startups are this sort of like daily iteration and you can adapt the strategy and you can adapt the product versus filmmaking which is like well you write the script and yeah you can change it along the way and you can change it in post but then every decision is just so much more expensive exactly yeah i mean i've been thinking about it as a like you're paying you're playing in a jazz then and you're improvising and so you're you get a shot you see something and then you can go anywhere you want it's like whatever you want to film and like take it there um it's very fun it's it's something i haven't filled for in a totally in really a while how do you think about the the the kind of shift in or the ratio or breakdown of creative video content as it relates to fiction versus non-fiction like youtube and the iphone has exploded the amount of creativity and we've gotten hours and hours of like non-fiction content everything from like Mr.

2:41:30Beast videos to video essays histories podcasts all this stuff that's non-fiction fiction stuff hasn't really taken off on YouTube in the same way that's still kind of the domain of Hollywood of course Tangerine won Academy Awards it was shot on an iPhone it took a number of years to get there from you know the iPhone being introduced but it did eventually democratize filmmaking to the point where an iPhone filmmaker could win awards. Do you think Runway is more of a benefit to people that want to tell fiction stories? Or do you think we'll see kind of the same breakdown where there'll just be a ton of nonfiction content that goes through Runway and you're seeing like illustrated video essays or I can just pull up history of Yerba Mate and all of a sudden I just get this beautiful story that's visually intriguing, but it's still nonfiction.

2:42:21I think Runway is made for stories. That's what the ethos of the company is. And I think fictional or not fictional, like stories are just stories. And I think we're, humans are like obsessed with stories and will consume anything that tells a story. And so I think, I think you're right. YouTube hasn't been like yet good at perhaps like AAA content or like professional, like fictional content. And it's better to do this like more informal, like blog type of things because it's just cheaper. But I think this is particularly interesting because like what gen models like can forward do is break that barrier now you anyone basically anyone out there who has a youtube channel right now can create a type of content that looks like someone made it in a studio with 50 100 million dollars um and i think that's the that's like the last unlock i would say media wise the first one was like everyone can do like this everyone can just like tell like create their own like channels and tell their own stories but then professional great content has been for the last couple of years, the last kind of like stage.

2:43:23And it's only been accessible to this very sort of companies. And now it should be everyone. It should be basically anyone who has great stories. It's like creative meritocracy. Whoever has the best stories will win. What happens to movie stars, right? You can imagine early on with some of these AI generated films, somebody is planning to generate a film using entirely AI. They don't plan to have any on-site shoots or anything like that but they want to go they go to brad pitt and they say we're going to give you 10 million bucks to start this film because brad pitt if they just completely use his likeness and his voice and whatever uh but then over time people start realizing like hey it's actually much more valuable if i kind of create my own essentially like ip or own superstars from the ground up because i can instead of paying brad 10 million for one movie i could invest a lot of money in this sort of yeah i can make my own giga chad yeah and uh i don't have to pay any of the fees and maybe there's some R &D-ish to sort of generate that character.

2:44:21But what happens to our stars? Yeah, well, I don't think that changes that much, to be honest. I think we had the capacity of doing that for a long time. I mean, animation and CGI and anything that's not live action, we've been able to do that for quite a long time. It's more the cost of it. It's really expensive. I think you can still have a space and a world where you can have both. You can have characters and actors performing the way they do, and you can have this new medium. It's a new format. it's like a video game that allows you to create infinite worlds maybe the two might overlap you may bring some elements of the real world into this you may create some elements generated into and it will swap i i don't see them as a replacement i don't see them as like oh we're gonna suddenly like stop hiring actors actually for most of our shots you actually require actors to do like act one or performances that you can transpose into the way you want and so um i think it's more of like a change more than like oh we're gonna suddenly like stop doing everything we're doing and do this new thing now.

2:45:18That makes sense. Do you have any type of sort of like long-term cost target for a 90-minute movie in 4K? Is it something that in a decade it's a dollar or is, you know, how long do we go? Just compute-wise or like just people talk? Yeah, ignore the sort of like... The prompting. Yeah. Well, look here. The iPhone is effectively free. Yeah, just the op-ex of like... Like you used to have to pay by the foot for film. Yeah. And it was very expensive to get it developed. Now anyone with a phone can record 200 hours of footage for a couple hundred dollar used phone. So I think maybe the way I need to think about cost is to think also to the point, to understand how like the models and the optimization of this model is like progressing.

