Anthropic Drama, Meta Now Tokenminning, Fox's $22B Roku Deal | Diet TBPN

Obviously Mistral is in the news. Incredibly bullish for Mistral. Obviously very unfortunate drama happening in the Anthropic world. We'll go through all of that. But if the, if the sovereign AI thesis could not get any stronger right now, this weekend really showed a lot of countries, hey, maybe we do want a national champion. Hey, maybe we do want someone that will provide us models forever. SpaceX IPO, we have Gavin Baker joining in 30 minutes and we're recapping UFC Freedom 250. It happened at the White House last night. Jordy, take us through your reactions. Introduce our first guest to the show.

Okay, quick intro on the event overall. Yeah, I am a UFC fan. Yes, it's one of the few sports, maybe the only sport where I actually know the names of the people that are competing. Our guest in just a few minutes is one of those people.

Now, are you more of a fan of the UFC or more of a fan of the government? Because a lot of people were watching just because they're fans of the government and they said, oh, I will watch anything. I watch C-SPAN all day long. I watch anything that happens if it involves the government. This was an interesting cross.

No, more, more seriously, I, as a UFC fan and as an American, I'm excited for the UFC to do well. I'm excited for the fighters to have a platform. I was a little bit wary of this event overall. Sure. Just because the aesthetics of putting on this crazy party. High stakes. In this moment when there is a ton to black pill about. Yeah.

Not after—

Interest rates, wars, all this other stuff going on. And yeah, so war raging, gas prices are high, house prices are high, rates are going up. A lot of people in America are not having a good time. And it really felt like a bread and circuses moment. But the bread and circuses, John, it worked on me. Top notch. It worked on me. Some of the best ever. It was—

Why was it so good compared to a normal UFC event?

Well, you didn't watch, so I guess it's hard.

Was it the aesthetics or the actual card?

Well, so one, the production value. One, the card was absolutely stacked. It was 8 fights, no filler. Wall-to-wall, uh, really, really, uh, fights that had a lot of, um, they carried a lot of weight, right? So like the, the main, uh, title fight, Gaethje versus Ilia Topuria, uh, that changed both fighters' careers, like full stop, right? Uh, and even with Bo Nickal joining, uh, uh, in just a few minutes, you know, Bo, that was the biggest possible stage that any athlete could be on, and especially in the UFC, at least here in America. So anyways, going into it, I thought the aesthetics were bad. I was excited to watch the fights. Turned out it was just really cool. And sure, there's still gonna be people that are upset that the president and the admin is like platforming something like this. Turned out to be really fun. It was a cool night. And I think America kind of needed it. Yeah.

A lot of people are saying it's not unprecedented because Teddy Roosevelt, held a boxing match in the White House almost 100 years ago, around that time. And apparently he fought in the boxing match, which is crazy.

Yeah, he was a big, big boxer.

He was a big boxer. And so imagine, imagine. So this is like a light, a light craziness. Anyway.

But anyways, yeah, I, you know, the event was, was incredible. And I came away thinking that big parties and sporting events are potentially the most underutilized political strategy, right? I would actually expect, uh, the, the admin's like overall popularity to go up, approval rating to go up based on— interesting, uh, based just on this event because there were so many iconic images. Funny enough, the most iconic image is of a Brazilian named Diego, Diego Lopez, which we had up on the screen there standing up on that. Um, but, uh, but anyways, uh, really, really wild.

So the other big, uh, topic over the UFC was the fact that Dario Hamade was not in attendance. He wasn't front row. Mark Zuckerberg was. And it sort of underscores this odd situation that Anthropic's in where like they just got dealt this weird hand where like they're doing so well and the admin could not be more misaligned aesthetically with Anthropic.

Their, their whole brand.

Yeah, but the whole brand, they're very, very different.

