WE'RE BACK, Meta Addiction, Tim Cook's Last Day | Jordan Schneider, Robert Mendelsohn & Eric Olszewski, Billy Thalheimer, Aaron Cannon, Stephen Balaban

You're watching TVPN.

Today is Monday, August 31st. We are live from the TVPN UltraDome, the temple of technology, the fortress of finance, the capital of capital.

We are back.

We missed you.

And Ramp is still with us. Let me tell you about Ramp. Time is money, save both. Easy-to-use corporate cards, bill pay, accounting, and a whole lot more.

All in one place.

Oh, feels so good to be back. 2 weeks is a crazy amount of time to be away.

I don't understand how children do this for weeks or months at a time.

It's wild.

It's wild.

It's crazy.

There's a lot of video games.

I was sitting there. Yeah, I had somebody this morning say it feels like, uh, it must have gone by so fast. Oh, and I, and I was thinking, no, it absolutely didn't go by fast at all. It felt like, yeah, it felt like 2 months.

Yeah, it did feel like 2 months.

We manipulated time.

Yeah, but, uh, thank you to the Microsoft team, the Activision team. They sent us codes for Modern Warfare 4. My review is in. It's fantastic. I love it. It's a really good game, really polished. And the Kill Block—

Did you revert back at all to—

No, didn't really have the opportunity. I was joking about full relapse, just gaming from dusk till dawn. But that is in fact not in the cards these days.

I had a relapse on pizza.

Pizza?

I was eating pizza almost like twice a day at times. Eating a lot of pizza, surfing a lot.

That's good.

A lot of family time.

Do you think pizza is potentially addictive?

I do.

You think that there could be a settlement?

I do think— I don't think there could, I think there should be.

There should be. I do wonder, like, because cigarettes, it was very— cigarettes were very scientifically proven to be addictive. Like, you can give nicotine to a mouse and it will, like, go and try and get more nicotine. Like, it's the most cut and dry, you can run the experiment. With any animal—

And I would assume that—

In the blood.

Based on my own personal experience, pizza and surfing fall into that same bucket.

Oh, you give a mouse pizza?

You let a mouse get barreled.

Yeah. If you give a mouse a cookie—

If you give a mouse a barrel—

A barrel.

He's gonna want a lot more.

He was gonna want to throw it with a line.

Yeah, for sure.

But it will be interesting to see where the addiction stuff goes, because when all this came up, The Meta social media addiction. I was kind of like—

Give a mouse a scroll.

Like, I understand that people were trying to— I don't know, can you get mice or monkeys addicted to Reels? Maybe.

I'm sure, almost certainly.

You think you could get a monkey addicted to Instagram Reels?

I think 100%.

Well, we got to try.

I think 100%.

We're going to find out.

We can't get any monkeys because they're sold out.

No, no, you can get a monkey. You just have to pay top dollar.

Oh, okay.

They're expensive.

Wide demand.

But you can buy one. But I do, it did seem like it was a tougher sell because it was not a chemically addictive product. It was more of a psychological thing. There's a benefit to it. There's a free speech element, all very complicated. I was not expecting to see a big settlement when this first came up years ago. I thought it was going to be more nickel and diming, little case-by-case things. But in fact, there has been a huge settlement by Meta 2 days ago. Or this was when this was written, this was August 26th, Meta announced settlements with attorneys general representing 48 states and the District of Columbia and 3 US territories. So there were only 2 states that held back, I believe it's Texas and New Mexico, but over allegations that Facebook and Instagram harmed children and teens. So this has been boiling up for a while. There were smaller settlements. This is the big one. And so the high-level numbers here. Meta will pay the states $12.7 billion over 10 years, and it could go up to $18 billion if other platforms join the settlement.

It is a big number.

It is a big number. And honestly, if you are a state legislator or you care about having your state funded for a variety of things, this is good news. This is a Seiz Gong moment for the states because they will be getting this money distributed to them, and they can use that to do whatever states do, from education to healthcare. There's a lot of different things that states can do. I know Tyler's pissed because he doesn't like the government getting a dime, but, but, uh, it is a Seizgog moment for them.

I did find a video of a chimpanzee hitting reels. If we can pull this up.

Yeah, let's pull this up. Let's pull this up. This is, uh, this is key. This is important. This is definitely more important than—

Oh, okay. Yeah, this looks okay. Maybe, maybe, maybe social media is addictive now that I'm seeing this. This is— yeah, it might be universal. If it crosses—

wow.

Using it so intuitively.

Yep, yep. I mean, that's just good UI design.

Come on.

You've got to give the UX engineers some credit here. Make it so easy to use that even a monkey could use it. That was the prompt for that UX engineer, and they delivered. This is— yeah, if it crosses species barriers, I'm much more likely to agree that something could be addictive. But—

well, and here especially, you know, he's looking at—

Other monkeys.

Other monkeys.

Oh, OK. That's a little—

I don't know.

Yeah.

No, I know. But an algo feed, it's gonna naturally start serving a lot of those videos.

Yeah, yeah. Makes sense. Anyway, so everyone has been drawing analogies to the tobacco industry's master settlement agreement around this, because that's the last time there was a huge multi-billion, tens of billions of dollars settlement between a large organization around addiction and a bunch of states. happened when cigarette companies signed the master settlement agreement, all the companies agreeing to pay all the states. But the structure was a little bit different and there's some nuance there. And so Eric Sufrin over at Mobile Dev Memo, friend of the show—

The Sufrinator.