2:46:06You're going to get really soon to a point where you can run these models in real time. And so by the time you run them in real time, you're streaming the pixels as you're watching them, interacting with them. And so if you do so, then the story, first of all, never ends. There's no 16, 90 minutes. It can be days that you can just start generating. And it's a cost of inference. It could just be as cheap as however NVIDIA wants to optimize their GPUs to make it faster for everyone and cheaper for everyone. So I think it's just a couple of dollars if you want to make it. Yeah, it's like a LUT file that you just import and boom, you get it.

2:46:37And you're just sort of wandering around. It's amazing. I mean, what does the business of Runway look like at that point? And how are you thinking about the importance of like value accrual in the application layer? Are you thinking you go more B2B and try and get big contracts? You want to go B2C and just get every consumer on some$10,$100 a month plan? How are you thinking about pricing and customer base? so far for now our focus is being professionals mostly making sure that like those who make a living out of telling stories can use runway to accelerate like their work um i think there's still a lot of like creators and individuals that will like start becoming professionals because of this like i mean i've i've i now could do things that will require like an army of effects artists um and there are many other people who can feel the same right now um our plan is mostly to make that those who want to tell something can actually tell a story.

2:47:32And that's a wide spectrum of folks. So we work with like Lionsgate, major production studios, to like independent studios, indie studios, like filmmakers themselves, just groups of like two, three people making amazing stuff. And then we have a noise phone app, so consumers can also make stuff. I think like the mission of the company is like, look, this is the most important technology for like art. Like I started working on this because I had the idea that neural networks applied to art will like eventually become any form of art. and I want to be part of that. And I think the consequences of that go beyond just B2B and B2C.

2:48:03It's like everyone out there should be making great stuff. Are you optimistic about any of the hardware developments kind of speeding up runway rendering or generation? Obviously, there's like new NVIDIA chips coming, but then there's also startups that are focused on transformer models baked into silicon like Etch, there's Cerebrus, There's Grok, not the XAI thing, but the chip company. And then there's like the DeepSeq approach, which was like, it seemed like they mostly just worked on algorithmic improvements to speed up inference. Where are you most optimistic about getting the next 10x improvement in your models and your systems?

2:48:45I think those are all like interesting developments. We've been watching some of them really closely. I think right now cost is mostly for us at the scale of what the business is and the research that we do, the main factor. And so if you give me like X performance, but it's unproven or like I will need to spend like six months trying to like debug into like much performance and whatever, or I already know, that makes it really hard to switch. and so I think it's cost mostly what cost in the sense of like how cheap can I get just like my compute and for how long can you guarantee me that price and maybe a price better after and then performance improvements if they're like significant will matter a lot but I think we've seen something so significantly yet that will make us like either switch or move to other like hardware I do think it's important like I do feel like all of the chips that are being made to compete with NVIDIA are relevant but But for us, it's still very hard to see where we can switch, to be honest.

2:49:44Makes sense. Last question? This has been great. This has been awesome. We'd love to have you back. This is really informative. Thanks so much. I've been a fan for so long. I wanted to talk to you. I'm really glad. Thank you for having me. We make some pretty cinematic videos at some point. And I really want to do the style transfer thing on some of our older content. You can. Now you can. Now you can just create. Now it's animated. So yeah, we'll definitely be racking up those runway bills when we get to hitting render it all. I want every piece of B-roll we've ever filmed in studio Ghibli style.

2:50:18I don't care what it costs. I want a folder full of it all. It's worth it. Yeah, it's awesome. It's such a fun time and there's so much that you can do with this stuff. So thanks for powering it and thanks for stopping by the show. Yeah, and congratulations to the whole team. Yeah, congratulations on the new model. It's really impressive. Perfectly timed. Talk to you soon. Thank you, guys. Take care. Bye. All right. We can do some timeline and then opening eye drop some news about a new open weight model with reasoning in the coming months. That's fun. And I'm going to use the rest of it. Okay.

