Eddy Cue Joins for Apple's 50th, Artemis Launch, Jamie Dimon's Plan, Save Snap Now, SpaceX IPO | Aaron Terrazas, Pratap Ranade, Abhishek Das, Garry Tan, Brannin McBee, Sam Yagan

22:55Eddy Cue:Yeah. So SaveSnapNow is recommending a thousand person riff to get fit and competitive and to empower your highest performers. It's again, if you're if you're Evan reading empower your highest performers, you're probably thinking like, oh, geez, I never I never thought of it. But again, they do have 5 ,261 employees as of late 2025. So this would represent roughly a 20 % riff, not block level cuts, but more in line with what we're seeing at Oracle, Meta, some other tech companies that are going through a transformation. On monetization, the recommendation is to improve monetization, which I think is a good idea for any business.

23:36Eddy Cue:But they say AI will massively accelerate product development and enhance advertising monetization tools. Again, everyone by this point should be well aware that Meta has done a fantastic job in leveraging AI, ML, to just make a better and better and better ads product. And they say product-led improvements across users, advertisers, and subscriptions to break out of Snap's monetization ceiling. Then they say deploy AI properly, monetize Snap's proprietary AI data sets, and then concentrate AI partnerships on clear winners like Gemini, OpenAI, and Anthropic. Again, Snap partnered with Perplexity.

24:23Eddy Cue:And it seemed like Snap got a fantastic deal out of that. It was something like a$400 million deal, if I remember correctly. some of it was stock and perplexity yeah but there was a huge cash component yeah and uh unclear if the other companies mentioned gemini openai or anthropic would have been able to match that how aggressive perplexity was getting and so it's possible snaps logic was hey we can we can do this deal with perplexity and then it's it's it's native in our app it's easy enough to swap it out at a later date. Let's like basically take the cash while we can get it. Yeah. I saw this post from Sean Frank that somewhat relates.

25:05He was talking about one of our sponsors, AppLovin, which I will tell you about in a second. But he said, in less than 12 months, I've spent$2 ,872 ,000 of my own money profitably on AppLovin. I don't own the stock. I don't trade the stock. This is the net amount that left my bank account. And he shares a couple other points. And he says he's spending 17 ,000, so 2.8 million on app loving, clearly like a scalable, large platform spent 17 ,000 on Pinterest, 266 ,000 on Reddit. You have to imagine that meta ads are up there. But the question is like, um, for a lot of advertisers, snap has not become this like, Oh sure.

25:43Maybe you don't get, maybe it's not gonna be your number one platform, but it's like in the marketing mix very regularly. And I think a lot of that should start working. even if the pool isn't super deep, even if you don't have 99 % of your customers there, maybe only 20 % of your customers are there. But even if they're there, you should be able to find them and AI can help that. And so I would expect that if this works, you'd see like really solid data from Ridge saying like, oh yeah, we're spending on Snap.

26:10Eddy Cue:Yeah, and they've been experimenting with Snap as far back as 2018. I'm sure. When I was hanging out with Sean and Connor back then, they were getting results on Snap, but there was a ceiling. Yeah. So then finally, they want to, on the governance side, commit to investing in safety and capital return, use newfound cash and profitability to further invest in privacy, safety, and parental controls, and allocate new cash flow generation to capital return and demonstrate conviction in SNAP's creation. Yeah, it's interesting on the parental control and safety side. I think SNAP has been able to stay out of at least Lanier's targets.

26:50Eddy Cue:Oh, they ended up settled. Yeah, they settled before it went to trial. Oh, okay. Meta and Google fought it. Interesting. And so, yeah, I'm not exactly sure what that means, but I think that they've been trying to sort of like step back from all of that. Yeah. And then on the corporate government side, giving shareholders a vote can unlock a multiple re-rate through broader index. Oh, interesting. And enabling one vote per class A share still preserves Snap as a founder-controlled company. Yeah, I think it probably is like 10x voting power or something like that. Anyways, there is a 70-page slide deck that they put together with all these different recommendations.

27:25Well, the market's reacting really positively to this. And I think Evan Spiegel has shared some statements that sort of mirror this, actually. It seems like there's maybe a little bit more reception than you might expect. The stocks have 14 % one day after publishing this piece, says Bose Weinstein. Adam is a rock star in the making. So smart, definitely worth a follow. And that's IrenicCap.

27:52Eddy Cue:Kerry, no, interest gave some feedback on Irenic. He said, a few critiques feedback. Daily opens are not equal to time spent on app. App love and meta clearly have higher time spent on app, but therefore parody on monetization is flawed. Arguing that Snapchat can hit targeting levels of meta is farciful. The amount of data that meta has on me versus Snapchat is astronomically different. I guess I'm open to being proven wrong since you compare it to Applovin, who IMO has always used other targeting sources. Three proprietary data slide is a one-time flash in a pan moment. Sure, some companies are selling deranged amount of data, but that's not lasting MRR, ARR monetization, although it is fair to call them out on it.

28:32Eddy Cue:I like the monetization per user slide. I think my feedback around Snap's ability to monetize relative to those peers' stands, literally screen time is much lower and you don't have data for targeting the way peers do. And what else was relevant? Calling out the founders' net worth growth was either God-tier, petty, or brilliant, or some combination of both. Fun presentation. Fun presentation. Well, you can go check it out. Irenic Capital has posted it. Save Snap now. You can listen to our interviews. I think a TBPN slide or quote made it into one of these presentations. Yeah, where was it? I think we're in there.

29:10Eddy Cue:Yeah, it was on the slide. AI should be an accelerant for Snap's core ads business. So Zuck had said on Q4 earnings, he said, we're also working on merging LLMs with the recommendation systems that power Facebook, Instagram threads in our ad system. Our world-class recommendation systems are already driving meaningful growth across our apps and ads business, but we think that the current systems are primitive compared to what will be possible soon. Adam at AppLove and said, if we believe that AI technologies are gonna be two times more efficacious in five years, just based off of that, if we do our job right, our system is going to be two times more predictive for its task, the sequence of problems that it's predicting in five years.

29:49Eddy Cue:And Evan said, our smart campaign solution suite, including smart targeting and smart budget, uses AI to identify incremental high value audiences and dynamically allocate spend across objectives, reducing the need for manual setup and ongoing optimization. And then in our interview, they highlighted Evan saying, as you look at glasses in the near term, I wouldn't expect AI to be a major accelerant. Well, since we mentioned them, let me tell you about AppLove and profitable advertising made easy with Axon.ai. Get access to over 1 billion daily active users and grow your business today. And let me also tell you about Figma.

30:26Agents, meet the canvas. Your AI agents can now create and modify your Figma files with design system context. It's in beta already.

30:35Eddy Cue:Yeah, so anyways, closing this out, Yeah, Irenica is clearly thinks that he says specifically snap is a special asset. He thinks it's it has a ton of potential. He's overall positive. He's just thinks like, you know, he really wants them to get in the game. And I think that honestly, a lot of people have felt the same way over the years, but have just ultimately been frustrated because some of these things that seem somewhat straightforward just haven't been done. Okay. I got to go back to the moon. We're going back to the moon. I'm going back to the moon because Artemis 2 is launching in three hours and 52 minutes and four, three, two, one seconds.

31:12Because Brandon Gurell wrote the op-ed today in the TBPN newsletter about some of the technology that they're using to document the trip. And it's a very different take, very live streamer coded of us. We only care about the camera equipment that's on board. Obviously there's a lot more that goes into it, but it's a fascinating deep dive. So let's read through this so that everyone has the update on how you can actually experience this, because there's a bunch of interesting deals that went into documenting this. So as you know, today, the NASA Artemis 2 mission will launch, sending the Orion spacecraft carrying a four astronaut crew on a high energy free return trajectory to get to the moon and back in about 10 days.

31:52It's longer than the Artemis 1 mission, which was six days.

31:56Eddy Cue:By the way, can you imagine this? uncrewed yeah can you imagine the stress when you're just like being sent straight out into space and you know there's a big turn coming up and it's pretty important that you don't miss the off yeah you can't be like texting and like miss the off the off ramp if you miss the off ramp you're going to saturn it's rough it's over for you um orion will enter a 24-hour highly elliptical orbit with an apogee 44 000 miles above the surface of the earth for context the iss orbits It's at 200 to 280 miles in altitude. So way, way higher. 100 times higher, 200 times higher.

32:32During the first day, the crew will test critical life support and communication systems. After reaching its apogee, Orion will essentially fall backward towards our planet. This will cause the craft to start picking up massive speed. If you scroll down, you'll see that the path is a little fishy. And I think a lot of the tinfoil hat crowd are going to be suspicious about the path that the rocket will be taking because it's fishy. It's fishy if you scroll down. Don't you think that's fishy? That's a fishy orbit. That's just a fishy orbit. I don't know. I don't want to be too conspiratorial about this stuff.

33:05Eddy Cue:It does look like a fish. It's a fishy orbit. It's fishy. Anyway, as it approaches its perigee, for those who are just listening on audio, it literally looks like a fish. as it approaches its perigee or the lowest point in its earth orbit the crew will conduct a systems review wake up the main engine system organize the cabin to make sure radiation shielding bags and water supplies are positioned to act as shelter in case of a solar flare put on their survival suits and strap in they're locking in so uh brandon said it's shaped like a figure eight i think that's very generous i think it's shaped like a fish and i think you should have just said fish there's a little bit of truth zoning that needs to happen right now but it's not bad uh it It is a figure eight.

33:45It is elliptical, but I'm going to still say it looks like a fish. After the burn, the crew will take more than four days to reach the moon. The craft just coasts there. The lunar flyby, where it will orbit the moon at a maximum altitude of 6 ,000 miles, a minimum altitude of 60 to 70 miles from the surface of the moon, is expected to happen Monday, April 6th. It will probably end up being the farthest humans have ever traveled from Earth due to the high altitude at which they'll orbit the moon.

34:11Eddy Cue:And remember, we still don't know if this is like Apple, another elaborate April Fool's joke. We could get to the countdown here and Jared Isaacman could say April Fool's. But let's continue. Imagine if that giant rocket, they cut it. It's just cake. That would be a good one. NASA is essentially aiming for a Netflix quality live stream on the flyby. And this is what the video creators, the content creators, the live streamers, this is what we care about. It will feature 4K UHD video streams that beam back to Earth with a three-second latency and some additional latency from encoding and terrestrial distribution using a frontier laser communication terminal that can transmit data at 260 megs a second.

34:51The stream will probably be compressed to 1080p for live video, but it will be saved in 4K. The craft has 28 dedicated cameras on board, externally mounted and astronaut handheld. externally can externally mounted cameras will be on the tip of each of Orion's for X-shaped solar arrays and they can rotate which will allow them to take selfies of Orion with the earth or the moon in the background we got selfie sticks in space space selfie sticks in space where this is sci-fi now these specific cameras will be heavily modified versions of the GoPro hero for black which is interesting the GoPro hero for is a pretty old camera but they probably had to start working on this.

35:31You know, it's a nine-year-old project, so they probably locked in the specs a long time ago, and then, of course, they started ruggedizing them. So that camera on the left -

35:41Eddy Cue:Making the GoPro even more rugged. Yeah, basically, they actually have to. There's a lot of radiation. There's a lot of pressure, and obviously no pressure when you're in a vacuum, which uses a 12-megapixel CMOS sensor and can shoot 4K at 30 frames a second. NASA contracted red wire space. So it's not all communism over here, Blake Scholl. There's some privatized companies involved in the process. Let's give it up for government contracts. Let's give it up for Red Wire Space. They ruggedize the cameras to protect them from the vacuum, extreme temperatures, and intense radiation of deep space. Some of the external cameras won't even generate imagery for the public.

36:19The optical navigation camera, for example, is a high-res monochromatic sensor that feeds image data of the moon and Earth against background stars to Orion's central computer, which runs machine vision algorithms to calculate the craft's exact position and velocity. They're doing slam in space. Inside Orion, astronauts use a tricked out Nikon Z9s, handheld cameras that can shoot massive AK video at 60 FPS. NASA actually entered into an agreement with Nikon to develop these for Artemis II. Astronauts on the ISS use unmodified Nikon Z9s, but because Artemis is going into deep space, the cameras needed to be ruggedized for the conditions out there and power optimized for the huge data transfers the cameras will need to make on the ship, which can be incredibly power intensive.

37:09Nikon even wrote a dedicated operating system for the cameras for this. Wow. Finally, NASA partnered with National Geographic to essentially record the footage for a documentary during the mission. The launch is expected no earlier than 6.24 Eastern Time, 3.24 Pacific Time, on NASA's YouTube. They are already broadcasting the live stream. The main program commentary starts at 12.50 ET. So they're going to start talking, and that'll be a lot of fun to follow along with. throughout the course of the mission national nasa will broadcast real-time coverage from ryan's cameras as bandwidth allows this will be on the agency's youtube channel so they're going to just like really embarrass us with this because 10-day live stream we've been talking about we will say oh we're doing a giga stream we're going to interview a bunch of founders from yc and it's like a four-hour five-hour maybe a six-hour show they're like 10-day 10-day live stream nice try tbpn yeah um italy rough for us yeah anyway the uh i i mean it's still

38:10Eddy Cue:it's still almost unfathomable unfathomable the amount of risk that these astronauts are taking on yeah and uh my thoughts are with them for sure for sure let me tell you about gusto the unified platform for payroll benefits and hr built to evolve with small and medium-sized businesses And let me also tell you about Vibe.co, where DTC brands, B2B startups, and AI companies advertise on streaming TV. Pick channels, target audiences, and measure sales, just like on Meta. So, the KitKat heist. This is the story you all have been waiting for. KitKats, the candy bars, were stolen and in massive quantity.

38:47The Wall Street Journal has a story of how the company reacted, how they turned a massive KitKat heist into crisis PR gold. We've seen this before. People were talking about Tucker Carlson having his nicotine pouch shipment stolen and how it sounded like the plot of a new Fast and Furious or Zoomer Fast and Furious movie. Well, something similar happened to Kit Kat and they took advantage of it and made the best out of it. So Kit Kat, of course, is owned by Nestle. But let's dig into what The Wall Street Journal had to say. Just how much are 12 metric tons of stolen Kit Kat bars worth? A lot of promotional gold, it turns out, says the Wall Street Journal.

39:27It was the brazen chocolate heist heard around the social media world. Oddly, this is the first time I'm hearing this. I don't know how I missed this. Had you heard of this before? Yes. You had. Yeah, I've seen this. I literally found out about this in the Wall Street Journal. I don't know why. I only heard about this in the reaction. But anyway, it's an interesting story. So over the weekend, Nestle confirmed that thieves had swiped 413 ,000 units of Kit Kats somewhere along their way from a factory in central Italy to Poland. Both chocolate bars and the truck carrying them remain missing, though no one was hurt in the theft, it said.

40:09with the swiss company lost in chocolate though it gained back in a public relations coup as did multiple other companies quick to hop on the meme bag bandwagon we need to pull up some of these memes i haven't seen any of them uh we've always encouraged people to have a break with kitkat but it seems thieves have taken the message too literally and made a break with more than 12 metric tons of our chocolate the company said in a statement i don't get it you you have

40:35Eddy Cue:to they would have had to steal the truck for it too yeah they stole the truck i mean it sounds like fast and the furious it sounds like they they stuck stuck it up and then you said get out of the truck yours you gotta call a cab the truck is missing the truck is missing the truck they took the truck wow they took everything uh yes it really happened uh a spokesperson confirmed that it wasn't an early april 1st joke uh taking their cue from nestle other companies soon joined in with some social media spoofing. We would like to share our thoughts and condolences with Kit Kat following their sad news.

41:09The account for Domino's Pizza in the UK posted Monday morning. Then it added on a completely unrelated note, we're pleased to announce that we'll be selling a new Kit Kat pizza. It's very silly. Charlotte FC, the major league soccer club in North Carolina, jumped on the same riff a couple of hours later on an unrelated note. We are happy to share that we will be offering roughly 413 ,000.

41:29Eddy Cue:I don't know about you, John, but I love when large corporations can just jump in on the fun. It's extremely millennial. This is like, this is my culture is not your costume. If you're not a millennial and you're the one posting this, like, stop. Only a millennial has the right to post jokingly as a corporate account. The discount airline Ryanair, meanwhile, simply posted a cartoon of photo of one of its planes with a face in the jet's mouth or five bitten off Kit Kat bars. Not long ago, most companies would have said little, leaving it to law enforcement authorities to disclose such a potentially embarrassing revelation.

42:03Now, any bad news is good news as long as a corporate brand can turn it into a viral meme.

42:08Eddy Cue:I wonder if this is a global crime ring that also came after ALP, the nicotine brand. The Tucker Carlson nicotine pouch, it's possible. It's possible, or it's possible that they're going to try to combine them, kind of one plus one equals three type situation where they think that merging Kit Kats and nicotine could produce incremental value. Who knows? Who knows? Typically, nicotine products are much more economically dense. So a single can of pouches might retail for anywhere between like$6 and$10, whereas a KitKat might be the same size but only retail for$1. And so stealing a truck full of nicotine products is typically like 10 times more economically valuable than KitKats.

42:54But who knows? maybe the thieves weren't thinking about economic density when they chose to stick up this particular truck. What do you think about this take? It's a masterclass in public relations. I agree with your intuition that, oh, this is not that funny. This is sort of just like corporate cringe. It's a little rough. I'm not getting belly laughs out of this, but just in terms of corporate comm strategy, this feels like the best of all possible worlds.