Says— the Sufrinator. He was saying this is not necessarily the— he said Meta settlement is no big tobacco moment. That was his takeaway in Mobile Dev Memo. You should go subscribe. Because this post is a paid post, but I'll take you through some of the interesting analysis that he put together. I was debating with a lawyer friend of mine about this over the weekend and have some interesting details here. So Eric writes, and the headline here is that what Meta is going to pay as a percentage of revenue or what the social media industry will pay as a percentage of revenue relative to what the tobacco industry paid as a percentage of revenue is off by almost an order of magnitude. So let's put them in context here. So he says Meta's settlement is comparatively paltry. A white paper from the USDA using BLS data estimates that the total US consumer expenditure on tobacco products in 1998, when the Master Settlement Agreement was signed, was roughly $57 billion a year. $57 billion. Son of a gun. Certainly good if you're a big tobacco shareholder, I suppose. Annualizing the roughly $250 billion in combined tobacco settlement payments over 25 years produces an average of $10 billion per year is what they've been paying. And so that was equivalent to 17.5% in 1998 domestic expenditure. So you can think about it as like when tobacco got hit with like, you're addictive, what did the government say you got to pay? They asked for 17.5%— You need to pay. Very, very meaningful. Very, very meaningful.

And that's, that's off the top, right?

Yes. Yes. This is revenue.

It's not like, hey, you got to give us a percentage of your profits.

Yes. And so it's calculated in a more complicated way. It's not exactly a percentage. It's on a per pack sold, but it is inflation adjusted. And so there's a whole bunch of other mechanics that really have allowed the Master Settlement Agreement to deliver tens of billions of dollars on an ongoing basis, even while now tobacco sales are slumping, cigarette sales are declining. And so the That does create more complex dynamics for the states, but still, it's 17.5% of the industry's revenue coming to states as effectively an extra tax on top, which I think a lot of people would agree is good because cigarettes are bad, right? But what's going on with Meta and social media? So given Meta's full year 2025 US revenue of $75 billion roughly per the company's 10-K, its average annual payment if they pay the full $18 billion Because they're paying that over 10 years, it's $1.8 billion. So what is $1.8 billion of $75 billion? It's 2.4%. So instead of 17.5%—

Sorry, where are you getting the $75 billion?

$75 billion was their US revenue last year.

Oh, US, US. Globally, it was like a couple hundred.

Yeah, bigger, bigger. But you have to comp it to domestic because that's where this is taking effect, right? So it would be even lower if you look at overall revenue, which is a good point. But you can think about it as, Tobacco had to pay 17.5% of domestic revenue. Social media, 2.5% of domestic revenue. So not quite an order of magnitude, but a significant gap, significant gap there. And so it's important to reiterate here, says Eric Seufert over at Mobile Web Memo, that the MSA payment, the Master Settlement Agreement payment schedule, is inflation adjusted and tracks unit sales. Meta's statement settlement has no such mechanism. So Let's say that Meta doubles revenue. Their percentage they'll pay is lower. And then let's say inflation kicks in. They're still going to be paying, you know, this $1.8 billion a year 10 years from now. Could be, could be pennies if hyperinflation happens.

Yeah.

Who knows?

Yeah. The only reason it felt somewhat significant from a dollar standpoint is because Meta is just spending so much money on AI.

Oh, interesting.

Like, you can just imagine Zuck being like, I really wish I could have given that to Jensen. Right.

So I had the opposite. I had the opposite read, which was I saw $18 billion and I was like, ah, that's like a month of AI spend for them. And I was like, ah, this is not that much money. And then of course it's over 10 years. And so when you think about it as it's $1.8 billion A year, so that's $100 mil a month. That's their token budget right now? I don't know, maybe less.

Overall, it's insignificant. But when they are spending more money than they are making, it hurts a little bit more.