2:50:45Well, while Jordy is doing that, I want to take a look at this video from ISAR Aerospace, which is a company I hadn't heard of before. But they had the most cinematic rocket crash I've ever seen from nasaspaceflight.com. drone and pad footage from ISAR Aerospace Spectrum launch. You can see it avoided the pad when it came down. Andrew McCallop, friend of the show, says, insane footage quality. Sorry about the loss, but serious credit to your media team. Top to your work. I couldn't agree more. We'll have to pull it up because it's the most cinematic just footage ever from this drone. It looks like they color graded it.

2:51:26It looks amazing. And Ashley Vance also chimed in and and said, this launch site slaps because it was great. And it's just a beautiful, beautiful website. But speaking of beautiful things, go to getbezel.com, shop over 23 ,500 luxury watches, fully authenticated in-house by Bezel's team of experts. Get a luxury watch, folks. Anyway, how are we doing on the footage of ISAR Aerospace's Spectrum launch? I'd love to have the founder on the show. we did defense week or defense day. We're thinking about splitting that up because we kind of did manufacturing in there and then also weapons systems. I think we need to break it up further.

2:52:09Do space launch day specifically. Maybe we'll do space satellite day specifically. But I saw aerospace would be a fantastic company to include in the space day if we do it because I want to talk to these folks and I want to see what cameras are they using to film this stuff? because it looks great. And we have a video here. So check it out, folks. Is it paned down enough? I think you've got to pan down more, Ben. Oh, here we go. Oh, we're good now. So look at this. This is so beautiful. I can't get enough of this. Look at this. The rocket takes off. Very sad that it didn't make it to orbit, but very cool video.

2:52:48And look at this. The drone just going higher and higher. This is why you've got to watch the video, folks. Sorry for everyone on the RSS feed. but the rocket's going up, up, up and then starts flipping over and winds up crashing. I actually don't know where this was. It looks like it's in Antarctica or something but I think it might be in Europe. I'm not sure. But beautiful glaciers. They got to get Harry Stebbings over to the launch site and have him figure out what's going on. But this post came, look at that. Wow, big. Like look at the teal and orange. It's just the classic Hollywood color grade.

2:53:23looks like a transformer movie uh fantastic and so if you're doing something cool like the the takeaway is just like spend the money on the nice camera because you're gonna get a lot of attention and attention is all you need uh rad backwards this one it was unrelated but it felt related he says uh caring is cool you should care a lot and work really hard that's the coolest stuff ever and i saw those folks blowing up that rocket i was like you know what it didn't go well but They care. They care. I love that. Try again. And should we go to Gokul Rajaram? He says, AI shrinks the skill premium. A year ago, a portfolio co would ask me every week on tips to hire senior engineers.

2:54:07They found it really hard to attract seasoned builders. For the past couple months, they have stayed conspicuously silent on this topic. Today, I asked them why. Had they finally found their dream senior engineer hire? They said, we don't need senior engineers. our junior and mid-level engineers are able to use AI to solve every logic and system design problem. This is what we were talking to Patty about. ChatGPT is his kernel and cursor is his army or something like that. I like that model. But really it's like, yeah, you have these tools and systems that plug in that help speed up the autocomplete and the implementation of your ideas.

2:54:41But how do you reality check? And it's the rubber duck test. Can you explain what you're building, what your architecture is. And these reasoning models are fantastic at that. So Gokul goes on and says, this will be one of the pervasive effects of AI. AI will massively erode the premium charge by senior professionals in any white collar role. You might still need the 0.01 % of talent to build world scale systems, but the top one to 10 % are going to see massive competition from the middle 50 % as AI up levels the ladder skill group. I don't know if you had to take Jordy, but I thought it was interesting.

2:55:21Yeah, I mean, it's interesting. Last night I had a question that I wanted to ask a friend of mine who is very much an expert in the subject. And I felt like, and I reached out to him, I said, hey, can you chat about this thing? But I also asked the same question to a language model. Yeah. And like hours before my buddy responded, I got a like satisfactory or potentially even better answer than my friend would have given me. Yep. And I think that that scenario is basically similar to what what Gokul is describing here. Yeah. Which is good. It's sort of I mean, again, it's sort of leveling the playing field in many ways.

