TBPN - The AI Slop Debate, OpenAI’s $1T Web, 𝕏 Timeline Reactions | Shayne Coplan, Antoine Tessier, Rami Karabibar

Episode Date: October 7, 2025

(02:09) - The AI Slop Discourse (18:48) - 𝕏 Timeline Reactions (20:07) - OpenAI’s $1T Web (24:50) - 𝕏 Timeline Reactions (01:13:47) - Shayne Coplan, founder and CEO of Polymarket, discusses the recent $2 billion strategic investment from Intercontinental Exchange (ICE), valuing Polymarket at approximately $8 billion. He highlights the partnership's potential to integrate prediction markets into mainstream finance and emphasizes the synergies between Polymarket's consumer-focused platform and ICE's institutional expertise. Coplan also expresses excitement about collaborating on future tokenization initiatives, aiming to enhance the efficiency and accessibility of financial markets. (01:22:27) - 𝕏 Timeline Reactions (01:55:35) - Antoine Tessier, CEO of duPont REGISTRY Group, discusses the company's transformation into a digital platform for luxury car enthusiasts, aiming to facilitate online transactions and community engagement. He highlights the launch of duPont REGISTRY's live auction platform and emphasizes the importance of storytelling in marketing, focusing on the passion between drivers and their cars. Tessier also notes the younger generation's readiness to purchase luxury vehicles online, underscoring the company's commitment to evolving with its audience. (02:09:02) - 𝕏 Timeline Reactions (02:17:57) - Rami Karabibar, CEO and co-founder of EvenUp, discusses the company's mission to level the playing field for personal injury victims by assisting attorneys with AI-driven tools that expedite case settlements and enhance outcomes. He highlights EvenUp's rapid growth, including a recent $150 million Series E funding led by Bessemer, and emphasizes the importance of accuracy in legal AI applications to maintain trust and effectiveness in the legal process. (02:35:49) - 𝕏 Timeline Reactions TBPN.com is made possible by: Ramp - https://ramp.comFigma - https://figma.comVanta - https://vanta.comLinear - https://linear.appEight Sleep - https://eightsleep.com/tbpnWander - https://wander.com/tbpnPublic - https://public.comAdQuick - https://adquick.comBezel - https://getbezel.com Numeral - https://www.numeralhq.comPolymarket - https://polymarket.comAttio - https://attio.com/tbpnFin - https://fin.ai/tbpnGraphite - https://graphite.devRestream - https://restream.ioProfound - https://tryprofound.comJulius AI - https://julius.aiturbopuffer - https://turbopuffer.comfal - https://fal.aiPrivy - https://www.privy.ioCognition - https://cognition.aiGemini - https://gemini.google.comFollow TBPN: https://TBPN.comhttps://x.com/tbpnhttps://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235https://www.youtube.com/@TBPNLive

You're watching TBPN! I forgot we were going to start on horse camp. How could you forget? It is. The horse has a fedora. Tuesday, October 7th, 2025. We are live from the TBPN Ultradome, the Temple of Technology, the Fortress of Finance, the Capital of Capital.

Fantastic day-to-day. Jordi did get food poisoning yesterday. We went out to, can we say where we went? Can we say who we were talking to? I don't exactly remember the rules, but. I don't think we can discuss the discussions. Okay. But I think we can say that we are with a Mag 7.

But if we say the event that we were at, then it's obvious who we were talking to. We had a magnificent evening. Okay, yeah, we had a magnificent evening with a magnificent 7. And you can probably just figure out who it is by looking at what event. I mean, there's a lot of companies in the Mag 7. Seven of them. Well, we did get to hang out with David and Ben from Acquired as well, which was super fun. We're going to have them drop by the studio tomorrow, which I'm looking forward to.

I'm going to ask them to quickly summarize the full history of every technology company. And every luxury goods. And every luxury. And every company ever, actually. Costco, too. Throw that in quick. Retail. Just give it one liner on Costco.

Name every person on Costco's team. Yeah. But we did have a fantastic evening, although Jordy did get food. poisoning. Ryan, Ryan Walsh, how did you get food poisoning

while with a MagS 7CHA? Good question. There was two types of fish served at dinner. There were two types. And I indulged

in both. I skipped the fish. There was that crudo. Oh, yeah, I had a lot of the crudo. The crudo was not the problem.

Okay. I'm fine. Okay. I think it was because I had a glass of the Sanssère, the wine.

And you did not partake in the wine. And I think the alcohol it served It detoxified my stomach and prevented me from getting sick. The toxins from the alcohol. It was poisoning the poison food instead of poisoning. That's it.

Exactly. That's it. Anyway, what a day. It did, yeah, so we talked a lot about AI slop, AI video slop. And I had this thesis. It's been kicking around my head for a while. And I thought it would be interesting to kind of relay my thought process after talking to a bunch of creators and folks working in big,

tech. So we've been talking about the AI video stuff and the wave has been fascinating where vibes was like immediately jumping to like this is infinite just. It's terrible. And same thing with Sora 2. Like we jumped way past like any of the normal discourse. There was a little bit of discourse about the electricity stuff and the water. My water bill is going to go up because of this. There was some talk of AI. That discourse is still going on to be clear. It is. There's a post here. We can pull it up. I'll put it at the top of the timeline while you continue on. Yeah.

But I thought about there was a time when the narrative around this type of content, like fake content, doctored content, fake news, AI content. It says me watching Stephen Hawking drop into a sick half-pipe McTwist while my electricity cost triple. Honestly, so down. That's great, but it. I mean, that is one of the greatest AI videos. And that gets to my point, which is that I don't regard that video as AI Slop.

I would like to be shown that video. I don't want to be shown an endless stream of AI Slop that is bad, but there are obviously creative ways to use these tools. There are even ways to reappropriate the tools where the joke of the video is that it's an AI video. And it's like self-aware. And those are very interesting. I was watching Casey Nystad's video all about his reaction to Sora. and he used a lot of AI video generated from Sora in the video.

And so there was a time when the answer and the call from the public was, we need to build an AI detector and we need to cordon this content off in its own tab. And that's kind of what MetaVibs is doing. And that's what Ben Thompson was talking about. He was like, I have not been enjoying my Instagram experience because I'll be scrolling and then I'll see some AI content and I'm in this active mode again of like, is this real or not? as opposed to on meta vibes, to his point, he was like, I can just lay back and look at it

and it's like a screensaver and it's chill. And you've been using it a lot, right? You've been vlogging a day. Also, Ben Thompson kind of reversed his position. It was like, like, because he was saying he was bullish on vibes and like less interested in SORA, but then he was like, look, the charts speak for themselves, like people clearly like SORA a lot more than vibes, just looking at the growth of the app store.

And so the, but there were, was a call for like let's build a detector. I think it's a bad idea for two reasons. One is it's this red queen's race between AI generation and AI detection. Like the original, did you ever look at generative adversarial networks, Tyler? Yeah, yeah for sure. Like can you describe like the thesis of a GAN and why it works? Yeah. So it's basically that this was like the first kind of way that generally worked of producing images. So there's like two parts, and they're basically acting against each other.

So one part is the generator. So it will, like, create an image. And then there's the other part, which is the detector, and it tries to figure out if it's AI or not. And then they basically go against each other. They're trying to beat each other at all times. And then basically you just, like,

because you're training the detector based off, like, sometimes you give it a real image, sometimes you give it a fake image, Yep. And then slowly... And if it can detect the AI image, that gets negative points in the reward function. And if it can't, and if it thinks, yeah, this is real, then it gets upvoted from, hey, you did a good job generating that AI. So, like, effectively, a lot of these models that are AI generators have AI detectors, like, built into them. Like, it's part of the training process.

I don't know exactly how the diffusion models work. I think they work similarly. I think that concept from the generative adversarial network has been employed and repurposed. But it seems like it's kind of this like endless arms race between like you build a detector and then the generator gets better. And so I'm not even sure if technically it'll work. What do you think? Yeah, I mean, so there's like a lot of people want the detectors, especially for text. But the text is a lot harder right because there's just less signal. Like there's only in a sentence there's only, you know, like 20 characters or something.

Yeah, you just can't get that much data out of it. Where with video and photos, it's actually much easier to like encode some like meaning that like, Like, there's this thing about, like, printers, they have these, like, little dots. Yep. And it, like, traces the printer. Yep. But you can't see it because there's just, like, it's super faint and very small. You can do the same thing with images where you can encode, like, you just change the images, like, just barely up a little bit brightness, down a little bit of brightness.

And you lose it. Yeah. Yeah. So you can. Like a slight crop and a slight tweak to the contrast. And it's probably the same thing in text where if you just add one extra space or one extra period or change one word, all of a sudden, like, the, the, the, whatever you've encoded in. in the overall text is just not gonna read as much. Yes, so like with text, you're basically like seeing the,

you're seeing everything. But with an image, you can like kind of change stuff just a tiny bit to where you're actually like seeing something else, right? There's this famous example where with the image classifiers, you can adversarily train against them to where like you have an image of a panda, but it thinks it's a dog because you just updated the kind of the pixels slightly. Yeah, yeah.

And even with like the soror watermarking, like immediately I saw a watermark removal tool that just does in painting just on those watermarks. But what I'm saying is you should, in theory, it should be much easier to create a classifier that can figure out if an image is AI or not, like if it's directly from SORA. For a while, but you don't think it'll be this endless race

where the AI image generators will just get better and better and better, and like the AI detectors will just always be like a little bit behind. That could be true. I think also Open AI probably has some incentive to have something that's like a classifier. or at least something that's able to be classified on, like a watermark. And if you think people will mostly just be using Open AI or Google's image video generations instead of open source stuff, because it would always be like a little bit ahead.

Do you think that Apple will make any moves in terms of embedding anything in the metadata to like prove that this was taken? Yep. It's an unaltered video from an Apple device. Yep, for sure. Some of that should be fakable. but I think that for the most part, that's definitely coming. It already is there to some degree in the metadata

that's maybe a little bit hard to spoof, but there's always a way around it. But there's definitely some stuff that you can do with cryptography to like sign, basically sign the image as like from this at the hardware level, from this hardware device. I think that there will be this endless race between these two, but there will always be people that figure out

how to sneak in something because it's like, okay, well then, like you can prove that it's from an iPhone, but what if you film a screen that's AI generated? Well, then you have a video on your phone that says it was taken with an iPhone and it's signed, but it's actually AI generated. So there'll always be these ways around it.

So I think that it's a little bit of like, like somewhat of a waste of resources. We should probably do it to some degree, at least be able to classify it, but that's not really a perfect solution. The other side of this is that I don't know that we need to build AI slop detectors into large media platforms,

Because the problem isn't AI. It's slop. Like, people enjoy seeing the Stephen Hawking at the X-Games video. You don't, like, we already have tools to filter out slop. They're called algorithmic feeds. If you don't like slop, you will swipe past it and it automatically, and you automatically get served other content. And that just happens naturally. And so there is a problem with algorithmic feeds, though. They have, they're pretty far from their final form, and we should go through a little bit of the history. So they used to just focus on the initial action of choosing to watch a video. So on YouTube, for instance, they would rank videos in your algorithmic feed based on click-through rate.

So they'd show you thumbnails. If you clicked, that was the only signal that went into the algorithmic feed that the video was good because it got a click. And so what people would do is they do clickbait. And they'd be like, I bought a Bugatti. And then you watched the video to the end. And it's like, actually the person just went to the store. They bought a toy.

Bugatti. Exactly. Exactly, yeah. And so Mr. Beast had this thesis of like clickbait that delivers. If he says he's going to crush a Lamborghini, he's actually going to do it. And so YouTube changed to view duration and watch time. So they shifted to being more watchtime focused. And this did really solve the clickbait problem because videos that earn clicks, but not watch time were deprioritized. So you can be very high CTR but low AVD and you do not go viral in the algorithm. So this led to reliance. high user satisfaction, but there still is the problem of being glued to your phone watching something that is engaging. You watch all the way to the end, but after you're done and you're reflecting on like that hour that you spent watching that video, you're like, why did I waste an hour watching that? Because it was essentially slop, even if it wasn't created with AI, it was not good. And the signal to the algorithm right now is, I love that video. I watched 100% of it. But internally,

you know that that was a waste of time. And so you can churn. I wound up churning from TikTok at one point because I downloaded the app like years ago, scrolled and was like, yeah, this is some good stuff, this is interesting. But then afterwards, I was just like, this is candy. I'm not interested in this. And I just uninstall the app and I haven't been on since. And so there's what I think is like the third phase, which is companies are starting to do now. So phase one is click through rate optimization.

Then you get to AVD, average view duration, watch time optimization. Third is ARPOO, like average revenue per user optimization. So it's not enough to get someone to watch a full video. They hate because then they might not open the app again. I personally turned from TikTok for this reason. Recommendation algorithm that optimizes for average revenue per user over an entire year is going to focus on mitigating app churn. Because it doesn't matter that you just got one really great video.

They watched it all, but then they never came back. You want to optimize for, they came back a lot over the entire year. But that can still lead to brain rot. You can brain rot your users into being very high average revenue per user for a year for a while. But eventually they get so brain dead that they lose their jobs, drop out of society. They become depressed. They stop being good customers. They stop being good consumers, which is something we can't have in the American economy.

And so my answer to this and my pitch for these algorithmic platforms is to optimize for LTV, true LTV, the lifetime value of a user is usually estimated over a couple of years. If you're a startup, you look kind of crazy if you come in with like a five-year LTV. But if you think about how long I've been using YouTube, it's basically been my entire life, and it's been like 20 years or something I've been on that platform since it started in, what, 2005? And so really, you should be modeling the... Just saying YouTube is 20 years old really makes you feel old. Totally, totally. But you should be actually considering the full human lifespan. Like, algorithm video platforms should basically work like this. If I show this kid a brain rot video, I might be able to

show some cheap toy ads right now or some mobile game ads or something like that. But if I show this kid an educational video and I teach them how to program, teach them how to be a business person, then they go get a high paying job. And in 20 years, I'll be able to show them ads for Ferraris. And if I show them ads for Ferraris, I'll make more money discounted in today's dollars. And so you should have an incentive at the algorithmic level to actually upskill, actually engage people in having like a meaningful life where they wind up generating enough. So you're saying YouTube should actually create their own drop shipping course to help everyone become financially independent.

It could be a variety of things. Anything that leads to a fruitful life, right? So the current problem. I mean, I think it's just that tension between, you know, helping create a more of valuable user over time and extracting as much attention that you can get value from in the short term. Exactly. Short term versus long term. And then the attention there is that, you know, all these scaled platforms are public companies that need to, you know, show results today. For sure. And so there's, there's a tug award. None of the current recommendation algorithms are at all tuned to the

average Americans lifespan, which is 78 years. But I think they should be. They don't have enough data for this and I don't know if they will how they would solve for this before they have a full cohort. Like it's going to take YouTube's 20 years old, but the average American 70, so it lives 78 years. It's going to take another 50 years to get a full, just one cohort to be like, okay, now we know that, now we know what happens if you show kids this video when they're young or this video and we saw how their entire life played out. And turns out that this one was a better customer, so we should show them more of this. Like that's the feedback loop.

It's a 78 year feedback loop basically. It's crazy. Maybe not 78, but maybe 50 years. Like, it's a really long feedback loop. It's going to take a long time to get all that data. But that's how these algorithms should work. And so I have no idea how they will solve for this before they have a full cohort. It's going to be some rough estimation. But I continue to believe it should be the guiding light of all the stewards of large algorithmic content platforms.

Like, even if you don't have the data, even if you can't just train the algorithm on that, you should still be thinking about the type of content. Yes. People would talk about it in China. The Chinese version of TikTok shows like science and math videos. Why is that? While your children are getting shown, you know, slop. And they have like a 10,000 year plan over there.