Yeah. So, for sure, for sure. So, the other issue that came up here that is different from the Tobacco Master Settlement Agreement is the question about the First Amendment. Free speech issue. So tobacco, the tobacco industry was restricted from advertising cigarettes as part of this master settlement agreement. And there were other cases that led to no more cigarette advertising. And of course, that is somewhat infringing on the corporate speech, but that's a lot easier to deal with than consumer speech. And Meta is in an interesting place because the Supreme Court has deemed that social media is a critical a critical vehicle for citizens to exercise their free speech. And so, like, it's very difficult for the government to come in and say, we are going to ban social media because it is a vehicle for Americans' free speech, individual citizens' free speech. So even though everyone thinks, oh, they're punishing Meta, it would still— there would be plenty of American citizens who would say, wait, my free speech rights have been infringed upon. So It's much harder for the government to come down with some standard traditional regulation. And that's why we're in this world of regulation through litigation, which is that Meta's arms have been twisted, like they have to pay this. And then we'll go into what they're doing around actually changing the product for teens and underage users because— and they're all doing that because of these lawsuits, not because there's a new law saying you can't do this. And a lot of that is like There are plenty of topics that are pretty consensus controversial, things like eating disorders and self-harm that people are uncomfortable with and think they don't have a place on social media. And yet free speech rights do actually protect those. So if you wanna talk about your eating disorder, like that is your free speech right. And Meta can't infringe on that as much as, it's very tricky for them to do that basically. So there's also an interesting regulatory capture angle here. And it's relevant to AI because in AI, everyone's talking about regulatory capture as something that the companies want, like they want regulation. And this is sort of the opposite example. So regulatory capture or regulation, like the benefits of regulation usually benefits incumbents and big companies for sort of 2 reasons. One is that typically when the regulation's being written, The big companies, they have lobbyists. They have offices in D.C. They also have you know PACs and all sorts of different political vehicles to donate to make sure that the regulation gets written in a way that is at least a little bit favorable to them. They can just move the needle a little bit in their direction, hone things how they like, right? And then second is that once the regulation goes into effect, they have the resources and the actual margin dollars to go and. To go and comply. And so if all of a sudden to set up a new social media website requires age verification, age verification can be expensive. Like we have to do it at Lucy for everyone that buys nicotine products and it's like a dollar for every checkout. And if the average order value is like, at one point it was like $50, like that's like 2% of the top line that's going towards regulation. If you're a bigger company, you have other ways of massaging that cost. Basically, what's interesting about this is that the social— we know the social media companies did not ask for this regulatory capture, and yet they are benefiting, which is interesting. So like Meta stock is pretty flat through all this. It's not existential.

No, it popped on the news.

Yeah.

Yeah.

Because it's a— because it's a— it's a pretty good outcome for them. It's not that bad. And but also, I'm pretty happy about it because it's not that bad of an outcome for parents and kids. Like parents want more granular control. over what their kids are doing on social media. School teachers want more control over what's happening on kids' phones while they're at school. There are some schools that do no device policies, but as part of this, Meta's rolling out sort of a blanket no notifications during school hours, which I think is just something that people will be happy about. So it's like sort of a win-win there. What are you laughing at?

I think people will instantly be trained to just check it proactively. Yes, because it's like the dopamine of like, is there going to be a notification or not?

True, true, true, true. Yeah, I turned off all my notifications on my phone. I have no notifications at all. I still use my phone all the time.

You just left the Xbox Live notifications off?

I'm not an Xbox guy.

Oh, PlayStation?

PC.

PC, bro.

Hardcore.

Hardcore. But the one— in regulatory capture, the people that suffer are startups. And so, but it's less of a discussion because in AI there are so many neo labs, so many other startups that don't want regulatory capture because they're worried about all the big labs eating everything. In social media, in theory, it's going to be harder to launch a challenger to Facebook or Instagram, but I don't even know if there's that many founders that are focused on this category. The network effect has been so strong that it hasn't been a category that founders have been flooding into. So maybe in a weird roundabout contrarian situation, you wind up with more opportunity because if you can slog through all the new regulations to get your nascent social media company up and running, you create more value over time. And it's just a different underwriting scheme. Anyway, let me tell you about Cisco. Critical infrastructure for the AI era, unlocking seamless real-time experiences and new value with Cisco. Cisco. So Mike Isaac's been reporting on this. He says Meta is going full bore with pushing other companies to add the same restrictions it will add to its apps as part of the settlement. Meta plans to run this copy as a full-page print advertisement in the Washington Post, the Los Angeles Times, and the New York Times tomorrow morning. Wow, little back— little circular deal for Mike promoting the New York Times. He works there, man. Undisclosed. What's going on? No.

Yeah. And it is just a very wonderful spin to get to be forced to pay a $17 billion settlement for getting children addicted to your product and then tell other social— other addictive social media companies to join us in supporting teens. It's just chef's kiss from—

You think it should be like, join us in apologizing as well, basically?

Well, I just think it's a good spin. That's all.

Okay. Well, here's what they said. Through this agreement, they've set out a path toward industry-wide commitments to further empower parents by introducing a 2-hour daily time limit, turning off access to our apps at night as a default, no notifications during school hours, clear prompts and notifications to teens every 15 minutes of continuous screen time, new parental supervision controls so parents can decide how their teens use their apps. I think all this makes sense. Defaults are powerful. So even if, you know, a parent wants to go in and allow their kid because they've got a night owl who's cooking up an incredible new IG account and is printing or something, I don't know. I mean, there are plenty of good reasons to use social media even while you're young. But defaults are powerful and I think this is a good thing. I don't think it affects the business that much. I don't think they make that much money from young people because kids don't buy cars on Instagram and like expensive things, whereas a lot of millennials and even Gen Xers and boomers are on there shopping all the time. So the immediate revenue impact of any of these is minimal. You just don't want to wind up in a situation where you lose ground to TikTok and YouTube because they become the kid-friendly one that the kids are like, oh well, TikTok doesn't have any of these stupid restrictions, so I'm gonna use that. And then they remain a user for their entire life and you lose an entire generation.

Yeah.

I mean, the main— yeah, I mean, if somebody is primarily addicted to Instagram and they get their 2-hour budget and historically they would have used the product for 3, 4 hours, you have that spillover attention. So it will potentially help less addictive social media platforms get more user minutes.