43:23Eddy Cue:Yeah, I agree with that. I think it's like a reasonable thing. Am I entertained? No. Do I think that it was worth doing? Yes. Yes. It's like the least bad way to deliver this. Does it make me want a KitKat? No. Also no. Yeah, but like I was not thinking about KitKats and now I'm thinking about KitKats. And I'm thinking about them. Well, now I'm thinking, I just think, I think I've never really, I've never really had a Kit Kat and thought, oh, that, that was, that was so good. Yeah. And so now I'm just remembering that why I don't care about Kit Kat. Okay. I think, I think that, you know, the, the, the two options were, you know, put out some sort of like serious, sad statement about how you got owned, basically.

44:09Eddy Cue:I think that could have been funnier. That might have been funnier. It was just, yeah, maybe a CEO viral video would have done the trick that kicked off. This reminds me of that McDonald's burger thing. They talk about that. All the brands. Yeah, yeah. Where does that go? Nowadays, other companies want to profit from potential buzz from a rival's misstep too. After a video of McDonald's CEO, polite bite into a burger went viral this month. top executives from Burger King, Wendy's, and Wendy's pounced with similar videos in a lighthearted dig at their competitor. McDonald's said his new burger, the Big Arch, got a sales boost for all the attention too.

44:47So it worked.

44:48Eddy Cue:Wait, they got a sales boost from having the CEO? They claim that getting dunked on, all press is good press. That's their claim. I don't know. It doesn't seem crazy. We found out about the Big Arch. We talked about the Big Arch. We sort of processed the value prop, at least. Yeah, I keep catching you leaving early. You're like, oh, I got to get home. I got to pick the kids up from school. And I drive by the local McDonald's, and John is there. He's got, in the passenger seat, 20 big arches, and you're just plowing through them. I don't think so. I don't think so. It's not for me. I'll stick to the other.

45:28Eddy Cue:Microsoft is in talks with Chevron. Bring it down. Engine number one of her$7 billion Texas power plant. That's good news. Exclusive talks. Okay. An investment fund with Chevron and Engine 1 over a long-term deal for a giant power plant in West Texas to provide electricity to a large data center campus. The proposed natural gas-fired power plant is projected to cost just$7 billion and initially generate 2 ,500 megawatts of electricity. 2.5 gigs. Yeah. 2.5 gigs. That's a huge campus. Wow. That's really, really good. Chevron, Microsoft, and Engine 1 have entered into an exclusivity agreement related to a proposed power generation and electricity offtake arrangement.

46:06Chevron and Engine 1 had previously discussed some details of their proposed power plant, but not the end user of the electricity. A deal with Microsoft would secure a long-term customer for the plant's output and help finance construction. The project, which could be up and running before 2030, still requires tax and environmental approvals as well as agreement of commercial terms. Microsoft, a longtime backer of Chachapi team maker OpenAI, is doubling down on constructing data centers. The pause is unpaused. This is fully, we are in unpause mode. Satya is going all in on AI. Access to reliable baseload power is emerging as a key challenge.

46:47One of the Chevron and Engine No. 1 partnership expects to address, given its extensive natural gas production in West Texas, and contracts for large turbines. What is LNG doing right now? Is natural gas spiking? It spiked in November and December, but is actually fairly low and is down since the start of the geopolitical conflict in the Middle East. So it at least it at least does, you know, Chase Lockmiller from Crusoe was telling us that the although the energy markets are somewhat intertwined, oil and natural gas do not move in lockstep because of the production supply consumption. Are you a net exporter?

47:35Are you net importer? So the pain at the pump does not always translate into pain at the data center, which is, I guess, good for data center operators. The Permian produces so much natural gas, a byproduct of oil, that it often overwhelms pipelines. As a result, some gas has to be burned off because it can't be transported where it's needed, making the region an ideal location for power plants. I believe these are called peaker plants, where the offtake, you know, normally it's just getting flared off, just going into the atmosphere. You capture that and turn that into energy there. But then you're stuck with electricity, maybe at some weird mining plant where there's no houses, so there's no one to power.

48:15So the next thing you do is you build the data center there. And that's sort of the quiche Lorraine of why Crusoe started mining Bitcoin at peaker plants, at natural gas plants. They had extra natural gas. They flared it off, used the very cheap or almost free energy to mine Bitcoin. Well, yeah.

48:36Eddy Cue:And of course, quiche is a wombo for a quirkian niche. Quirkian niche. When you're trying to explain the lore, you say you're Lorraine-ing something. Quiche Lorraine is the quirky niche lore explained. Apple just removed anything. Okay. $100 million Vibe Coding app that lets anyone build iOS apps with prompts from the App Store, also blocking updates for Vibe Code and Replit. The founder tried moving code execution to a web view, still rejected. The walled garden versus AI builders war is here. uh so apple has been saying according to some reporting in the information yeah that they don't have anything against vibe coding specifically yeah but the apps still have to adhere to guidelines yeah and this and this particular policy has been in place since like the start of the app store i i remember facebook uh you know dealing with this when they launched their initial mobile app which was uh it was it was an app but it would load a lot of html and it was like And that was sort of slow.

49:40They went back and forth. But there was always a question about, like, how much custom code could Meta or Facebook at the time serve within the Facebook app? Like, could you have FarmVille exist as its own app on the Facebook mobile app platform like it did in the Facebook desktop app? So when you loaded Facebook on desktop, you could load FarmVille basically as, like, an iFrame. And the FarmVille developer, Zynga, could change FarmVille, add upgrades, do whatever they needed to, and that would all be vended through Facebook. And, of course, your Internet browser doesn't care what you're loading, but the App Store is different and has had different terms of service for a very long time.

50:25mostly because Apple's one of their primary pitches to customers is that, hey, we review all the code that will run on your phone. And so we don't want that code to change after we review it. And so we have a process for reviewing it. And that reduces the risk of bloatware or crypto mining behind the scenes or spam or, you know, viruses, all sorts of different stuff that is the value prop for why people choose to buy iPhones. And so Apple has maintained that for a long time. Let me tell you about Phantom Cash. Fund your wallet without exchanges or middlemen and spend with the Phantom Card. And let me also tell you about Shopify.

51:04Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now agents.

51:11Eddy Cue:Apple specifically mentioned guideline 2.52, which is the rule anything apparently violated. Apps should be self-contained in their bundles and may not read or write data outside the designated container area. nor may they download, install, or execute code, which introduces or changes features or functionality of the app, including other apps. Educational apps designed to teach, develop, or allow students to test code may, in limited circumstances, download code, provided that such code is not used for other purposes. Such apps must make the source code provided by the app completely viewable and editable by the user.

51:48Eddy Cue:Anything launched on iOS in November with no issue, And the tool has apparently been used to publish thousands of apps in the app store. Yeah, you can imagine Apple being like, this app is pushing so much, so many other apps into our review process. We got to make it stop. The app lets users create and preview VibeCode apps on the iPhone. And it raised$11 million at a valuation of$100 million back in September. Apparently, Apple had been blocking updates to the app since December. So it's been basically a full quarter that it was frozen. But clearly, users are excited about the product. So I hope they can work something out quickly.

52:35Well, global venture capital dollar volume has incredibly spiked. Look at this chart. A bunch of interesting things about this chart. Market clearing order inbound. Indeed. It's pretty loud. So Ryan Hoover shares the chart from Crunchbase here. Q1 2026 was absolutely massive with global venture capital dollar volume spiking from just over$100 billion to almost$300 billion. So a massive, massive growth. A lot of this is opening Ionanthropic. But even if you remove those two rounds, it's still up 40 % quarter on quarter. What's interesting to me about this is what has been happening for the last 10 quarters before Q1 2026.

53:27Like it's felt like a funding boom. It felt like we had the Chachipiti moment at the end of 2022. 2023 was the start of like, okay, let's, you know, lick the wounds from the Zerp era and start ramping up venture capital. We saw new funds raised. We saw more deployment, tons of up rounds. We're always hitting the gong. And yet that the prior trend line was pretty flat prior to Q1 2026. It feels like venture is still in this like K-shaped dynamic. I don't know.

54:01Eddy Cue:Yeah. I mean, the answer here is that this is not venture at this point. Yeah. It's a big factor, right? You have these the biggest companies in the world investing tens of billions of dollars into single companies. Yeah. And so, of course, the chart looks insane because this is the largest, you know, the largest private financing ever. Yeah. Is Julie Black getting the scoop of the day? She says new funding, new model, new policy push. I think it's a scoop. She says exclusive, exclusive. OpenAI preps policy push. she says open ai will begin releasing a series of policy proposals next week meant to spark conversation about how to rethink the social contract it's going to be an interesting year

54:53is opera she asked the question is open ai's chief futurist prepping for a major breakthrough or just another hype cycle? And you can go and read this on Vanity Fair.

55:05Eddy Cue:Vanity Fair, the tech publication? Yes. I mean, Julia covers tech very well. Very, very well. There's a lot of good stuff. Metab over at OpenAI says, we are excited to share a new paper solving three further problems due to Erdos. In each case, the solution was found by an internal model at OpenAI. Each proof is short and elegant and the paper is available here. So there's a breakthrough for you. Yeah. Prince asked, would you be able to comment on the delta between the unreleased model and GPT 5.4 Pro? I'm reading correctly. You tried fewer than 10 identical prompts for each of the three problems with GPT 5.4 Pro and that GPT 5.4 Pro got the first problem correct twice.

55:47The result, which you link in the paper, plus one other similar solution. The third problem corrected twice. Interesting. I saw another interesting paper from Google that apparently just repeating the prompt twice improves LLM performance. Did you see this, Tyler? So if you say, like, you are Shakespeare, write me a poem, because of the way the tokenization works, like when it first processes that you, it doesn't know, like it doesn't, it starts thinking and it doesn't necessarily know what's coming, right? It doesn't know that Shakespeare's coming. So if you just repeat the prompt, the exact prompt twice, it improves performance on a bunch of different benchmarks.

56:32And it's this funny paper because it's like something that anyone could just do at home. It's like the classic like prompt engineering thing. Yeah, that's interesting. I mean, you would think that wouldn't work, right? Because like still like, you know, you like attend to all the previous tokens at the same time, this is the thing you think so i don't know yeah but i remember like early on yeah there's all these things where yeah you would prompt like you are shakespeare write me a poem and you way better poem than if you said write totally or like pretend that you're terence tau when you solve this math problem yeah way better right yeah yeah whatever in the training data yeah you know you know whatever it just knows to go down that path as opposed to like the other paths of like oh it it doesn't pull any april fool's jokes because it's not being an april fool's jokester yeah There's these basins or whatever.

57:14Yeah, yeah, yeah. If you get the personality basin where you can get the model to act like a certain thing, and it does much better on that. At the same time, I think this paper was specifically on vanilla LLMs, and I think reasoning solves that even more because you're basically repeating the prompt a bunch through unpacking it, compressing it, doing whatever you need to do to actually set the LLM up for success in answering the question. And so it might sort of hydrate through that process into something that yields those results. But interesting how simple some of these hacks are. Well,

57:50Eddy Cue:In biggest number news, yesterday, OpenAI announced the closing of their latest funding round with$122 billion. Yeah, hit the gong. It would be an honor, John. Boom. Probably the largest gong hit. 120 hits. 122 hits. Oh, 122. They upsized the round, remember? They did upsize for a ton. Yeah, so take him, hold out some of the highlights. post-money valuation of$852 billion. NVIDIA remains the foundation of our infrastructure, our training fleet, and the majority of our inference stack to continue to run on NVIDIA GPUs. We are now generating$2 billion in revenue per month, raising over$3 billion from individual investors.

58:33Eddy Cue:ChatGPT is the overwhelming leader in consumer AI with more than 900 million weekly actives and over 50 million subscribers. Our ads pilot reached more than$100 million in ARR in under six weeks. Momentum is just as strong on the enterprise side, which now makes up more than 40 % of our revenue and is on track to reach parity with consumer by the end of 2026. Codex now serves over 2 million weekly active users, 5x in the past three months, with usage growing more than 70 % month over month. And they talk about the vision for a unified AI super app, one app that lets you basically access every product that OpenAI is making.

59:12Eddy Cue:So, um, chat GPT codex browsing and other agentic capabilities. Yeah. I'm very excited about, uh, linking codex to chat GPT in some way, because right now there's really, I mean, you can sort of set reminders, but there's very little that you can do if your question requires building some sort of like small system. Like the textbook case that I'm thinking of is like, if I want to be, if I want like a deep research report of, you know, how the Artemis mission did when the astronauts land, I'm probably going to see that on social media. but what about just automatically letting me know one year from today?

1:00:07Remind me that the Artemis III mission is coming. Give me an estimated, are we on track? Pull all the news that I probably won't be following that closely because I'm not a daily consumer of space news, but I might be interested later. There's all these different things where you could track things, but you need to build something that just sort of runs.

1:00:29Eddy Cue:Entering a new era. Our astronauts are now Twitch live streamers. Are they on Twitch? No. Okay. YouTube. YouTube. But it would be fun to get over on Twitch. Yeah. Well, there's some other news. Apparently new iPhones are being packed into the suits of the Artemis 2 crew. Owen Spark says there's something very familiar about the iPhone look that will make the moon feel accessible. We are going to see the lunar surface through the same lens we used to capture our own lives every day. Oh, that's cool. Take some phones. There's an article in the Wall Street Journal about smartphone proves too big for most.

1:01:05We were debating this earlier, and I'd like everyone else to chime in too. The original iPhone, launched in 2007, had what was considered the largest smartphone screen at the time, 3.5 inches. The biggest iPhone now is more than double the size of the original at 6.9 inches. Or is, yeah, is now more than double. Wow. In 2019, the 2019 introduction of Samsung's Galaxy Z Fold introduced foldable devices to the masses, opening up like a wallet. It has a 7.3-inch display. And Samsung Galaxy Z Tri-Fold in 2026, when fully opened, unfurls to roughly the size of an iPad. Can, and the journalists in the Wall Street Journal asked the question, can a smartphone be too big?

1:01:54Tyler, what do you think? Yeah, I'm in favor of a smaller phone. Okay. This is a 16 Pro. 16 Pro. But it's not Max. It's the normal one. And why don't you want a bigger phone? It's just too much. What if you're on a plane and you want to pull it out and watch a movie or something? Yeah, but I feel like if I want a bigger screen, I have a laptop. It works great. But you don't always have the laptop with you. I mean, you always have it in your backpack? What if you're on a train?

1:02:23Eddy Cue:You've never thought about getting an iPad with a cellular connection and then maybe like a strap on the back that you could hold and be able to hold it up to your head and use it like a normal phone? I was seriously considering the iPad mini for a while. Just as an everyday carry phone, you could do it. I don't know if you can get a phone number on it. Is that even going to fit in my pocket? It would fit in my pocket. I tested. Does it fit in your suit pocket though? It would definitely fit in the suit pocket. You should get all of your suits custom tailored so that you can fit an iPad mini in here.

1:02:52I don't know. The chat confirmed. What are they confirmed?

1:02:57Eddy Cue:ZYT says. They're restreaming. Is on Twitch. They're restreaming. Let me tell you about restream. One live stream. 30 plus destinations. If you want to multi-stream NASA, go to restream.com. Hopefully they're on restream. Ryan Peterson. Yeah, what do you say? Is joking around. He says, Elon filed for an IPO on April Fool's Day. and apparently they filed last night so off by just a little bit according to But yeah, on pace for the kind of June listing June listing, okay Yeah, we'll have to track that That will be fun, I think a lot of people are really excited Should be good Austin Larson says, our team at Google is releasing more More details on the recent NPM Axios supply chain attack.

1:03:52Eddy Cue:Notably, we now attribute this activity to UNC 1069, a financially motivated North Korean nexus threat actor, active since at least 2018. We made this attribution based on their use of WaveShaper V2, an updated backdoor previously used by the group alongside clear overlaps in C2 infrastructure. Check the blog. and they wrote more about it on cloud.google.com. And yeah, we also have an expert from CrowdStrike coming on the show tomorrow. Adam Myers is going to come on and help us break it down, which we're very excited for. In a note to staff this morning, shared with Max Taney over at Semaphore, Business Insider said it is giving out a new quarterly AI award for best use of AI at the company.

1:04:45Eddy Cue:The winner gets$400. As soon as we talked about this this morning, Tyler asked us to create a similar award, seemingly signaling that he's quite confident. You think you can win this one, Tyler? He could take it home every quarter. I'm feeling pretty confident right now. We'll see. Once the incentive goes out, right? Show me the incentive. I'll show you the outcome. We're going to have a lot of people in the office gun for this. I still do want to run that experiment of like the race to see the true leaderboard of how long it takes everyone at our company to generate 10 ,000 lines of code as fast as possible.

1:05:22Because there's a bunch of shortcuts that you can take. Tyler thinks he can do it in like five seconds. But for other people, it's going to be a lot of copy paste. It's going to be figuring out how to open a text editor. Like everyone's on a different learning curve here. So I think the results would be interesting.

1:05:35Eddy Cue:Yes, I'm feeling pretty good about my odds there. Okay. Yeah, well, head over to business. Anyways, I think this is smart. I would say that maybe add a zero to the number. But given that BI is a massive company. Matthias Doffner knows that money doesn't grow on trees, okay? He's running that thing as a real business. He's a businessman. Rebecca Torrance is scooping. Scooping. The scoop is that Valor Atomics, a nuclear energy startup backed by Palmer, lucky. PL, backed by PL himself, has raised$450 million at a$2 billion valuation. There's no investor named here, so that's still unclear. But Isaiah and the Valar team have been on a tear.