They're thinking in decades. And so you as the steward of an algorithmic platform should be thinking in decades. And you should be thinking about if I take this, you know, quarterly earnings bump right now, but I wind up ruining all my customers' lives and they wind up being bad consumers, that's a financial issue. That does work within the capitalist framework. And so I think you should build an algorithm that optimizes for your users leading more flourishing lives so they can ultimately afford more expensive things and make you more money.

It's capitalism. It's best. Perfect blend of techno-optimism and hyper-capitalism. Yes. And for what it's worth. In one theory. I like it. The response to this, I gave an abbreviated version of this pitch.

And the response is magnificent. I think it resonated very well, which I was excited about. Well, you know what else is magnificent? What else? Ramp.com. Time is money. Save both. Easy-use corporate cards, bill payments, accounting, and a whole lot more all-in-one place. Orramp.com.

Another magnificent thing, potentially future member of the MAG8, Sam Altman, will be joining the show on Friday to talk about SORA. We're very excited for this. This is a real picture of him. Yes, he's been in the gym. This hasn't really been in wide circulation. but we kind of expect this to be one of the default headshots that he uses going forward. He looks fantastic here.

We're very excited for this. Lots of questions that we have around SORA. The real plan is. And I'm looking forward to it. It'll be noon Pacific on Friday. And in other news, Morgan Housel, friend of the show, has launched his new book, The Art of Spending Money. Is he coming on soon? We've got to get it out here's down.

I just sent him a message. I kind of botched it. I was supposed to email him a long time ago. Me too. We'll work on getting him on hopefully tomorrow or Thursday. Very excited to read the book myself. I'm assuming a lot of the book will be about Ramp, which we obviously know quite a lot about,

but I'm interested to see what else he's writing about. Yes. This is the psychology of corporate cards. That's the next book. That's the next book. Is he doing SponCon now? What? That's what we should rate. Good stuff.

Anyway, restream.i.o, one live stream 30 plus destinations, multi-stream, reach your audience, wherever they are. Get signed up for free folks. Fabricated knowledge. Doug O'Loughlin, he was on the show yesterday. Had a great time hanging out with him. He says, oh my God, face plant, face plant, Oracle lost nearly $100 million from rentals of NVIDIA Blackwell chips, which arrived this year. That's partly because there's a period between when Oracle gets its, data centers ready for customers.

And he's bolding this part. And when customers start using and paying for them, the documents show. It's not clear what causes the gap or how Oracle plans to shorten it. And so, of course, he's referencing an article by the information this morning. An absolute bombshell. Bombshell. They dropped internal Oracle data show financial challenge of renting out Nvidia chips. And obviously, Doug is highlighting that, yes, it's possible to lose money between the time

when you invest a lot of money in something and when you can actually sell the service. Yeah. So not a huge surprise there. So Joe Wisenthal had a little take on this. He said, is this from him? This is from a Bloomberg terminal, but I don't know if it's an information article put into a Bloomberg terminal. There's another post here from Dolly Ballie. You want to read that one?

He says, information article is stupid. If you look at the beginning of a site ramp, obviously there's no margin LOL. So anyways, timeline. Putting this one in the truth zone. Indeed. I think we need to ourselves. Indeed. The Financial Times had an interesting breakdown of OpenAI's current corporate structure, which...

Not enough red string here. Yeah, every time you think it can get... It's like, oh, oh, well, OpenAI has 12 different organizations within OpenAI. It can't get any crazier than this. Well, they go and do deals with it. every single company in the world, apparently. So this... Yeah, every time I list off how many companies are riding with OpenAI and how many companies

current market caps depend on their partnership with OpenAI, I always leave off too. I'm always like, think of like, Nvidia, Oracle, and CoreWeave. And then it's like, oh, there's actually Broadcom and AMD. There's also SoftBank. And there's also Microsoft. Yep. And it's just all over. And the interesting thing that we had been talking about off air yesterday,

is this dynamic where you kind of have to assume there'll be some type of correction in the coming two years, not a super bold statement to make. Yeah, we'll get to that. And OpenAI could very well be the only company in this crazy web that's private, which could be an advantage in an environment where we're seeing massive drawdowns.

And just due to Open AI effectively controlling your share price. Yeah. Yeah, I like this. take, I thought, what's interesting is that, like, being public during a major correction is not a death sentence. Like, Amazon got through it. They saw a massive drawdown. But is being private during a massive drawdown an asset? Is that something that you should be optimizing for? Does it matter at all? Because in some ways, if you go out, you're public, you raise a ton of money, you have a fortress balance sheet. Yes, you might have employees who are,

are disappointed because their personal financials are being marked to market and there's and there's in their way down that would be emotionally rough right but at the same time it is possible that at least it's public and at least you know there's some liquidity and you're not sitting there wondering like am i is this company actually worth zero like you're being marked to market and it's like if you're at amazon in 2000 2001 and you're down, I think they went down like 80, 90%, right? Yeah, Oracle withstood a more than 80% drawdown between 2000 and 2002. I really wonder how that was as employee.

Like if you were there at Oracle and you were like, I'm about to buy a house. And then you're like, I'm about to buy a shack. Like, was it that, like, was it that rough? Like, did you, like, did that emotionally affect your work? But do you, but did it emotionally? I've talked to, like, family. Where are you like, I'm too depressed to do the database migration. Like, I'm too depressed to call.

I have a family, I have a family friend that retired in their early 40s, kind of after the, after the dot com craziness. Yeah. And they sold not at the top. They sold about, they exited the majority of their position. I won't name the company prior, you know, six months out from the top. Sure. And so for a while, they looked really silly. Yeah.

And ultimately, they were telling me all the peers that they had were still working to achieve the wealth that they had on paper during 2000, 20 years later. And so you can, again, going through these is, I feel like the venture industry broadly had to go through this somewhat, obviously, in 2020, 2022, 23, right? everybody had these crazy paper marks that suddenly were not, not living up to expectations. Yeah. Or at least. There are some crazy, crazy examples. I don't know if Accomi is one of them.

There's a few companies. I think Acmei was actually, it still is not at its all-time high of the dot-com boom. Because it IPOed at $3909. It was at $327 a share. And now it's at 76. I never know if the splits are accurate and whatnot. But there's been a bunch of companies that traded down. I mean, Acomai went down by 99% and then built back up and saw 50x gain over the next 20 years as they built back.

And it was always a real company. It was just massively overvalued in the dot-com boom. But it will be a, oh, this is a funny post you just put in the timeline. There's a bubble in people calling. for bubbles, calling for bubbles, calling for bubbles? Are you calling for a bubble? We pray for a bubble. We used to pray for a bubble like this. We did. We did. Anyway. I do remember in 2023 during, you know, post-SVB, I wouldn't go so far as to say I was actually praying for a bubble, but I thought it'd be crazy if our industry went through something like that, quite that crazy again. And sure,

enough. We got it. And AI, the current AI hype cycle, hasn't even got to dance with low interest rates yet. It hasn't gotten to dance with low interest rates or leverage, really. Imagine that. Or leverage. Yeah. Yeah, everyone's saying it's 1999. Not many people saying whether or not it's January, December of 1990. Because that is important. What date did the actual stock market peak in 1999? I have a search here. It was actually March 10th. 2000, March 10th, 2000. So if it's January and you're like, oh, I'm calling, it's 1999. I'm going bold. I'm calling the top. It's like, no, you're calling, if you're calling a January of 1999, you're calling yourself 15 months out from the top, which is a wild amount of time. A lot of room to make money. Who knows? Yeah, there was, who was it yesterday? Paul Tudor Jones was saying he feels like there, he was on CNBC yesterday saying, hey, things feel overheated, but he

he was basically calling for like a blow off top. Sure. Like he says, in his view, it can get a lot crazier before a correction. So, and again, that feels like the,

the general sentiment right now is, you know, Doug on the show yesterday was saying, like, look, we're not even, most of these companies are not even levered yet.

And like, that is like almost certainly going to happen. Yeah. Except dorkal. They're pretty levered. But not that crazy levered. Like there, There is some fake news around the 5X leverage that's going on over there.

It's not quite that big of a deal. But let's switch years, but first let's tell you about Privy, wallet infrastructure for every bank. Privy makes it easy to build on crypto rail, securely spin up white label wallets, sign transactions, and integrate on-chain infrastructure all through one simple API. Joe Lonsdale is celebrating one of his portfolio companies. He says he's blown away by the talent behind some of these great companies,

just a casual Nobel Prize. So Sonoma Bios, Fred Ramsdell, won the 2025 Nobel Prize in medicine for his work on regulatory T cells or T-Regs, which act as an emergency break on the immune system. T-Regs are a critical link in understanding autoimmune disease and advancing immunotherapy more broadly.

So congrats to him. He's the co-founder, former chief scientific officer and a current advisor to Sonoma Bio, a trailblazer and cell therapy. in an 8BC portfolio company that 8BC has proudly supported from the beginning. He led the team that discovered and characterized the gene Fox P3. Talk about a cool name for a gene, FOXP3. I like that.

Which plays a critical role in T-Regs development leading to a new field of immunotherapies. The thesis continues to be validated with portfolio companies like Orca Bio, which recently received FDA acceptance, a priority review of its BLA and Orca-T in hemologic malignancies. See below. Joe, just comments cool. I agree. Did he comment cool? No, I think I clicked in. I think I clicked it.

I was wondering if he commented cool on his own post, which would be some king, a king move. Anyway, another ABC portfolio company, Cognition. Boom. The makers of Devon, the AI software engineer.

Crush your backlog with your personal engineering team. Martin Screlly's been on a tear. Got your capital breaks down. So Screlly does it again. He picked a stock that went up 1,400. percent today. That is an insane pop. Martin Scroly says long. S-P-R-B. His price target is $500. It was trading at 25 at the time, or I guess even less. It was trading at 8, and it went out to 152. On Friday, you could buy one share of Bruce Biosciences for $8.78. It's now trading at $197.97.

dollars per share. That is crazy. Still sitting at a $111 million market cap. I still don't understand, like, he can't run a fund, but he can trade this. Like, is this financial advice? Like, he seems like he's playing it really fast and loose. But it's clearly calling good stuff and doing research. So appreciate the word.

He's been, I think, making a lot of people a lot of money. Yeah. So everyone's happy. He says his price targets $500. Rare disease is his specialty. the ERT for San Philippa will be approved, CMC notwithstanding, and be the new standard of care. He said, I was working on a presentation when the news was scooped,

but this drug restores the abnormally high levels of heparin sulfate to normal. Without this drug, patients will die at 18, the normal ranges of HS. They may live a normal life. Well, that's really exciting. That seems really good. TjP says, I'll pass, already up too much. It's just, who knows.

up a lot. There's no way it could go up more. There's no way. Well, let me tell you about Figma. Think bigger, build faster. Figma helps design and development teams build great products together. Get started for free, brother. Tyler, start playing around with Figma Make in Jad ChbPT. I want you to test out the integration. The integration. That sounds very interesting. At Dev Day. Yeah. Do you know if those agent builder tools are like general availability already? Yeah. So I tried to use the agent builder earlier. today. And then you have to,

I couldn't get it to work because you have to verify your organization. And then I tried to put in my ID and then it just like said failed. So I don't know. Oh well. Yeah.

But I'll try it again. But they are generally available. Okay. Well, that's good. Well, Max Meyer is breaking down the free press acquisition. This was announced yesterday. Barry Weiss's free press

was acquired by Paramount. Max says, seeing so much cynicism and bitterness on the timeline about the free press. So let me tell you why I love it as a contributor and why it means so much to many people. The FP is a match made in heaven between writers and readers. We writers can't live without readers. And the FP readers are amazing.

Savvy, thoughtful, opinionated, but kind. I can't tell you how many emails, texts, phone calls, DMs I've received every time I've put a piece in the FP. These are Americans I want to know. And the FP is their gathering place. I've written pieces about very random things. Warren Buffett, my brother. Warren Buffett's not his brother.

There's a comma. He says, I've written about Warren Buffett. I've written about my brother. I've written about a Missouri resort town, the Latter Day Saints Church and SpaceX, which is an amazing blessing. And a credit to their smart editors. The FP team and their readers care deeply about America,

even if they disagree about politics. Some just scroll some of the common sections on Trump-related articles, and you'll see the amazing bites and disagreements. What Barry Nellie and the entire FP team have done in three years is now the model for media, make something so good that people would walk over glass to pay their hard earned money for it,

not to get through some terrible paywall, but to be part of something. The free press is worth every penny paramount paid, probably more. How often does a talent like Barry come along? And people who work in the world of talent should know this rather than be cynical or conspiratorial. Barry and team have a long way to go from here.

This is the beginning. The complainers should watch and learn. Which is exciting. Yeah, I mean, I just, Looked at this as a way to make CBS news relevant again to a new important class of readers, a much younger demographic than the CBS audience today. Yep. And you look at it as like a trade deal, right?

This is a talent acquisition. Barry Weiss will probably create far more than $150 million of value just by being a part of CBS. Yeah. Also, I mean, I do think that like, I mean, we talk about. about this a lot, but like audience quality is like still deeply misunderstood. A lot of people just look at total audience size, impressions. They see every reader as the same value. And that was certainly true in the days of the three major networks, CBS, NBC, ABC, but we're just in a completely different, much more fragmented world. And so I don't know if they had a million subscribers or something

like that, but the people that opted into the free press's funnel are probably way more monetizable than the people that are, you know, turning on late-night CBS for the, you know, 50th time. And so they definitely pulled in interesting people, pulled in me when they signed Tyler Cowan, and I've been enjoying his column there. He wrote a wild one about how his favorite new actresses AI, very controversial. He's really leaning into it. It's great. but Alex Broszini, Bronzini vendor is a perfect example of someone who's a little bit of a free press hater. So the free press on October 6th said the free press is joining Paramount and they share a link to the FP.com. They put the link right in there. And Alex says, this post has accrued 147 likes in four hours,

nearly one for every million dollars Paramount thought the free press was worth. And honestly, 147 likes on eggs with a link in there to a substack is, it's a miracle. It's a miracle. But, yeah, people are laughing. But, yeah, Nick says the New York Times gets similar likes on Twitter. You're not saying anything meaningful. And, yeah, I don't think anyone, it believes that Barry Weiss has as big of an audience as Mr. Beast. But that doesn't quite matter because it's a different audience.

and hitting certain audiences is more valuable than other audiences. Where else should we go? The entrepreneurial tech blazes. The hat is back in the news. The hat is back in the news. Anthropic hat. Yeah. Jordan says the tone of the, quote,

entrepreneurial tech glaze makes me sad for the whole world in all of history. That is a bold statement. I don't know. I mean, did we talk about this yesterday? Yeah. I think that Anthropics campaign is good. I don't think it's, and I can see why people want the hat. And Anthropic is a great brand, and they have a great team, and they have a unique culture.

Yeah. And they've been able to retain people. It's the place that many of the brightest minds in AI want to work at. Yeah. That said, I don't think this is an award-winning campaign. It's like they put a verb on a dad cap, right? We've been doing this since, like, 2016. 2016 is kind of how I clocked it.

They're selling coffee. A lot of people in tech have done this. I think this is more just excitement around anthropics overall positioning right as like the anti-slop AI company. Sure. So people are choosing to side with them and they've built up a cult around what they're doing, which is very powerful. But again, I'm not blown, I'm not blown away by the, I love the anthropic. team and I think their products are amazing. But personally, at no point was I thinking, I need a liner up around the block to get this hat, you know, and I actually wouldn't,

wouldn't wear the hat. But, um, um, so anyways, I would be caught dead in that. I wouldn't be caught dead in the hat. No. Don't even try to put one of these on my head. I, I, I think the hat is kind of, no, they're cool. They're cool. It's fine. It's just that you put a verb on a dad hat. Yeah. Do you want an award? No, they didn't ask for award. People are just excited about that. No, I know.