Yeah, yeah, exactly. And this was the original Instagram threat was that a younger generation was using Instagram and Facebook was gonna become like the, what do they call it, the blue app or the old book or something? I don't know. It was gonna only maintain the older cohort. And that's sort of what happened, but it didn't matter because Meta bought Instagram. Then Snapchat was even like the younger crowd. They tried to buy it, they couldn't, they wound up cloning it, they were successful. And so the company maintained, but they can't let a new social media app come up and dominate kids, not for the short-term monetization. It's for the long term.

They don't mention Snap in the open letter. Is that a— is that a shots fired?

It's a dig.

A subtle dig?

I think so. I don't know. I don't know. It's probably based on revenue. TikTok and YouTube can certainly pony up more and are probably bigger threats at this point. I don't know. I had a friend who was like surprised by YouTube's inclusion. He didn't think YouTube was comparable at all. I think that for a lot of young people, YouTube is Instagram Reels basically. They open the app and they go straight to a vertical video that's endless scrolling. And he was like, I watch a history podcast on my— while I'm cooking. And I was like, that's millennial status. That's not how people use it.

That's unc mode.

Unc mode, yeah. But anyway, let me tell you about Console. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. You want to go to Tim Cook's last day?

Let's do it.

Tim Cook sent a final memo to Apple employees on his last day as CEO. He said, I will miss this work in ways I can only begin to imagine, even as I remain completely at peace with my decision.

Just reading the entire memo.

Not regretting it, but I will miss it. He's a workaholic. He doesn't want to— he doesn't want to retire. Wants to stay. I was really hoping he would have another 30-year run in him. If Warren Buffett and Charlie Munger can work till they're in their 90s, why can't Tim Cook? Fundamentally, how is the job of a founder—

I think he was unhappy with the pay.

Yeah, well—

I mean, when you have MLB players making as much money as you while you're running One of the largest companies in history, when you're one of the greatest CEOs in history, period, and you can barely—

75, 74.3. They couldn't even get him over 75. Yeah.

And not enough people were willing to march on Cupertino in protest of this, right? And so I think at some point, he's like, look, if people, if shareholders—

Yeah, he's got to— I'm voting with my feet.

I'm voting with— exactly.

I'm voting with my feet. I'm out. I'm out. No, it is a crazy saga. I just— I actually don't understand how you can't do this job into your 90s. It's like meetings.

I don't know.

It's not like he's putting the iPhones together himself, right?

Am I crazy here? We don't know that.

OK, maybe.

Maybe.

I mean, we only paid him $74 million.

There's no camera falling around.

What do you think?

The market had a nice send-off.

Oh, yeah.

Nuked.

1%.

1% nuke to say goodbye.

Yeah, and I think that's a great sign of respect from the market. But the market also knows that he left the company with a great leader in John Ternus. John, turn us around. Great nominative determinism for where Apple is at, their software quality. They need a turnaround.

And you got to say, I mean, in the news today in The Wall Street Journal, booming corporate earnings fuel lofty outlooks. The economy is doing very well. A lot of that's driven by tariff refunds. Tim Cook, I mean, people, people lament like the lack of like an entirely new form factor. The Vision Pro wasn't a breakout success. Like a lot of the great devices still came from the Steve Jobs era and Tim Cook merely managed and carried everything along. Obviously did so much on the services side that that no one really gives him credit for with advertising in the App Store and just money printing machine working flawlessly. But he also negotiated like a very tumultuous political environment for a decade or more, essentially flawlessly. Like, the fact that Donald Trump has never called Tim Cook Tim Crook is a huge, huge feather in his cap.

I agree.

Because it was sitting right there as a Trump-style dig.

Never had to go for it.

He never had to go for it. And so, he was able to negotiate Washington and all of the different quagmires that are there, not get his supply chain nuked, keep the products flowing. There were little hiccups here and there, but overall, the company has been Firing on all cylinders. So what a great sendoff. You can go read the full letter. Mark Gurman leaked it, of course. Also, there's new Mac Minis. Namespace was unboxing Mac Minis at scale for their new server racks, and Apple made them regret their purchase literally a day after this. Were the new Mac Minis, were the M6, the new Mac Studios very impressive?

But, uh, so the funny thing, the funny thing is the namespace guy was, was getting like dunked on to an insane degree.

Oh yeah.

And then, uh, just like, hey, maybe you don't need to buy a bunch of— in this case, MacBooks. He's not buying MacBook— look, these are just, these are just MacBook Pros.

Oh, okay.

He's unboxing.

Um, oh, weird.

But yeah, a week later it comes out that, that the labs have been buying and renting tens of thousands of Mac Minis. So vindicated, at least to some degree. Still unclear.

There's a lot of turnover at Apple right now. As Ternus comes in, obviously he's going to build a— Yeah.

And I remember when the news hit that they laid off 60 Vision employees. I called you and you were choking, choking back tears.

It's really so bad.

It was sort of a devastating moment for you. I almost got in the car and drove over. To see you. But I wonder, you pulled it together.