And, and I was, I read the article yesterday that Anthropic staff were going to Washington, D.C. And it's funny, some people were like, oh, this is bad. Like, if they could solve the issue easily, they could do it over the phone. Not how this works. Serious issue. Everybody goes. Yeah, you would obviously go. But going to Washington, flying to Washington, D.C. to try to solve an issue on a Sunday, there was not a worse—

Well, they'll be here.

They'll be there. No, no, I know. But you can imagine you land there, you're like, OK, where is everyone? Like, meet, this is a big issue, let's meet, figure this out, this is important, the national security is on the line, like, let's figure this out.

I'm busy.

And they're like, where is everyone?

And like, if you take safety seriously, cybersecurity seriously, biosecurity seriously, export control seriously, like, UFC could not be less aligned with that, the stakes there. And so like, there is this big gap between the administration and Anthropic. And I think that everyone should, Like every American should care about bridging this gap. Like private enterprises should be able to flourish in the United States. Everyone agrees with that. And the government should be able to enact reasonable regulations to protect America's interests. And both of those need to find a way to play together, but there is just a really, really broad gap between these two organizations. And so everyone's hoping that things resolve quickly. And it seems like the right, you know, wheels are in motion. The flights have been booked or already taken and things are moving along. But just to recap on the play-by-play, March 4th, Anthropic received a Department of War supply chain risk letter that of course got rolled back eventually. April 7th was the Mythos preview announcement alongside Project Glasswing. June 9th was the Fable 5 launch last Monday. Mythos with guardrails around cyber, bio, and AI research. And on June 12th, just 5 days after the end of the week, Fable 5 gets suspended after the Commerce Department issues an export control directive. So Anthropic explained in a blog post that the US government is restricting both Fable 5 5 from being used by any foreign national, whether inside or outside the US, including foreign national Anthropic employees. And so that's a huge compliance burden, very sudden. And that led Anthropic to completely suspend the use of those models for all users, not just, not just foreign national employees, because they need to, at least if they want to comply with this, do all the KYC, understand how different organizations are using the model, who's using the model, who at those organizations has access. If something's being vended through an API and it's showing up in another product, does that product do all of the proper authorizations? And so it's a big wrench in the gears of the rollout of this model. Then the information reported that Andy Jassy at Amazon had raised concerns with the senior administration officials about jailbreaking, the risk of jailbreaking Fable and getting access to all of Anthropic. Anthropic countered in their blog post saying, quote, we reviewed a demonstration of this specific —technique being used to identify a small number of previously known minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly available models are able to discover them as well without requiring a bypass. So they're saying, this isn't that big of a jailbreak, this isn't that big of a deal, we can move forward safely, we've taken safety seriously. But that is the debate that's going on between the admin and Anthropic. So as everyone knows, the rollout of Fable 5 has been a bit rocky. Incredible demos, incredible benchmarks, but offset by the odd decision to silently degrade the quality of the answers related to frontier AI development instead of just refusing the request like cyber and bio prompts, which that was the main thing that everyone was really confused about. The actual rationale behind AI development refusal is pretty sound. Like people are upset about that if they're working on machine learning systems, recommender systems, anything that requires, you know, like instantiating an AI product, let alone if you're just, hey, I'm just doing open source AI research and I'd love to use the latest and greatest models to help me. Now it's not an option. But at the same time, just think about the logic. If a model doesn't let you hack a system, just have the model build you an unrestricted model that does let you hack your system, that system. And so, Clearly in that scenario, it makes sense to restrict. If you want to restrict hacking or bio, you also have to restrict the tool that makes the tool that hacks the system or develops the bioweapon in theory. And so, but the choice to silently degrade responses was not well received. These are important issues. And there's another timeline where AI leaders are cut from the same cloth as America's elected officials. But we are in a world where Anthropic has to make their case for safety in an age of RSI during a UFC fight on the White House lawn. It's a very bizarre timeline, Here we are. Most of the big tech CEOs, they have figured out how to work with the administration effectively, showing up to large dinners, photo ops, the occasional movie screening to make their case about hot topics. Many of these CEOs also have business relationships with Anthropic, either on the customer side or on the investor side. And that's what's really complicated about the Amazon relationship. Obviously Amazon uses Anthropic models, also is an investor in Anthropic, also provides compute to Anthropic.. And so they have a very hairy, complex, circular relationship. And so that clouds the discussion over whether or not is Anthropic going to try and help Anthropic get through this? Are they doing everything above board? At the same time, you know, people were sort of putting out that like, oh, did Amazon like spike the ball on this? And at the same time, if the government comes to you and asks like, hey, you're a big tech company, you've been using this model a lot. Like, were you— They're a close partner to Anthropic. They're a close partner. Were you able to jailbreak it at all? And your security team, just issued a report like, yeah, we actually were able to jailbreak it a little bit. Like, you're not going to withhold that from the government. You're going to turn that over. That's different than calling and saying, hey, I really think it's dangerous. You got to put pressure on them. So there's this wide range of levels of interaction that can go both ways. You can be a lab partnered with a big tech company. That big tech company can have a really established government regulation— government relations group that is actually helping you get through. To the admin. On the flip side, you could have a big tech company that is competing with you or frustrated by your business and is actively trying to, you know, put their thumb on the scale the other way. So I'm sure we'll learn more in the future. Many of these CEOs have business relationships with Anthropic, so we'll see how all of this develops. And they're always renegotiating these contracts too, which makes things more complex. And they're competing across multiple markets. So expect the 4D chess to be on full display all summer long in Washington. Anyway, what else is going on? Meta is doing a 180 according to Amir over at The Information, trying to be the vanguard of token minimizing. I think the token maxing, token minimizing thing is so silly because do you know where maxing comes from really? Like it comes from gamer lingo usually and—