2:56:06The Internet as a technology leveled the playing field globally where like you didn't have to be in the United States to compete with U.S. workers. your jobs. It could be a web developer in UK or India or whatever competing for work in the US. Yeah. I want to go to this post by Dylan Patel. He says, data center CapEx is going to zero. Grok struggling deploying. Anthropic extremely low limits for paid plus cursor. Google not offering 2.5 on API. OpenAI not serving 03, struggling to serve 4.0 image. Azure still not serving many models, including 0103 Deep Research 4.5. And this is so funny because he's the most respected analyst in semiconductors.

2:56:52Dylan Patel, obviously the founder of Semi-Analysis. But the first line is a joke. And if you don't have a good irony detector, you would just read this completely wrong. And unfortunately, Brad Gerstner chimes in and says, bizarre how the GPU doomers all somehow understand true demand better than the companies who all have massive compute shortage and are melting down their GPUs and scrambling to find supply. And so, I mean, yeah, if you look at the, uh, how long it takes to inference some of these models, it's all very slow. It does seem like demand is very high. Just anecdotally, I think we saw that everyone is using all of these models all the time and they are intense from a GPU demand perspective and you can just see it.

2:57:35Gary Basin actually has a good point here saying they're looking at two different timescales. The short-term squeeze is priced in, but like that capacity for the short-term squeeze is in many ways coming online or being built out. Yeah. There's also this interesting dynamic where some of the founder mode companies do have an incentive to overstate the level to which they are over strained. You know, it's like the classic stage founders, like I can't sleep. My phone's ringing constantly. I'm the most popular startup. We're going too fast. Like you kind of know that if you signal that to the market, you're going to get more term sheets and stuff.

2:58:14But I don't believe Google is in that mindset with like, oh yeah, like if we don't offer 2.5 on API, like the market's going to see that and price us higher. It would be very odd if all of these dynamics were fake. It would be reasonable if it were like just one was, is oh, that person's kind of overstating, or they're a challenger, and so they say, oh, we're so popular. But when they're all struggling, clearly there would be an economic incentive for one of them to break rank and say, actually, we have plenty of GPUs, so come over to us. We would love to have your business. When they're all struggling, it's definitely a sign that data center CapEx is not going to zero, in fact.

2:58:53And speaking of 4.0 images in ChatGPT, there was an interesting post by Technium saying, 4.0 image generation can do calculations during its image generation somehow. This was community noted because the first LLM prompt, so it's a picture of a calculator and the prompt is make an image of a calculator app for the calculation 53 times 88. And the result of that calculation is 4 ,664. And this would be shocking if the LLM had just been able to, or the The image model had just been able to do this at image generation time. But in fact, it appears that in the prompt reasoning step, before the image is generated, there is an internal prompt interpretation that basically hydrates the prompt and adds extra context.

2:59:52And so someone else shared that and said that when you, if you give the exact same prompt in the screenshot, you can see that they pre-process the prompt and that prompt actually does the math, which is still impressive. It can get it right. Let's not be wrong about that. But it is interesting. And this is obviously just like a huge unlock and obviously pairing LLMs with image generation. this is the future and this is I think why this new image generation the studio Ghibli vacation of ChatGPT it didn't happen in Dali 3 or Dali 4 like a separate app it just happened in ChatGPT and it takes advantage of all the other ChatGPT features and eventually you're going to see consolidation so you can upload a PDF have it search the internet do some deep research generate an image with images in chat to gpt it's all going to happen in the same app and same window which is great um but but interesting like kind of technical deep dive here um is there anything else you want to run through on the timeline because we are past 2 p.m we've been streaming uh i wanted to take an opportunity to talk about another new partner which some people have called out in the ticker which is polymarket that's right and uh we're very uh excited to be partnering with Shane and Matt over at Polymarket.

3:01:16I'm very happy to have them as a partner. Thank you. And the reason that we were so excited to partner with them is that the Polymarket platform is news, right? It is a way to understand the world and understand news. And so it's a natural fit to integrate into the ticker. It's like everything that's happening in tech, like here's how to get an understanding of it uh we're going to be sort of updating the ticker making it um yeah we're going to be improving it basically like improving it uh day by day like we do uh and we're also going to be creating some of our own models yep um and they already just have markets uh one of my favorites that's going right now is ai wins imo gold medal in 2025 at 61 that's really high uh which is really high and what that would mean is that everybody basically gets the average ramp employee or average cognition employee on their team.