So Jordan Castro is reacting to the reaction. It's not like Anthropic went out there and was like, please, like, glaze us. You know, it's just like, they got glazed. And then Jordan is like, stop at the glaze. And it's not really their fault. But I think the bigger question is like, is like, is I, I just think the question is, can they get a true foothold in consumer? Because this campaign. That's a great question.

This campaign is very consumer focused. I saw these keep-thinking billboards on, I don't know, not Melrose, Sunset Boulevard in West Hollywood. Sandwich in between two friend.com billboards. Yeah, probably. I'm surprised they could even buy them. John and I drive, John and I drive around. You see friend billboards everywhere. You can't go to, you cannot go two blocks.

They're literally everywhere. It's like every other block. It's every block. The most massive billboards. I saw, we saw a friend.com billboard that is like basically behind the building. And I was like, Avi. Why didn't even pay for that one? Did you just really hit market buy?

He bought everything. He went on adquick.com. Out of home advertising made easy and measurable. Say goodbye to the headaches of out of home advertising. Only ad quick. Yeah. Get on ad quick because they will help you pick billboards that are not directly up against walls. Yeah.

No, you know in no circumstances. would an entrepreneur with limited resources look at this billboard and think, yep, I got to have this one. Because it's like more than two thirds of it was covered by a building. You had to be at the most insane angle. While you're driving by, you have to peek around out your window. Yeah. So I mean, but so here's what I'll say about enthrobics campaign. Yeah. Breath of fresh air. Yep. It's quality. Yep. It's not award winning. Yep. And I think that the people that like it, I totally get why they like it, breath of fresh air. I understand why other people that are not in tech are thinking, this is the best that you got. Yep. This is, this is your hero? Yeah. It's like the bar

is just pretty low, right? The bar overall is low. And one of the challenges is that AI, the brands of AI companies get tied to the outputs. Yep. And sometimes the outputs are great. And often, and most often, they're not so great, right? And so I'm interested to see how aggressive OpenAI gets with that video campaign they did with the, what's the tele, euphoria. Yeah, the euphoria. I thought those videos were great. Yes.

What's interesting is that Anthropic did a video series that was very similar aesthetically, warm tones, highly cinematic, shallow depth of field, very emotional, like just a few weeks or months earlier. I can't even keep track at this point. So both of them were experimenting in the same brand landscape. I personally would like our sponsor, Vanta, to do a hat that's actually just a huge llama helmet. It says automate compliance, manage, risk with trust continuously on it. And it should also say, you know, the anthropic hat is so simple, just as thinking.

It doesn't even say anthropic on it. I think the Vanta hat should say Vantas trust management platform takes the manual work out of your security compliance process and replaces it with continuous automation, whether you're pursuing your first framework or managing a complex program, like all the way around the hat. Just put the entire landing page on the hat. I think that would be innovative. I have a bunch of responses. I was thinking somebody needs to make a, nobody's made the AI smart helmet yet that you can wear all the time.

From the moment that you wake up to the moment you go to bed, potentially even when you're sleeping. Think about how much compute you could put in a helmet. You could have all day battery, all month battery. Also, wearing a. motorcycle helmet is incredibly cool. Hardcore. I mean, I know you haven't seen Kill Bill, but the bride, played by Uma Thurman in Kill Bill, Tarantino, a Tarantino classic, wears this motorcycle outfit with this motorcycle helmet and just

looks so sick. Yeah, so next time we talk with Zach, that's going to be my pitch. Yeah. AI motorcycle helmet that you can daily. Imagine you're just like, just rocking in the workplace, and you're just like, I'm locking in, I'm locking in. I don't, yeah, we're joking about it. but I could see this actually being created.

It's like a rugged helmet. Dude, it's more comfortable to wear a ski helmet or a motorcycle helmet than Applevision Pro. Let's just be honest about it. And so why not strap that whole thing on your head? I love it. It's amazing. I have a couple rebuttles to you.

First, I'm going to talk about graphite.dev. Code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster. But so, Anthropic in consumer. They're clearly behind, but it's very interesting because they have a fantastic model that can instantiate code very quickly. You could do all of this on-the-fly app development, basically. So when you ask a question, it actually builds a piece of software for you that's interactive. There's so many cool things that they can do with their foundation model. Also, they have my Krieger. And so I just feel like... So I was trying to push him on a little bit. What can you do on the social side to catalyze Claude's growth? Yeah, like when I think of the backlash to meta vibes, the backlash to Sora 2 and Sora. If I were to try and build a team that would not face backlash to a consumer product, to a new consumer product,

I would back Claude with Mike Krieger, right? Because he's going to build something that's like delightful, just like the early days of Instagram. I would think that he would inject that kind of like simplicity and beauty into the product that they build. Who knows? Maybe it's coming. Maybe every foundation model comes. company needs an answer to the AI slop TikTok feed. And I would be very excited to see what comes out of Anthropic.

The other take on this hat is, hat aside, I think that keep thinking is a good phrase. I think thinking is a good phrase. I think it's a good word. I think they picked that word well. And it excites me and it actually hits, like,

you can talk a lot about it, like, oh, just a word on a hat, But of all the words to pick, it feels more human. It feels less cyberpunkty. It feels less like super-human. It's very human.

Personal superintelligence feels like, okay, maybe that's opposed to my interests, like fast takeoff, all these scary AI doom things, which Anthropic is like honestly famous for. I think everybody is loving this campaign, but thinking machines laugh. That is true. Absolutely brutal. But I thought it was interesting that the message is keep thinking.

If you're a human, keep thinking. Don't stop thinking. Like, what we do here at Anthropic is not think for you. That's an interesting message. Because for a long time, people were saying, they say, we hear for you. A lot of the, what's that problem? Waystar, Roy, Co.

Oh, yeah. So a lot of, like, a lot of the AI, like, a lot of, lot of the AI rhetoric has been, the AI will be able to think for you. And Anthropic is specifically putting on a billboard, whether they believe it or not, or whether that's where AI actually goes. They're encouraging you to say, hey, keep thinking because that's not our job. Our job, you do the thinking, and we do the coding and all the boring stuff. Are you sure they're not just worried? It's like, hey, you guys are relying on Claude a little too much. Keep thinking, buddy. Maybe. Maybe. And so that is the question is like, is this Daria's new view?

Because the view that Dario has espoused on many a podcast is, yeah, you're not going to be thinking because there's like a 45% chance that everyone's dead. Because, you know, he comes out with some hardcore rhetoric that does not feel extremely warm or human or anything about like, hmm, hat. Like, are you going to be wearing hats if we're all paperclips? Like, no. And so, like, is this a, is this the marketing team in Anthropics saying, like, let's step away from Dario's viewpoint a little bit? Or is it more Dario has actually shifted his thinking and is saying, like, no, I actually feel like what we are building is complementary to humans. That would be exciting. And I feel like that would be inspiring to me and I would be excited about that.

And so I don't know where all this goes, but it's, I think it's a step in the right direction. There are so many other campaigns that could have been less human, more clanker coded and. would not have hit. And this one clearly hit to the point where it got so much positive attention that there was a backlash to the positive attention, which I think is it's kind of the best case scenario for a really positive ad campaign is that people get sick of people talking about how good your ad campaign was. So I think it was pretty good. We need to react to the AI homeless man prank on someone's dad. Yep. We've been instructed that we must do a re- Your silence has been deafening.

It has been deafening. Wait. Can we pull this video up? While we pull it up, Avi Schiffman is hit the timeline again. He says every single available bus shelter in L.A. Now has a friend ad. So apparently there weren't enough ads. That's what people were saying.

They said, Avi. They said more. We need more bellboards. You got to push this campaign further. There just aren't enough billboards. People aren't aware of your campaign. You're barely baking through, buddy. Step it up.

Let's watch the AI Homeless Man prank on my dad video. Okay, so you take a picture. Send it to the dad. I can't read any of this. This is too small. Is this like nanobanana adding the dad to the pictures of the actual house? Seven phone calls from the dad. So he's pitching.

I'm bringing in the homeless person into my house. No. Oh, in the bed. Do you think this is at all real or do you think this is all theater? Do you think this is all created? All theater. All theater? But it's mildly entertaining. Yeah.

It is an interesting case study in like AI slop in the sense that like it's using AI to create, you know, Photoshop text messages, I suppose. And, uh, and, and getting a laugh out of that. Yeah, it's like using AI to lie for entertainment. To lie, yes. But the lie is like a meta-commentary on the nature of AI slop, right? Because it's the joke is on the dad or the theoretical, the hypothetical dad and the character of the dad. Anyway, let's talk about Julius. What analysis do you want to run?

Chat with your data and get expert level insights in seconds. Ask Julius to answer. to analyze your data. Images that make you feel pain. There's a post here from Ishwarya, who says, lo-key, the anthropic NYC pop-up hype

is proof a lot of tech people have never been invited to a fashion week event or a real brand activation before keyword being invited. That's probably true. I mean, these are wildly different worlds.

Hayden Johnson has a post here that says every AI ad is like, hey Gemini, what would I have for lunch? And then the phone is like sandwich. And the guy is like, wow. That's so funny. Wow. Okay, we need to watch the video of Ken Griffin, uh, given some hot takes about the AI. This is, this should be the most near term bearish thing you watch today. Let's see. From Bourbon Capital, Burbin Capital highlights it. Finally, someone isn't afraid to speak the truth. People even speaking the truth or, you know, give it hot takes for months.

Let's just make that clear, but let's hear it from Ken Griffin himself at the future of global markets 2025. Ciddle Securities Conference. Large language models in particular have been in existence for a few years. I think there's a real chance that when we think about the forms of AI that are alluded to by many of the markets leaders today, that their dream for the future may not take three to five years to play out. It could be 20 years. It could be 30 years. You know, it's like the Internet, and I, people hate this analogy,

like the dot-com moment back at the start of this century. Did anyone have any doubt that the Internet was going to change the world? No. Has the Internet changed the world? Absolutely. But did it take longer than anticipated? And was there a period in which there was a real sort of sorting of the winners and losers that took place, unequivocally, unequivocally.

And I think that the AI story will have many of the same components. There's a bubble and people calling for bubbles. Tyler. But I think this is just ultimately a very sound analysis, and it's coming from somebody who is highly intelligent, highly accomplished, highly invested, and yet doesn't personally, to my knowledge, have a billion dollars of equity in a lab who's incentivized to keep the part.

Well, he might have a billion dollars like this second and then not tomorrow. And then no, no, the next second and then this second. And then, Ken's going to make money no matter what. Yeah, yeah. He doesn't, you know, I'm sure he's making a lot of money now. Maybe in a correction he makes less, but he's going to do fine no matter what. And so it's important to- -Suderdale got roasted in the financial crisis.

Yeah. They drew down 50%. Yeah. He almost like wind down the front. Yeah, it was really, really rough. But it's important to not just listen to the predictions and opinions of people who have the heaviest bags. Sure, sure, sure. That are dependent on keeping the party going as long as possible.

Tyler, what you got? So it could also be that his competitors are invested in the AI labs, right? I think Jane Street is big in... Anthropic. Yeah, big anthropic. Oh, okay, yeah. So maybe it's really... Lack of bags, problem with the AI labs.

lack of bags, so he's bearish. Sidelined. Lack of bags is as much of a conflict as not having conflicts is a conflict because you don't have bags and so you're incentivized to be bearish on people that have bags. Yeah. This is the fact. He wants to crush his enemies. Tyler, how many days until the singularity?

We're still at 3650. I guess it should have. There wasn't a lot of big news since I changed it last. So I guess it should have gone down every day. So 10 years until the singularity, we need to be updating that day. So at least we're ticking down. But watching that King Griffin clip, are you adding days? Are you removing days?

What are you thinking? Griffin is like kind of boomer-coded. So I'm saying we're closer to AGI after watching this. Okay. You think he's a contra indicator for AGI progress? Yeah, yeah. Okay. So you're removing days.

So we're down at like 3,000 days now, just nine years. I mean, 600 days, I think that's almost two years, right? That's a little too big of an update. Yeah. But I would say- Did either of you watch the Dworkash Sutton reaction, the apology video? Someone said, this is so funny. It was not apology video.

Dworkesh did apologize in the video, but only saying, I apologize if I'm mischaracterizing your statement. And I have some of the ground truth wrong. But I thought it was interesting. I felt like I came away, originally thinking that the bitter lesson was poor endless compute behind any algorithm that's working and scaling when Sutton kind of reclassified. it as like, like, the game of the AI researcher is to hunt for the next algorithm that will scale with compute. And he's not saying that any algorithm will necessarily scale infinitely with compute. And so he, like, my, my takeaway was that he was, he's, if he looks at LLMs,

he's like, good progress for four ooms, maybe need a new thing to go for the next five ooms or something like that. Was that your, was that your read? Yeah, I mean, the thing about LMs is like, like the scaling, like you, there's just no more data to scale with. So it's like not kind of bitter or lessened built in that sense anymore because you can't just keep throwing more compute at the same LMs because if you just make the models bigger or whatever, it's not like it's not going to get better. The model won't be better because the, because the data is is constrained. Yeah, yeah. It's this weird thing where like on like you can zoom out and look at like computing power over the past 50 years. see a very, very smooth curve. Then you can look at like the AI winters and see that there's like

these huge steps in progress, the transformer paper, the, I don't know, there's been a whole bunch of like image net, right? Reasoning models. Reasoning models. Big, big step forward. And those are like, those are like a decade apart sometimes. And so you get these like qualitative bumps where you're like, oh, wow, like this feels different. But then in that interim, you might not see a lot of like qualitative jumps in clear progress, clear new capabilities, but the diffusion of that economic value is immense. And so we like technically pre-LLM, we were kind of in an AI winter, I suppose, where like, you know, self-driving cars were like 10 years behind and nothing was really like popping in AI.

AI was not a buzzword. People weren't using dot AI domains. But in that, in the, in the period of time from like calling, it 2010 to 2020, like, what was the economic impact of AI? Like, it was insane, right? Because Netflix was doing recommendation algorithms. Amazon was doing recommendation algorithms. The Facebook algorithmic ad feed, TikTok feeds.

Like, AI wasn't transformer-based LMs, but AI was a massive driver of economic growth. In Gwila created the AI hype cycle? Probably. In order to sell more. i.i domains. Isn't like 50% of their jobs? GDP AI domains now. 50% of their like budget

basically like budget comes from it's remarkable. It's remarkable. Well, if you want to generate some media head over to Fall the generative media platform for developers, the world's best generative image, video and audio models all in one place, develop and fine-tuned models with serverless GPUs and on-demand

clusters. If you want to bring our soundboard into your team meeting, team zooms, your team hangouts, go to TbPN.com slash sounds brought to you by fall. Yes. Chubby says, I double-checked. You can already buy Unitri's G-1 at Walmart, only $21,000 shipping in one week. Tesla must hurry to deploy their optimists. Do we buy one? Do we buy one? We got to, right? We have to buy one of these and have Tyler fight it.

It's such a wild thing to let a fox into the hen house, but sometimes you've got to do it, right? The global economy let Sam all in the fox into the hen house. and we'll see what happens. The fox and the henhouse analogy is amazing. But I don't think of the TPP and Ultram as a henhouse. We got to get the UD3 G1 Basic. This is the CCP Fox into the Ultradome and just... I think of the TBP and Ultram less as a henhouse and more as a bull ring.

It's more of a foxhouse. We're still on the farm. But over here you have the bulls jacked fighting. You get the horns. You mess with the bull. You get the horns. Unitary you're on. notice if you send over a unitary g1 basic for $21,000 shipping in one week and you try any funny

stuff you're getting the horns because this is not a I would say this is like letting a clanker into the fox house we're all foxes now I think we I think I this is a hen in the fox house in the foxhole I'd be kind of worried leaving the G1 basic with Tyler the you know presumably Tyler would be tasked with setting it up, but then you're thinking about, okay, the G1 is going to be here with Tyler. Potentially after hours, and I just should be a little bit worried about it.