Adopted by the looks maxing community?

I don't know if looks maxing community was the first one, They certainly made it really, really viral.

But they popularized, like, I feel like through, effectively through Clav's clipping, that language. Yeah, that language. The idea of like doing anything like, oh, I'm computer maxing today.

Yeah. Isn't it from like min-maxing? Min-maxing.

And that's what I'm getting at. Like, you shouldn't be token maxing or token minning. You should be token min-maxing. And the idea of min-maxing is when you're in a video game, you want to get the most resources for the lowest cost. So if you're playing League of Legends, you're going to put just the right amount of gold into this particular item or this particular stat point in an RPG, and you're going to try and find the optimal build based on your resources. You have confined resources. And somehow it feels like these, like, Tac and all these companies went through some sort of hallucinatory period where they lost sight of that they should always have been min-maxing because That is the goal of business at all times in all things. You would never say like, oh yeah, we're ad maxing this quarter. We're gonna buy as many ads as possible. Oh no, it didn't work. We spent way too much on ads. They were not effective. We need to admin, spend as little as possible on marketing. It's like, nope, you always wanna be spending, you always wanna be getting the most return on ad spend, the highest leverage output.

Yeah. There was, and again, unsubstantiated post. Yes. But somebody was saying they talked to somebody at Meta and there was one engineer that spent $90,000 in one day. And was quickly terminated. No, really? Terminated within 3 days. So again, could just be entirely fake news, but yeah, it seems somewhat informed.

The crazy thing is that, is that like the lesson from that—

this guy's like, I'm not getting fired, I'm going to use AI more than anyone else in the whole company.

I'm writing loops.

Uh, $90,000 bill. The—

I mean, so, so the lesson that will be learned from that is, is don't spend $90,000 a day. But the Super Bowl is literally, for a marketer, the Super Bowl is spend $5 million in one day and it works sometimes. And sometimes Super Bowl ads deliver $20 million back.

Yeah, but what do teams do when they run Super Bowl ads, right? We ran our Super Bowl ad. Yeah. We trained our own model. Yeah, yeah. I'm kidding.