1:06:26Eddy Cue:They're talking about building clusters of small nuclear reactors to power data centers. And they got some access to some government programs and government resources. And it seems like there's been a bunch of blockers that have been knocked down one after another. He's been on absolute terror. We'd love to have him back on the show to chat more. Augustus Dorico is having fun with it. Ooh, I'm going to invest in another energy company besides Velar Atomics for boobah, boobah, boobah reasons. That's what you sound like. Do you see how far they've come? Do you understand that they will make the world's energy?

1:07:03I love that he's pumping up his buddy. Always riding with the crew. Anyway, let me tell you about Vanta. Automate compliance and security. Vanta is the leading AI trust management platform. And let me also tell you about public.com. Investing for those who take it seriously. Stocks, options, bonds, crypto, treasuries, and more with great customer service.

1:07:24Eddy Cue:In other nuclear news, apparently Utah is going nuclear to keep ski runs cold. What? For the 2034 winter game. How does that work? Explain that to me. State says, this is in townlift.com. Okay. State says neutron-based atmospheric technology could lower ski-run surface temperatures by up to 20 degrees and offset drought impacts through enhanced snowpack retention. Interesting. Interesting because when you've given your case for Alaska historically, you've talked about how with abundant nuclear energy, you could take an otherwise cold area and just heat up. Yeah, you warm it up. But they're thinking about cooling things down.

1:08:00Eddy Cue:State officials confirmed Wednesday that Utah has entered a$1.2 billion agreement with a Vienna-based energy consortium to install three micronuclear reactors along the Wassock Mountain Range as part of an unprecedented effort to stabilize snow conditions on Utah ski slopes ahead of the 2034 Olympic Games. the project designated the waspac atmospheric retention and mitigation initiative or warme could make utah the first state in the nation to deploy nuclear-powered atmospheric cooling infrastructure at altitude utah is not going to sit back and let the climate dictate the terms of our 2034 legacy said a spokesperson for the governor governor's office of economic opportunity who confirmed the project has received preliminary approval under SB 114, the legislature's infrastructure acceleration statute, which removes local government approval authority over projects designated as state economic priorities.

1:09:04Eddy Cue:So apparently Juniper Peak selected as the first nuclear test site. I'm sure everyone at Juniper Peak will be thrilled about being able to ski around a nuclear test site. The three micro reactors are proposed at elevations above 9 ,000 feet in what project documents describe as thermally neutral subterranean containment pods. I like the sound of that. Designed to blend with the natural terrain. So, anyways. Well, we have some other news. Will Depew, I think he won April Fools with the joke that he says he is joining DeepSeek. He says, like many of my friends, I set a goal at the start of 2026 to become more Chinese this year, and I plan to fully follow through on it.

1:09:52After nearly three years at OpenAI, I have decided to leave and pursue new opportunities after much deliberation. I'm thrilled to announce that effective today, I'll be relocating to Hangzhou to join DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. dedicated to building Chinese-style AGI. I simply couldn't pass up the opportunity to become a billionaire in RMB. He had a lot of fun with it. We had fun putting up a card, and we're big fans of Will DePue. Well, without further ado, we have Eddie Q in the Restream waiting room. Let's bring him in for the TVP in Elsterdome. Eddie, how are you doing?

1:10:24What's going on? It's great finally to be here. Thank you so much. I wanted to be on this show, and I've got to tell you, I've gotten more text messages from friends about being on here, including my kids, than probably anything I've ever done. So it's great for you to have me. That's amazing. And what a special moment. What an amazing time. I would love to just start with some reflection. I want to hear particularly about your first decade at Apple. What was that like leading? What led you to the company? What were some of the first projects you worked on? Sort of take us through some of the early history.

1:11:05Yeah, I was lucky. I was a junior in high school when the Apple II was out, and I wanted to be an architect. And when I discovered a computer, I realized I wanted to be a programmer and engineer. And I said, there's two things I want to do. I want to work at Apple, and I want to meet Steve Jobs. Wow. And dreams come true. Here I am 38 years later at Apple. I came in as a programmer and was working on HyperCard and sort of the precursor to Blue Links with lines underneath. linking and I've been done so many things here at Apple I've had a amazing team and continue to have the the I'm working with the best people in the world at what they do yeah what was the lore of Steve Jobs like when you first sort of heard about him because you know my generation knows like the iPhone keynote there's videos online there's interviews there's whole books there's multiple books written but what was your experience learning what drew you to steve early on in your career i just think it's it's the innovation of creating these products uh that let people do amazing things uh and i felt that way when i was using the product the attention to detail of those products there was a connection that you could just feel yeah um and so it was more than just what you could see.

1:12:29And then it let me do things that I couldn't imagine doing before. And I think that's something that we've continued over our 50 years. Yeah. Can you talk about the launch of the original Apple online store? I feel like a lot of people assume that this always existed. No, it was a Herculean effort, I'm sure. What was the inspiration? What was the backdrop there? What was the mood like as you entered into that market yeah it was a crazy time because people forget but in those times we sold all of our computers through channels like comp usa and and local computer stores and the idea of building an online store and selling direct uh there were a lot of people inside of apple even that felt like if we did that the channel's going to walk on us and they're going to stop selling and uh Steve and we wanted to move forward and be able to do custom configurations so people could order exactly what they wanted.

1:13:28And we thought it was something that customers, you know, it was just beginning, but it was something that customers really wanted. And Steve and I and a small team worked on it and built it and launched it at the same time that one of our best products we've ever done was the iMac. The Bondi blue one with the clear. And so we launched the store and the Bondi Blue iMac at the same time. And I remember at the end of the day, we were wondering, you know, Steve, I came by his office and he's like, well, how did we do on the first day? And we had sold a million dollars worth of iMacs and we were high fiving each other and going, this is amazing.

1:14:07How did you drive people to the, did you just have apple.com? Were people already typing in apple.com? Like, how do you tell people that a website is launching before you can go viral on social media or do live interviews on, you know, how do you promote this? Yeah, we were lucky in that we had Apple.com already. And so some people were coming from that. And so it was that part was a little bit easier. And in those days, you relied a lot on press interviews in print. Sure. And so we did a lot of, you know, you'd want to be on the cover of, you know, a magazine and the front page of the newspaper.

1:14:42And so we had all of that pretty well. And I think our design when we did this, it was called Good, Better, Best. You could buy different configs and change them. But our design for shopping for a computer and a Mac at that time was something no one had ever seen. It had all of the things that we cared about. The simplicity, really easy to check out, easy to buy, all of the specs and the questions you would have. Things that were difficult when you went to other sites, I thought we did a great job, and it really resonated with customers. Yeah. Can you help me understand, like the services division of Apple is massive.

1:15:21It's a huge growth engine. There's so many interesting pieces of that. I want to go into a lot of those. But when was the first time in your career that you realized that there was something that you could sell or actually turn into a business line that was not a physical product and would live in this services category? When did services even become like a division or concept or an opportunity at Apple? Yeah, I think we started as a hobby. You know, there wasn't a lot there. It was very early days of the internet and we're doing things like email and things like storage in the cloud. But it was very, very early.

1:16:00The thing that was a big change for us was really music. And it was iPod plus iTunes. And that was something that was, it truly revolutionized music. And it really gave us a whole different perspective of what services can do when you take the hardware product, in a sense, the operating system and the software and the services and you tie them together, which is something I think we do better than anyone. it really showcased when we did iPod plus iTunes. And so all of a sudden we did that. Not only did we do it for the Mac, but we also did it for Windows. And so it opened Apple to a whole new ecosystem of customers that had never used our products before, but were using iTunes and iPod for the first time.

1:16:47That was my first Apple experience was iTunes and iPod on a Windows PC. Yeah, and now I have 25 Apple.

1:16:54Eddy Cue:My first is I was so loyal to Apple products that I refused to get a PC for gaming. So I worked, I probably reffed like 300 soccer games, like some absurd amount to get the maxed out MacBook Pro at the time. That's amazing. Because I was just so, so loyal. And I was like, I've got to, if I'm going to play video games, I'm going to do it on Mac. Yeah, it's great. You know, when we launched iTunes on Windows, I remember we did a poster and Steve called it on the presentation. It was like hell froze over. What was actually getting iTunes off the ground like? And how was it different than the other just motions that Apple had developed?

1:17:39Because it's not only a software product, but it's deeply linked to rights holders and agencies and musicians. And you have to get so many different groups. It's it feels much more permissioned than just building a computer and selling it Of course you need manufacturers and you need a lot of people on board to build a computer But it's a very different go-to-market or building motion like how is that different? Yeah, it was painful because I Think there were three three pieces you had us you had the label and you had the artist. Yeah We were really good with artists. So which is something we've always been about the creators And I think when you look at all the things that we've done, the two primary people that we focus on and think about are people that are the end customers that are using it and the creators that are creating all these incredible products.

1:18:32So we had a good relationship with musicians at the time, but we really didn't have any relationship with labels. And ultimately, they did control the environment. And at the time, they had a different perspective. It was really the beginning of Napster and piracy. And instead of thinking about how to move forward into a future, their view was to lock things down and really stop it. And as you know, when you have something that's better like that, there is no stopping it. And so we went to the labels and we had this idea of selling songs at 99 cents. And they kind of told us to go pound sand. They weren't really interested in us at all.

1:19:12and their idea was they were going to build uh some music services so there were five or six major labels and they built two music services and we told them like what you guys are doing is not going to work they had different pricing for each song they had different rules sometimes

1:19:26Eddy Cue:you could buy so they would like price a hit higher than like some random song on an album yeah i mean it was all over the map yeah and and part of part of the pushback against like just $99 a song. 99 cents, I mean, is like, you know, typical Apple style. It's just like, let's just make it simple, easy to understand. But was their pushback like kind of concern that people, you know, hey, we're used to getting people to just buy an entire album. And maybe what's going to happen if people just buy, you know, a song here or there? Yeah, the problem was whether you sold it at$1.29 or 79 cents, that wasn't going to change that.

1:20:05The key to it, there were two keys to 99 cents that we really believed in and people didn't see. There were two primary things. Number one is, when the price is 99 cents and it's consistent, you never have to think about price. And so you would preview a song, decide whether you like it or not, and if you did, you bought. And so there was never any transaction, a billing transaction that you had to think about. Because you knew it was 99 cents, it's not a lot of money at the time. and it was really easy to do. The second thing was that people could never do that because at 99 cents, if you're charging a credit card, you would lose money because credit cards have a fixed fee and they have a percentage that you pay.

1:20:50Well, the fixed fee and the percentage on a 99 cent song was like a quarter. And the vast majority of the money went to the labels. So every time we'd sell a song, we would lose money. And so nobody wanted to do that. And so no other service did that what we decided to do is as we were building this And I remember it was a huge discussion because we would lose a ton of money obviously if you're losing on every song We said look this thing is amazing. You're not gonna buy just one song You're gonna buy a lot of songs when you go on there. Yeah, and when you do that instead of closing the transaction on every single one Why don't we just combine them over a period of time?

1:21:29So let's keep the you know, let's keep the transaction open for a period of time Let's call it 24 hours or eight hours and everything you buy We're just gonna give you and then we're gonna charge you at the end and so therefore That's exactly what happened very few transactions were just 99 cents. Most of the transactions were multiple dollars and The fixed fee didn't matter interesting How how important was it to position? iTunes as sort of a step up from the status quo from like the Napster era and a positive, because I feel like anytime the economics of an industry change, there's natural uncertainty from artists.

1:22:11And iTunes did represent a change in the economic structure, but it was such a great countervailing force. What were discussions like at that time about positioning the economic opportunity that that artists would have in the new regime yeah I think we wanted during that time the music business was cratering okay from an economic point of view yeah and our feeling has always been the vast majority of people want to do the right thing yeah and they want to pay artists yeah and so but what they don't want is they don't want to be forced into something that doesn't make any sense or isn't really friendly or isn't the right way to do it and so we That's part of the 99 cents.

1:22:56It was part of like, in those times, you were burning a lot of CDs. They had limitations on the number of burns. We didn't want any limitations because that's not something a customer would understand. And so our feeling around this was, if you let us do this, you're going to grow again, as opposed to cratering. And I remember Steve asked me once before we had launched, he says, well, what is success around this? And I said, honestly, I don't know. I'll go ask. And so I went into Universal Music and I asked them, what's success for you guys in this business? And they said, well, if you could sell a million songs in a month any time in the first six months, that's success for us.

1:23:42So I came back. I said, okay, that's the goal then. We sold a million songs in the first six days. That's amazing. I love it. That's great. Yeah, so it's like that's what we, you know, obviously it surpassed even our expectation, but it was an example of if you give people the right way, people are willing to pay, but it has to be done well. So talk about the shift to subscription, because it feels like a much more natural experience for all the Apple service that I subscribe to. What was the thinking? How long, like what were the hurdles along the way to get to the current situation with Apple TV Plus where you can consume everything?

1:24:29Was this just a market dynamic? Was it something that you saw in the future early on and it was more of how do we get there smoothly? What was the process? Yeah, the key to this is it's hard to remember this now because we're so used to it. but it's having internet connectivity anywhere you are. Sure. And all the time. And pretty much, it's almost impossible now to be anywhere and not have actually fast internet. Yeah. And so that allowed a whole different thing because before that, you didn't have, one, you either didn't have it, or two, you were paying by usage in a sense. So you wanted to limit the amount that you actually used.

1:25:10And so things like downloading and keeping things on device all the time was really important. When you have unlimited, in a sense, internet access or a network access, then you can provide all these capabilities and not have to worry about whether you have it downloaded or not. It's now invisible to you. You don't even think about it. Most of the time, we put things on device just to cache them or whatever, but you don't need to worry about whether it's on your device or not. Yeah. We have a question from the chat. it's a bit random but i'd love to know your uh your favorite keynote moment throughout your career that's great i'll say look there are two there's a personal one which was the first one when we launched the imac and the apple store because that was the beginning of turning apple around and it was a big moment for apple we were it's hard for people to imagine this but apple was going bankrupt at that time.

1:26:08And Steve came back. And that moment was the beginning of a change where at least we knew now that we weren't going to go bankrupt. And so it really gave us life. And so it was an incredible moment. And I remember going backstage with Steve after it was done and hugging, actually, because it had gone so well. And we knew that was a big step. the second one and and honestly now in hindsight i was i completely underestimated it was the iphone launch it's the only time i made my wife and my kids my two kids come to the event they were eight and eight years old and i was like this is a historic moment because i had had the ability of using the iphone uh for a few months before we launched and played with it and and and it was just amazing to it's it's the coolest best thing i had ever seen in the world and so i thought this is going to be amazing now i completely underestimated it because now you look at it and go it's like i don't even know what the world is like what would we what would you do without an

1:27:13Eddy Cue:iphone yeah what what lessons from steve or kind of memories do you find yourself coming back to the most in the present day of Apple? Well, I think something that people take for granted, but nobody worked harder than Steve. These things don't come easy, and he was the hardest worker of anybody I know. How did that manifest? Like long hours? Just de-focus? Yeah, it's focused because it's focused and long hours. What it was was there are only two things that mattered to Steve, And I think when people ask me what's the difference between Tim and Steve, the reality is that's not the right question.

1:27:55The question is what's the same things between Tim and Steve. And, you know, their work ethic, they worked harder than anybody. They were completely focused on two things, their Apple and their family. Those are the only two things that mattered. And the third thing was the attention to the products themselves. It was about the products and what we delivered to customers. believe it or not, not the financial results. That was a secondary function that you obviously needed to keep going, but it was never the primary thing. And so those three things are something that I still, you know, take to heart and I feel I, you know, that's what I try to do and how I feel.

1:28:36Can we talk about F1? I love that there's a movie and also you can watch the, you can watch the actual races. This feels deliberate. What's the strategy? It seems to make a ton of sense, but how long has this been cooking? What's the thought process?

1:28:52Eddy Cue:I remember it must have been last year, John had talked about this on the show, wanting this to happen to see it come together the way it has is amazing. And it seems like soccer or football sort of faced a similar strategy, but I'm very interested in how you see different media properties connect together. Yeah, look, the F1 thing is personal. One, I've been an F1 fan for a long time. You know, I learned about F1 by going to the library and reading magazines because, believe it or not, F1 just wasn't televised at all in the United States. So you didn't know anything about it. So I knew a lot about it.

1:29:30Stefano, who's the CEO of Formula One, is somebody who was at Ferrari and then later on at Lamborghini, and I've known him. So when he took on Formula One, I remember meeting with him in London and saying, you You know, we're not quite there yet, but someday I hope we can be working together on F1. So I always envisioned that there was things that we could do that no one else could do. The movie came about separately, not kind of related, but this idea of doing a movie and Jerry and Joe Kaczynski. It was really Joe's idea. And I just love the idea because there hasn't been a huge racing movie.

1:30:10Most racing movies have not done that well. Yeah. And I thought there was a real opportunity with F1 to tell an incredible story. And Brad Pitt. And the cars and the excitement. And that we would, for the first time, had enough technology to show what it was actually like to be in an F1 car. Because when you watch on TV, it kind of looks like they're on a Sunday drive. It looks pretty easy. You don't get the G-Force. And so we had these ideas of taking the iPhone camera and putting them in all over the cars and different ideas that we thought would give that experience. Now, the movie took a lot longer because we had to go through COVID, strikes, all kinds of things.

1:30:48But it turned out spectacular. And when we when we would show the movie, one of the questions we would ask to people in the U.S. is how many of you have seen an F1 race? And the truth is very few hands were ever raised. And then after the movie, you asked him how many people would want to see an F1 race. And, you know, every hand went up. And so we thought, wow, if we did this together now and these ideas of how we can really innovate on the on the whole experience of what it's like to watch an F1 race, we really can make a difference here. And it's been great. We've done three races so far. The ratings are way above what they've ever been in the U.S.