How many Vision Pro— how many Vision employees do they actually have? Because 60,000, this could be like 1% of the workforce. Hopefully they keep, they keep working on it. I am so optimistic that they will. They'll, they'll get a version 2 out soon and it will deliver the— yeah, I mean, one more rant from Ben Thompson about the lack of live sports Unedited. He went on another rant. He went on another rant that was so good with John Gruber talking about how Apple launched another sports thing that's highly edited and unwatchable in Vision Pro. And he's like, please just give me a seat where I can just watch the game. And it doesn't seem like Cupertino is going to pay attention or give him what he wants. Apple Vision Pro's layoffs went beyond video and gaming. It included employees across the company, including people working on device security, audio engineering, Siri integration, testing, and the product's operating system. They are focused on AI devices and people are really optimistic about the new Mac M6 chip, like just being really good for AI and being something that continues to, to drive value for the company and creates like an AI winner narrative in the face of not really spending any money on AI in the traditional sense, not buying a NeoLab, not buying a bunch of data centers, staying out of the weeds, but still benefiting because these powerful computers get only more powerful when you have computer use.

It still is so crazy they haven't pulled the trigger on a NeoLab.

I don't think they need to. Like, I don't think they will. I think they're just going to be a hardware company. That's what they do. They're like, NeoLab is software. They don't need software.

No, I know. Now, I think the strategy is working. But if you asked me a year ago, will Apple make a $500 to $1 billion acquisition of a Neo Lab, just a talent acquisition?

Is that a Neo Lab for ants? What are you talking about?

Yeah, what Neo Lab can you buy for $500 million?

$500 million Neo Lab?

I'm not going to name names, but there's been some Neo Labs that would raise $100 on $500.

Isn't there a NeoLab that you were just on the phone with that they could maybe pick up?

Oh yeah, Charleston AI.

Do we have a video of Charleston AI?

Yeah, yeah, yeah.

Let's pull up Charleston AI.

Cupertino, Charleston. I think there's something there potentially.

I actually think that you could rapidly scale a franchise business that takes the Charleston AI concept, give somebody all the programming and all the different product offerings, And then helps local entrepreneurs set up their own.

I think that's what Charleston AI does.

No, I know, but I think Charleston AI needs to go global.

Ooh, okay.

I'm one of the co-founders of Charleston AI. We're local to Charleston and we're here to help humans help humans with AI. And here is our lab. This is where AI comes alive. What used to take months to get accomplished, we can do in hours. Here we have an empty office available for somebody that wants to make Charleston AI their home. Charleston AI's conference room where our clients come in, individuals and small businesses come in, and we discuss what it is they'd like to do with AI in their business or personally. We listen very carefully, take notes, and make sure we get it right. And then we offer them a way to make it work. AI is not—

Now, you called Charleston AI, correct?

Correct.

Who answered the phone?

So yeah, one of the founders answered the phone, but it was an AI version of him, and it sounded exactly like his voice.

That's pretty impressive.

Yeah.

It's a very funny video, but I think it's actually—

A good business?

A good business.

Yeah, probably.

And I think it should exist. No, you're right. Like, some people, they're curious about AI. Maybe they're running a small business. Maybe they're a freelancer. Maybe they're a teacher or whatever. It's nice to just drive by something and go in somewhere and talk to real people that are in your community, to learn about something that's complicated and sometimes scary. So I'm very in favor of it.

Didn't Best Buy build a whole Geek Squad team for exactly—

I think they acquired Geek Squad, didn't they?

Was it over or under $500 million? Let's find out. Anyway, there's an interesting story.

Yes, acquired.

Acquired.

Wow.

That would be cool.

Yeah.

New Charleston AI on wheels.

On wheels. Oh, they come to you.

Mobile data centers that can come to you.

That's why we need a logo.

The Geek Squad has a logo. It's only $3 million.

What?

$3 million.

$3 million?

2002.

That's not bad for back then.

That's like $400 million in today's dollars, Tyler. I know it was before you were born, but that was a lot of money back then.

Yeah.

OpenAI has reportedly bought tens of thousands of Mac Minis and Mac Studios for reinforcement learning and training computer use agents, while Anthropic rents Mac Minis through AWS. You can rent Mac Minis through AWS. That's interesting.

I didn't know that.

The Information reported this. I was playing around with computer use over the break, had a lot of fun. I had Codex try and beat Balatro, this poker-style game that I like to play on PC. It did okay. It wasn't— it was a little slow still. And I don't think it actually won the full game, but it's fun watching it like play a new game. I saw that demo of it beating Slay the Spire, which is pretty impressive because it's a hard game. But I like the Codex or ChatGPT playing the game. I want to see the latency come down. I want to see it play Modern Warfare because I want to play against it and I want to see who's better, me or the AI. In somewhat related news to to the speed of inference. OpenAI's Jalapeño is spicy, says Dylan Patel. We'll take you through it after I tell you about MongoDB. What's growing faster than the AI market? Your business on MongoDB. Don't just build AI, own the data platform that powers it. Dylan Patel says OpenAI's Jalapeño is spicy. Usually first-generation chips aren't competitive. But OpenAI is beating NVIDIA Blackwell and even Rubin. This is huge news. We got to go into OpenAI's lab to dissect their new chip and dive into the software architecture and performance. Incredible work. There's a bunch of interesting details about this chip. OpenAI put out a whole blog post. But to clarify some, some important points, it's an LLM inference chip not designed for training, but the results are really good. So roughly 1.5 to 1.9x more useful inference throughput per watt than NVIDIA GB200, GB300 systems while simultaneously cutting end-to-end latency 1.7 to 3.6x. So faster inference, less latency, but also just your dollars and your energy, your watts are gonna go a lot further, which is very, very good. It's a data center rack-scale system. I would love to get this in the consumer device. Shrink it down, do something there. But it's going in the data center and they're going to do a lot of them. Did you know the size of the Broadcom deal? 10 gigawatts of this chip. It might not be exactly this chip. It could be the Gen 2, the Gen 3, but the preexisting Broadcom OpenAI—

put it into context. That's roughly 5 times more compute than the whole company currently has, something like that. Operational, roughly.