No, but teams do run ads. No, we made sure that ours was ROI positive and we only bought the ad in a small market and we did it before the game.

Right, right. So that we already knew our audience was.

And so For a very low cost, we were able to get a very high value. So I believe we got a very high return on investment. We mid-maxed the Super Bowl. Now there were some companies that sort of just Super Bowl maxed and they just spent a ton of money. And when you looked at the data on how they were received by audiences, it seemed like, oh, they definitely didn't get $5 million in brand value back. In fact, maybe that ad was poorly received. Some of the AI-generated ads.

I gotta give it up to the UFC for the, for the level of advertising during the event yesterday.

They're running full, full ads in the middle?

No. So one, the interesting thing— Just in— With the new Paramount deal, there's a lot of ads. OK. But one, Dana White was in a lot of the ads that were running. So like Dodge was doing a TRX ad. It was Dana White driving. That's cool. There was another ad. I forget the brand. I guess it was a bad ad, but Dana White was they faked a podcast set where it was with an insurance company and Dana was having a conversation about business insurance with the CEO of the company. Interesting. So it looks like a podcast. So I thought it was interesting they were casting Dana and all them. Then second, the level of advertising just in the octagon was out of this world. There was so many different ways to gamble. Every— there was like, there's no like UFC cannot do like a gambling exclusivity deal with anyone, right? Because it's like they have It's Crypto.com, it's Polymarket, it's Bet365. Yeah. Any possible way that you could want to, like, gamble.

Bet365. So the only day you're taking off is Leap Day. Every 4 years you take 1 day off of gambling.

That's kind of like the ethos of the company.

That's kind of like the mission.

1 day off every 4 years. And then, and then Kalshi's running ads during the event.

Oh, so Kalshi and Polymarket are doing that?

So Polymarket is in the octagon.

Okay.

But Kalshi's running a bunch of ads. Against it? Against it.

And they're buying directly through Paramount because you streamed it on Paramount, is that correct?

Yeah, I don't—

Did you have to pay for it in addition to— No, no, no. Okay, so you pay for Paramount on a monthly basis and then you get UFC for free. Yeah. Gambling is just min-maxing done poorly. That's true. So the news around Meta is Meta is doing a 180, trying to be the vanguard of token minimizing. 2 months ago, Meta epitomized token maxing, on trend, on track to spend billions a year on cloud, etc. We've seen an exponential increase in AI usage and we are tracking to spend billions on internal use alone in 2026. The memo said, at the same time, individuals and teams have limited visibility into and control over how they use AI. In 2027, we expect Meta will move towards managing AI tokens in a more structured way with bullet budgets, allocation decisions, and supporting tools. That makes sense. You have a marketing budget. You should also have a token budget as part of—

I also saw something they have like Meta Code, which is their internal.

Yeah, that makes sense. I mean, they're, they're like, you know, catching up to the frontier with—

Yeah, I mean, like you talked about this, there's a bunch of like these leaked calls where Zuck is like, you know, you should imagine that Meta employees are gonna be better than like the average open source contributor. So it's like the basically quality of coding data within Meta should be super good. You can train some model on that.

It's the same pitch as take a frontier model, bring them in through some sort of consulting partner. They fine-tune on your business workflow, they automate your business process, but your business process in this case, if you're Meta, is generating code that makes your social networks better.

Yeah, it's especially good because like in Meta, all of the like, you know, internal business, like whatever data, it's already in like code format. You don't have to like scan in a bunch of papers. Yeah.

Into, you know, yeah, like digital. I still, I still don't know what Meta is actually building. And I think there, there comes a time soon that they should probably communicate that to the market.

Personal superintelligence, baby. What's there not to understand?

No, it's just— So $1.5 trillion company. What does that mean? It's 75% as valuable as SpaceX, buddy. Come on. It's crazy. SpaceX, well— SpaceX is $2.3. Well, if things keep going the way they're going, we'll see even more of a divergence. No, but I think it's so unclear. You should be able to put a lot of the SpaceX bull case on Meta, right?