1:31:29and uh so and we're just getting started but things like multi-view 30 percent of the people watching f1 races uh are watching with multi-view so they can get different cameras uh see their favorite team yeah uh so it's it's it's definitely changed a lot of how people are experiencing it

1:31:47Eddy Cue:that's amazing product request apple apple racing sim you'll have two buyers

1:31:57we just did uh vision pro yeah uh with uh sim racing so you could do that okay there you go

1:32:06Eddy Cue:we got it for you yeah the racing and the automotive world has a man on the inside so we're i'm the strongest supporter of the vision pro i watched another movie in it this weekend jordy if i call if i when i call john at 10 p.m on a friday night he's always he's always I love the product. I'm a huge fan. Anyway, thank you so much for taking the time. It's truly been an honor. It's truly been an honor. Congratulations on 50 years. What an amazing accomplishment. We'll talk to you soon. I appreciate it. Thank you. Thank you. Have a great rest of your day. Goodbye. Let me tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk.

1:32:43Secure every agent. Secure any agent. And let me also tell you about Finn. The number one AI agent for customer service. If you want AI to handle your customer support, go to fin.ai. And without further ado, we have Aaron Tarazas from Gusto coming into the Restream Waiting Room. Now the TV PN Ultra. Aaron, how are you doing? I'm doing great. Thanks so much for taking the time. We were very excited to dig into some of your research and reporting. But since it is the first time on the show, would you mind introducing yourself, some of your background and what you do for a living? Sure. No, I'm in a condense by training, currently working with Gusto.

1:33:19So previously been through the tech industry at Glassdoor, at Zillow, at a company called Convoy. Started my career at the Treasury Department. Okay. So maybe start with like what is unique about being an in-house economist? What data do you have access to that others might not? What data is anonymized? How do you do your job and what are you actually looking for? What are the questions that you're trying to answer? So most economists play with this public data that the government produces. That's great. It's been around a long time. The really exciting thing about being an in-house economist is that there's all of this data behind this private walled gate.

1:34:00We get to understand what is happening in the economy. You know, sometimes we pick up on things before the official data say. Sometimes we pick on trends that the official data aren't necessarily showing kind of what's happening beneath the surface. So it is this really exciting place to be as an economist. Okay, let's go into some of the findings. But first, I want to know about data skew and how you account for the type of customers. Gusto is a sponsor. We talk about small and medium-sized businesses. So I imagine that you have some sort of calibration step where you're adjusting your internal data to some sort of benchmark to make sure that you're not biased towards your particular sample set, right?

1:34:40Absolutely. Such an important point. Gusto is like a small business payroll platform. We do reweight our data so that it reflects the broader labor market, the broader small and medium sized business ecosystem. And how do you do that? Do you have ground truth data that's public that you can reference against? Yeah, exactly. So, you know, the government produces this census of all small businesses based by an industry and size. And so we take that distribution and match our internal data to that distribution. That makes sense. And then, so, I mean, the big question is the small business jobs report.

1:35:14What is the health of the American economy? What is the health of the small business community? What questions were you trying to answer and where did you wind up? I mean, every month we get this headline jobs report from the BLS. We're watching for this Friday. I think what we're trying to do with this small business jobs report is get a much firmer pulse on what's happening with small businesses, in particular, so businesses with less than 50 employees. You know, that is the majority of businesses in the American economy, but just a small minority of total employment. So it often gets buried in the headline jobs report numbers.

1:35:50And we know they're a driver of job growth. They're a driver of innovation. They're a driver of kind of the bread and butter work that gets done in our economy. Yeah. And where does net hiring stand? Give me some of the headline numbers. What are expectations? What should we, by default, assume is happening? I mean, there's a lot of doom and gloom in the economy, honestly, broadly, because of geopolitical conflicts and nervousness around AI. But what are you actually seeing? You're right. There is a lot of doom and gloom. Our headline jobs number for March showed that small businesses created about 120 ,000 new jobs.

1:36:25In March, that's an exceptionally strong number. That's the strongest number we've seen since 2022. And, you know, I think it's important to note that's not an outlier. Other kind of payroll providers are showing similar strength. So I think this narrative that we all have in our minds of the great freeze kind of businesses paralyzed by uncertainty. That's in some ways last year. I think 2026 is shaping up to be the great unstucking after being stuck last year. That's amazing news. Do you have any thoughts on, I can throw out a bunch of random theories off the top of my head, but this is paired with just yesterday, Oracle is slashing staffing over costs.

1:37:07There are big headline attention grabbing job cuts that are happening at name brand companies. And that news will fly very far because everyone knows Meta, everyone knows Oracle. And maybe they don't know that the small business just down the street actually increased headcount by 20%. But they only added two people, so it's not going to be on the cover of the Wall Street Journal. But is there some sort of, you know, how crazy am I to think about, you know, the large companies maybe being overstaffed and maybe changing their business model. So they need to recalibrate, whereas small businesses are benefiting from artificial intelligence, maybe trying to do more, trying to expand, trying to compete.

1:37:54And so they're actually expanding their labor force, even while the bigger companies have much larger questions to ask about what happens over 20 years. Yeah, big business grabs big headlines. And you think about the shocks that our economy has experienced over the past year, the past few years. Big businesses are kind of, I say, like a big freighter. They're a little bit more resilient to the tsunami, but they have a really wide turning radius. Small businesses are like a schooner. They have to adapt and pivot really quickly in response to these shocks. And I think we saw that. Big businesses are just catching up, whereas all these shocks we experienced last year, small businesses are already ahead of the curve and are thinking about what comes next.

1:38:40Eddy Cue:Yeah. If you were tasked with doing job reports for the federal government, how would you approach it, given your exposure to having access to data sources like Gusto? Everyone by now is kind of used to seeing a jobs report from the government. And then a few months later or six months later, you get these massive revisions. And it's sort of confusing. And it feels like we have all this real time data and like the data of people like adding an employee to Gusto and starting to pay them is like much, much more obviously accurate real time. And so it feels like the process could evolve. But what's your view?

1:39:22Yeah, I was a government economist. I have enormous respect for government economists. They have a big job. Their job is to look at the 40 year trend in the economy. These are tools that were designed over multiple decades. The reality is there is so much data exhaust, real-world exhaust right now in our economy that we can capture. And companies like Gusto, we don't have data from 1963, but we have much better data from what's happening right now. And so I think the two go hand in hand. They complement each other. One's going to give us that longer, more stable view of the economy. The other is going to give us these deep windows into what's happening in different parts in real time.

1:40:04Do you have any insight into subcategories of jobs, how job titles are changing? I keep using this example that we're a small business. We have 10 people on staff. And we have Tyler, our intern, who's basically a full-time software developer, as well as a co-host of the show in a hybrid role. We have other folks who have vibe-coded whole systems. But in a previous business, if I was building a media company, I would have had to think extremely hard about developing like a software engineering organization. And I would have assumed that would have been extremely costly, a very big build versus buy decision.

1:40:44Instead, it was just, oh, we hired Tyler because we liked him. He can do a bunch of different things. And he starts vibe coding. And all of a sudden, we have six systems. And they all work really well. And so we've become like this hybrid tech media company. We're developing video games now, too. That happened randomly, but we're not a game studio. And so I'm wondering about just are you seeing any data that suggests that more people are picking up hybrid technology jobs or software engineering jobs or even if the titles are changing yet? Anything you're seeing on that side? Yeah, everything's in flux right now.

1:41:15I think all the traditional lines that used to be bright lines of businesses are being blurred. People are taking multiple hats. Founders are doing more things. You raise a really good point. You think about how does a big business hire? You ask for a head, the head goes to your manager, the manager says it to finance, the finance says it to recruiting, et cetera, et cetera, et cetera. And that's a week here, a week there. You're talking about months. When the small business hires, when you hire, they decide, oh, we need someone yesterday. Let's hire them next week or the week after. Yeah, yeah.

1:41:48What economic data points are you tracking that are potentially upstream of small business hiring in particular? Because, you know, we all know, you know, gas prices surge, consumers feel pain at the pump, they cut back on travel, the economy potentially decelerates. There's these like logical chains and then, you know, in the big companies, it might be like interest rates or something geopolitical risk or tariffs. So there's all sorts of things that could drive, you know, hiring at the large scale. What are the biggest needle movers that you would watch that might lead to some expectations about what might happen over the next couple months in the small and medium sized business market?

1:42:31Yeah. I mean, obviously, you look at things like kind of hiring and job openings, that's plans that are already in place. But almost more important than that, I think it's sentiment, you know, expectations, particularly among decision makers who have the power to shape the future. So, you know, things like kind of CFO expectations, CEO expectations, that's how business leaders are thinking about the future. And, you know, future is partly prediction, future is partly creation. Yeah. How good are those surveys? I feel like I've been a hiring manager for maybe like 15 years now. I don't know that I've ever answered a survey.

1:43:08Maybe I just did it and I forgot about it. But is it broad-based? Are you confident that enough CFOs are actually picking up the phone or answering the survey? Because you can imagine that some of the most high-performing managers are too busy hiring people to answer questions about whether or not they're going to hire people. Totally. This goes back to the conversation we're having about the strengths and weaknesses of different data sources. When the government goes out and surveys people, there's a lag in response, bias in who responds. When we're looking at payroll data, that's not kind of responding or not.

1:43:43I see what you're doing. And so in some ways, you know, that just gives, particularly for smaller businesses where people are busy, they're not picking up the phone. They don't have time to twiddle their thumbs, kind of responding to a government survey. that's capturing a real phenomenon. Yeah. How are you thinking about feeding that into Gusto's actual product offering? Because I could imagine if I'm in there hiring someone and I sort of get like the survey after I hired one, I plan to hire more people next quarter and it's just, you know, like leave a five-star review type thing on an e-commerce website.

1:44:19That would probably have very high, you know, survey completion rate. Are you looking at doing more to get like qualitative or quantitative data that's more optional from the user base? That's a really cool idea. I certainly have not thought about that. You got to pitch it. Yeah, serving people in the platform, that's something we haven't really done a lot of, but cool idea. I would love that because even just more qualitative, just, you know, with LLMs, there's so much that you can do to compress down qualitative data and whatnot. Like there's so many interesting things you could do.

1:44:47Eddy Cue:How is AI changing your approach to work? It's a great question. Yeah, obviously kind of it's completely transformative. It's transforming the way, certainly I work as an economist. I think it's transforming how everyone works. It's transforming how businesses start and hire. I think there's just a growing body of evidence that AI is making it a lot easier to start a business, to incorporate, to do all of that front leg work. And on top of that, it is changing kind of the skill profile that you need when you hire. You need more of these people who manage processes, particularly at the start. Yeah.

1:45:27And then day to day, I would randomly predict like IPython notebooks a few years ago. Is it all like command line prompts today? How is that changing for you? Yeah. I mean, kind of, I still work a lot in Python, but, you know, I'm doing more things at once. And I think that's true for most economists, kind of anyone who's, you know, kind of in some ways it's lowered the barrier to getting that data. And that's great because that accelerates decision-making that helps these businesses adapt in real time and pivot to a rapidly changing world. Yeah, it's very cool. What a fun position to be in. Congratulations on the success.

1:46:09Eddy Cue:I'm sort of laughing in my head about imagining you if you had like a secret LLM subscription back when you were working as an economist for the federal government. They're like, how is Lisan Al-Gayy? How does Aaron do what he does? And you're just sitting there. That's true. Anyway, thank you so much for taking the time to come chat with us. We'll talk to you soon. Have a good day. Goodbye. Let me tell you about 11 Labs. Build intelligent, real-time conversational agents. Reimagine human technology interaction with 11 Labs. And let me also tell you about Cognition. They're the makers of Devin, the AI software engineer.

1:46:44or crush your backlog with your personal AI engineering team. Scoop. Scoop?

1:46:49Eddy Cue:From Charles and Business Insider, Meta is forming a new elite AI lab. What? The tech giant quietly reorganized its recommendations division. Okay. Rexis. It formed a top AI team run by TikTok's former head of growth. Reorgan. Rexis has been luring OpenAI, Google, and Amazon talent. So just one more lab. One more lab would fix me. That does seem like a great place to apply AI. So yeah, I feel like the recommendations on Instagram are pretty good these days. I do feel like I get caught in, well, what's that called, Tyler? Like the local minima or like the basin or whatever. Yeah. Is that, is that the term you used, used?

1:47:32I said basin. Basin. Yeah. Like I feel like it's like, okay, yeah, I watched like a few snowboard videos. Like it doesn't need to be all snowboard videos. We can go back to some vibe reels about, about airplanes or something like that. it goes back and forth but in general i've had a good experience and i expect it to get better um let's pull up this photo uh from kirk evans uh showing an advertisement for for tiger woods it says step into the mind of tiger woods and kirk evans says i think i'll pass oh very very unfortunate situation uh anyway uh we can move on um well what what else is going on let's see there's news from perplexity

1:48:17Eddy Cue:yeah I can't find any actual reporting on this Zero Hedge said perplexity was accused of sharing information of its users with Meta and Google but there's zero source here. Yeah and that might be within the terms and conditions it could be anonymized the sharing personal information I mean I guess it's personal if it's de-anonymized the quote here from the All In podcast is every single penny we make, we make profits on that, but the overall company is still yet to be profitable. So I think he's talking about inference margin. The way that's phrased tells me we make money on subscriptions, saying for every single penny we make, we make profits on that.

1:48:59Eddy Cue:To me, it kind of ignores free users potentially, right? Because they've tried ads and they killed it. Sure. So they're basically saying, like, on our consumer subscription business, we make money. Yeah. But the overall company is not profitable. Yeah. And you would expect that. Like, for users, if there's not ads, like, it's going to be negative margin on those. You just hope that it's not that bad. And you hope that it all blends out across your subscription profits. Well, new job alert for Omar Shaheen. Yeah, this is interesting. He says, a new job at Microsoft. He's bringing open claw plus personal agents to Microsoft 365.

1:49:36My goal is to help usher in a new generation of workplace proactive assistance, ones that lighten your load by taking on tasks end to end and that can also step in proactively when they can help. Interesting to see this counterposition against the Facebook Manus deal. Isn't Manus open source as well? Is it? I don't remember. but you know it's like every company will need to have you know a partnership or an in-house LLM to vend into every service it does feel like the next couple years are going to be just like the chatbots popped up in every single app with varying degrees of success some of them are great some of them I wind up using and I'm actually glad that I didn't need to export the data and I can just ask the question directly in that app and I have faith that they're using a good model and so they'll give me a good result.

1:50:32And mostly I'm happy that I know that the data is going to be clean and also secured in that app, in that ecosystem, in that walled garden. I'm fine with that. It does feel like the open claw agents actually build some software to answer our question is going to be vended into another box or another button with a star or something. Maybe it'll be a claw, just like how every app got a star during the AI boom or the chatbot boom. Now the agent boom will need a claw next to the star. Star will be chatbot. Claw will be, okay, you want to do something where it actually goes and works through this. I feel like Google launched something similar, not necessarily a full open claw personal agent, but they did announce something along the lines of an agent for your inbox that processes all your emails.

1:51:22They said goodbye inbox zero. Was that the news? I don't know if we have it pulled up. I can search for it. Was it Sunday? Who posted it? Goodbye inbox zero. Let's see if this pulls anything up. I saw a friend of the show posting about it, but I will try and figure it out.

1:51:41Eddy Cue:Open claw setup that just deletes every email as soon as it arrives in your inbox. Yeah. It's the archive all. You know, were you ever an inbox app user? No. This was like the pre-Superhuman era. It was like the hot email app that a team of software developers, iOS developers in particular, built a very well-designed email app for iOS. And this was at a time when a lot of the other native iPhone apps for popular email services were still a little clunky. They didn't have all the great features and the inbox team sort of went and reimagined it using inbox zero as this philosophy And so if you're not familiar with inbox zero basically, it's the idea that at the end of the day You should have nothing in your inbox which to some people Sounds impossible because they have like 50 ,000 emails in their inbox.

1:52:39How many emails are in your inbox right now? How many you know you're you're a 10 ,000 text message guy you don't even do inbox zero on your I don't have 10 ,000. I have 20 ,000.

1:52:49Eddy Cue:20 ,000 in the inbox. I don't have 10 ,000. I have 5 ,750. That's really close. That's closer to 10 ,000 than zero. Anyway, the whole idea is that email is a metaphor for a literal office. It's supposed to be a desk, and you receive mail that comes to your inbox. If you want to send something, you put it in your outbox, and then your team will send that out. But once you've actually taken something out of your inbox read it processed it You can physically throw it into the trash or you can physically file it in a cabinet or you can physically Write a response and put it in the outbox Like this is the way people did business back before email in like the 60s when they were just sitting at their wooden desks Their mahogany desks and so the inbox team said let's copy that Paradigm and this was popularized in books and management philosophy for a long time But let's copy that and have the idea that you should get to inbox zero by the end of the day and so they made it very easy to Inbox to basically archive all your emails like there was this big long waiting list it took like months It was really really hyped in Silicon Valley.

1:53:57They eventually sold a Dropbox It was very good outcome, but it became a very popular product that everyone like had to install Because it allowed you to get through your emails much faster So the main feature that is now in basically every email client is that you could just swipe and archive the email Which does not delete it you can always search for it later But it says it just it's just a way if you're in a to-do list It's saying that this email I don't need it like I don't need it on my to-do list like the inbox is your to-do list If you archive it you can always search for it later. It's not deleted.