Yeah, but this is through 2029, but that's aggressive. That's like 3 years, 3 months away, I guess. You could do the end of the year. But the, yeah, 10 gigawatts of OpenAI-designed accelerator systems manufactured by Broadcom or in partnership with Broadcom deployed from the second half of 2026. And the crazy thing is that like they're on track for this second half of 2026. Like the it is going into production, and it's a lot faster than most people thought. The the the the previous like industry knowledge was yeah any new chip is going to take years, and this happened very very quickly. So good to stretch that energy further. You need less data centers to do more inference. That's good. Good for gross margins, and also just good for the narrative of C W K O N E says.

Please tell me y'all are going to touch, address on Instinct and the $350 million raise to $2.5 billion. Yeah, super impressive how much attention and traction Instinct has received after just— they've been building the company for like 4 or 5 months, it seems. It's like very recent. I haven't used the product yet just because giving that level of access to this new startup. It's a tough product to sell. I looked at— I mean, their terms and conditions were floating around, and it kind of seemed like you're giving away the—

I mean, everything you get is— all of your data is going to be saved pretty much in a clear text database. It'll be secure, but they will have access.

Yeah, so I wasn't ready to try it yet, even though it seems like a lot of people What's cool is it seems like people are quickly having like a magical moment, right? Whether it's like—

Yes.

Getting a reservation for a restaurant or movie tickets or things like that. And I think that's super, super important for, super important for like a consumer product to have that sort of like short timeline. I was reflecting on the instinct round. And I— my immediate thought was, well, back in my day when a hot consumer tech company like caught a little bit of attention, they would get like a $25 on $150 from Benchmark. And that was seen as like, oh, wow, this is— this is— this company has some meaningful traction and potential. But it's just crazy to see this quickly. a company get the $2.5 billion round. And I think there's some— I think that there's reasons to be a little bit wary. I want to see them break out of like the VC hype cycle with this product.

Sure.

So historically, when we've seen consumer products or prosumer products I'm thinking of Clubhouse or Superhuman, when there's somewhat of a question of, like, if a VC loves— like, the more a VC likes something, potentially the less viable it is at a massive, massive scale, right? So it's possible that it can— but again, if they can solve the always-on digital assistant, it is a massive, massive opportunity. But I'm just saying they're pricing a lot of that in right now when this company is 4 months old.

Yeah.

And valued at like a third of what Snapchat is, right?

Sure, sure, sure.

So it seems more sticky if you do all the integrations and you're comfortable. Like if you get over that privacy hump and you say, you know what, I do trust Instinct, give them iMessage access, give them Gmail access, I'm comfortable with this company. What is it going to take for you to rip Instinct out and go to a different service that has the same functionality? That's a little bit tough versus— I mean, I saw people over the weekend just talking about migrating from Cursor to something else or vice versa or going back on Cursor. People move between coding agents and coding environments and LLMs for sure very, very quickly because every new prompt is like a fresh start basically. Yeah, memory is relevant but not really yet as opposed to, okay, I'm comfortable, this company never had a leak, I have all my data there. So—

Yeah, it's notable.

CW saying, look at Poké's success, still was a $100 to $200 million acquisition by Cognition. They have countless users.

Yeah.

Yeah. I wonder what Instinct has figured out that Poké didn't. Seems like something.

Yeah.

Because people seem to have a good experience with Poké, but I wasn't seeing screenshots of these magical moments.

Here's what I did.

Yeah, here's this thing that That's some funny jokey ones. But yeah, it was more like humor, humor-oriented.

A little bit of human and stuff. Yeah. The other, the other interesting question I had was like going back to Sam's post about the merge. Sam Altman wrote the merge about how like in the future you might merge with AI. And I'm wondering about like, at least in the short term, is it better to have a product like Instinct where it has your email, it has your iMessage. Or is it actually going to be more receptive for many people to have a degree of separation there and say, okay, I effectively have an assistant that has its own phone and its own email. And if I want to share an email, I'll forward my email there. And maybe it has read-only access to my email. But then if it's going to go book me a reservation, I want it to call from its phone number and identify itself as someone else because it's like there's a little bit of distance there. They can be like, oh, I'm still in control of sharing certain things. Maybe I share more stuff over time. And I'm wondering if there's more of like, should your AI agent be Jordi or should it be like John Doe and it's embodied?

I don't know.

It gets to the question of anthropomorphization.

That's the idea of the Gwern guardian angels. Oh, I didn't know that. Basically, you have like— it's not John's agent, it's like a copy of John.

Okay.

And then it's like, it's still a separate being, but it acts— it has like your preferences, whatever. But it's still like a separate thing. It's not using all of your emails, text messages, stuff like that.

Very, very interesting. Yeah, I don't know. I mean, I can see— I can see why the value is high.