It's like, sure, maybe Colossus was built a little bit faster, but Meta's pretty fast for building hyperscalers. I think they're wildly—

They couldn't be more different. With Meta right now, it's the bull case is it's potentially the greatest business in the history that is not threatened by AI. Totally, totally. That actually is accelerated by AI. Yeah. And I don't think, anyways, but I do think I would expect some communication from Meta around what their real strategy is over the next month or so because it's killing them that they just keep saying, oh, we're personal superintelligence, but wait, they're building like a coding harness and they're working on coding. Like, are they gonna sell that externally or is it gonna be internal? Coding? Are they gonna try to?

Coding seems like it makes sense.

Like they're spending so much money on that.

Are they gonna try to sell their model on an API and get into enterprise? Like the big question is like, do they take the leap? They kind of did that with Manus. Yeah. But that's getting unwound. Fox is buying Roku.

The streaming service for $25 billion. It's in the Wall Street Journal. 25 or 22? 25 billion in the Wall Street Journal here. Let me refresh the page, see if they updated it. It still says 25. Weird.

Bloomberg is reporting $22 billion.

Okay, well, who knows? Somewhere in the north of a billion, sub-hundred billion. Let's be very conservative with our estimates.

Yeah. CNBC also $22 billion. Okay, well, the Journal is—

The Journal may have inside information given that it's a News Corp property. Related there, right? Anyway, let's read through the Journal piece. Let me just give you some facts and then you can sort of give me some takes and contextualize it with more information. So Fox Corp said it's acquiring Roku in a deal valued around $25 billion. So the Wall Street Journal saying around is $22 billion around maybe making a major bet on the future of ad-supported streaming. Ad-supported streaming, obviously Netflix, Pivoted to ads after saying we don't— we're not going to do ads. Turned out to be a great business. Ads are undefeated and more people want—

Down with ads.

They're just down with ads, especially if they're good ads. Anyway, the Fox— the, the deal, the deal is Fox's largest to date. It brings together a media company known for its live news and sports programming with the biggest provider of streaming platforms for connected TVs. It will add scale to Fox's streaming business, currently home to free ad sporting streaming service Tubi, which the company bought for $400 million in 2020, and subscription-based Fox One and Fox Nation.

Uh, in addition to— Tubi, from my understanding, has been a home run. Yes. Like, I don't remember exactly what the revenue is, but they bought it for $400 million.

It's in the billions now. And all the metrics about actual audience size are huge. And, uh, getting to tens of millions of dollars in revenue, hundreds of millions of dollars in ad sales against a big audience like the Tubi viewership when you have the Fox machinery behind it makes a ton of sense. Yeah. In addition to distributing other streaming services through connected TVs and devices, Roku has its own ad-supported Roku Channel. The combined company will, will better compete with the likes of Amazon.com and Netflix for ad dollars. Quote, bringing these two companies together will really help define the future of television in the United States and in many other markets around the world, Fox Chief Executive Lachlan Murdoch told investors on a call Monday. FOX will pay around $160 per share, $96 in cash and 0.99693 FOX Class A shares. The deal is valued at around $22 billion on an enterprise basis. FOX expects to fund the cash portion of the transaction with $12 billion in new debt as well as cash on hand. The combined company expects to cut around $400 million in annual costs. FOX plans to keep Tubi and the Roku Channel as separate offerings. Consumers are increasingly opting for free and lower-cost ad-supported streaming options as the cost of subscriptions marches ever higher. Ad-supported streaming plans now represent almost 50% of all premium subscription ad, uh, video-on-demand signups in the US, up from 39% just 2 years ago. More than 100 million global households stream with Roku. Roku's the largest streaming platform for connected TVs with 25% market share. Samsung's Tizen is number 2 at 23%. And then I'm sure, uh, probably Amazon's Fire TV is not far behind. Fox remained on the sidelines during the heady early days of streaming boom, pouring money instead into programming for its cable channels and buying up sports rights. It launched Fox Nation in 2018 and introduced direct-to-consumer Fox One, which includes sports and Fox News, last year. Fox said last month that Tubi had nearly 100 million monthly active users. And its revenue had grown by 23% in the fiscal third quarter. And so Fox's stock is down today. Roku is down 0.2%, something like that. And the market doesn't like it. Take Him has a take here. We can read through his opinion. He says, I don't understand how this Roku deal makes any sense, bringing together the most valuable live—