So we can train it to swipe ramp cards or something? What can we do with this? You could automatic gong hits. Ooh, automatic gong hits sounds good. Put me out of a job. Wow, real, real black fellow economic. Dario was right. Automate.

White collar. white collar work, a gong hits. White collar and I hit a gong. That is remarkable. I can't believe we're doing this again like we did with drones where we're just going to let Unitree just flood the market. Yes, where is, where are the hawks? I'm normally, normally, you know, against government intervention.

But two-hour battery life? Look at this thing. We got to get one. Tyler wants it so badly. I wonder what else you can do with it. Like how strong is the API? Does it have like the ability to issue voice hands? Okay, this is what we do.

Can you like wrap the hands with ramp cards? Swiping everything. You get dangerous. Anyway, we are being joined by Shane from Polymarking in just a few minutes. In the meantime, let me tell you about what Bucco Capital bloke is saying on the timeline. He says, we had a good thing, you stupid son of a, we had 98% gross margins and LTV to KAC ratios above five. We were considered safer than first lean debt.

Nobody questioned our terminal value or stock-based compensation. We had everything we needed, and it all ran like clockwork. You could have shut your mouth, funded a few new SaaS companies each year, and made as much money as you needed. But you just had to blow it up. You and your pride and your ego. Talking about AI. Lots of people having fun. Well, in the rest of AI world.

Obviously, Open AI sent around a bunch of basically deal toys for people that spent over or used a trillion tokens with Open AI. I, Notion got one, Ramp got one, a bunch of others. Lauren Good says, this is sort of like spending a million bucks gambling and the casino gives you a free hotel room for the night. It's close. I think that's the next leg up in the cycle. Cognition got one trillion. Hacking.

Ramp got one. And I was trying to do the math on how much a trillion tokens actually cost so I could clock how much these companies are spending. ChatGPT clocked it as if they're doing GPT 3.5 turbo, that's the cheapest, cheaper case. That's a million bucks. If it's GPT4 standard, that's $45 million. So potentially hitting, I mean, Notion has talked about how this hit their gross margins. It's not an immaterial cost, but a lot of value to the customers.

And Tyler, what's the cheapest way? Do you have an update on how cheaply you could acquire one of these deal toys? Yeah, well, so the main thing you would do is you can do batched API calls. So instead of just doing like one and then you get the response, you send like a ton and then it, I think it's like within 24 hours, you'll get the response back. Do you actually do that when you run your examples? No, none of the stuff I do is... Oh, so money just doesn't matter you around here. You just don't care about cost optimization at TV. I got to be timely. Oh, yeah, that's true. But I think that is... That is money. I think that's either 50% or 90%. And then the other thing you do is you can do cashed inputs, which is like, I forget which one.

is which 50 or 90%. But if you combine those, then it's like massive savings. So then I think if you combine those with, is 3.5 triple the cheapest? Or I feel like it's GPT5 Nano. Yeah, something like that. But I think you could probably do it for a couple thousand dollars. I think I saw someone, maybe they were talking about 10 million tokens.

10 billion tokens, but they said $25 for 10 billion tokens. So for two grand, you could potentially get one of these deal toys that are pretty rare. You might be able to flip these. Who knows? This is a piece of lore. Did you see Matt Levine's summary of AMD in opening eye? Yes. Martin Scrolly put this in the truth zone to be clear.

Let's read it. But first, let me tell you about turbopuffer, search every byte, serverless vector and full-text search. Build from first principles on object storage, fast, 10x cheaper, and extremely scalable. Read Matt Levine's summary. How do these negotiations go, like schematically? Open AI, we would like six gigawatts worth of your chips to do inference. AMD, terrific.

That will be 78 billion. How would you like to pay? opening i well we were thinking that we would announce the deal and that would add 78 billion to the value of your company which should cover it amd dot dot dot dot dot amd no i'm pretty sure you have to pay for the chips opening i why uh amd i don't know it just seems wrong not to opening i okay why don't we pay you cash for the value of the chips and you give us back stock and when we announce the deal the stock will go up and we'll get our 78 billion back amd uh here we go amd yeah i guess that works So I feel like we should get some of the value.

Okay, you can have half. You give us stock worth like $35 billion and you can keep the rest. Martin Screlly didn't like it. He said, this is not at all what happened. And Matt Levine is stupid. You shouldn't read him. Zero Days Wall Street Experience Word Sell. He's not happy about it.

Mogged. Mogged. But who knows. This is interesting. I think it's a funny take. From Jay Bull, Tard on X. How hilarious is this Oracle move? Oracle makes all the AI names rip on some made-up CAP-X guide using revenue that doesn't exist from OpenAI,

and now it sinks the entire space with bad news regarding AI CAP-X not working. What a joke. I mean, how quickly do you expect the CAP-X to pay off? Oracle's sold off hard this morning. Went from $290 a share to $270 to share, but it's climbing back up. We may be back to where we were. I think that's because people are reading into it. to that article about the margins.

The new Hindenberg research. Potentially, but I think people corrected the record pretty quickly on that. Yeah, you were expecting high margins, what it's like two weeks out from the Oracle News, maybe a month. I mean, I guess we were at the stock exchange for the Klarna IPO when the Oracle News broke, right?

Or what, it was then, right? Yeah, that week. So that was just a couple weeks ago. And I mean, it was definitely like the most aggressive deal ever and it was a crazy deal, but people are potentially reading too much into the near term financial impacts. Because it just like takes time to understand. Like if the demand shows up and more people pay and the ads monetize and more people are, you know, seeing ads on Sora or paying for credits, like there is a world where the economy

keeps flowing back and forth, and consumers are putting money into the system. And so even these circular deals work out. You want to know something bullish? Please. I'd love it. A. Bolshe's flagship AI product is called Gigachat. That's bearish, because Russia is the bear, right? Their mascot. That's a crazy mascot to pick, Russia. Just instantly bearish in your entire country. Wait, what is it? It's called gigach chat. Gigach chat. That's a pretty good name. All the different countries are really leaning into their sovereign AI.

Yeah. You don't hear much about Russian AI, though. No. I was wondering when they would get around to announcing something. Because they have all these incredible math, like math experts. You'd think that they would be like really, really good. But maybe they've been preoccupied with a bunch of other stuff going on. Certainly.

Yeah, I mean, they lost Nebius too, which would have been a national champion had they not gotten into a cold war. That is a crazy story. Yeah, we got to have the nebius CEO on and dig into that. It's what a wild, wild time. Good morning, Nat, Ellison. Is it Eliasson? How do I pronounce that? Matt Eliasson. Good morning. Matt. Remember the name. Good morning. Do we have time to get through this Wilmonitis post? This looks a little long. Yeah, this is the new meta, I think. I think the algorithm has changed on X. It used to be all about short, pithy, punchy posts, but now the 500-word long post is doing really well. And so we're seeing a lot more of those. Jeremy Gaffon has dropped a couple that have done very well. Well,

Manitus is now getting in on the action. Let's read through it. He says every few weeks now, a fee drunk allocator declares a company going from zero to two million is not interesting. Only the ones that go from zero to 500 million. This is, of course, a very silly idea, but it is proximal to an important one. We are returning to a much older model for growth of internet companies. Welcome to the deal yuga. Following the dot-com crash, the modern internet company was defined by sales as its fundamental motion for growth. Scared of the business development of the 90s, companies would hire farms of thousands of SDRs to smile, dial, and build pipeline. ACVs across the businesses fell steadily over the decades as we realized if we lowered the contract

pipe, the contract price, collective delusions like product-led growth allowed increasingly lower and lower individuals on the totem pole to sign software contracts when cold-called and thus increased the speed in which a business could grow, albeit in smaller and smaller steps. As the discretionary spend capabilities of a random software engineer increased, this is like the bottom-up adoption of software, so did the viability of product-led growth, and thus we cargo-culted a set of growth metrics built around it, time to $100 million. cloud 100, et cetera.

Some good businesses were built alongside many bad ones. It's easy to think of PLG as a trauma response to the dot com and subsequent SaaS bubbles. If you lower your engine of growth to only grow in human-sized chunks, you can both grow increasingly predictably that revenue and decrease the damage of any given customer imploding has on your entire book. The best way to think of PLG perhaps is to perhaps imagine it as the golden calf of Exodus 32. If we had the real God enterprise, and we had the real God enterprise, and we decided to take the easier path after he stopped, when he stopped speaking momentarily.

The shift we are watching today with many companies growing quickly, one to $100 million in 48 hours, and many companies growing in massive chunks, OpenAI, owning AMD, etc., is not a new model, but instead a return to the origin of this industry. Deals. Deals are, of course, the origin. of the internet business itself. The internet business, as originally built, raised money exclusive to conduct deals.

The 90s were awash with channel partnerships, leasing agreements, customer book buyouts, IP sharing and washing, and all kinds of beautiful deals. Deals. Deals, in all caps. The incredible Alex Danko, speaking on Jackson Dolls podcast,

put this well. Quote, one of the interesting hallmarks of that era was that much of the genuinely forward progress a company had to make was by doing deals. The incredible stable, surfaces and platforms we enjoy today from which you can just build and get growth didn't exist. Companies were ultimately dependent on business development and a much earlier and more critical stage. The path was much more milestone-driven, involving major deals, for instance, you had to secure

vendors for compute, buy computers, buy a database from Oracle, and also arranged distribution deals priced in than it ought to have been. Sound familiar? It certainly should. Even Oracle is still the same. These 100 to 5.5,000, to $500 million stories are not software companies growing fundamentally quicker than their predecessors. They are software companies growing in fundamentally different ways. As we return to the deal yuga, we should expect the shape of these companies and the teams that run them to change even further. The teams and investors that learn from history from the first great deal age will avoid the many pitfalls that come from this kind of growth. Chief amongst them, the idea that

partnerships and infrastructure providers much larger than you are somehow win-win altruistic arrangements and not having you the device of your underdog, of your undoing and hoping for the best, welcome to the second deals era. We've never been more back. It's fun. Very similar to what... Yeah, I mean, this is what people say about, you know, in another life, Sam Alton would be on Wall Street, being as a banker, right? Doing deals.

Yeah. So it is, even his critics should admit that it's remarkable. Fantastic deals guy. remarkable. And like, I think people have been so skeptical of the early deals, like the Microsoft deal. It seems like that deal completely saved OpenAI. They were never going to get compute as a nonprofit. They had to do this insane deal to create this for-profit entity. Very complicated, something that requires an immense amount of, you know, greasing the wheels. And we're just seeing more and more of these like founder-to-founder deals.

no investment bankers involved. It really is the era of the high agency dealmaker. Jeremy Gaffan was talking about this. I asked Patrick Hoshanasi about this on the show, this idea that in the future, folks with private equity backgrounds, investment banking backgrounds, might be better as founders

because they are more set up to be deals guys on day one, as opposed to the more product-focused, PLG-focused, engineering-focused founder of the YC-2010 era, that was focused on building a product that was sticky, had a viral loop, would grow and kind of, you know,

just compound. And then eventually they would learn to do deals at some point, but that was something that they would add on later, like learning financial engineering in the second decade of your company. Now companies and founders seem to be adopting these ideas on day one. For sure. Well, there's a video that went quite viral yesterday that we can pull up in the timeline,

another illustration of the deals between Nvidia, AMD, and Open AI. While we pull that up, let me tell you

about profound, get your brand mentioned in ShotGBT. Reach millions of consumers who are using AI to discover products and brands. And we will pull up

this video. Let's see it. Where did this come from? You just saw this on the timeline? This is... Is this AI generating? A lore. Okay.

We need some audio here. Let's see. AMD, Nvidia. Isn't the three stooges? I think it is. Uh-uh. You owe me 20. Here's 10.

I owe you 10. Here's a 10 I owe you. Speaking of money, how about the 20 bucks you owe me? Oh, yeah. Well, I only got 10, so here's 10. I owe you 10. Thanks. Hey, Mo.

You owe me 20. Well, here's 10. I owe you 10. Uh-uh. You owe me 20. Here's 10. I owe you. Here's a 10.

I owe you. Here's a 10. I owe you. Good. Man, we're all even. And speaking of money, how about the 20 bucks? Just keeps looping. Somebody quoted this and said a cup he says 25 years ago, this was called round-tripping and was considered illegal.

After investors suffered trillions and losses, it will probably be considered illegal again. So we will see, it's kind of the Wild West right now. Well, in non-AI news for once, we have Shane from Polly Market joining the show. He's in the stream waiting room. We'll bring him into the TV. in Ultradome. Shane, how are you doing? Look at that setup.

Shane, can you hear us? I can hear you guys. I cannot see you guys. Okay. Okay, well, trust that we are looking at you and smiling at you. You look fantastic. You look fantastic. What a day.

Today's about you. Break down the news. It's hard not to look fantastic with this backdrop. This backdrop. Yeah. Right there. I had to face time on my family members be like, look at that.

You know, now with all the AI stuff, you got to show that it's true. You got to break through. From the bathroom office in 2020 to the NICC in 2025, anything is possible. What a run. Take us through the deal today. Take us through the news. What are the key points in the announcement? Yeah, I mean, look, I've never been really like a deal guy or a guy that's very focused on, you know,

I would say a lot of entrepreneurs, they're very optimizing around fundraising, et cetera. I really like just like building things. And usually what I've found is when you get traction and when you get motion and when people start taking notice of what you do, the money usually follows. And in this case, there was a lot of interest from people who wanted to invest. And I'm very grateful and humbled by all of it.

And then Jeff reached out to me and it was kind of like, we got to talking. He has key nose markets and, traditional finance better than anybody. He runs his own company that he started solo founder like myself. And there were all these ideas that we had that we could do together and we were very like-minded on it. And he was a big fan of polymarket. So when he started saying, hey, let's do something, there were a lot of synergies that came together quickly. So it's very exciting. And this is not a token strategic investment. I mean, this is very, very meaningful. I think

I correct me if I'm wrong, but this is like the single, this is the single largest investment in a crypto company ever. Okay, no. Yeah. Give us the actual numbers. What's the structure of the deal? Is this, this is a strategic deal? It's not like, are you giving it a series number, like a letter?

Like, how are you actually describing the deal? I want to ring the gong for you. I want the numbers. Yeah. they're investing $2 billion. Let's go. Hit that gong. There we go.

Oh, the gongs from your side. I thought the gong was like a periodic New York Stock Exchange thing. The gong is from our side. We have it in the studio. What can you share about what the partnership will actually look like over the next 12, 24 months and beyond? Like what's most exciting about it to you? I mean, look, it's right now such a moment for prediction markets. It's been exciting for us to sort of be, like, you know, sit there.