1:54:28It's there. You could also Shuffle it away in a folder you could delete it But then they also had a snooze functionality where you could swipe to the left or to the right or something and say, hey, I'm not ready to deal with this email right now. I'm not responding to it right now. But I, you know, this is a receipt or a ticket for a flight. And I want this to come back into my inbox the day before the flight or something like that. And so they made it really easy to both archive and snooze. And when you archive and snooze every email, you can very quickly get to inbox zero by the end of the day.

1:55:01They did a bunch of gamification stuff to make it fun and be like you got inbox zero like take a screenshot of it People were sharing the screenshots of their inbox zero lineups update. What's the update from?

1:55:11Eddy Cue:NASA Administrator Jared Isaacman. What's happened the Artemis 2 crew is boarding Orion. They're boarding. Okay. Good news video I'm on the edge of my seat. I didn't know what you're gonna say

1:55:23Can we turn up? Reed or Jeremy. So it looks like Andre is with Reed. And this is the locked in image. This is the fully locked in with 25 ,000 buttons. While Andre is buckling him in. That's a five-point harness.

1:55:41Eddy Cue:Tyler, would you stay in a little box that small for... Two weeks. Eight days. Four. Getting the displays ready so that they can monitor... $400. If I went to the moon, then... For sure. What if you're just sitting here in the Ultradome, you have access to a laptop, but you cannot move out of... Yeah, should we do a simulated moon mission where he has to live in a pod for 10 days in solidarity with the astronauts? That's good. That's good. In solidarity. In solidarity. You will... Wait, we can send you to the moon in VR. We can't afford to send you to the moon, but we can send you to the moon in aggregate.

1:56:20Well, we have our next guest in the Restream waiting room. First, let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Just do it. And let me also tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web apps, servers, databases, and more, while Railway automatically takes care of scaling, monitoring, and security. And without further ado, we have a proud top form. Arena Physica. He's a CEO and co-founder. How are you doing? Hey, Jordi and John. Great to see you again.

1:56:55Great to see you again. Great to have you. Welcome back. Yeah, it's been a bit. It has. Please reintroduce yourself. Give us a little backstory and then we'll go into the company. Yeah, perfect. So yeah, I'm Patak Pranade, CEO and co-founder of Arena Physica. We were on your show announcing some fundraising about, I'd say just about a year, a year and change ago, maybe a year and a quarter. And companies have evolved a bit since then. We started with bringing AI to hardware engineering. So AI hardware engineers, that's evolved actually quite a bit, as we've seen the whole AI space evolve, which is, if we think about modern hardware, it's software defined, right?

1:57:32That's happening everywhere. But software-defined hardware is really like electromagnetically governed. Like the key problems are the electronics interfacing with the firmware, interfacing with the mechanics, and it needs to obey physics and behave well. And it turns out humans can really use one of those forces of physics really well, electromagnetism. And it's a rate limiter for progress. And so what we released yesterday is the first foundation model for electromagnetism. Think about it as a large, similar to a large language model in architecture, but the training tokens, they aren't words.

1:58:05They're materials, geometries, and electromagnetic fields. So learning us, helping us speak another language, basically the kind of language of the universe. So what's the data set for that? Like, is there, see all the website? Yeah, for sure. So the April Fool's, it's actually just a chat app. April Fool's, it's a new LLM. It's a new chat box. Yeah, no, for sure. The data set doesn't exist. Okay, yeah. Which is why this has been hard. Yeah. So what did you do? Was it simulated or did you have to go collect this? How does that work? Yeah, that's a component. So the reality is like what we've done is basically build a data factory.

1:58:43We spent about a year and a half building that. And, you know, how do we get all of the data that a model of this sort of architecture, a transformer type model needs? Yeah. So it starts by creating random patterns. So that's layer one of the cake, if you would, is large amount of random generation. And that's different geometries with different material stackups and then running them through simulation to see how they behave. So that's sort of think about it as layer one of the cake. But, you know, the number of possibilities here explode. You add more layers. You think about modern design. It's happening on silicon.

1:59:13How does, you know, how do they actually behave in the real world? You can't explore that whole universe. You can't randomly explore it. And this has been one of the big bottlenecks is there's very few of these experts in the world. You know, I think Starlink is a great demonstration of what you can build when you've actually mastered such capabilities. But few people, you know, have achieved that outside of SpaceX. And it's bottlenecked by experts and simulation. So the second layer of the cake is our experts have been creating designs that work. And then the final step is we actually fabricate those designs.

1:59:45And then we pipe that back into the system. And so I think you guys might have something over there. We do. We do. We received something in the mail. But we have some weird news because it was intercepted. Somebody opened it and stole maybe half the package. I don't know if there's two things. I only have one. But I do have a screwdriver. and this metal box.

2:00:05Eddy Cue:It was presented in quite an amazing way. I think somebody basically was like, okay, this looks valuable. I think that's what happened. So I will open this. You can tell me what this is. And then if the other piece of the puzzle is missing, you can break that down and we can go find the thieves at some point. I will talk through the other piece. That is a booklet that describes the background of what you're opening. They stole the booklet? They stole the booklet. I know, weird. That's extremely weird. for the metal box. Yeah, the metal box feels like... It's probably a nation state. I have no idea.

2:00:38That's just so weird. We're not making this up. Like, we opened it. We received it, and it had been ripped open, and the booklet had been stolen. But, I mean, I hope that's not, like, intellectual property or something. But inside, it says, caution, electromagnetic super intelligence. Handle only if you are ready for the future. I love it. There we go. And so let's open this up. All right, so break it down. Yeah, break it down. What is this? Okay, great. So I'll fly through a bit some of the booklet for background and then I'll tell you what that is. But if you look at what the booklet is focused on.

2:01:09So what can we do with electromagnetism? What can we do with structures like the one you're holding? What you're holding is an RF circuit. So a circuit used for radio frequency communication. And if you think about what's happening right now, communication and sensing is kind of the big bottleneck for robotics, for satellite communication, for a lot of the bottlenecks in data centers. And that's gated by this small group of experts and these really slow simulators from about the 80s. And so what we've done here is actually create a design from scratch from nothing but a prompt. It's the way it works.

2:01:46And it's actually live on our website, on our research page, arenaphysica.com slash research. You can read the technical blog and you can play with the product. You can literally type in a prompt. And that goes to an LLM. And then the LLM basically passes it to our foundation model. So the LLM interprets the prompt. It's like, oh, John's trying to build a 10 gigahertz bandpass filter for his new satellite company. And it passes it off to our model. Because LLMs are not good at that.

2:02:12Eddy Cue:You guys sort of like hydrate the prompts to make sure it's structured and has like the necessary information? Or what does that pass off look like? Yeah, it's a great question. So the prompt, that's where we actually will rely on the LLM. So let's say you give an incomplete description. You're like, hey, I'm trying to build a space antenna, but you didn't specify that it's a filter. The LLM, you will make some good assumptions and say, are you looking for this? And then that prepares sort of the inputs that we need. And then it calls our model. So in a way, what I'm really excited about is I see a future where you have multiple AI foundation models working together.

2:02:50And, you know, we've got one that speaks English really well, which is your interpreter, which is your front door. but now it needs to go and create a structure which is physically valid so that's where it calls heavyside our em foundation model and so what you're holding is a 10 gigahertz bandpass filter which is one of those key components inside a modern phased array system which is used for radar okay for satellite communication um uh and so you think about there's a lot of conversation about data centers in space we're not going to fiber optically cable them together uh you know you're going to need to transmit that data again between them wirelessly.

2:03:23And so that industry basically started off in about the sort of 60s with phased array radar. Actually, I think the first one was back in World War II. And so what you've got in your hand is like, if you look at it, it's an alien geometry. It's a strange structure. You know, it looks, it doesn't look like something that came from a human brain, right? It's like, it doesn't look like a nice geometric pattern. It's not a line. It's not a coil. the crazy thing is that it works and Jordy just to be clear you're not supposed to touch that with your bare hands

2:03:56Eddy Cue:yeah it's over it's highly radioactive I'm kidding no I'm not convinced this isn't alien technology but cool story no it's very very cool so I mean I appreciate the example of like the connectivity in space data centers obviously that's like frontier but walk me through the actual like market map industrial scale of like i imagine there are like certain power law buyers of uh of rf equipment like it's everything it's in everything from like that bluetooth connected toaster to your phone to your car it's everywhere but what is the actual shape of the industry where what when i'm thinking about like subdividing the industry and getting down like the satellite industry is here, the smartphone industry is here.

2:04:46Like walk me through the map of the potential customer set. Totally. Yeah, if you think about that, there's a few different buckets. So if you can harness electromagnetism in RF, which is let's say application one, you've got satellite communication. So phased rays for a SATCOM. So that's your user terminal, your satellite. So if you think about a Starlink satellite, I believe it needs to track 64 different locations on the ground. So make 64 different beams. Sure. You have to do that with a phased array. Okay. So anyone launching, anyone in the space industry, which I think is exploding right now, needs communications between Earth and space.

2:05:24So phased arrays are key there. Yeah. Radar. Radar is another really big one. So if you think about that, you know, we've obviously got Raytheon as the incumbent here. But if you think about advanced radar, they're incredibly expensive today. And we have few of them on ships. You know, we should have backpack-based radar for soldiers, for counter drone. We should have a whole variety. We want to de-bottleneck this field and basically democratize it, almost like AWS-ify that for companies building for space, for radar makers. Also for companies, if you think about data centers, chip-to-chip communication.

2:05:58So these channels that let chips talk to each other are incredibly difficult to get right. In some way, you're trying to build a very bad antenna there because you don't want it to pick up random interference. You want it to send the signal you want. So just those alone are massive markets. Actually, we looked at the forecast for the phase ray and RF components, and it looks like this line. And I feel like it's something I would have done right out of school as, oh, great, let me take the historic numbers and propagate them forward. but it doesn't seem to account for the fact that we're expecting a boom in space.

2:06:28We're expecting growth in AI data centers. And if you think about imaging, I heard something from our chief scientist here that blew my mind. I always thought about radar as detection. You're trying to detect something, right? And he said radar is an imaging platform. So as you get to higher frequencies, like LiDAR, you've got this high frequency beam that's echolocating things like a bat. And it's useful for all weather imaging. So when light's getting blocked, you can actually use a radio wave to kind of create your LIDAR map. So if you think about that being used for robotics, which is another application.

2:07:02So each of these pieces we feel haven't been sized into the forecast. And so that's, I think, where a lot of the opportunity lies. And when we talk to customers, so, you know, we work with AMD and Andrel as examples. So our mission is partner with companies at the frontier that are trying to challenge existing cost models and push frontier technology harder. Those are sort of the places where we're seeing early adoption. And do you want to be like selling intellectual property like ARM or software for chip design or the actual finished chips? How vertically integrated do you want to be? Yeah, I would say like realistically as of where we are today, our model is closer to the model of like a Palantir or an ARM.

2:07:47Obviously, my own heritage, last company was bought by Palantir. And so partnering with companies using our whole tool chain. So this is the latest product, Heaviside, the EM Foundation model. But Atlas, our current product, is an agentic product that lets you use LLM agents to debug parts of the hardware process. So that plus our FDEs that are not just software engineers, but also electrical and RF engineers for deploy as partners to drive major outcomes. What we are seeing and what you're holding is a piece of this is, you know, the existing IP industry is let me license you IP. And this is where it's an open question.

2:08:22We haven't figured it out. But you now have an IP generator as a machine. And so we're really excited about the ARM kind of model as I think probably the most proximate to where we are today. and our way of having the most impact because we can empower more teams. Like our mission is to empower more teams to build more advanced stuff. If we think about a lot of the AI doom and gloom, we're like, where do the jobs go? You know, the conclusion we come to is companies just need to be more ambitious. Like that's how we create jobs. Let's assume we're working with AI. How do we enable you to do that?

2:08:52How do we enable others and more companies to do this? The last thing I want to say on that is if you look at what happened with language models, it was all possible because of scaling laws. And like, you know, before language models, There was a lot of machine learning where we were trying to go and solve problems at that use case level. We were solving like language translation, summarization, spam detection separately. But once you pointed a huge model at language as a substrate and open and published that seminal paper in 2020 on scaling laws, you could scale it up and it started to generalize.

2:09:23Now these things can do remarkable things. We believe we will see and we're seeing early hits of scaling laws for electromagnetics. And this might be true for physics in general, electromagnetism being one of the four fundamental forces here. But that's sort of what we what we think is possible. So each of the industries you talked about doesn't need a specific solution, we think, in the future. As you scale this up, the central brain could power any industry that's bottlenecked by electromagnetism. You know, phased array radar, SATCOM and data center interconnects being a starting point. Yeah.

2:09:54Eddy Cue:Wow. What's the reaction been from the experts that you've mentioned? Are they experiencing a chat GPT moment with the new model yet or you still need to iterate? How are people responding? Yeah, it's a great question. I think their reaction has definitely been really positive in some pockets and deeply skeptical in other pockets. As you can imagine, getting this stuff right is really difficult. I would say we're at like a GPT one moment here. We're not at a chat GPT moment. We have about tens of millions of data points in the training set. As I mentioned, this data doesn't exist. We had to make it all in the data factory.

2:10:31So that's sort of one of the keys. And so we're finding as the models get larger and as the data set gets bigger, they're capable of more. But today, what you're holding, a human could design an equivalent component like what you're seeing with a normal structure, and it would work. We're not yet at a superhuman point with the model. But what we are is transforming the cost structure. This is 800 ,000 times faster than a commercial solver, for instance. and the expertise is not locked up in an expert. You could go and just get that kind of on demand from the AI. So that's what we're seeing today.

2:11:04And I think it will take some time to move up that stack. But that's why I'm super excited about the scaling laws. And I think from, as I mentioned, AMD and Andrel, I think we're fortunate enough to get partners and customers that I think are really on the frontier that are leaning in and believing in this. And so we're super excited about that. Can you talk about the FDE model in this case? I mean, I feel like a lot of people don't even understand the typical engagement of having Palantir FDEs inside of an organization. They might be more familiar with like McKinsey coming to the organization, doing a de-layering or some management consulting engagement.

2:11:42Like how long is a product lifecycle, development lifecycle? How long do you imagine that a typical engagement with FDEs might be? Is it going to be the same as Palantir, different? Like how is this unique? Yeah. So right now, you know, we definitely see the engagement models being, you know, you're partnering on an array of problems. So we start with one, for example. Let's say that new product innovation cycle is, you know, for something like a new chip or a new data center that might be in the number of like a year. It might be a couple of years. But what we're seeing is then those architectures are used across multiple product lines by a lot of these companies.

2:12:19And so could we become a partner long-term across all of these electromagnetically limited products? That's sort of where we see sort of that outcome driving, you know, that driving really big outcomes across the stack. Yeah. Yeah, that makes a lot of sense. Well, congratulations, Jordy. Anything else?

2:12:35Eddy Cue:Amazing progress. We will put this to use. We will. As many of these as you send to us. The entire show will run through that. We'll figure it out. We'll do it. It might take a lot of queries for us. Incredible progress. We might need our FTE. You have to send one over next time. And let's have you back before another year passes. Yeah, we'll talk to you soon. Have a good one. Thanks, guys. Great to see you. Great to see you. Cheers. Let me tell you about console. Console builds AI agents that automate 70 % of IT, HR, and finance support, giving employees instant resolution for access requests and password resets.

2:13:07And let me also tell you about Lambda. Lambda is the super intelligence cloud building AI supercomputers with training and inference that scale from one GPU to hundreds of thousands. and I think we're in the Lambda lightning round now or are we in something serious? Let's fire up the Lambda lightning round. We love the Lambda lightning round. The cloud is working overtime.

2:13:28Eddy Cue:Kik has acquired FaZe Clan. Oh, that's interesting. It announced a strategic acquisition of FaZe Media. Yeah. Was FaZe a public company at some point? There was some clan that was public. They were public. Whoa. They were taken private. Okay. Are we sure that this is real? It's not April Fool's? Oh. I'm just throwing that. I don't know. Okay. Yeah, kick. I don't know. Can we dig into this? Is this an April Fool's joke? Somebody already asked Rock. Yes, 100 % it's an April Fool's prank. The press release is dated 4-1-2026. There's zero real news of confirmation of any kick phase deal, and it fits the classic nostalgic streamer org acquisition troll vibe.

2:14:06Nice catch.

2:14:07Eddy Cue:I'm so unk that I didn't realize this was – is this supposed to be funny? It's on their Wikipedia, though, so I don't know, actually. I think this might be real. I don't know. I mean, it seems like if FaZe was taken private and is looking for a new home integration into Kik, this doesn't seem like that crazy. It's not like if it is an April Fool's joke, it's not hilarious. It's not a very good one. I'm not like, oh, wow, I never would have imagined that Kik and FaZe teamed up. There are people that are making that joke. There was some company, a high frequency trading firm that was joking about being acquired by Anthropic or something.

2:14:42Eddy Cue:It was like a crypto wallet company. crypto wallet company. Like that feels a little bit more non sequitur, right? Well, Adam Faze has a good story. He says Faze Media owns the trademark for his born last name Faze and sent him a cease and desist in 2023 when he originally named his production company Faze World. He says, I want my last name back. I'm rooting for you. I'm rooting for Adam Faze to be able to use the brand. Well, you should have bid on the IP, buddy. You should have bid. I mean, also, what were your parents doing? not trademarking your name in every possible category as a child.

2:15:17This is deep alpha. I need to lock up Coogan rather than later.