I'll tell you how it makes— How does it make sense? No, so I think it makes sense. One, Fox, they have a great product in Tubi, right? It's again, there are certain people that are happy with their HBO or Netflix or whatever, but Tubi has a great niche. It's been a great acquisition for Fox, north of $1 billion of revenue. That is a product. Roku is a platform. They have over 20% market share in connected TVs. I was looking at the data. They have over 40% of the engagement with connected TVs. So when you think about it from an advertising potential, right, it's not just that they have 20% of the connected TV market. They have over 40% of like the watch time effectively. Right. So looking at the potential from an advertising standpoint and how many dollars are still yet to shift from traditional television to connected TV. So from that lens, I think it's, I think it's great. Roku's also a much younger demo. So like a lot of media, a lot of media properties, right? Especially television related are gonna be, you know, older demos that are eventually gonna be aging out at least of like prime consumers. But these are consumers that are in their, you know, prime earning years, prime years as consumers. And so ultimately, I think it makes sense. I think it makes sense for Fox. Next, the thing that I hope that they can do as a consumer is rebundle, right? We've just gone— we've gone through this era of unbundling and the fact that you get a new TV and you're like, okay, great, let me log into, you know, 15, 10 different services, something like that. I would like to see them do I would like to see them do something in bundling for, for consumers. So again, looking at it from the lens of like having north of 40% of like watch time, right? That is a super valuable property. And like with the right ad-driven approach, I think, I think it'll pay off.

I wonder how windowing fits into this. Like I, when I grew up, I didn't have cable, but you would get a prestige movie played on ABC, like once a week. And for a while, that was the flow on Netflix. And HBO was like its own premium subscription. You could open up the HBO app, it's all premium series and movies. And then some of that would trickle down, but then it feels like the ecosystem sort of branched a little bit where Netflix stopped getting like the James Cameron library for a while. Disney started hoovering up some of that, keeping some of the really popular movies, the Marvel franchise, off of Netflix, even for later run, a year or two after, because there was more competition from the actual different pieces of the different studios. So I wonder if Roku will be able to provide some sort of like neutral territory in some ways, where viewers can pick up something that's in the premium end, something that's, you know, a sequel to a franchise that they know and love, but it's not seen as as competitive as, as, you know, Disney versus HBO or whatever. Well, we will see.

Overall, real platform. Yeah, they already have a product that sits on top of it. Tons of advertising potential. And I think that regardless of whatever the market reaction is today, I expect that people will get it over time.