And even two years ago, it really was a category that people kept writing off and saying, oh, yeah, it's an idea that makes so much sense, but it's never going to work. We're really excited that we were able to buy QC and be set up to go to enter the U.S. market. I think it's lining up perfectly. The stars are lining. the timing's about as good as it could be. So on that side, there's a lot in terms of data and in terms of distribution and sort of who can have access to polymarket, particularly in the US, but also globally, because

ICE has, you know, exchange assets all around the world. But also, Jeff in particular, as he mentioned on CBC earlier, is like he's a big believer in tokenization. He's a big believer in all assets are going to be tokenized. It's a superior technology for the exchange of apps. assets. And Polly Market is the biggest consumer product that's built on tokenization under the hood. For every prediction market is actually tokenized. There's a yes token and no token. So there's a lot of ideas we've kicked around there. And I think ICE has incredible savvy

and Jeff is incredible savvy about how you can do these things. Utilizing existing assets and understanding sort of the existing matrix of U.S. regulation. And then we're where there could be pockets that there may have to be collaboration between the regulators and the innovators for how to basically do things that exist, but in a more efficient way, utilizing new technology. So that's the thing that I'm definitely going to be leaning on him for it. That's not the thing that I'm, that, you know, I can't say that I'm the best of that like he is. But in terms of the consumer side in building the products, ICE doesn't have any consumer properties, right? So that's where we come in, and that's where I think him and I, we have this mutual understanding common ground of how

these products could work. It's a very innovative guy, very forward thinking. And we can kind of see the vision in terms of polymarket.com, how it can look there, and he can understand the infrastructure and how to make it happen from the regulatory and infrastructure side. Did you ever expect prediction markets to be in South Park to have an episode? Was that on your bingo card for this year? Was that still a surprise? Yeah, I mean, I love South Park. Like, I grew up a South Park. I grew up a South Park Kid. I remember watching the original movie. I remember being a Best Buy and buying season two, season three on VHS. And look, I think that it's funny because the geopolitical markets on Polly Market have proven to be extremely useful, like in the WhatsApp groups

from people in the Middle East and people who are into open source intelligence, OSENT, they really watch Polly Market like a hawk. And the sort of crux of the episode being this idea of like, you know, this sort of, will the mother in the Jewish family, you know, bomb Gaza? Like that kind of being, it's this play on those markets. Like, it's definitely pushing it, but it's also just so South Park. And, you know, it's obviously all in good fun, but the way that they interpreted the situation

and distilled it into a South Park episode, you know, you wouldn't expect anything less from them. What about just the Intercontinental Exchange, ICE, founded in 2000, backed by Goldman Sachs, Morgan Stanley, Shell, Deutsche Bank, Sock General. Like, it is the ultimate bridge to Wall Street. How important are those connections? Are you already plugged into all the Wall Street firms? Is this a new pipeline for you to just Wall Street generally? professionalization, like how does that all play out for you?

Absolutely. Look, I've historically been more of a cold email guy, so I don't think that would get me that far on this street. And yeah, you know, when I talk to Jeff and Coe about this, like they know everyone, and everyone's a text or an email or phone call away. And that's really exciting because there's a lot of interest. Every time I meet somebody who's, you know, from the traditional world high up, they love polymarket.

They love what we do. They're obsessed with the data. And they oftentimes track it. And, you know, it's something, the Polymerk's saying that they look at frequently. And, you know, now, one, having the bridge, but also having the credibility of such a great partner, it's probably hard to underestimate how important that will be in terms of us bridging the gap,

especially as we ran into the U.S. market and we opened up for institutions. Fantastic. Anything else? Well, absolutely iconic image today. Yeah. you outside the exchange. Congratulations.

You know, it's funny, like, we've had so many misconnections, right? Like, there were so many times we were going to hop on, and I was going to be in a little booth or, like, a little chair in the back. And I said, you know, one time, if I'm going to go on, if I'm going to go on your guys' pod. It's got to be special. I got to figure out of mug one way or another. Well, Mog, you did. So here we are.

Thank you for everything that you do to make this show possible. I appreciate it. I'm a huge fan of what you guys do, and it's all. awesome to see right there that the polymarket ticker. There it is. You guys are a real, real champions, pioneers. Thanks so much for having me. Have a great right to have you.

Congratulations to you and the whole team. We'll talk to you soon. Cheers. Those cute look fantastic. Absolutely. Mogging. We have another post here. Blake Robbins said yesterday, Open AI is operating on a different level.

The amount they have shipped in the past few weeks and months is in incredible. Eric says, indeed impressive, but the scattershot nature raises questions about the company's discipline and ability to support these desperate initiatives. Is OpenAI Frontier Research Lab, a social network operator, a commerce engine, a hardware company, because it's hard to do all of that well. I would say they are a hyperscaler. If you look at any of the hyperscalers, they have a frontier research lab. They usually have a social network. They usually have a commerce engine. They usually have a hardware division. Yep. And it is hard to do all of them well. I think that

Sam is playing for keeps, right? He's, you know, there's a lot of founders that wouldn't have launched ASORA as a standalone app, right? There's a lot of founders that wouldn't acquire a company for six and a half billion dollars that doesn't have a working products yet and just kind of an idea of what they want to build, right? But he is incredibly aggressive. He's fast becoming too big to fail. And, you know, I think that put differently, if you asked, okay, this company has 800 million weekly active users. Do you think it's

fair that they would get experiment with a social network? Do you think it's fair that? So there's, there's a lot of companies with far less scale that are doing this kind of range of activities, but you have to, my framework for Open AI now is looking at them as a, you know, as a hyperscaler that's going to launch products all the time, some of which will work, some won't. And that's totally okay. They're just, they're experimenting, their shipping, and they're taking big swings because at this point for them to take, for them to create something that's meaningful to their business, it has to be a billion dollar opportunity.

I agree. I have a rebuttal, but first I'm going to tell you about linear. Linear is a purpose-built tool for planning and build products. Meet the system for my or software development, streamline issues, projects, and product roadmap. So back in GPT3, launched 2020, I was playing around in the playground, and I was writing video essays at the time about companies, and I wanted to create a list of companies that could potentially profile. And so back in the day, there was no prompt, there was no RL, there was no RL, there was no RLHF. And so you had to, as Rune told us, like the foundation models are complete hallucination. And so you had to write a prompt that would really set it up to just continue what you were saying,

because it was just guessing the next word and it was not very good at it. And what I did was I went to GPT3 in the playground. And I typed out like some of the most interesting business stories in history. And I put one, Theranos, two, we work, three, Lehman Brothers, four, space, and then it would continue to fill in based on your prompt that you had written, and it's just trying to guess the next word. And it actually did come up with some interesting companies that weren't on my radar, and I was able to go and Google them and fact-checked them and understand those stories.

I never really became part of my workflow, but I thought it was interesting. And at the time, I was thinking, like, this is a search engine. Like, it's just something that I would normally have gone to Google and search for, like, top 10 most interesting companies, and I would have gotten a listicle, but GPT3 just gave me the text itself. And so for a while, my model for chat GPT and Open AI has been like they're just the new Google. And my wondering is like how much of the culture of Google has, you know, as a natural outgrowth of just having a fast-growing consumer, you know, web app. And if they're, if chat GPT is growing, then will they have a bunch of 20% time projects? And so the social network operator,

the commerce engine, the hardware company, like Google's tried a bunch of those. They tried Google Plus. They tried Google Glass. Those didn't work. But YouTube's a great social network loosely. And their Android phone ecosystem is a successful hardware bet. Commerce engine, certainly that's cooking at Google. And so I do wonder how much you, like, I agree, the scattershot nature raises questions about the company's discipline. But we've seen it play out before where if you have an engine that's growing, growing, growing, you can actually go and do a bunch of 20% time projects. Some of them hit, some of them don't. Also, Sam Altman has never come out and said focus is our highest virtue, and we're just going to work on chat, GBT. It's like,

okay, their research, they're becoming an infrastructure player, right? They're going to experiment in all the different areas. They have an API, cloud business. I think I would be interested to see if they do, again, like do they acquire someone like a Snapchat. That's the point. Does that happen? Yeah, I don't know if they'll do. At some point, they have to do a bunch of acquisitions, I would imagine.

They're already doing some of them. Didn't they buy? Stats. Statsig, and obviously other companies as well. They announced an acquisition of a company called Joy today, the other day, which was like a finance tool. Oh, yeah, yeah. So they're adding a bunch of stuff. I mean, other hyperscalers have done this in the past.

I wonder, you know, obviously there's always a risk of being too scattered, but it is a large company. And it's also interesting to look at the time frame. A lot of people are viewing OpenAI as a company that started in 2023, essentially, even though ChachapD launched in 2022. They really feel like the Open AI story started in 2023. It did not. The company did start in 2015. And it was a frontier research lab for now a decade. And so if you look at the timeline for Google, well, Google search launches in 1998.

And then six years later in 2004, they're launching Gmail. that's a great product and it worked out. And so should you give them credit for the time they were wandering in the wilderness as a non-profit or should you comp them just to when they got product market fit on chat GPT? Maybe something in between. But there's certainly a lineage of companies that find really strong internet product market fit, scaling, growing, and then doing other projects and not having those, not getting over their skis. A social network operator? What's the real cost if SORA blows up and it never goes anywhere and they light a bunch of GPUs on fire. Well, if it fails, they won't be lighting that many GPUs on fire because nobody will be generating

anything. If their hardware project goes south, like, yeah, that's probably multi-billion dollar investment, but a company of their scale at $500 billion with billions coming in from all sorts of different deals can probably take that hit on the chin and keep moving and keep refocusing on the main, on the main, main project. So I don't know how, how, how big of a worry it is, but they certainly don't seem to have been, you know, massively falling behind in the core product. We're not seeing, you know, huge turn or anything. Yeah.

Doug, from semi-analysis, says OpenAI is tweeting about real interest rates. We get a new Fed share that's whole job as to lower rates to as low as possible. O B, B means no taxes. Money becomes really cheap. You've got to invest in America at all costs. Wow, it's so clear. Yeah, people are excited. DJT certainly wants you to invest in America.

Yeah. Well, let me tell you about numeralhq.com. Sales tax on autopilot, spend less than five minutes per month on sales tax compliance. Peter Wildaford has a forecast on how AI infrastructure spending goes over the next four years and what kind of models will see as a result. This is a very bullish take on infrastructure that I thought would be interesting to read through. I continue to assess that we are on track for models that can replace a wide variety of human labor, sometimes four to 15 years from now.

It's a big gap. Right in that singularity calculator we have going. He says, currently the world does not have any operational one gigawatt plus data centers. However, it is very likely that we will see a fully operational one gigawatt data center before mid-20. 26, this will, this likely will be a part of a 45 to 60 gigawatt of total computer cross-meda, Microsoft, Amazon, ABUS, Anthropic, OpenAI and Oracle, Google and DeepMind, and XAI. So everyone's building huge clusters. They're all investing a ton in CAPEX, and it will net out to something between 45 and 60

gigawatts next year. His median expectation is that these largest one-gigawatt data center facilities will hold between 400 and 500,000 Nvidia Blackwell chips and be used to train 4 to the 27th flops or 4E27 so 4 times 10 to the 27th

flop model sometime before the end of 2027. Such a model would be 10x larger than the largest model today and 100x larger than GPT4. Each individual 1 gigawatt facility would cost 40 billion to manufacture with 350 billion total industry

spend across 2026. Now the big question is how confident are we are, are we on scaling up those models? There's a wide range of takes around how easy it is. Clearly, we saw with GPD 4.5 and Lama behemoth, just scaling up, you have to be very careful about what you're scaling up. The actual underlining design of the model matters a lot. By the end of 2027, he expects a fully operational two-gigawatt facility with total AI compute across all companies, raising reaching 90 gigawatts. That'll be a lot of inference. These two gigawatt facilities would cost

$95 to $100 billion each and total industry spend would reach $500 to $600 billion. By the end of 2028, I expect the largest single facility to be a three gigawatt facility holding 1 million Nvidia Blackwell's Rubin Mix.

That's roughly nuclear reactors worth of energy. Something like that, yeah. Costing $150 to $165 billion to build, capable of a 1E28 flop training run. A 1E28 flop training run represents a computational effort thousands of times greater than what was used to create GPD4, allowing the AI to process and learn from vastly more information.

Total AI data centers would reach 130 gigawatts combined with 900 billion to a trillion spent by the AI industry over 2028. That is a lot. But that matches up with what everyone's saying, everyone's marshalling the capital, everyone's planning to do this. Yeah, I think we, we want to look to meta's earnings, next earnings call October 29th. Yep. Yep. Zuck is going to feel pressure to say the biggest number. Yep. And we should probably,

we should probably stream that live through a companion stream. It'll be fine. I'm looking. I mean, I just think that's going to be a big indicator of do they push all the chips in? Literally. Yeah. By 2029, it starts to get very fuzzy to predict in his forecasting past. powers break down, how AI continues to scale whether we've encountered data related or other algorithmic bottlenecks, what AI capabilities have already emerged, and how economically valuable those AI capabilities are will be key to whether the economics favor continued scaling. Needless to say, building a $150 billion individual data center campuses and spending a trillion on AI infrastructure would get very difficult to sustain financially, let alone to increase

dramatically year over year. I feel like the trillion dollar cluster is within reach. I mean, it needs to be probably distributed across all the different companies. But these numbers don't, they're not as scary as I think. When you think about like across the hyperscalion doesn't scare you? No, because across the hyperscalers, you have almost 20 trillion of market cap, maybe 10 trillion of market cap. And so you're looking at, okay, so they need to move 10% of their market cap into this project.

that's not 50% of the market cap. Can they liquidate 10% of their market cap and put it towards this? It's at current prices and they would have to lever up significantly. No, not significantly, 10%. Like they would need to take 10% of their equity, of their market cap. But if you have a massive correction between now and then it starts to be quite a bit more than 10%. Sure. But I mean, a trillion dollar spread across the Mag 7 isn't, it is.

isn't actually that much. I mean, Nvidia is a $4 trillion company. They're all, they're all a trillion dollars, right? So, I mean, I agree with this conclusion, 2029, it gets crazy. And then the real question is like, okay, what about the $5 trillion training run or the $10 trillion training run? But just purely, like, is the money there? Like, the money feels there. The less clear thing to me is this question of how AI continues to scale, whether we will encounter data-related or other algorithmic bottlenecks. And then the question of, you're spending a thousand times more money,

are you actually getting a thousand times better results? But when you look at this, it's like it's more likely that it'll be an energy bottleneck, right? Yeah, yeah, yeah, that's another good one. Where you're going to get three nuclear reactors worth of energy in one year. In one year. In two years from now. Yeah, yeah, when it takes us 30 years to build them. So, yeah, there's a big, big question there.

We got to check in with the nuclear bros. Yeah. Because a lot of, wasn't it the last time we had at Isaya? Yeah. He was saying he was going to have a first reactor in like 12 months or something like that. Yeah, he's working on it. So, I mean, a lot of the, he was going kind of mid-scale, I think 10 megawatts. And so if you're like, you're talking about him not just making the first one, but then making 10,000 of those, which is a lot to generate this much electricity.

we did talk to the nuclear company at Palantir Debtay, and he was saying that he's much more encouraged by bringing nuclear, like the original GE plants online, the one gigawad, like the large-scale nuclear reactors. Yeah, Doug's point is like energy, when a premium is placed, when there's more demand, there's more capital-chasing energy, energy will come out of the woodwork.

Yeah. So there's... Gabe in the chat says Buffett is good for $400 billion. He does have a lot of cash right now, right? He has, I think, like, 350-ish billion of cash on hand. Can you imagine? He just gets super AGI pills? If Sam does a deal... At age 99.

If Sam can do a deal with Buffett, I mean... I mean, is the final boss? It's the final boss. There's no one bigger that he hasn't been able to do a deal with. I mean, he even did a deal with Elon back in the day. He's done deals with every single person at various times, pretty much. Yeah. What if Buffett ends up owning all the data centers because they just get super levered up and just go bankrupt?

He buys them for pennies. I can see it. I mean, he loves rail. He loves infrastructure. I am compute. I become compute. Yeah. Berkshire just becomes the next hyperscaler.

It would be so insane if Buffett's vindicated and just like his model holds forever. I mean, it's pretty much It's held for his whole career It's pretty Lindy at this point Can you imagine? He just winds up owning Every power plant in America It's so crazy

So, so Berkshire is up 2% Over the past six months Let's go So 2.2% to be clear Hit that gone Congratulations, 2% gain Sending a lot of cash A lot of cash

Should be in gold The market cap is just basically three times their cash on hand. Wow. Three times their cash on hand. In other news. Elon names Anthony Armstrong. Good name.

That's a new X-A-I-CFO. Strong name. Armstrong, strong arms, the financial markets during complicated fundraise at X-A-I. The New York Post headlines write themselves. Let me tell you about not X-AI, but Finn.A.A.I. the number one AIG for customer service,

number one in performance bench marks, number one in competitive bakeoffs, number one ranking on G2. Mark in the chat is asking about TBPN merch. This first merch run we did, we didn't sell, we just gave it away somewhat randomly.