2:15:21Eddy Cue:Max, Tawny on a roll today says, it's never been a better time to be in the fake wood panel industry. If you sell fake wood slats that can be quickly installed in a room somewhere, business is booming with no signs of slowing down. I love the slat walls. We got to get one in here so we can just like step back and go and hang out and do a little slat wall podcast. Wait, that would have been the April Fool's. oh what were we thinking you know we struggled with april fools this year we think it's a little bit overplayed it's not that it's you know it's not that funny uh and kind of every single day is april fools yeah yeah we do agree yeah and there's a lot of it's a good day for us to just get serious yeah right get the white lock in focus on the market the market's up we're in white suits dylan really serious but you know the the actual play would have been to rebuild like the typical podcast set with like the bookshelf, the slot walls, but then make it miniature, right?

2:16:15Oh yeah.

2:16:15Eddy Cue:And so we're sitting there, we're sitting there, we look like, you know, giants. Giants of the industry. Do you know why the slot wall is so popular? Isn't it Huberman? Yes, but do you know why he selected it? I do not. So when you are recording a podcast, you want good audio quality. You don't want reverberation. You don't want a lot of flat walls that will bounce the sound back to you and create echo and distortion and hollowness in the sound. You want a nice bassy response. You need a nice microphone, but you also need a sound treated area. In the Ultradome, we have some sound treatment over there.

2:16:54We have some sound treatment over there. We do have a hard floor. Maybe that would make the audio quality better. Who knows? But we, for a long time, people would put up that sort of, it's an egg crate style. I don't know if we have a camera. Maybe the reverse PTZ would do it. But the egg crate style soundproofing works incredibly well. And if you're ever in a professional recording booth as doing voiceover for a film or recording a song, you will be in a booth that has a lot of spike yeah it has a lot of spiky yeah you can see right here so this sort of egg crate it creates a lot of little holes for the for the for the sound to get caught in effectively and uh very fun and uh and and the problem is is that that makes it look like you are in a sound booth it's not very aesthetic it doesn't look like a natural environment and so there were companies that said let's get the best of both worlds something that's aesthetic but still sound treating.

2:17:56And so they launched these wood slat walls. Of course, the space in between the wood captures the sound a little bit, acts as a little bit of deadening. And then of course you can also hide, um, you can also hide soundproofing material behind the wood slat wall. And so it was a very logical way to have an elevated aesthetic while still getting the benefits, at least some of the benefits of sound treatment. Then Andrew Huberman did it, It went mega viral and everyone sort of copied it over and did the same thing. And it's become a huge trend. He wound up painting his black, I think. And who knows?

2:18:33Eddy Cue:I think they got sick of being. Now, you know, if you're in the black paint industry, that's where the money is. Because everyone with the fake wood panels is going to be painting it black. We have yet to do this. I've looked at this in the past, but I've never actually done it. Anyway. Do we have the right person? Guys? I believe so. Are you sure? Sorry. We are figuring out who... Oh, yes. Okay. Got it. Sorry. I'm... You thought I was... Yeah, I thought we were jumping ahead to Gary Tan. I'm sorry. Anyway, without further ado... Let's run it back. Let's run it back. Abhishek. I messed up. Abhishek.

2:19:11Great to meet you. Thank you so much for taking the time. I got a little mixed up on the schedule, but it's great to have you here with us today. Anyway, please introduce yourself and the company. Hey, thanks for having me. I'm Abhishek Das. I'm the co-founder and co-CEO of Utori. And introduce the company. What are you building? Yeah, so at Utori, we're building web agents, so agents that can take actions and complete tasks on the web. So we think that the future of interacting with the web is going to look quite different than it does today, where we're not manually navigating web pages, clicking buttons, filling forms, et cetera.

2:19:46We're going to be operating at a slightly higher level of abstraction where we have AI agents who we delegate tasks to that carry on a lot of these tasks in the background on our behalf. And does that mean like virtual machines loading the full web page and clicking on it like a mouse and keyboard interaction? Or are you interacting with the APIs and reverse engineering the routes to sort of just build CLIs on the fly? Like what are the different strategies that are working and not working right now? Yeah, so it's everything you mentioned basically. Okay, both. We should probably think of it in layers.

2:20:21So there's the core LLM that's looking at how web pages are laid out, how to click buttons, navigate websites to take a task to completion. Then there's the agent harness, which surrounds that LLM. So on the web, we have APIs available. It makes use of APIs where there's no APIs used, this kind of a visual linguistic model. And the agent harness takes care of things like persistence. So if it makes a mistake and it has to backtrack and try something else that it knows how to do that, the agent harness would take memory, orchestration, breaking a task down into subtasks, all of that. And then the third layer is putting all of this together.

2:21:00What does the product experience look like where this works for everyone? Makes sense. How are you thinking about the divide between consumer versus business to business, enterprise? How do you see the customer mix evolving? Yeah, most of our users are prosumers. So individuals who are using it for work, so like small to medium business owners or individuals working at bigger companies. The model layer that I talked about, we do train our own model in-house. It is as accurate as Opus 4.6 and GPT 5.4, but it is 2 to 3x faster and much cheaper. So that we make available as an API as well. Sure.

2:21:42Eddy Cue:What was the process creating the model? did you take an open source model and tune it? Yeah, exactly. Exactly. So most of our focus is on mid-training and post-training with a mix of SFT and AUTO. We started off with an open source Quen based model and we collect our own data both in simulation and actual websites and we came on those. Yeah, what are you talking about? So sorry. The big labs don't scare you. Talk about how you see the competitive kind of environment evolving. This is something that I think everyone should assume that all the different kind of consumer prosumer and enterprise products do maybe not super well right now, but will do.

2:22:26Eddy Cue:How do you see the market evolving? Yeah, I think the market for non-coding knowledge work, digital work is massive. There's a lot of work to be done there. It hasn't quite hit the kind of inflection point that coding agents have. So, yeah, there's a huge opportunity there. The area that we do peak on and we care about is tasks that happen on the web. So browser use capabilities specifically. Anthropic, OpenAI, et cetera, have models for computer use, which is more general purpose. But for browser use, currently ours is the best model that's out there, both for accuracy and latency. How do you think about the textbook use cases?

2:23:09I'm sure you get asked about booking a flight. That's more of like agentic on the web, go do something for me. I can also imagine there's a huge value in just, you know, monitoring websites, scraping data, putting things together. We were talking yesterday about how, you know, will we see an explosion of token consumption among financial professionals like we've seen among programmers? And my bull case that we will is that, yes, you're just maybe building one financial model. You're not necessarily building 1 ,000 financial models a day. But that one financial model might interact with thousands and thousands of web pages to collect every possible data point to create aggregated data sets that then can be compressed and compressed down into a 12-tab Excel sheet that eventually results in, you know should you buy the stock or not or whatever whatever the financial analysis is yeah i mean so there's a there's a lot to unpack there yeah um the first thing you mentioned were like uh logging into a website and booking uh a fight or like ordering food or something uh we actually shipped a bunch of a big upgrade today and it's basically possible today like you can connect your favorite websites and apps to our product which is called scouts yeah um and you can just give it a task.

2:24:32A bunch of us internally have been using it to automate our Krillz orders, Instacart orders. We've had people externally try it out as well for LinkedIn, other websites as well. But a lot of that is possible today. And it makes use of our code model, the agent harness, all the tech components coming together. Now, in terms of how I see the token consumption and usage of this going forward. I think it's still quite early in this space. Like a lot of non-coding digital work hasn't quite gotten the kind of attention and hit that inflection point yet. So the cheaper and more reliable and more accurate it gets, I think we're just going to see an explosion of usage of this technology.

2:25:16Yeah. It feels like there's a little bit of a capability overhang. Like you can do really deep research and you can do deeper research with a coding agent in many ways, but that workflow and that just, it just hasn't broken through to everyday consumers that they should even think about, you know, asking an LLM or an agent a question like, you know, build my financial profile from every data source all over the place and analyze everything. They come to it with like, you know, how much should I invest in the stock market? Like a basic question that's basically just web search. Yeah, so there's two or three aspects here.

2:25:56One is that we actually had someone try this recently on our product where they gave it access to their email and they had a bunch of expense reports that had come in on their email and they asked the agent to prepare sort of a nice categorization and spreadsheet of all their expenses and they were able to one shot it. Wow. So that's one part of it. The other part of it is that the previous version of our product was primarily meant for agents that can monitor anything on the web. So kind of like Google Alerts, but on Settler, it's like an AI native version of it. But with today's release, one of the things we released is the capability to build live artifacts.

2:26:39So now these agents don't just monitor, they can prepare a single sort of spreadsheet on a website or a dashboard that stays updated as new information comes in. So if you want to track any time a startup comes out of stealth or a startup announces a fundraise, you can now use Scouts to maintain a single spreadsheet for it.

2:27:02Eddy Cue:Where is the company at? When did you start it and what were you doing before this? Yeah. So we started the company in 2024. I'm an AI researcher by background. I grew up in India, moved to the US in 2016 for my PhD. That was at Georgia Tech. After that, I spent some time at FED at Meta as an AI researcher. There's two other co-founders. All three of us are AI researchers by background. We're about 15 people. We're a little beyond our safe zone. That's where we are. Are you compute constrained at all? massively massively well i mean what's the plan uh is is is the best practice just like hunt around for cheap gpus or slide around different services like or or use use your product to monitor when when availability yeah online on on the neolabs yeah neoclads yeah yeah so we we definitely do a lot of that and we are i think quite compute efficient uh from that from that point of view.

2:28:05But compute is one part of it, the other is data. For a lot of tasks on the web, it's not easy to collect and generate data as, let's say, computer-use tasks, where you can just hire a bunch of annotators to, let's say, simulate certain tasks on their Microsoft Office apps and so on. On the web, many of these are irreversible actions. If you actually buy something, then there's a real cost associated with it so it's not as easy to collect data do a mix of like simulation and sort of using our product to visit web pages especially websites where there's a few clicks and navigation steps in board so you can't like index or crawl them in a naive manner so are you building the RL environments for particular websites is that programmatic or are they like handcrafted, like how many, like what's the scale?

2:28:59I imagine that you need to build a lot of these for generalization, that's very cumbersome and there's gonna be like flaws with every single different system that you try and build. Yeah, so we wipe code a bunch of like simulated websites and use that to generate data and like evals. There is a good amount of in category generalization that we see, like for example, if you imagine how Amazon is laid out or how any e-commerce website is laid out there's not that much variance between them but if you compare that to for example how zillow or redfin is laid out what the ua looks like um yeah so just being intelligent about like how we're sourcing data and then are you like stack ranking those like is amazon more valuable than zillow because consumers will demand that or prosumers will demand that over zillow or vice versa and how do you evaluate that i think it's a it's a mix we do um keep a close eye on making sure that makes looks good and like as we want it to be a lot of odd users use it for more work related things sure as opposed to in their in their personal life so we keep a focus on that yeah so let's log into my erp my payroll system pull stuff from all different dashboards i can imagine us pulling analytics from all the different analytics providers and they all have separate websites, but they all have some similarities in the design philosophy and the best practices and what color the CTAs are and whatnot.

2:30:28Anyway, this is fascinating. Congrats on the progress and thank you so much for taking the time to come chat with us.

2:30:33Eddy Cue:Thanks for breaking it down. We'll talk to you soon. Cheers. Have a good one. Goodbye. Let me tell you about a very related company, Labelbox. RL Environments, Voice, Robotics, Evals, and Expert Human Data. Labelbox is the data factory behind the world's leading AI teams. And let me also tell you about graphite code review for the age of AI graphite helps teams on github ship higher quality software faster And we should have Gary tan joining us in just a minute But we can go back to the timeline. We can pull up about this post from Marek has on he says we just rebuilt every startup and YC's latest demo day batch Here's what our agentic founders pulled off and what it means for the future of startups fully usable products at the bottom of the thread below.

2:31:19Eddy Cue:He's like, I one-shotted all of your companies. I built everything. What a crazy time. Let's play this video. Yeah, let's play this video. We built every startup in YC's latest Demo Day batch. At Felt Sense, we build agentic founders that source ideas, build product, and go to market all on their own. Source ideas from the YC Demo Day page. So we asked, could our killer agents compete with these cracked founders? On Demo Day, our agents swarmed the YC website, found every startup in the batch, and locked in to reverse engineer each build. They reconstructed their own PRDs based on public product specs and started to rebuild each application.

2:31:55When they hit snags or needed input, they called in a human to help resolve every problem. Core technology that took founders months or even years to create was rebuilt, rebranded, and ready to go to market within 24 hours. So what does this mean for YC and the startup ecosystem more broadly? It means AI replication risk is becoming more and more of a threat to every business. So what makes a startup more or less replicable? This is what we learned. Several obvious protections exist for startups. Building things in the physical world and owning data that others simply can't get to are a few examples.

2:32:32But there's a third one that caught us off guard, and it's more interesting. Most people think the best protection in an AI world is human creativity, meaning, or ingenuity. the positive parts of humanity. But that's not what we're really seeing. The real protection seems to come from the messy stuff. Industries filled with politics, lack of trust, turf wars and bureaucracy, markets that are painful to work in due to complex or even failing social dynamics are exactly the ones that are hardest to deploy into, and therefore the most protected. A company that has learned to pick apart that mess and embed their solution has something few competitors can copy overnight.

2:33:13Difficult markets aren't bad markets. In an AI era, they may be the safest ones to spend your rare time and resources. You're saying that hard companies are hard. This is sort of a cool launch video. It feels like a blog post as a full video. Like it has a lot more, there's like more of a thesis and meat on the bone.

2:33:35Eddy Cue:Wow, he ratioed himself in the video. He got way more likes in the video then. Well, we have Gary Tan here. We can ask him. Okay, fantastic. If all of these new companies are going to be in. Gary, how are you doing? Doing great. Is it over or are we back? Is it over or are we back? Everyone's saying it's over. We're back. We're back and forth. How are you feeling today? Oh, we're so back. Okay, break it down. How are we back? Oh, my God. How can we not be back? Okay, explain. What's exciting? We are at the dawn of, like, being able, like, software is totally fungible now. Okay. And then basically, the tricky thing now is like, you just got to spend the money on the tokens.

2:34:15That's like one of the things that I think people are very, very afraid to spend money on tokens. And the things that you can do right now, like you can literally make open claw. Yeah. And the reason why Pete could do that was he's actually like really, really successful previously. Sure. And he could let it rip. Like I think one of the key things that people have to understand about like the current moment is you cannot be precious about your like Claude Code Max account. You have to like really just let it rip. And when you let it rip, you can create like these things that like set the world on fire, actually.

2:34:54What about all the other things? I mean, so YC is interesting because there was a, you know, for a long time it was about technical founders, you know, writing code oftentimes. Still is, right? But it was like, okay, you're technical. You are going to write code while you're here. But also, you're going to talk to users. You're going to figure out a go-to-market motion. You're going to find a hard problem. You're going to figure out secrets of how this business works, what makes your moat deep and whatnot. And it feels like, is AGI capable of that or are we in the software on the Singularity? Oh, no.

2:35:28We still need people. I mean, this stuff is like Jiminy Cricket, man. but you still got to be, you know, Jiminy Cricket's got to stand on someone's shoulder and it might as well be yours. Okay, so don't be precious about tokens.

2:35:43Eddy Cue:So obviously if somebody sent you an investor update, a YC Portco and said, and the only number they shared was how much they spent last month, based on what you're saying, you might be like, okay, you're on the right track. But that's clearly not a metric. Yeah, let's see what you're saying. It's not a metric to optimize for. what should founders be optimizing for when they're locked in? Making something people want. It's crazy. I don't know. I feel like I'm at the center of a lot of weird controversy. Did the guy lose his mind? What's going on? I've been thinking about it and it's not actually about me.

2:36:17It's about people's relationship with their craft and that has to change. I think my response to the haters is like, have fun coding at 1x speed, bros.

2:36:31Eddy Cue:Well, I don't know that the haters are not also using the tools and not generally excited about it. How are you interpreting? I think it's just primarily... Like, I'll give you a concrete example. Please. Like, I noticed yesterday, I had a throwaway comment, because I created GStack, it's got 60 ,000 stars now, like, I'm actually growing faster than OpenClaw by stars on GitHub right now. and they had their moment and I'm on my way. About 30 ,000 people use G-Stack every single day. I've been getting emails that are pretty awesome actually, like people trying to start consulting firms and they just want to feed their kids and maybe they lost their job and then they literally are like, starting this consulting firm, I've never done this before, I sit down with a client, I just open office hours with G-Stack And as they talk, I type in what we're talking about.

2:37:27And then I'm live talking. And they sign a customer on the spot within 20 minutes of talking to someone because of the G-Stack office hours skill. To build custom software for that client? Yeah, exactly. Okay. Yeah. And maybe the thing is custom software. I was at Palantir when Stephen Cohen and Shyam Sankar basically invented the idea of forward deployed engineer. So everyone's an FDE now. That's what you're doing. Everyone can be an FDE. And the thing is, if you don't know how to do it, this thing is the training wheels that will teach you. And then it was the training wheels for me as I was learning how to do it.

2:38:09The original prompt I had posted on Twitter right when I was working on Gary's list. Both my engineering prompt to shake out all the bugs and get to 100 % test coverage. That was my plan eng review that is in G-Stack today. And then the other one was a plan CEO review. I call it my Brian Chesky review. It's like having Brian Chesky sit on your shoulder and be like, I mean, that would be a heavy person to sit on your shoulder. But he's going to ask you, are you in founder mode? Are you moving fast enough? What's the 10-star experience? If you've ever seen him talk about the 10-star experience, Like that's what having CEO, planned CEO review in the G-Stack skill feels like to me.