Tyler Cowen shared some more thoughts on the recent Mythos brouhaha. It's a good term. This was all over, we talked about this earlier, but he says, according to not yet confirmed but likely true reports, it was shown that the model could be jailbroken. There's debate over this now. The released Mythos already restricted bio and AI improvement queries rather than strictly in fact, rather strictly in fact. So now we are back to the model not being available. Here are a few of the constraints on the US government, the only ones, I might add. One, it needs for the main companies to stay in business. On top of that, it wants their IPOs to go reasonably well. And it's now much harder for the top companies to recruit foreigners, which is a significant share of their highest quality workforce. Denis, Ilya, Andrei, for a start, are all international engineers. It is much harder for the main competitors to drum up foreign business in a credible manner and sustainable manner. That's interesting. We haven't seen many breakout charts of how much revenue is domestic versus international, but we have seen reports of OpenAI and Anthropic doing big deals with international companies. Some of these companies, Fortune 500s, they just have global footprints, so they gotta use the models everywhere ideally, unless they do all of their software engineering in Silicon Valley. 1B, how are American multinationals operating abroad supposed to use top systems moving forward? That's, Literally what I just said. I just predicted the next line. Whoops, I am a stochastic parrot. It wants to use model access as a tool of both hard and soft power. So model access has to be possible at some level, but it is very hard to control what foreign agents will do with partial model access when they get it in the future. 3, the US needs to stay ahead of China in the AI race. This is a big question. We have Aaron Ginn from Hydrohost joining in just 10 minutes. We'll ask him for his perspective on this. For the US needs to issue restrictions that are actually enforceable and US citizens only doesn't fit that bill. That's a good point. Furthermore, Markets and Everything, he, it is easy enough to hire a traitorous American to access tools of wrongdoing. That's a really good point. How can you verify that every American is loyal or for a matter, it is not that difficult to fake citizenship in various ways.

Yeah, there's been people over the last 50 years that have gotten caught, like, passing on or basically selling secrets around our various military aircraft, for example.

Yeah, there's a very controversial story of Virgil Griffith, a cryptocurrency researcher who I actually met once. He worked at Caltech, was deep in math and very excited about cryptocurrency. And he went and gave a talk on cryptocurrency in North Korea, explaining how cryptocurrency works and how cryptography works. And that was seen as a violation of export controls, and he was sent to prison. And the crypto community is very upset about that, because they maintain that math should not be illegal. This information should not be illegal. The government side is helping— What was his defense?

Was he thinking that he was going to liberate the North Korean people by giving them access to effectively, or explaining how code can be used to create a store of value that is unsanctionable, essentially.

And so it was seen as a violation of sanctions by the US government.

And more importantly, they have a North Korean stablecoin.

I don't know. I don't know.

They are calling it the most sanctioned stablecoin potentially.

Potentially. This one, number 5 by Tyler Cowen, was a little bit controversial. He says the US government cannot nationalize these companies. And then proceed to run them effectively. And people were debating that, saying that it might be possible for the US government to own the labs in some way and leave them to their own devices to actually roll these things out. What do you—

Yeah, I mean, if you're like super RSI-pilled, right, you don't need the employees at all because the AI just improved itself.

So the government just says, continue to improve. Yeah, basically. And that's it. So you don't actually—

That's like the extreme scenario.

Yeah. Yeah, and then on the other side, it's that, okay, these systems are actually extremely complex. You look at the uptime of some of these systems, the guardrail, the research taste. In fact, you need so many employees to make these products effective. You actually need the international employees as well because they come up with great ideas. And so yeah, those are the two sort of sides of the debate. Six is Chinese and open source models do in fact improve at some reasonable pace. Even though they are right now considerably behind them at best proprietary models. And that's something that I wanna dig into more. It does seem like the pace of open source development in China is slowing. I'm not entirely sure what's driving it. It feels like there's a number of different things. Gavin Baker said it's about compute and they're turning their backs on the Blackwells that they had the opportunity to buy. I think it might be research ideas that are more locked down now in tighter controlled circles and so on. San Francisco. It also might be the data, might be, but you'd have to imagine that they have lots of people that are writing code, lots of people that are inferencing those models and then full data and information capturing of those rollouts. There are a number of different reasons why that pace might be slowing and we'll see. Maybe there's a big jump in the graph. Maybe we have another DeepSeq moment in the next couple months. We'll see. Clearing order in. Leave us 5 stars on Apple Podcasts and Spotify. Sign up for a newsletter at tbn.com and we will see you tomorrow at 11 AM Pacific. Goodbye. Technology's Daily Show. Technology in motion.