3:02:16And so, yeah, there's a bunch of other stuff. Some of this stuff feels like, I probably can't say this, so this is not an ad, but it's just commentary. Some of this stuff feels like free money, just because it's like, you know, you look at the probabilities on some of this stuff. But I think it'll be cool because Polymarket's been known for political prediction markets, but we never talk about politics. And genuinely, this is not a political show. And so I'm just excited to build markets around tech predictions. I love, you know, will the next Starship make orbit? How many Starship launches will there be in 2025?

3:02:51Like these are the markets that I've always cared more about. And I'm excited to highlight on this show and kind of, there are plenty of other shows that are partnered with Polymarket that can highlight the important political markets. I want to drive the tech markets forward for sure. So that's what I'm excited about. Yeah, there's so much other stuff. I mean, which company has the best AI model by end of April? Google is sitting at 51%, which again is not what I think the average person on X would even... Tell you dangerous. It's forbidden knowledge. It's forbidden knowledge, but look at this.

3:03:27I mean, Sundar Pichai posts 2.5 pro ship today to everyone. Find it in the dropdown here, gemini.google.com. Still, he puts up 5K likes, a million views. Built different. Built different. He does it. He's so confident. He throws the link in there. He's got the hilarious image. Sundar's growing on me, man. I like that he's going direct here. I own links. Yeah. I am the internet. I'll post this, whatever. It's also hilarious. It's in this dropdown on a subdomain. You're taking it seriously. send our, but like, how seriously are you taking it? Uh, should this just be baked into the Google search box?

3:04:05Yeah. Right. Um, but hopefully soon. And I'm excited. I mean, it seems like they're, they are hearing the message. We need better productization. Google legendary team in terms of AI algorithm development. Uh, attention is all you need. That's a Google paper. They created the transformer time to capitalize guys. You got the TPUs, you got the team, you got the The Polymarket has a Polymarket for Google's whiz acquisition being blocked before July, sitting at a very six, only 6 % chance. But that's an interesting one because normally when a big acquisition is going on, you can actually determine, you can extract that market from the stock price movement.

3:04:46So like you can see how Activision is trading and understand, okay, the market expects that Activision will be acquired. This was what happened with the Twitter acquisition. Like it went right to 46 billion or 44 billion or whatever Elon was going to pay. And then it would drop by a little bit based on the probability that it would not go through. But Wiz is a private company. So you don't have that real-time data on how the share price is trading, which you can use to address whether or not the deal is going to go through. Polymarket comes in there. Very cool. We got it. We're also sitting at a 25 % chance that X relaunches Vine before July.

3:05:23Do you think they need to do that? I feel like they have their secondary apps are rough. Yeah. They have one app and they're, they're very good at getting you to open up one video. It autoplay is then it scrolls you to the next one. Pretty soon. It's just all slop, but you're watching it anyway. Those, those unregretted user seconds are dropping to zero, but we'll see. I mean, I, I'd love a vine comeback. Uh, the vine team, one of them just legendary crew of, uh, entrepreneurs yeah the other thing that we're going to be paying attention to as april starts is just what's happening with tiktok oh yeah polymarket says today there's a 30 chance that the tiktok acquisition is announced before may wow that's pretty high yeah i mean it's high but that's also what everybody's been orienting around this early april deadline to get a deal done so we'll see um and where would it go oracle walmart ben thompson says walmart should buy them you'll think there's only a one percent chance that mr beast is the winning bitter so no faith in the beast man no faith in jimmy oh well uh well that's a great show for you guys today folks thanks for watching leave us five stars apple podcast spotify stay tuned repost all of our clips like and fave everything do it do everything uh thanks also uh last shout out yeah chris amadon We saw you got an eight sleep.

3:06:51We're competing with you now. Yep. I want to see you put out some good numbers, Chris. Let's see those scores, Chris. Let's see it. That's it. Thank you, folks. Thank you. We will see you tomorrow. Have a fantastic rest of your day. Bye.