2:38:52Okay. And I really think that that's awesome. Like I want everyone to have that. And so I was like having a throwaway thought. Like, you know, one of the core things I made was a browser plugin that wraps Playwright. And so, you know, the reality is like I'm just solving problems for myself. I was building Gary's List, and I found myself, the plan CEO review worked well, the plan end review worked really, really well, and then I found myself, I'm using Conductor, I have six or seven windows open at that time, I'm up to 20 right now, and I found myself running between windows, just doing manual testing, and I'm like, great, I worked so hard to build all these skills, to get way faster, and then now I'm just a black box QA engineer for the robots.

2:39:42This is so boring. How terrible. Can I automate it? So I opened another conductor window, and I was like, man, why does Claude in Chrome MCP suck so much? I don't know if it sucks now. It's been a month or something. Maybe they fixed all the bugs, but it literally couldn't do it. It's like two or three seconds, five seconds, crazy context bloat. It was unusable. And then I said, okay, well, I know Playwright exists. Can we make a CLI that's like a thin shim over what Playwright does? And then basically it did it. And then that was what the first V1 of GStack was. It was those two scripts and a Playwright browser plugin that allowed me to not QA anymore.

2:40:23I could just say, okay, now slash QA. You know what we did in this branch. We have all these plan files. Go line by line and act like my QA engineer. And it did it. And I like the first time I used it, it was like a hundred milliseconds. It was doing it. It could log in. It could like click on things. It like took screenshots of things that were broken and then it would fix them immediately. And I was like, oh my God, this is how to do it. Like, this is how you build a software factory. It's like, I automate each of the steps. Like I've built so many pieces of software over the years and I've help so many people do that and it's like this is the process and so now there are 30 of these skills and then you know now you still like i think we're still in manual mode like i built an auto plan so that now like all of these different pieces like there's a recommendation in each of them um and like if you really want to be in the weeds and understand what's happening you can just like read the recommendation you can talk with it you'd be like well i like option a but like option b sounds better for x y and z reason like let's talk about it right um so yeah it's interesting because like i'm arriving at similar things to what i think like devon or other automated software factories are doing but i'm doing it in the open open source mit license like anyone like if you don't like something just fork it man and like you know yeah just make it yours and i have like 200 prs i have to look at right now they're you know it's amazing it's amazing we'll take care of it uh that's right what was the what is is gary's list uh more of just an experiment to put this to work?

2:41:57Because, I mean, it's...

2:41:59Eddy Cue:Because in some ways, like you didn't have to build Gary's List. You could have used like a... Substack or WordPress. Or fork something that's already open source. But building it from scratch, like are there clear benefits where you're like, I'm really happy? Or was it more like the processes? I mean, one of the experiments I'm having right now that you can't, you know, if you use package software you can't do is all the emails come from me directly. Oh, okay. The front line is for me. And when you reply to it. You know how you sign up for New York Times or something like that? You get an inbox, a daily digest, the same digest that everyone gets?

2:42:35You get a personal one. You do have to apply to get into Gary's List. And you have to tell me what you're into. But all the emails come from me directly. And I don't write them. The AI writes them, but I do read them. interesting so everything that and you know when you reply you're it's going to my own inbox wow and ai is helping me read it yeah and it's like going into the personalization system i built so that we can send you things that are more relevant to you but like in some sort of weird parasocial way like gary's list is my experiment and um like communication at scale personalized like i can be your personal assistant on like, hey, I'm mad about politics or like my kids getting, you know, can't get a good education in school.

2:43:21What should I do? And like, obviously I have a chat and it's like, you know, state of the art, like Opus 4.6 thinking like all that stuff. Like, you know, it has data sources. You know, there have been users on there that have gone online who are like just engineers, like living in Pleasanton. Like there's this one guy who never got involved in politics before and he's sending letters to all of his uh representatives based on the policy things that are like screwed up in his life like his kid you know yeah his kid's school is messed up like there's all that you know and you know he's having you know a hundred conversations a month with uh our agent about like this you know and then he emails me about it i'm like oh cool well i like this and maybe you should send this email to you know such and such person um and so I don't know.

2:44:07I think all of these are just fun experiments, but I'm learning a lot. I am trolling on people about the lines of code thing, but I will say that there's a grain of truth here where I am not about lines of code. I don't work for anyone, really. I work for the institution of YC as the steward of it, but aside from that, I'm not here to max lines of code to good heart law something. What I've learned is that the machines aren't good heart lawing lines of code either. So that means that if 100 % of your code is written by the machines and you are earnestly trying to solve the problems of your users and yourself like I am with GStack in the open.

2:44:56You can look at it and I'm doing 36 ,000 lines of code a day and not a single one of them is like for me to be able to say that it's actually like, I'm just having the time of my life building software. Like we're about to launch something really awesome at YC. Oh, I was about to ask, can you tell us the story of Bookface and then where you think software development will go at YC? Yeah. I mean, Bookface, we have, I mean, we have, you know, 20 engineers, uh, at YC and, you know, book, one of the funniest things we, uh, we found out was like, actually a lot of people seem to have access to Bookface.

2:45:28is an internal social network used by we have about 16 ,000-17 ,000 people who are YC alums now who have access. About 40 % of them use it every single day. I've started to realize actually anyone who has a friend of someone who works... People basically share their logins. Which is cool actually. The reality is we should probably open it up. We probably should have a... If you're a builder, I don't think I want absolutely anyone. I want techno-optimist people who are psyched about making the future. And I think Bookface should be open to people. That might be one of the things we work on this year.

2:46:10There's so much interesting wisdom and stories in there. Some of the stuff is maybe private and people don't want it to be screenshotted and shared. But there's definitely a process to open it up.

2:46:21Eddy Cue:How are YC founders in the latest batches adapting to the change of pace of software development? There was a guy yesterday who made the claim that he rebuilt the whole batch with agents. Yeah, good troll. Congratulations. God bless. Big respect. Incredible troll. I'm not sure he made anything that people want. He did a good demo for his product. He got some attention. I love it. Shout out to Tim Draper. You know, big respect, man. Like, you know, you got to do what you got to do. I love it. Yeah. But in a world where a YC founder might have, in a previous era, been very proud of the system that they built, the code that they wrote, even if they're not tracking lines of code, they said, look, this piece of software did not exist.

2:47:12I spent three months grinding with my co-founders. We built a piece of software that solves that problem. And now that's, you know, five minutes of work. Where will the moats come from in the future YC companies? I mean, basically the future, like what's funny is basically all the people who already did that, like you got to speed up. Like it's time to let it rip guys. Like software is too, we treat software like it is so precious. And in the new world that we're going to, it is going to be much more fungible, but that's the good news. That's where taste and agency and trust matter a lot. This is something that only crystallized recently.

2:47:51We've been talking about agency and taste for the longest time. Agency is being able to prompt, and taste is being able to do evals, right? And then the third part that I only started thinking about, I think there was a tweet I reposted today that like said it really, really well, that like trust is the third thing. And so, you know, that, I mean, that's probably the thing that's the most important. Like when you have an enterprise company that is selling to real businesses and they're relying on you, like that's your moat, your moat is actually trust. Like they went to the effort to try you, to get you onboarded and to incorporate you into how, you know, they work.

2:48:30And then, you know, I think that either you can hold on to that and that's a real moat and you can build something over the long haul. Someone gets promoted or someone avoids being fired because you exist and they're never going to switch and they trust you. That's even more important than those other things. What is this story about? It's not about if someone came along to you, this impressive Draper company troll on YC companies is a good case study in that, right? Like, I'm impressed by the volume of software, and they clearly built a software factory. But, you know, maybe you should open source it.

2:49:11Like, that would be cool. Aside from that, it's like, are people going to trust things that are like trolls? Like, I don't know. I mean, I hate to bring up, you know, let me state plainly, like, I actually really respect Roy. He's always been really cool to me on the internet, and I feel like I've been unfair to him. I think that clearly legit is like we were talking about it at YC Partner Lunch today. It's like, oh yeah, the guy actually has real revenue and the idea is good. The troll marketing part we have problems with. We talked about that a lot.

2:49:46Eddy Cue:Yeah, the interesting thing with the product from the beginning I think some of the criticism was okay, if this product works there's going to be some big companies that roll it out and so that's like this other question, right? That's the trust thing again, right? It's like there are things that can get you here, but they're not going to get you there, right? Is sleep overrated in this moment? Is it a waste of time? Are you adapting? I don't know. Sleep is great. I'm starting, like last night around, I don't know, 1.30 a.m., I was like getting foggy, and I probably could have kept going, but I'm like, let's go to sleep, man.

2:50:28Like I'm going to wake up in the morning and I'm going to dream about a bunch of the things I want to do. And then I'm going to wake up and I get to wake up and go to my computer and go to conductor and Claude Code and, you know, let's see what the workers are up to. I love it. I love it. I have two, uh, two questions. One, uh, what was your stack like back when you designed the Palantir logo? What was the, was that a tasteful exercise? was that like what went into designing the Palantir logo? Yeah, I mean, basically what's funny is, yeah, that was such a weird formative year. Like I'll paint the picture.

2:51:01Like we're in Page Mill Road in Palo Alto, like the iconic downtown Palo Alto office. We were a year away from even moving into, this was like across from where I think Tesla headquarters in Palo Alto is now on page mill. Um, we only were 12, maybe 10 or 12 engineers at that point. And, uh, I was working on the hiring site for Palantir dot, you know, palantir tech.net or something. I think we didn't have the.com yet. And, uh, the original logo was created by my high school buddy, Stefan Cohen, who's still at Palantir and a billionaire many times over. And, um, you know, basically I was like, how do I get engineers to want to work here?

2:51:44The logo kind of, it was like six hexagons I think, and people came and they thought it was like a biotech company or something. Oh, interesting. And then we just basically I took maybe, I don't know, a week, it wasn't more than a week of like working on it, but we made like a thousand different versions of it, which is really funny because like... And that was pre-GStack. Yeah. Yeah. Now in GStack you can go slash design dash shotgun and you can say, design me a hundred of these things and then it'll pop a window. It'll use codecs. That's amazing. So yeah, I mean, how do you think, do you think AI is changing design?

2:52:22Do you think there's still room for taste in this world? Do you think that all the same rules apply or is it somehow different? Oh no. I mean, the machines don't have a point of view, like they'll give you lots of options. And you know, I think if you don't have taste and you're using these tools, you'll make something that's marginally better, but it'll still just be marginally better than average, whereas I think it's clear that there are tools like Variant UI that's like a YC company that we funded. It's awesome. You just go and then it's like, that's the real version of Design Shotgun. You will have design tools that have taste and if you use those, you will have a leg up.

2:53:03But I don't know. I guess it just goes back to the agency part again. It's like, you know, one of the principles I built into GStack recently that has served me really well is before you start coding or before you start making technical plans at the engineering level, I tell it always search the web. It's like figure out what's, you know, you want to, someone else has figured it out. Like there's something open source. There's something that's out there. Go all the way to the edge. And then, you know, basically don't try and reinvent the wheel. Right. Interesting. Is there a world where it re-implements Gary's List in Jekyll or something?

2:53:44Who knows really? I mean, we'll see. It might. I mean, it's using a lot of open source software. Obviously, it didn't write its own programming language. It probably didn't write its own database. It's using off-the-shelf tooling there. But the core app, I mean, I guess with the right flexibility and the tools that you wanted and the features that you wanted, it makes sense to... it's not build versus buy anymore. It's like fork versus inference or something like that. There's like a new trend off. I mean, increasingly you don't even fork. I mean, it's funny. All of these things are very strange.

2:54:19We're in bizarro land right now. We're in mid-transition, and all of the transitions are not in distribution yet, which means if you try to make a product right now and you ask it to estimate how long the agent is going to take to make it, it says human terms. So it'll be like, it'll take a human a month to do this. And then you have to ask it explicitly, well, how long are you going to take? And it's like, well, I'll probably take like 15 to 20 minutes. And it's like literally across the board, I am coding 90, like my output in terms of like real products, not just lines of code is like 90X that of what I did when I was in 2013.

2:54:58I made Post Haven that year. I made Bookface that year. and I was at a CEO conference last week at JPMorgan 100 and I was talking with a bunch of former tech very technical CEOs who now are full manager mode and I was like, good news man, it's time to come back you're Peter Parker right now, time to put on the Spider-Man suit because this is the time and I was blowing their mind and they're like, oh no, I shouldn't do this, it's not worth my time and I was like, dude, it is worth your time. You can still be the manager during the day, but by night you can be Spider-Man or Batman. You can don the suit.

2:55:41Pull out the G-Stack. Pull out the G-Stack. And my pitch to them was like, look, I am 90 times. I am 90 of myself. I'm really good at manipulating time. Yeah, that's remarkable. And you can have 90 of yourself working on things like tonight and be the CEO, right? It's fantastic. Well, you won a ton of people over in the chat. They were skeptical and they're very fired up. So thank you so much for coming on and trolling the haters. Good to see you, Gary. Keep building, keep shaping those lines of code, and we will talk to you soon. Have a good one. Goodbye. Good to see you. Let me tell you about Plaid.

2:56:24Plaid Power is the app to use to spend, save, borrow, and invest, securely connecting bank accounts to move money, fight fraud and improve lending now with AI. And we are running over time, but let's bring in our next guest. Bren from Core Weave is in the Restream Rating Room. Sorry for keeping you so long. Thank you so much for taking the time to come talk to us on TBPN. How are you doing? I'm doing great. Thank you for having me on today. Of course. And could you introduce yourself and kind of take us through some of the news? I mean, there's so much going on, but I know that there's some big announcements more recently.

2:56:55Yeah, absolutely. My name is Bren McBee. I'm one of the co-founders of CoreWeave. I'm the chief development officer. So I lead all things debt, equity, M &A, ventures for the business. This is where all the money's raised within the organization. Yeah. And this new project, this new vehicle, how are you positioning it? How are you explaining to people who are familiar with CoreWeave as a stock that they can buy and they understand that you might use debt to finance the purchase of a facility? They're familiar with a mortgage. How are you introducing this product? Yeah, so this product is one that we brought in the market about three years ago.

2:57:32And it was with our first facility, DDTL1. And what we've introduced today is the continuation of just continuing to build upon and execute within that original product. We're effectively taking this GPU infrastructure along with revenue associated with the GPU in these long-term take-or-pay contracts, put them together and then take them out into the debt market and finance them. And I would say it's a financing strategy that has occurred across many different sectors. Right. Like this is how LNG facilities or power plants get financed. Right. You have the infrastructure in one hand. You have these long term contracts in the other hand and put them together, marry them.

2:58:13And you can stand up very scalable financing mechanisms around. it. What's really important about the one that we announced yesterday, this is our largest transaction at 8.5 billion. And it's our first transaction being done. Thank you. Yes, it is. Awesome. It's a big moment. Huge number, massive. Sometimes forget the numbers that we're working sometimes. And it was done at an investment grade cost of capital. So this was done at SOFR plus 225 or, you know, approximately 5.9 % cost capital, which is incredibly important to us, given the capital intensive nature of our business. And what, what, what gave the market, what gave the rating agency or you the confidence to put this in the investment grade bucket?

2:59:01How are you, how are you providing the data in the back and the, and like the backbone of that strength? Yeah. To your point, this all comes down to market validation. The clear signal out there is pricing and confidence grows as cost of capital comes down. It doesn't really work the other way. So being able to drive down cost of capital here just shows the amount of demand there is for the core weave credit product in the market. How much of this is, how important was sort of beating the depreciation gate allegations, to put it sort of silly? But a lot of people were wondering, like, oh, those H100s, they're not going to age well.

2:59:47And what we've seen is that there's a lot of data points out there that seems like older. They're aging like fine wine. They're aging like fine wine from what we've seen. But take me through your, did you always know that that was going to play out the way? Was there any unexpected surprise to the upside? And then how important has it been to have more data about how these GPUs monetize over a longer length of time? The depreciation subject has always been very interesting to us. We've been very consistent. Six years is the depreciable life of this infrastructure. And I would say, if anything, we're expecting for the infrastructure to have useful life beyond six.

3:00:23I mean, you see that today in like late 2010 SKUs that are available in market on clouds. It's like they're not running that infrastructure unprofitably. Yeah. Right. And it means that there's client demand for it. But as we look towards like Ampere and Hopper generations, you know, I think in our last earnings, we've noted that our pricing on Ampere went up during 2025. And our pricing on Hopper stayed within 10 percent or so of where we started the year. Why is that happening? I think it's largely driven by inference. And ultimately, this concept that not every workload only needs the latest, greatest piece of compute, which we're incredibly well known for running.

3:01:03But there are different types of workloads that need different types of infrastructure. And the market is simply efficiently pairing the right compute with the workload that they need. So we see incredibly robust demand all the way back to the Ampere generation, which I believe is six years old at this point from its original SKU launch date. And we're really not seeing any deterioration of these older SKU demand profiles while also seeing accelerating demand for latest generation infrastructure as well.

3:01:36Eddy Cue:I think everyone's heard the stories of electricians being flown around in private jets and kind of the chaos of actually building out data centers over the last two years. Any funny stories from the CoreWeave team recently around all the effort going into building out these clouds? Yeah, it would take a lot of private planes to fly all those electricians around. I mean, we're talking hundreds to thousands of engineers on these sites. And, you know, I think that's it's a great point you raise because I believe it's an underappreciated aspect of what is being done here in the true scale of the engineering effort out there in market.