We're making more. We actually have a design meeting right after the show today to get into it. So it is coming. This next time, we will make it available for the world. and the chat will get first access to it.

For sure. Thank you for your patience. I got to stop wearing it. I know, you're teasing. Everyone's up teasing. They're on the suit. David Holes, the founder of Mid Journey, says, my previous startup, Leap Motion,

made the best hand tracking to this day and an amazing open source AR headset North Star. Now the tech is getting deployed by Globus. Great name? To surgical theaters around. Oh, I work for Globus Corp. Globus. Oh, what?

Oh, I work for the Globus Megacorp. I work for the Globus Center of International Business. Globus is a great name. That's so good. It's like, well, we do business around the globe and in the United States. Globe U.S. Globus. Globus.

Globus. Globus. We've got to get the CEO of Globus on. This sounds like a great company. They are deploying these AR headsets to surgical theaters around the world. I mean, globus is an $8 billion company undervalued

just on the value of the trademark alone. Globus. Great hit. Let's hear for Globus. We're huge Globus fans here. This is a very cool tech. It makes a ton of sense

in a medical setting. Obviously, you're not wearing the headset for fun all day. It's a professional environment. There's always been enterprise applications here. And, of course, it's a medical device

and it's adding a ton of value. to a high paid doctor's routine. You can probably put the best, most cutting-edge technology in there. If it's 10K, it's probably cheaper than the chair that you sit on in a medical setting. So, congrats to David Holes for being on an absolute run. The Leap Motion Story is fascinating. I dug into it a little bit. It seemed like they were about to sell the company to Apple,

and then Apple had printed, like, welcome letters to everyone on the Leap Motion team, and then something happened. The deal fell through at the last minute. David was like, you know, in and out, had a really rough run with venture capitalists, then went back to Mid Journey. Didn't raise the dime. You'd love to see it, bootstrap founder.

And him and Mark Enderese were going back and forth. Did you see that yesterday? Yeah, I missed it. So David Holes posted something to the effect of like, hey, any other bootstrap founders in SF? It's like to call Mid Journey a bootstrap company. I mean, it is a bootstrap company. They haven't raised a venture.

Yeah. But it's like the biggest company ever. And Mark, oh, I did see this. Mark and Mark's like, me, I bootstrapped A16Z. Like, I own the GP. Hundreds of millions of revenues. Love to see it. Yeah.

I didn't see this yesterday either, but the U.S. is taking a 10% stake in trilogy medals. I haven't heard of this company. It was a $480 million, Canadian dollar market cap company. Okay. Yesterday, it is now $1.63 billion. Canadian dollars. Well, any metal, so that's exciting.

How many, I don't know what the exchange rate is probably pretty rough. I have no idea. Well, let me tell you about Adio, customer relationship magic. Adio is the AI native CRM that builds scales and grows your company to the next level. The next level CRM. Tim Cook is likely to step down as Apple's CEO. According to Bloomberg, John Turnus, who I'm dying to have on the show, got to make it happen, get to him before he becomes CEO potentially.

Apple's senior vice president of hardware engineering is expected to take over. Interesting because you don't see him in all the keynotes. He's not the biggest face, but he's clearly doing something right at Apple. And for all the discussion over, do they get AI right or whatever? Like the hardware is good. Nominative determinism. Turn us around. Turn us around.

Okay. I see what you're doing there. Yeah. Like it. Bloomberg notes that Apple will now likely be. Johnny Apple. see. Tim Cooked.

We know this. Tim Cooked. I saw a graphic of like all the different products that that Tim Cook has like six, you know, overseen. And it's actually remarkable. I think he will be, uh, while he certainly isn't being celebrated in the AI era, he will be celebrated in the fullness of time. Yeah.

There's two very, and while I think he's dramatically underpaid. Yeah. I hope that the history books will remember his name for just exceptional execution on Steve's vision that was Steve's, Steve's visible vision. Totally. So. Yeah. Well, whatever your view on Apple, whether you're long, you're short, get on public.com investing for those that take it seriously.

They got multi-asset investing, industry living yields. They're trusted by millions. I have some-odern. And they just launched direct. indexing, which is very cool, specifically because it allows you to, you know, if you have outperformance with one stock, underperformance of another, you can actually recognize those losses. Oh, interesting. And so it can be more tax-efficient way to get exposure to a broad swath of companies. Interesting. In one go. Cool. I have a bit of cognitive dissonance about

Apple. On the one hand, apparently they have this like belief or something that Bloomberg was reporting that they don't believe, like, chatbots are going to be a dominant form factor. Like, they've really leaned back in terms of chat bots. Siri has kind of been slow. I've been trying to use Apple intelligence to query chat GPT, and it's still, like, it's, I have to click a button every time. It's very, like, it doesn't seem like they're really moving quickly, and it's been, like, years now since chat GPT has come out, and I would have liked a deeper integration on day one.

They could have bought something. They could have trained their own model. Like there are so many companies that have been able to catch up. It doesn't feel like it's complete rocket science. The narrative of like the foundation model commoditizing, even if they have something that's not at GPT5 or Claude 4.5 Sonnet or even Gemini 2.5 level, like it would still be better and it would improve my experience, but they've really lean back on that.

So that's like a really, really clear like bear case for like they're not winning AI. And then at the same time, whenever we talk about hardware and the next devices, we just keep coming back to, well, not only is the hardware extremely good and sticky and, you know, just like the best devices and the best supply chain. And Tim Cook's done a fantastic job of managing the tariff battles and whatnot. But they also have iMessage, which feels like an incredibly powerful social network that is very, very hard to rip people out of. Social graph. social graph. And so when we, when we demo smart classes or interviews, somebody was posting yesterday, Apple better start moving faster. And it was in reference to chat GPT's like booking.com integrations. Yeah. And it was fairly, you know, it was getting quite a lot of engagement. But I was surprised by that, considering that Apple's never been able to really directly tax like retail. Yeah. That's they don't, they don't get a percentage of every dollar. They don't take a cut of. Ben on meta. Well, they do take a cut of booking.com in the sense that booking pays Google, Google pays Apple.

Yeah, but it's way less. That revenue is not necessarily directly threatened because if Google has more, you know, they're happy to keep paying their version of the Apple tax. But again, it's not like Apple is getting billions of dollars a year from Amazon, and they would suddenly not be getting that if ChatGPT starts dominating and, retail, right? Yeah, yeah. Maybe what the move internally or the mood internally is at Apple is something like they don't see AI or LLMs or chatbots as disruptive to Apple's hardware plans. They think that even in a post-AGI world, people will still be carrying iPhones. And so they can

just sit and wait until they get some foundation lab to come to them and say, we're going to pay you $10 billion a year to be the default. And they're just like, Yeah, we'll just wait. And you can come to us today. But we're not doing the deal until, and so it's going to be uncomfortable for customers until someone ponies up. Yeah, it still feels hard for me to imagine a world

where Siri is just replaced by another brand, like what Siri does functionally in terms of being an assistant layer on top of the hardware device. It feels, it feels very unbelievable. to me that Apple would say, like, we're going to let this other assistant directly integrate throughout the entire device. That feels like a big step. It does. Because the phone in many ways already was operating as like an assistant, right? It's like the center of your world. It can do

it can do anything you want to do in your digital life on your behalf. It's sort of, and so giving up that, giving that up feels major. But yeah, I think we have to see more announcements. out of Apple this year on this front. I know based on the last, they still have longer time horizons internally around their like AI turnaround, right? They're not saying we're fixing this this year. Yeah. It's very much like we're fixing this over the next 24 months.

Yeah. And again, I still stand by Apple and believe that the competitive threats, Open AI just doesn't feel like a legitimate hardware threat today. No, not at all because of the lock-in around the Apple ecosystem. And it does feel like Apple might be looking at AI more as a profit pool. And once that profit pool gets really big, then they're going to want to go after it. And they will have a ton of leverage at that moment and still it'll be fine. But just chat interactions are not generating tens of billions of dollars in profit yet.

But Google search is. And so Apple says, hey, we got to go get a cut of Google search. They did, and they've been very successful there. Once the profit pool of chat interfaces becomes really big, Apple will probably say, hey, we got to get a cut of that. And maybe at that point we will invest. But we understand the economics and we're not supply constrained on data centers. and so we don't have to pay a high margin.

We can do Apple Silicon chips for our inference or something, or our own custom model. There's a million different ways that they can get their fairish share of that market when the time is right, because they will still have the leverage. And they don't see that their leverage is decreasing because people are not churning from iPhones based on the lack of chat GPT integration or the lack of LLM integration, even though it's like a pretty crazy feature to be missing. in 2025. It just still hasn't really hurt them. I bought a new iPhone. I didn't sweat.

I'm still waiting for my iPhone that I order. I'm waiting for my iPhone that I ordered weeks before John just walked into a store and bought an iPhone off the shelf. Skel issue. Skill issue. Unbelievable. Unbelievable. The iPhone was delivered to me. Here I am thinking like, wow, there is so much demand for Apple, the new iPhone that they can't deliver me at one. I mean, to be clear, like, there was only one in stock that I was able to purchase, and it just happened to be a model that I was happy with, but I didn't get to pick a color or anything like that. I was just like, yeah, just give me what you have. Anyway, speaking of iPhone apps, open up your eight sleep app and tell me how you slept last night because you had a tummy ache from lack of alcohol consumption.

I was boozing it up with the white wine at the luxurious event. And I said, and I said, slept for seven hours and got an 84. I did pretty well. I slept for five hours 56. Five hours. Wow. Oh, 56 minutes. Okay, that's close to six hours. Yeah. And my eight sleep is very concerned about my health. Oh, no. Well, did you get over? I got a 70. I got a 75. And you know what you got to do, George? I know what I got to do. Let's go. I win. Credit where credits do. I've been on the run. I've been doing very well recently. Anyway, Mick franchisee shares, I remember this Happy Meal promo back in 1991 for Earth Day. You could ask for a free tree at McDonald's, no purchase necessary, and they would give you

a sapling, I guess. Pretty cool to see. Apparently, some of these Mick trees, over three days later, 30 years later, have grown into behemoths. And so after 50 years, this man's family is moving out of his child. at home. In the 90s, McDonald's gave out pine seedlings for Earth Days and look at this tree. It's now 30 years old. What a beautiful tree. My happiest childhood memory with McDonald's, they had a deal with Sugar Bowl that if you bought anything at McDonald's and you brought the

receipt to Sugar Bowl, you could get a $5 lift ticket. And I was doing that all day. I mean, that was like the greatest trade of all time. You get a McClurry and then you get a $5 lift ticket. You used to be able to take a Coca-Cola can and go to six flags for cheap. There were a bunch of these cross-promotional deals that were going on. I think this is a great merch idea. I think some startups should do this. If you're super not AGI-pilled, give away a tree sapling and be like, yeah, this tree is still going to be around. It's not going to get paper-clipped.

I'm long humanity. Keep thinking, folks. This also reminds me of our idea for merch, the bonsai tree. I really like the idea of sending everyone a bonsai tree to remind them that progress is measured in decades, and you should never stop working, keep grinding. Anyway, what else is going on? Do we ever recover this? This is something that's probably worth mentioning.

Jay Bull says, here's the thing with today's Open AI Pump. Elon is B-U-T-T-H-U-R-T. He hates Sam. Sam. Sam is getting all the attention these days, and nobody cares. about Elon anymore. I fully expect the mother of all pumps from Elon sooner than later because he's a narcissist and no way he doesn't do it. What a funny way to frame like what happens in the capital markets. Like, oh yeah, like they're going to do a pump because there's spite or something. It's like

spite pump. What about just like building something that delivers value? In Tesla's at an all time high right now. I mean the roadster. The roadster video dropped yesterday over the weekend. You're going to get I'm 100% going to get one. I'm so excited. It's actually having the aesthetics of a supercar, but having great self-driving, that's a recipe for a hit. Wait, we have the perfect guest to talk about the Tesla Roadster.

We have the, is it the CEO of DuPont Registry joining us? I corrected Tesla's not at an all-time high peak last year, but getting close, getting close. Well, we have our next guest in the Rooest Room, waiting room. I am absolutely pumped for you. Oh, look at this. What's happened? Fantastic background.

Great, great setup. How are you doing? We couldn't have said yes faster to having you on the show. I think I'm a daily active user. 100% of the product. And yeah, super excited to get the update from you. Thank you for having me, guys. Yes, please, guys kick us off with an introduction on yourself and actually describe the company

for anyone who might not know. Yeah, sure. So I joined the company two years ago, as you will hear from my accent. I'm not from Paris, Texas. I was born and raised in France. I came to the US 16 years ago, started my career in the luxury and tech world with Louis Vuitton and moved two years ago to lead DuPont Registry Group. We're going through a major transformation.

We're disrupting the luxury automotive industry. The group has been building for the past 40. years, amazing trust towards the audience of car enthusiasts and car collectors. Today, we're the only one place where any car collector can find the product of their passion. And we're transforming the industry. We're building a brand new tech platform to have a, for our ecosystem of dealership and clients, a way for them to transact online. We're very excited about it.

The market opportunity is huge, and the group is moving in the right direction. What's the shape of the business now and then how much do you want to move to online platforms? I want one-click checkout. Yeah, like how far does this go? I want one-click check-out too, and this is what we are working towards. So the group has been the foundation of the group is, again, 40 years of legacy and trust. We have today amazing brand within our portfolio, events that are gathering all the enthusiasts and car collectors, some of them the most affluent car collectors in the world.

And we want to use that foundation and the data that we've been accumulated for years to build really a non-channel tech platform that will allow anyone when they see a product and a car that they want to be able to purchase it online and get delivered at their home. This is what we're building on right now. And the recent investors that we announced today, that brings capital to the company. We're going to use that capital to build that tech platform.

So expect a lot of new changes in the next few years. What is, you know, what is, what I kind of, you mentioned events, more seamless transactions. That all makes sense. Are you guys planning anything on the auction side or you feel like that is well served by the industry today? No, no, no, it's definitely something we have in mind. It's part also of our, where I came from in the luxury world, it's all about Omnichannel and being taking care of our clients wherever they want to buy and whenever they want to buy. So if you look today in the car space, the auction piece is quite important.

And we are launching in a few weeks our own DuPont Registry Live auction platform that will support. our audience and allow us and our dealers to sell their cars. It's going to be a disruptive one. You will see a couple of announcements are coming in the next few weeks. How do you think about defining the tone, the vibe around what type of cars make it onto the platform? It feels like, you know, dollars a dollar. You could have every car on there, but you have to define the brand somehow. So how do you think about, what is an appropriate car to work with? And how do you let someone down

if they bring something that's just not quite there a little bit too not special enough? To me, we don't let anyone down because at the core of our audience, there is passion. Sure. So whether you're talking about, and we had someone the other day

from Eastern Europe that rebuilt the Dacia from his father, there was a strong passion about that Dacia. And we actually talked about, it and with Petrolicious, one of our brand, we did a movie about this guy. So to me, it's all about that passion component. Of course, for the past 40 years, we've been the leading marketplace and our positioning is luxury and high-end luxury, and we want to continue to go there. But my job is not to say no to my clients, it's to be able to welcome them in our ecosystem

and serve them and make sure that they evolved with us throughout their collector's journey. Give us a high-level view of the current state of the automotive industry. We obviously are car enthusiasts ourselves, but cover tech about a thousand times more closely than the automotive world. But it seems like we're at a very interesting time right now. You have pressure on the Chinese market. You have trade wars. You have... Back and forth on electrification.

Yeah. Electrification that enthusiasts have certainly... not, you know, overall not been huge fans of. You've also seen, you know, the massive depreciation on that front. If you're an automotive enthusiast, nobody wants to buy a car that might look pretty and drive well. But if it's going to depreciate 40% in the first year, it's just not, you're never going to really get that enthusiast love.