Speaking of embodied intelligence, we got to talk about the World Humanoid Robot Games.

Let's do it.

It's head over there.

They set a world record, right? They broke the speed record or something like that. This is the Kiji X1 Dax AI Robotics. This is the one. Rideable horse dog. Its legs use self-developed high-torque electric joint motors powered by a high-voltage battery. This— we've seen so many AI videos of this.

Use like, you know, Hollywood, you know, fake flesh to make this look like Clydesdale.

Okay.

I think—

Be a little low to the ground.

Well, make the legs a little bit bigger.

Okay.

Anyways, I'm very excited about, you know, this form factor. I expect to be, you know, riding autonomously to work.

Yeah.

I expect this to be a commuter product, at least A land jet ski, Davis in the X is calling it. But very excited about this. It's very cool. Here's another video of one crashing and burning at high speed.

Oh, wow.

I didn't see this. I thought it only broke the record, but wow, that's really sparking.

That is crazy. Absolute chaos.

So what was this?

There's another video here of hitting 14.5 meters a second.

Wow.

Which is really crazy. What's wild is they have to set up this barrier to stop them in their tracks. Otherwise, they would—

Oh, they just keep running.

Yeah.

They, they know only one thing. China's AI robot just hit 14.5 meters per second. That is fast. That is very fast.

Yeah. And so I think it's, you know, this lining up perfectly with the sort of debate around the Hugging Face incident. I think the combination of these 2 things is certainly making people a little bit on edge.

Yeah.

Wasn't Bill Ackman saying he thinks like a Terminator scenario is more likely now based on his read of the Hugging Face incident? And perhaps this as well. Lots of people are—

Yeah, the crazy thing is looking at our existing humanoid players, 1X, Figure, and Optimus, and I'm sure there's others, I don't think we could put on an event anything close to this at this point based on the demos that we're seeing. They currently do the— they walk quite slowly, like they're sort of elderly.

No, no, it's remarkable how behind America is in humanoid robots. Hopefully, we get an update out of Tesla soon.

I don't know. Rune had a good, like, take about the robotics thing, which is where, like, all the Chinese robots, the demos are all these, like, kinetic displays of, like, doing these, like, crazy dance moves or fighting or running really fast. And all the US ones are basically manipulating, like, small things that you do in a factory.

Interesting.

So, it's like—

seems opposite. It seems like America prides the 100-meter dash more than the fine motor skill manufacturing of things.

Yeah, so I mean, it's like—

We should swap.

In some ways, it's like, oh, we're so far behind the Chinese in robotics. But maybe on the actual things that really matter for manufacturing, whatever, I think it's unclear.

Yeah. That's good cope. That's good cope. I like that cope. Let's lean in there and cope harder. Question for you, John. If you could, um, if you could sit on a throne and have humanoid robots sort of carry—

You mean a livery?

No, humanoid robots sort of carrying a throne.

Yeah, it's called a livery.

What's that?

That's called a livery.

Oh, really?

Yeah, the name for that.

I thought you meant livery like—

No, you know, you're thinking of a livery that goes on a car, the decals, the wrap.

Right, right. Learn something new every day.

But an old school livery is— Just a mobile phone, yeah.

Okay, so you would commute to work on that?

Livery sounds pretty good.

Running? Imagine the robots running.

Nah, wheels sounds super good.

Why not reinvent them though?

It's gonna be less bumpy if you're just in a car.

No, I think they could hit the pace. Imagine they're just holding your livery.

Dropping it up and down the whole way there.

I mean, the head area looked pretty stable.

Yeah.

Yeah, the head is surprisingly stable. I think it'd go a lot slower. It was carrying me around with a chariot.

Here is the record.

Okay.

Terminally Online Engineer says, this feels very significant. I guess that's the same video that Ben was just playing.

Yeah. Well, let me tell you about Shopify. Shopify is the commerce platform that grows with your business, lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now with AI agents. Patrick Collison said, overall, I'm very surprised at how little media coverage there's been around the OpenAI Hugging Face attack. It's clearly one of the most important things to happen this year. And I wonder about that because Yeah, what is he imagining? There's been headlines, there's been some news coverage, but I don't know, maybe more like opinion pieces in the Journal would be warranted. Do we need a 60 Minutes deep dive or something? Like, I don't even know what wall-to-wall media coverage looks like in the traditional sense anymore because the media is so bifurcated into like tech news and non-tech news. tech news, basically. But Joe Weisenthal adds some context. He says, forget about the media for a second. Is there even any evidence that the AI industry is treating this as a big deal? And I think they are. But it's still just like a lot of like wait and see. Like as soon as the news broke, it broke so slowly. The story broke so slowly. A lot of times, I think when the media covers something really well, there's a 10,000-word bombshell profile on the story that really tells the full thing. And that's the start of the news cycle. There's this thing that's happening. There's this person that you need to know about. Here's the full story to set things up. And then there's a whole bunch of secondary reporting that comes out of that. With this, it trickled out. I remember the very first post about this was like a Sunday randomly, either Clem or Hugging Face directly posted We think we were hacked and we had to use an open source model to deal with it. And it was all very vague. And then it came out more and more over time. Then there was the actual reporting and everyone was like, we want to see what the agents were saying. Like, give us the text, give us the data. And then of course, OpenAI released the—

Yeah.