3:02:19You truly just have to go see one of these sites. They are massive, massive engineering feats that are being executed on within incredibly intense supply chains, all the while moving in an extremely high-velocity technology sector. It really boils down to risk management, which is the DNA of our management team. It feels like in terms of inference demand, we've seen a number of sort of exponential jumps when we went from LLMs to reasoning models. The number of tokens generated to answer a question increased exponentially. Saw something similar with agents, coding agents. Are you guiding or expecting for just a smooth trend to continue?

3:03:07Are you thinking that this is more discontinuous? Is there some sort of next, next generation that you're already looking forward to in terms of increased inference demand? It's really hard to predict where the technology goes since the technology itself is meant to be so disruptive. I think what's super important for us is our client base. And our clients represent the organization that are at the front lines of pushing this technology as far as it can go. And we're consistently supporting them. And it gives us this really interesting information flow, understanding where they're going in their organization so that we can be orienting ours to be moving with them in lockstep.

3:03:52What do people misunderstand most about CoreWeave right now? Um, you know, I would say there's three real pillars to our success. It is having the best product in the market for running this infrastructure. Our product is purpose built for AI and supports today the leading AI organizations on the planet. You have to have the right physical infrastructure capacity, the ability to navigate the supply chain. Right. And that that is not something anyone can just stand up and do. It's incredibly challenging and detailed. But today we have over 43 data centers in active operation. That that's a key component of how we've been able to pull down our cost of capital so aggressively is just simply through execution.

3:04:43And that final piece is financing, right? You can have pieces one and two, but if you can't raise the capital needed to bring that infrastructure online, it's sort of meaningless. And so we've spent an immense amount of time focused on the credit market and driving an ability to navigate it effectively and scalably while driving down our cost of capital. And thus, why this instrument that we announced yesterday is so incredibly important to us and will serve as a blueprint for future transactions. Are you in a rough agreement with the broad consensus that I'm feeling in the AI industry around a chip bottleneck being more of a gating factor in scaling AI progress over the next four years than power constraints?

3:05:39You know, I would qualify it as cloud connected infrastructure. Incredibly important. So it's really marrying power with chips, getting them online delivered to clients. That bottleneck within there is delivering power shell capacity, at least within our business. We can bring online more capacity to our clients as we have more power shell within the market. So we're consistently scouring the market for incremental capacity to be able to market to our consumers.

3:06:10Eddy Cue:how uh how have the sort of mom and pop data centers done over the last two years i'm talking about people maybe like 18 months ago that that were realizing hey this ai thing is going to be pretty big we have we can find some power here let's throw up a shell let's get some gpus and try to bring it online how have they fared because i think there was a lot of there's obviously so much demand, but a lot of, there were a lot of question marks. Post-neo clouds or like micro-neo. Yeah. Yeah. Just, just somebody saying like, Hey, we can get some power here. Let's jump on this. How have they fared? I imagine some of them end up being kind of rolled in to a platform like CoreWeave or the other, other, uh, neo clouds or, or they end up, um, but, but how have they, uh, fared?

3:06:58You know, it's really tough to comment on, on their businesses. We don't see them in market too frequently. I think our product looks a lot different than the other offerings that are available out there. Well, yeah.

3:07:09Eddy Cue:And these people were like, oh, it's like it's like real estate. You just like put up a box, throw some GPUs in it. And I would say like no one, no one, none of the leading neoclouds are saying like, oh, it's just like real estate. You just like throw up a box. So there was that disconnect there. I would agree with that. There is a chasm of execution between signing a contract to deliver infrastructure and actually bringing it online and making it revenue generating. And that's where the CoreWeed product sits. Yeah. Well, it's fantastic progress. Thank you so much for coming and breaking it down for us.

3:07:44Yeah, great to meet you. I really appreciate it. Great to meet you guys. Congratulations on the deal. We'll talk to you soon. Thanks. Have a good one. Goodbye. Let me tell you about Gemini 3.1 Pro. With a more capable baseline, it's great for super complex tasks like visualizing difficult concepts, synthesizing data into a single view, or bringing creative projects to life. Tyler, do you know how to count cards in poker? In poker? In poker. Do you know how to count cards? No. You have exactly one day to learn because you have to go to Sam Blond's Monaco Invitational. You have to bring home the$100 ,000 cash prize.

3:08:19You got to win. It's a no-buy-in poker tournament. Is counting cards a thing in poker? That's for blackjack. No? Am I? You're going to learn how to apply counting cards from blackjack to poker. You're going to invent new card counting technology. I think it does work in, oh, maybe you're right in blackjack. Maybe it doesn't work in poker. I think it does work in poker. Anyway, you're going to cheat, and you're going to win. That's the plan. You're going to cheat. You're going to win. You're going to get thrown out of Monaco. You're going to get thrown out of San Francisco. Sam Blod will be putting hands on you, but it'll be a good time.

3:08:55Eddy Cue:put out an article that says the Artemis moon-based project is legally dubious. Okay. Cain says, oh no, someone call the moon police. Well, we have our next guest in the Restream waiting room. Let me tell you about MongoDB first. MongoDB, what's the only thing faster than the AI market? Your business on MongoDB. Don't just build it. Own the data platform that powers it. And without further ado, we have Sam from CoreZone. Welcome to the stream. Thank you so much for taking the time. Hey, guys. How you doing? All right. Great setup here. Great setup. Great background. Great polo. Nailed it across the board.

3:09:31But we'll let you introduce yourself. I don't look like you guys. Huh? Well, white. Market's up. I got the memo. So we're all in white. Yeah, we're happy. How are you doing? Please introduce yourself and the group. Hi, I'm Sam Yegan, co-founder of Corazon Capital. Spent most of my life as a founder. Sparknotes, OkCupid. uh it's involved with tinder grinder so you know hopefully got you through school and hopefully got you some dates and now uh investing in the future of consumer ai do you want to focus on the dating market with this fund do you think that there's opportunities for new platforms in ai yeah i mean

3:10:08Eddy Cue:the the biggest question i have for you is are you surprised that we're you know a few years into the ai boom and when you look at the top 25 chart in the app store it's still like pretty much just lms and incumbent consumer products. Yeah, that's going to change. I think the entire consumer stack is going to get rebuilt. I think most of the consumer incumbents are at risk. I do think dating is a pretty protected category because you need liquidity. The big apps are always the big apps. But I think for almost everything else, I think the way we message each other, I think the way we connect with each other, I think the way we find jobs, the way we find friends, I think all that's going to change and it's going to be AI powered.

3:10:52At the same time, on a relative basis, isn't there maybe counterintuitively more opportunity if you're building a consumer AI, if you go into something that needs a liquidity pool, like the dating market, and you can maybe use AI or just come up with a unique wedge, get over the cold start problem, get some liquidity on the platform, and then maybe you're better positioned because you can't be fast followed by a copycat that's just like cloning all your features because they won't have liquidity, but you do. If you can get past the cold start problem, that's right. But I actually think in the case of dating, I think the incumbents are much better positioned to leverage AI.

3:11:29But I think there are lots of other places where, and I'm surprised so many VCs are running from consumer right now. But I think AI is going to change, transform the way consumers interact with technology in a way that has never happened since the dawn of the internet itself. Yeah. So we're talking about some opportunities.

3:11:45Eddy Cue:Yeah, well, I mean, I wanted to ask, like, what are you actually looking for in an early stage consumer founder? Because it's very rare, you know, your case, you've had like multiple big hits. It's hard enough to have one. In my consumer, you know, the random consumer tech investments that I've made, doesn't matter if I'm investing, you know, pre-product or after they've gotten to a million users, they usually end up as zeros. So it's an absolutely brutal game. Welcome to venture capital. Well, and I would just say the handful of zeros that I've had have concentrated in consumer. Yeah. I do think consumer tends to be much more variable than B2B businesses.

3:12:32You tend to either find product market fit and find a way to scale, or you don't. It's very hard to end up, I think, in the middle range in consumer businesses, which I think is why it's so attractive as a venture capital investment. If you can get them right, you get them really right. And that's what we're looking for. You asked about founders. I think AI founders are no different from previous types of founders. In the early stage, we're really looking for founders who understand the process of finding product market fit. That's all that matters in the early stage. And I actually think AI can be a little bit deceptive that way because you can get the first few users to adopt because people are so curious right now.

3:13:08But I don't think that means you've necessarily found product market fit that's scalable in the way that it used to be much harder to get that first million users. And I think it's a little easier to do that now.

3:13:17Eddy Cue:Totally. Yeah, just this year, I've seen like the exciting kind of one aspect of why now is I think exciting to make a consumer fund is you're just seeing like a thousand X the number of apps. I'm talking to a family friend later today that's like built their own app, right? They're very excited about it. They wanna get my feedback. And I just think there's so many people that were kind of gated by not having access to engineering talent or didn't know how to code themselves. And now they're building apps. So I think like the bar, the bar is going to go up pretty dramatically. I wanted to ask, like, do you think there is room for like taking like the capital intensivity of deep tech and bringing it to consumer?

3:14:06Eddy Cue:and I'm specifically asking because of like TikTok. TikTok spent how many billions of dollars on customer acquisition? They had a product with product market fit. They did some interesting things around kind of juicing engagement from at least my point of view. But then they also spent like hundreds of millions or billions of dollars on acquiring users. Is that something that you think you'll see in the fund in some kind of breakout companies? Yeah, I think that's a great and very nuanced point that you make, which is, are AI companies going to be more effective at acquiring customers or not? And I think AI hasn't really, you know, your example earlier about an AI-based dating app, you just kind of assumed you'd get over the cold start problem, but you couldn't really explain how you were going to do that.

3:14:51And so I don't think, other than there's a little bit more word of mouth early on, because people are so curious about any new AI app, I don't think they've really figured out how to leverage AI differently to scale these apps. And so I do think your point about bringing some of the capital intensivity, I do think that will all be reduced, but I don't know if it's going to be reduced on the marketing side specifically, for sure on the engineering side. Okay. So yeah, does that change anything about the investment strategy? When I hear 100 million fund, I have some conception of the type of checks at various stages that you might write.

3:15:26But do you think there will be anything about running a$100 million fund in 2026 that will be different if I just look at the size of checks and spaces that you're occupying versus, you know, five, 10 years ago? I think people are going to have to spend a lot less of that early capital building the infrastructure, getting the code written. I mean, when I started SparkNotes, we had to go lease servers and plug them in in, you know, colo facilities back in the 1900s. Right now, you're going to be able to get your app into the app store so much cheaper. And the real separation among founders is going to be, can you really figure out product market fit?

3:16:03It's not going to be about writing the code. It's not going to be about doing the marketing. It's about, can you actually build a product that consumers love? and that's really what we know how to underwrite here at Corazon and what I've done throughout my career.

3:16:13Eddy Cue:Consumers love and has positive unit economics because we're seeing this whole other dynamic right now where there's products that people love, but if they were actually charging as much, I'm thinking like image and video gen products where if you actually price it the way you would need to to be building a business, would the demand still be there? And if you think about, you know, when the internet started, the beauty was that everything became free to publish. And so you had all the first business model that came out was like Sparknotes, just take Cliff's notes and make it free. And tons and tons of companies did that.

3:16:47And now you actually have marginal costs associated with, you know, the tokens and the AI that you're using. And so now all of a sudden, for the first time, you have cost of goods sold in a consumer tech business, which has really never happened before. And so that's one thing that founders aren't necessarily thinking about so much is what the end unit economics are. But I don't worry about that early stage. I worry about let's make it let's make a product people love and let's figure out the business model later.

3:17:11Eddy Cue:What have you learned about pivoting? That all the best founders have to pivot multiple times. It's you know, it's very rare that you hit the product on the on the head, you know, the very first time you do it. And I think that's why any memorable any memorable pivots, you know, across your founder journey. Yeah, I mean, we, you know, the original idea for Sparknotes was a humor site that was very much like the onion. Probably probably older than you. Yeah, it's called the spark.com. In fact, the corporate name for Sparknotes was the spark.com. It was a humor site. And then we kind of realized that that was it.

3:17:49We made a website for jokes. I love the onion and I love spark. Yeah. We basically tried to knock off the onion first. We realized that was a really crappy business to be in. And then we pivoted to do study guides and that ended up being the business. How did you think this is a odd question? I'm not sure to be aggressive because I loved spark notes. It helped me through the high school. But how did you think about like the moral implications of potentially like helping you cheat? Because we just went through this like moral reckoning in Silicon Valley last year around Cluley, this company that was very, very in the marketing, like cheat on everything, very aggressive about that.

3:18:27I don't remember Sparknotes ever marketing to me as like, cheat. It was very much like supplemental, additional. Sometimes I'd read the Sparknotes chapter instead of the full chapter if I was running late or late on sleep or wanted to hang out with some friends. But how did you process the trade-off or the impact of that business? That's such a funny question. Here's the thing. People who want to cheat are going to cheat. Yeah. And so what we focus on is let's make the best product, both for the people who want to get A's because they're doing the work. And if you're going to cheat, you might as well actually learn something in the process.

3:18:58And so we didn't give you the paper that you could turn in. We gave you the best path to learn the material. And honestly, you know, if you didn't read Romeo and Juliet, but read the Spark Note, you probably understood the book. Yeah. Yeah. Yeah. And so that's probably better rather than. And I think CliffsNotes never really did that. CliffsNotes was always about just trying to get you through. And we really tried to make sure that you learned what you were doing. Yeah. Yeah. Did you ever, did you ever think about like deliberately how you position it in the marketing piece? You clearly had that philosophy, but you could have had the choice to like flyer every college campus with like cheat.

3:19:31And like, maybe that would draw people's attention. Did you ever think about like click baiting, rage baiting, any of that? No, no. And in fact, our approach was the opposite, which is try to actually get the teachers to recommend us as the preferred study guide. So there were a bunch of teachers who would write us and say, hey, I've told my students you can't use CliffsNotes, but you can use SparkNotes. And then now all of a sudden you've got the teachers endorsing you because you've made the better product. And so what seems like a cheating product gets repositioned as a study aid. And that's what we wanted to.

3:19:57Yeah. And now teachers. Look, there are moral questions. There are moral questions in every business. Of course.

3:20:03Eddy Cue:How do you think about investing in anything kind of education focused when it feels like LLMs are just naturally. quite good at that, but still there's possibilities to build harnesses and stuff, the application layer. No, no. I think AI is going to transform the way we learn for sure. There's a company we're invested in called Brilliant, which is a native AI tutor. And in the 1900s, you had a graphing calculator. That was your like, to get through high school. And I think the next generation of kids, you're going to have your personalized AI tutor that's going to get you through school. And, you know, we're funding that in our fund.

3:20:44Yeah. Yeah. It's very funny. I completely agree with your intuition that like, oh, AI has already solved education. But then you look at like Andre Carpathie is building something in that space, which is just like fascinating because like he has the most insight and has worked at OpenAI and Tesla.

3:20:58Eddy Cue:Like I wonder if there's I wonder if there's actually like an AI hardware device specifically for learning, because like think about the chat. Like if a kid is sitting there and they're like trying to use their phone to learn. But I could ask it about anything. But then you're getting a push notification from your friends, right? There's potentially some of these AI hardware devices that have flopped because they were trying to replace the phone or maybe replacing it in the wrong context. And you should be thinking, like, what is the next TI-84 that every student has that they're actually learning?

3:21:28Eddy Cue:That's cool. Cool. Well, congratulations on the fundraise. We want to ring the gong for you. It would be an honor. Oh.

3:21:37Eddy Cue:Thank you. Great stuff. Great stuff. we'll hopefully get many of your founders on in the coming weeks, months, and years. And we'll talk to you soon. Great to meet you. Thanks, guys. Have a good rest of your day. We'll talk to you later. Deep into the fourth hour. Deep into the fourth hour. We got a one-hour countdown. We're coming up on one hour, one minute, 14 seconds until the launch. We have an interesting post from Andrew Reid sharing on the 50th anniversary of Apple, the investment memo. This is the victory lap of all victory laps for a venture capitalist, writing$600 ,000 into Apple. On a real piece of paper.

3:22:22On a piece of paper. And there's some handwritten notes on here. It said, will be tough to do this deal. Small amount, minuscule, high price, second position. CMS wants to guarantee. It's a very interesting document. Proposed financing structure, 600K to A and 60K to AG, friends and company, 480 to one venture investor, invitees, Venrock.

3:22:44Eddy Cue:Julian Weiser says 600K buys 10%, very rich deal. Like I'm used to buying 60 % for my 600K. It's actually crazy, crazy lore. Well, if you need something to do, it's not out yet, but maybe tomorrow if you need to to kill some time. Donald has made a documentary about a bunch of friends of the show, Will DePue and Riley Walls and the creators of the JML Mail Suite. It's an attempt to capture what it feels like to be a young person in San Francisco right now. He's dropping that tomorrow so you can go check it out. We have one last video to play for you today. Let's pull this up, production team. What you got?

3:23:28Eddy Cue:It is in the chat for you. We also need to send... Profanity warning. apparently someone was swearing on CNN let's get the audio on okay cover your ears why do you want to be here why do you love space why do you love being a part of history we're going back to the freaking moon this kid rocks

3:23:54Eddy Cue:glad I warned everyone it was worth it we don't normally do it but we're in the fourth hour and we got to sneak one in because it doesn't hit without it. So, sorry for the profanity, but that is a fantastic moment, and we're very excited for the launch, which is happening in just one hour. So you can head over to NASA's livestream, YouTube, Twitch, probably other places, and check it out. Leave us five stars on Apple Podcasts, Spotify, sign up for our newsletter at tbpn.com, and we will see you tomorrow at the Lettingham Pacific. Can't wait. Goodbye. you