So I'm curious, what kind of trends are you tracking most closely and kind of new activity over the next 12 months? Yeah, it's a very good question. So all you mentioned does not really affect us today. We're in the use cars, luxury space, and we're in the business of our clients are looking for something they want, not something they need. So from a pure purchase funnel standpoint, they are not really affected today by tariffs and otherwise.

You will see in the next few weeks we're partnering with Boston Consulting Group and we're going to release a luxury car market report. Oh, cool. Nice. We've seen first few things. in that report, we're talking about $100 billion market, so it's going to be it's quite interesting. There is a huge market

opportunity, not at all digital today. What's interesting, too, is that we see that electrification is not there. There is nothing. If you're looking at the luxury car market, people are looking for performance, they're looking for fun. They're not looking at their daily driver, so they will look

more for that car behind me. At 2001, Ferrari 550, Maranello. than an electric car. Yeah. Yeah, the other thing I think is worth calling out, part of your guys' opportunities, is like I have bought,

I've purchased a number of cars' site unseen online, and maybe that's just because I was born in the, yeah, I was born in 1995. I grew up purchasing everything on the internet. I have no problem seeing a video of a car. I might get a remote PPI done on it to have that confidence, but I continuously am happy to purchase cars out of state, you know, wherever.

I don't need to see them. I've done that, I think, at least three or four times at this point. And I expect to do that many more times in my life. Do you think that that is part of your guys' opportunity and that the new generation of car owners and enthusiasts are just willing? Again, it doesn't matter if the car, if the spec you want is out of state, great. I have no problem kind of purchasing. Yeah. It's actually very interesting.

Today we see that about 99% of the purchase decision starts online. It takes a couple of months for anyone to decide which cars they want to buy, and there is a lot of online research. Yet less than 1% of those transactions are happening online. So there is a lot of potential to disrupt the market and disrupt the industry. When you're talking about buying a car site unseen, and this is the power of our brand and our ecosystem, time we have the trust from millions of people that have been, you know, looking at our brand and

coming to our brand for the past 40 years and they trust the brand to show them the right product with the right condition and be transparent about the product. And this is what we're offering at the national audience for the U.S. You'll be able to buy a car from California, get it delivered in New York. And behind the scene, we will be the intermediate to facilitate that transaction and make sure that you trust the seller and the buyer and trust both parties. How do you think about the next generation getting into cars? What's different about Gen Z or millennial car collectors? Are they going as, are they going less far back into historical or retro cars?

Or does every generation just love the car that they had on the poster in their bedroom forever? and you're kind of locked in a time capsule? Or is there something unique and different about the millennial generation or the Gen Z generation as opposed to the Gen Xers or the boomers? I think there is something unique is that they are discovering the luxury car market through experiences and community. And that's what we're doing at the group. We're organizing many events every year to build that community

and to continue fostering the dream within the community. From a purchase standpoint, as any other car enthusiast, you will start with the car that you had as a poster in your bedroom when you were 14. And that will be your first car. But then you're going to evolve throughout your collection. The good thing is that today we're seeing a transfer of wealth between generation. The passion is also transferring. So we're actually seeing that our audience is getting younger.

And definitely, to your point, the audience is ready. And they are used to buy those products online and luxury products. products online, and this is where we're going to offer them that ability. How are you thinking about marketing and content production? I engage with DuPont Registry on X, actually. I see every new auction, and there's a lot of facts and interesting details in there. It's very interesting to me. But there's so much that you can do with cars when you're marketing a platform because

the cars themselves are interesting. Like that car behind you, I could watch a 20-minute video about it. It's fascinating. I probably have by other creators. Have you how are you thinking about partnering with creators? Either sponsoring content or bringing people in-house agencies. Like, how do you think about the correct shape for just raising the awareness of DuPont? It's all about stories to me.

We need to tell stories between a driver and his cars. That's what we're doing with one of our brand petrol issues every Friday. We have a new movie about a car story. Cool. This is what we want to do. We want to make sure that we tell the story. of the enthusiast community. We tell the story of the car collectors

and how they build their collection from the first purchase until their latest investment. And this is really how we're today, how we're targeting our content strategies, really focusing on that. Very cool. Jordan, anything else?

Amazing. We'll send you some posts after this, but we've featured DuPont thousands of times. Thousands of times in our early episodes. It was a core part of the content. Please, like never stop posting on X. It is the best little drip feed.

It just sprinkles itself in with all the other tech news, and I see it, and I love it. And I usually tag a friend, and I say, you should get this car. Well, yeah, I can't. I'm very excited you guys have. Again, you know, it's just a daily active user. I'm very happy that you guys will have more, you know, investment. And let us know if we can help you recruit anybody in on the technical side. Did you share the terms of the deal, how much was raised, anything like that?

Do you have any numbers to share for us? No, unfortunately, we don't disclose financial numbers. What I can tell you is Francois is one of the top collectors in the world as well as a very well-known race car drivers. He brings a lot of validation of our strategy. He's a big tech guy and he loves tech as well. He sees what we're building from a tech and AI standpoint with the data we have. And we're very happy that he's on board. But one number, you guys reached unicorn status.

That's correct, right? Yes. Yes, there we go. Hit that gong, John. There we go. You'll appreciate this. We're waiting for the right car to put this on. So this is just came in the mail.

Maybe it's the 599. Yeah, yeah. Well, thank you. But we're not for sale, those. No. Yeah. Another one. They're there these days.

Awesome. Thank you so much for coming on and congratulations. Thank you for a miles. We appreciate it. We'll talk to you soon. Cheers. Bye. And if you're looking for a luxury watch to go with your luxury car, you know where to go, get bezel.com.

Your bezel concierge is now available to source you any watch on the planet. Seriously, any watch. Add cars, bezel. Tyler, what do you think about this GPT Image 1 mini post? Angel is saying they didn't fix consistency, apparently. The prompt was remove all the. ingredients from the burger and keep only the top and bottom buns leave a gap between them keep keeping the same spacing as if the fillings were still inside why are you laughing sorry taylor with a

deep cut he says dupont registry is going to love my Nissan morano cross cabri of lay let's hit the gong this is most underrated car it's so good uh what else convertible SUV two-door are you getting a rangerover evoke we really botched it not there's one for sale right now with almost a hundred thousand They're available. It's for $10,000. It's honestly a great production car. You lean out the side, take some video of the

other cars. It's a great production. It's a most wild car. It's a V6. Give it up for V6. That's not bad. Underrated. That's a four cylinder. But Tyler, did you see this post about the remove the toppings from the burger?

What do you think is going on here? Does this push your AGI timeline out or pull it in? What's going on here? I mean, I think just from this, like, I think it's, we can probably just assume that, like, nanobanana is there's some kind of, like, architecture difference where it's not just, like,

doing diffusion on the entire image. Yeah, I mean, maybe it is. It seems like it might be, it might have, like, tool use. And I feel like tool use has got to come to image generation soon to just be like, for this prompt, you actually just need to cut one half of the image out and cut the other half and just stitch them together using, basically like Figma or something.

Like a basic lasso tool is like what you need. So just give the AI image generator that instead of trying to re-render the entire image. Yeah. And like maybe that like kind of tool use was only in the training step and it's like kind of been distilled into the weight somehow. Yeah. But it does seem like there's some it's like it looks like it's just like segmenting something out. Yeah. And then it just does kind of in painting on that part.

And then it just gives you the new image where GBT is like very clearly. It's like you're it's doing the whole image. Yeah. exactly. Yeah, Sean says vision reasoning is far from a solved issue. I'm very excited to see where that goes. It feels like tool use and reasoning. All that's coming to image, but just slowly.

It's going to roll out slowly. I demoed one image generator project that was clearly cutting out pieces of the image because the grain was even the same and it wasn't re-rendering the face. And so you could say, make this a gigacad or a bodybuilder. and it would add the muscles, but you'd zoom it in the face, and it would be pixel perfect. And it's like, that's not re-rendered.

Like, I can just tell. Anyway, GROC is still on a tear on open router. One trillion tokens every three hours. That's a lot of plaques. They're doing well. We already dug into the GROC stuff. That's fun. And KSA says,

the dumbest person you know is currently being told you're absolutely right. By chatyPD. This is a repost. Someone else posted this. We talked about this. somebody just found a banger and just re re uh re appropriate you that yesterday you said you did you said i finally post it's and suddenly i'm craving pumpkin everything binge watching horror movies

and planning the ultimate costume who's with me on turning this month into pure spooky magic i said you're absolutely right john yeah uh compound 248 says if i am sundar pachai i will bury all competing foundational models in negative gross margin tokens for eternity i will never take my foot the pricing gas. To merely compete, I will force you to dilute your equity into oblivion.

Sam will drown in my tsunami of cogs. So a very well-written piece. The dilution is real. Delusion seems real. Certainly, yeah, it's interesting. I talked to somebody a long time ago who was like there was a moment

at maybe AWS or maybe GCP or one of them where there was the option to get in a price, war and they chose not to. And so all of the hyperscalers, all of the clouds have 30% margins. Right? If they got in a price for, it would suck all the profits out of the industry.

For sure, there's that. There's also just, the way I heard it described to me was, like, if you're a CFO, you're just kind of like, but I want 30% margins. Like, that's my goal. So like, why would I do,

why would I do low margins when my goal is 30% margins? And so it's almost just like kind of hard-coded into the, fabric of these financial institutions or the finance functions at the large companies is that it's like, yeah, there's like this world where we get into this price war and there's risk, like maybe we could win, but we could also just continue to get 30% margins. So I don't know if this will play out this way. Maybe, maybe on the API side, but Tyler, have you noticed any, any like real price fight going on in the foundation model layers? They're all pretty price competitive, right, on a per million token basis, right? Yeah, they're all pretty close. I would say maybe Anthropic is a little

bit more expensive. More expensive. But it's supposed to be a better model. Yeah. And it's like, it's all this parade of French. If you look back, it's like, oh, the models are staying the same price, but it's just because you keep updating to the new model. Like if you're using GPT 3.5 turbo, it's like obviously super, super cheap now compared to what it was. Yeah. So there's some something there, but I don't know. Well, Barry Weiss dropped a letter to all CBS news. employees. And there's some debate in the comments as to whether or not this was AI generated. Really? What was the call-out? There are M-dashes, but that's it. And so I'll read it. And I think she wrote it herself. But she says, dear colleagues, I am thrilled and humbled to be writing you as

the new editor-in-chief of CBS News. Growing up, CBS was a deep family tradition. Whenever I hear that tick-tick-tick-or trumpet fanfare, it sends me right back to our den in Pittsburgh, the opportunity to build on that legacy with you, hyphen, and to renew it in an era that so desperately needs it, hyphen, is an extraordinary privilege. Right now, I imagine you all have some questions. I do too. My goal in the coming weeks, days, and weeks is to get to know you. I want to hear from you about what's working, what isn't, and your thoughts on how we can make CBS News,

the most trusted news organization in America and the world. I'll approach it the way any reporter would, hyphen, with an open mind, a fresh notebook, and an urgent deadline. She drops the Oxford comma. feels extremely human written to me. And of course, it should because she's been a writer for a decade, way before LLM's were a thing. So why would she lose that ability to do that? She has a couple, she has 10 points that she defines journalism around.

It reports on the world as it actually is. It's fair, fearless. And factual. You can say journalism. Journalism uses all the tools of the digital area. Journalism that understands the best way to serve America is to endeavor to present the public with facts first and foremost. I look forward to meeting many of you in the days of head

and listening and learning to you. I am profoundly honored to join you, M-Dash. And I can't wait to get started with gratitude and excitement, Barry. Seems very, like, well-written, I don't know. Gold is up at $4,000 for the first time ever. It's a record run for futures. Comes and made concerns about the economy's outlook.

Gold has been on such a run since 2000. A little bit of a sell-off post-GFC, but has been on an absolute tear going from $2,000 per trie ounce to over 4,000 in just a few years. Goldsword to 4,000 a troy ounce for the first time, signaling an investor rush into alternative assets at a time of concern about the outlook for the U.S. economy. The price of precious metals has surged this year more than it did during some of America's biggest crises, raising more than 50 percent futures. run up in 2025, has outpaced rallies during the pandemic and the 2007-2009 recession,

not since the inflationary shock of 1979 has gold catapulted so much higher in a year. I think I look at the gold chart and I think I'm in danger. The U.S. dollar is in danger. Yep. Bitcoin was at an all-time high last year, or yesterday too. Everyone is flying to other other assets, whether it's AI stocks, energy, Bitcoin, gold. Everyone wants something else. But what a run for gold? We have our next guest in the re-stream waiting room. We have Rami from Even Up. Let's bring him on in. Let's bring him in to the TV. What's happening? How are you doing? Hey guys. How's it going? It's great to see you. We're doing great. We got our gong ready for you. Yes. Introduce yourself, the company, and any of the news today.

My name is Rami. I'm the CEO and co-founder of Even Up. We are happy to announce our $150 million series E led by Bessemer. Congratulations. Yeah, breakdown the company history. Yeah, how you got into the category. Yep, we started Even Up about five years ago. This is before Legal Tech or Generative AI was all that interesting. And the premise was to ultimately even up the playing field for personal injury victims in the U.S.

So every year there's about 20 million injury victims where often these cases take a lot longer than they need to to settle or settle for a small fraction of what they're worth. And so our goal is to even up that distribution by helping ultimately their attorneys. So it's a B2B startup. We're helping them across the whole life cycle of their injury cases. Think of it as doing the heavy work for these attorneys for them and what I call superhuman quality. we're helping them settle their cases faster, helping them sell their cases for larger amounts, and removing a lot of the manual work

as we go through a lot of these workflows for these attorneys and their staff. How are you guys already running up against the oppositions or the defenses AI themselves? How is the battle evolving? Yeah, we don't really see too much of that. The crazy thing is this is our fourth. round of financing in the last two years.

Wow. And we're on this really rapid growth trajectory where we see about 10,000 cases a week. But when you think about just the TAM here on 20 million cases, we're about 1% penetrated today. So in a way, even though it's a big number, it's still really early days in our space. So we haven't seen any of that come through. What we have seen is ultimately being able to change some of the operations of these law firms for the better.

So one of our firms has scaled revenue by about 70% year on year on a base of over 100 million of revenue without adding headcount. Another one of our firms have used one of our products in mediation in real time where they turned a $50,000 offer into a multimillion dollar offer. So it's kind of crazy to see the impact we've been able to have, but it's still early days. Is plaintiff screening a different business? So yes. Not fully. The way you can think of this is the way we help. is, you know, imagine somebody gets injured.

They would typically call an injury attorney. They would sign up. They see a billboard. They see a billboard. And the billboard says, Injured, call this number. They call that number, and they describe their injury.

And then someone needs to qualify that. And it's probably not the most highest paid lawyer. And it feels like those intake forms could be processed by AI. But is that something you're looking at doing? So, yes, that's really interesting. So we don't do the intake call themselves. Yeah.

But as soon the intake call is done, the transcript of those calls is something that we would analyze. Sure. To say, hey, law firm, you need to prioritize this case. Or this might be a case you want to draw. We don't do the signing up with the case,

but as soon as there's any information, whether it's audio or document-based info in these cases, we read that, we scan it, and we ultimately make decisions off of that. Obviously with the attorney hand in hand. How are like hallucinations and how are you thinking about like lawyer in the loop? Jordy was talking about this. There was some case about a,

a consulting group that turned over an analysis to a city and had to give them a huge refund. Yeah, it was Deloitte basically got caught delivering a study that was just inventing, like, judgments that no judge had actually. Yeah, it wasn't grounded in, like, the ground truth of case law. And I imagine that, like, I mean, the stakes are pretty high. The model just had a really great imagination. But also, I mean, you could wind up in a situation where, like, your firm is like forever tainted by a certain judge and the judge is like, I remember what you did. That was really sloppy. I imagine that your clients and your customers care deeply about that. Like what are the, what are the strategies for avoiding those types of situations?

You both brought up really great points here. And it's a really funny thing because folks are always like a generative AI and legal tech are all shoe ins for each other. But folks generally tend to forget accuracy really matters these cases. You can't be 90% of the way. there in a legal letter. That's how you get this one. Right. That's how you get those, again, some of those headlines that you guys just mentioned. The way we think about this internally is you need to have done a lot of cases because every basis point of accuracy matters. It's another way of thinking about it. We do a number of different legal workflows for these attorneys, but if you think of one letter called the demand letter, that's one of the documents that repair for our firms,

the quality from this in us being able to repair this two years ago versus today was very different. Yeah. When we started, keep in mind, this is before GPT3 and the big, you know, all the craziness around AI two years ago, we had to have a human in the loop to QA these letters. Yep. And if you think about the inputs in these cases, and we're at a point now where we're going to be millions of pages of medical records, medical bills, police reports, raw underlying documents, we've had to fine-tune these models, one, to be able to get a much higher level of accuracy

than what you typically see off the shelf. And so what's a simple example of this is if you think of something as intuitive as, hey, list me all the medical visits that a plaintiff has seen in a case. If you just go into any of your models today that are publicly available, ask it to do it for you. You'll see the accuracy is just not good enough. And so this is where we have to attain our own data service model. Just to do something as intuitive is to be able to pull out a date. When you look at these case files, you get to see timestamp dates, signature dates, data first, There's so many different dates.

It gets pretty complicated with even just one piece of extractions. And that's how we had to build this with time. You've had to do a lot of, you got to get your reps in to get the quality of where it needed to be. How are you growing the business? Are you running billboard ads? If you're an injury attorney, call this number. Or are you doing steak dinners?

You should really do that on the 101. That would be great. Or are there like legal tech conferences that you're like, What does the top of funnel look like for you? Yeah, and this is the crazy thing as well about our space is this is such a massive space that folks just don't necessarily to know too much about. But when you mention those billboards, you folks are saying this in just, but when you see one of these firms has a thousand billboards in California, I don't know any of any SaaS business that has a thousand billboards. No right. I mean, like the massive marketing machines and they're based on our map is around 300,000 of these injury attorneys.

And about 20 million cases, and they're all over the place, right? One of our largest customers, as an example, is in Alabama. You would have never thought it would be in Alabama. We have massive firms in California, massive firms on the East Coast, and it really follows population. Wherever there are people, there's going to be injuries, because it can mean anything from a motor vehicle accident to a dog bite case, even police brutality cases. There's many different ways to get injured. And so as long as there's people, there's going to be personal injury attorneys, as long as it's personal injury attorneys, they're all B to C, they're all kind of like they want to be found.

And if you think about a go-to-market approach, we have over 100-person to go-to-market team that's territory-based. So this is where we would go local in L.A. and S-F, wherever it might be, to be able to get these folks. And that's where, you know, a lot of this is outbound as opposed to in.

Did you ever think about the Atram-C-Pier-Belder model of like bolting an actual law firm onto your software company of an LLC, bolted to the, to the C-Corp with like an MSA? That was a model that's been tried. a few times I think people are trying it again, but did it ever occur to you to try and try your own cases, team up with a lawyer, build a firm? Yeah, for us, like we want to, the way at least I see this space evolving is I see this

is very much a winner take most. And even though we're in vertical SaaS here, there's like elements of Uber and Lyft that feel true to our space where the first startup that gets 10% of LA will win LA. Yeah. Because every time you see these cases, the better this whole system gets in. I'll just elaborate on that a little bit more, and that should hopefully explain why we're doing this in a very B2B passion. But the one thing you folks should know is, even though there's 20 million cases,

99% of them are settled privately. And so there's a real wide gap in performance in terms of how long these cases take to settle and how much they settle for. And generative AI alone can't solve that problem. You need to be able to see a lot of cases for you to be able to do that. Because most of this data is sitting in the inboxes of a very, fragmented industry. And so for us, the goal is how do we expand or go to market as quickly as we can to be able to go from 1% to 10% to 30% market share? And if you want to think about it as a key

metric, more cases we see the better of the system ultimately gets. And if we're just very focused on, I mean, we will never do this, but like competing against our own customers, you're only going to get over the cases that are out there. Yeah, that makes 10 sense. Do you think that legal AI is underhyped as a category. Obviously, there's a ton of funding flowing in, but there's not, you know, on the code generation side, it's sufficiently hyped, right? Every researcher at every company, you know, is talking about the leverage they're getting. Every startup founders talking about the leverage they're getting. Every CTO is talking about the leverage. They're getting public company CEOs. Legal is like, you know, just happens, you know, we certainly have plenty of lawyers in tech, but they're not, you know,

they just don't tend to be in a position where they can talk about stuff in the way that regular software engineers can. So it feels like you could be seeing the same type of revolution happening, but it just would be much, get much less coverage. No, for sure. I think that there's obviously been, when we started even up a couple of years ago, you folks could imagine it's like legal tech sucks, AI sucks, personal injury sucks. And my background is I used to work at Waymo, and I used to work as like a tech investor before that. And so it's definitely a lot, there's a lot more attention today in legal tech than there's ever been before. Is it, again, is it overblown or not?

I think in the end of the day, the thing that makes our space so different, when we're selling to these injury attorneys, the thing you folks should know is it's a very fragmented, highly competitive, and most importantly, commission-only industry. These folks make a percentage of the value of the cases that they settle. So unlike other parts of, you know, AI or legal, there's no experimental ARR. These are not like big firms where they can afford to spend a million dollars and something. And because of how competitive it is, like you better believe this thing can actually help their case operations.

Otherwise, they'll turn. They don't have the luxury of, you know, being able to have big IT budgets, etc. And so in that sense, if you think about our 2,000 customers, it wouldn't be around if they're not seeing real demonstrative ROI in terms of increasing case outcomes or reducing the duration of these cases. And in that sense, you know, I do think it is. call it, it's still early days. And if you think about the mission of the business,

it's not, I think it's just going to be, it's inevitable in terms of those 20 million cases being able to be settled, not again all over the place, but as fast as they can for the fair amount. And so in that sense, when we're just 1% penetrated, it's not over-hyped, at least in my mind. It's like clear ROI. Well, yeah, and there's that customer alignment where they're instead of to adopt as much AI as they possibly can because they're not billing hourly, right? They're just getting paid when these cases settled versus, you know, traditional law firm. If you, if something they used to be able to bill 10 hours for suddenly it takes 30 minutes and they had a lot of margin in that 10 hours, there's not, there's not the same incentive to

adopt as what you're seeing. You nailed it. And the other thing that's also different about this wave and it's funny because I was having a conversation with one of our customers about this. Some of these folks are still on-prem, but they're one of our biggest customers. And so when you go to a law firm and you're like, hey, you got to be on the cloud, they're not going to understand what you really mean. When you're going to these law firms and you're like, here's the legal letter, it was

drafted instantly. You can evaluate the quality of this thing relative to your own staff because you're a subject matter expert. You know what this legal letter does. And that's what allowed the space to adopt a lot quicker than anything that we've seen before in legal tech. And again, I think you really pinpointed, but they really genuinely care a lot

about being faster, settling quicker, spending less time on review, because anything that they save, they get to keep right as their own margins. Totally. One of my first jobs ever was doing paper filing at a law firm as an intern. I was like a kid.

You're a paper boy then. You're kind of a paper boy now. It's still paper enthusiasts. I hadn't put it together until, I actually just had this conversation and reflected on like how crazy that was that. I mean, I guess the cloud like barely existed then, but yeah, it was paper back then. It was wild. I mean, computers were around, but like my job is sort of like, yeah, inventorying this stuff.

So yeah, law moves slowly. Like a lot of the law firm partners that get like effectively tenure are like my process works. And I like it printed out. Or I like, I like, you know, having the network. The intranet at my office is reliable. I don't know about putting it in Google. And so, like, people have got to figure it out. When do you expect the first AI legal company to IPO?

Hmm. Still feels like we're probably a couple years out, but I imagine the revenue for you or some of these other players could get there, you know, get to that $500 million mark or wherever it needs to be to be interesting to Wall Street. Yeah, so my, just on the paper piece real quick, and I'll answer the other question. And it's so funny. The way you're going to kind of think of this, there's been so much innovation and more of the consumer, like how to get leads, like the consumer part of personal injury where like, you know, attorneys are advertising on TikTok, on social, et cetera.

But on the operation side, there's been very little innovation back to your paper days. They're still doing stuff on people. we've had some of our folks asked to pay our annual SaaS subscription with monthly checks to give you a sense of the ops piece can be in this space. And so this is the first time really in history that these operations of these firms are changing pretty substantially, right? Because you now can do a lot more with a lot fewer staff. And if you think about where we're going with this, you're going to see these firms be

just be able to handle a lot more volume. And if you think about the amount of cases that personal attorney's take on today, you'd be surprised at how many cases they're not taking on because they just don't have capacity or they can't make the math work because, again, of their model. Yeah. Sorry, I think I missed the other question. What was the other question again? Oh, I just kind of, yeah, IPO timelines for the category. When do you expect the first companies to get out? Yeah, it does feel like an entirely different, like, market. Like, there's no, like, clear comps that everyone knows. It's not just like, oh, another, like, you know, SaaS company that we can just immediately comp to. So it would be like a new. Yeah, and the legal industry law firms can't go

public, right? Yeah, yeah, exactly. And financial's got to be different. Yeah, I would love to hear some color on like how the public market may receive some of these. Yeah, I mean, in terms of timeline, I'd say, again, the 500 million of ARR, median ARR business going publishing, the threshold there, I think we're a couple of years away, but, you know, we're scaling super fast. for doubling year on year on a meaningful base. So I do think we're maybe two, three years, not us, but for one of the companies, maybe us as well, two, three years out before going public. In terms of how the markets might receive it, it's so interesting, right?

These are such different businesses where you're going to see a lot more Tim than you've seen in traditional software. And so I kind of say this, one of our biggest customers is paying us $4 million a year, and they have 100 employees at the firm. Wow. Think of the implied Yeah, $40,000 in value per head. Yep, exactly. So you're delivering a ton of value to this.

Yeah, it's because one of the things that comes up with some of our investors, hey, is like you're focused in a very niche part of all personal. And 300,000 attorneys out of, you know, over a million is maybe 30% of the base in the U.S. But when you're making 10x more revenue per head than traditional SaaS, your tent is actually meaningfully bigger than a lot of these SaaS companies that are out there. And so, again, I do think. it'll be interesting to see much more rapid growth, much higher base SaaS businesses than you've ever seen before and how the market would be able to evaluate that because it is going to be

different than traditional SaaS. Totally. Well, congratulations on the funding milestone. I'm sure we'll see you back here again soon. Maybe one more round before the end of the year. You've already done a couple, you know. But looking forward to the next conversation. Have a great rest of your day. Thank you so much for hopping on. Cheers. We'll talk to you soon. Thank you. And now it's time for our time.

our song. Find your happy place. Find your happy place. Book of Wander with inspiring views. Hotel Great Amenis, Dreamy Beds, top-treeing, and 24-7 concierge service. It's a vacation home but better folks. Go to Wander.com. When Lisa Sue took over

as chief executive of Chip Company Advanced Micro Devices, AMD, in 2014, the company's market value was just under $3 billion. A decade later, today, it's worth more than $330 billion. A little 100x.

There we go. The more than 100-fold increase that reflects how deftly AMD has pivoted its strategy from mainly producing graphics cards for gaming and personal computer processors to more tightly focusing on the data center chips that power the artificial intelligence revolution. The share price rose 24% on Monday after the company announced a partnership with OpenAI to build six gigawatts of power for inference. The deal has given rocket fuel to AMD share price and the company's ambitions to compete to compete with rival chip designer, Nvidia, which is by far the dominant competitor in the AI semiconductor industry at more than 10 times the valuation.

Now, the AMD share price is kind of up and down, and we're now at all-time highs, but just barely. March 8th of 2024, the stock was trading at $207. Now it's at $211.21. And so they've already ridden an AI boom from $50 a share, $60 a share, up to $200, then back down to $80 a share, and then back up to $211 shares. $211 a share. So Monday's deal specifies that OpenAI will be issued warrants for $160 million shares of AMD stock at a marginal price of one cent per share once OpenAI hits certain deployment targets,

and once AMD share price rises. the final tranche of shares will be granted only if AMD's stock hits $600 a share. So it needs to three-x and become a trillion-dollar company, basically. For now, the market capitalization, with a market capitalization of $4.5 trillion, Nvidia is nearly 14 times the size of AMD. And most analysts estimates peg the market share for graphics processing units, or GPUs, as they're known. Thank you, Wall Street Journal.

That power AI training and inference at more than 75%. So 75% of the market is Invidia, and they're trading it 10 times the price of AMD. AMD faces pressure from companies such as Broadcom, which produces application-specific custom chips for customers such as OpenAI. And so the Open AI deal might have shifted the balance in AMD's favor somewhat. How AMD got to this inflection point is a combination of careful strategic planning and being in the right place at the right time. Quote, over the last few years, what's been important to us is to understand the workloads

that would really drive the next generation AI training and inference. Sue said in an interview, this deal is a huge expansion of the work that we're doing. For much of the past decade AMD's arch rival has been Intel, the troubled chip designer and manufacturer that recently received major investments from Nvidia and the US government on the strength of popular designs for the graphics chips used in the PlayStation and Xbox gaming systems

and the CPUs, or main computer brains. Thank you for explaining what CPU is, Wall Street, German. using consumer PCs, AMD has steadily eaten away at Intel's market share for years. Intel's been bogged down by a costly effort to turn around its chip fabrication business. AMD, however, spun off its manufacturing business, now known as global foundries in 2009,

while Intel has continued to pour money into its unprofitable foundry segment, even as it fell badly behind the more technologically advanced rivals such as TSMC. And so this is where everyone always says Intel should have gone fabulous. They should have split the companies off. That's what AMD did. They split off global foundries, became a fabulous semiconductor design shop like NVIDIA. The business is similar to NVIDIA. And now this seems like they're catching up. I mean,

the stock's up a hundred dyes. Somebody asked Lisa Sue, if they are going to partner with Intel, Bucco Capital says they gave a very, she gave it very awkward and funny response. She says, well, as you know, the supply chain is something that we work on, you know, very, very meticulously. I think we have a very strong supply chain. We're certainly partnered with, you know, TSM across the supply chain, you know, just to that earlier question, we're absolutely prioritizing building in the United States because I think that's super important. This is the USAI stack.

We want to have as much of it in the U.S. as possible. And, you know, we continue to really look at, you know, how. How do we ensure that there will be a strong supply chain, you know, going forward? So when you have it typed out, it really, you can tell she's trying to... It's always tough to stuff a mic in someone's face and then transcribe it. Who knows? We've been going back and forth on the AI bubble. I think we should close out with the clearest evidence that we're not in an AI bubble from Trunk Fan.

He says he's not worried about the AI bubble bursting until corporate social accounts start having these exchanges again. And back in 2021, December of 2000, December of 2021, meta says, this is going to look great in the Metaverse. And the official Pepsi account says, you know it, friend. And Budweiser beer chimes in says, welcome brand friend, wag me. And then Pepsi says, thanks, friend, wag me with the rocket emoji. And this was a...

That was a moment. This is a dark time. That pretty much was, that was actually two weeks or about a week after the actual top was in the top of November. Yeah, Keith had called it. Keith called it perfectly. Everybody should have post notifications on for Keith. Yep. Probably going to, if he makes a call again, I think people will be listening.

Yes. Anyway, keep monitoring the situation, folks. Stay sharp and we will see you tomorrow. I cannot wait. We'll keep us five stars on Apple Podcasts and Spotify. And tune in Friday for our interview with Sam Altman, the CEO of OpenAI. Samma. It's time.

Thanks for watching. See you tomorrow. Bye.