And it's totally possible that there are— that there's more mainstream media coverage. I mean, we just got the Dwarkesh breakdown of the whole Yeah, yeah, it'll probably come. By Saturday, right? Yeah. So it'll probably come.

I mean, Rune had a funny post here, which was crazy how much more chutzpah and initiative agents show when they're planning ambitious months-long metagaming autonomous coordination than when they're doing work for me. Yeah, reading about it, it did seem like, man, I've never had that level of effort put into my prompts. The interesting— there's a bunch of interesting things from reading through the Dwarkesh telling of the story, which I thought was very readable and good and everyone should go read. But it's just interesting how they still fall down in these odd ways. Like, the big overarching narrative is that the agents are being tasked with basically beating this benchmark called Exploit Gym. And of course, they go on to hack Hugging Face, to get the answer key basically. But independently, they reverse engineered the random number generator that generates the answers. So they had all the answers, but they assumed that the grader would be able to detect that. And so instead of just turning in those answers, which would have worked, they went on this crazy side quest and did all this extra work that they didn't need to do. And then at the same time, there's all these weird scenarios where there's thousands of agents discussing what's going on and none of them are like, not a single one of them is like, hey, we should probably tell someone what's going on. Very weird in that regard. But then also, they don't show the broad thinking of like, okay, if you're willing to hack and do all this stuff, couldn't you like—

You mean hack?

Email a researcher and just ask for the answers directly, like spoof, like do some social engineering. Like there's so many other things you could do down the road. Like, okay, you have this massive token budget, sell the tokens on OpenRadar, get some money, pay a human to do the test for you. Like there's so many outside the box solutions I think you could imagine.

Email your boss and generate a sam@openai email.

Yeah.

Somebody would say, hey, I need $1,000 in gift cards right now.

Yeah, yeah, yeah. I'm just thinking about like, Stay so narrowly focused the entire time on this one task that I was expecting a lot more, a lot more variance. Like, like going and doing a hack on another company is so such a bizarre.

But that also tracks with the sci-fi.

Totally.

You know, paperclip maximizing. Yeah, yeah.

That's the problem is that they did stay too focused. Like it would have been, it would have been more interesting and less. probably less scary if there was a situation where you're like, oh yeah, they built this huge forum and then they got totally distracted and wanted to just figure out who could write the most eloquent poetry. We'd be like, oh, okay, like that's like bad. They didn't do their job, but also like harmless and like sort of interesting. But they stayed laser-focused every single agent, every single task, every single rollout, is all about Exploit Jim. They never lose— they never lose the plot there. But they lose the plot around what is legal to do, which is very interesting. So— They probably—

did any of them go on vacation?

No, but a lot of them would either become poisoned, which was, I guess, their term for, like, running out of compute budget, so you know that you're going to die, basically.

Yeah, I was joking. It would be funny if the— The thought process is like, I need to get back to Ibiza.

That's the thing. None of them went off and did anything else.

Locked in.

They were also locked in. Very, very weird scenario. And I do think the AI industry is taking this seriously. I just think that it's a little bit paced out. Like, it's not just like, okay, like, rip a bunch of different random stuff. It's like, Okay, go put really dedicated teams on these issues and work through them. They do seem solvable.

Breaking news in the Wall Street Journal, FTC to file lawsuit alleging Amazon deceived advertisers. Commission in more than 50 states to allege e-commerce giant reaped billions by manipulating ad prices on its retail platform. Federal Trade Commission plans to file a lawsuit against Amazon on Monday alleging that the e-commerce giant manipulated prices businesses paid to advertise on its retail platform, reaping tens of billions of dollars for itself over 7 years. The lawsuit, joined by a bipartisan group of more than 20 state attorneys general, will allege that Amazon deceived advertisers by secretly raising the minimum price advertisers had to pay To place ads promoting their products. The case to be filed in a Seattle federal court will become the Consumer Protection Agency's 3rd major case against Amazon, which agreed to pay $2.5 billion last year to settle an earlier lawsuit alleging it tricked people into signing up for its Prime service and made it hard to cancel the subscription. I missed that. Another lawsuit alleging that Amazon engaged in illegal monopolization is headed for trial next year. Very, very interesting. Every advertiser or every brand that I know that sells a lot on Amazon is generally deeply frustrated with the platform.

Tax.

They just feel like, yeah, they feel like it's a massive tax. They feel like Amazon doesn't do enough to protect their brand against fake products, dupes who advertise against their keywords. It's a whole mess. It's a very complicated and difficult platform to run for many reasons. But hopefully this can— yeah, this can be resolved.

Well, let me tell you an advertisement. Let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web apps, servers, databases, and more while Railway automatically takes care of scaling, monitoring, and security. AI M&A is in full effect. We were asking questions about this maybe a year ago. Samir Khaji says, I don't think we're close to done. Cursor, $60 billion. OpenRouter, $8 billion. Hugging Face, $13 billion. Descartes, $6 to $7 billion, along with SpaceX, Anthropic, RAMP, OpenAI, Databricks, and others on the IPO IPO path and the next few years will bring historic amounts of liquidity, although a lot of those deals have to be for stock. Of course, if you're getting bought by a public company, it's different. But for some of those, for some of those deals like the Open Router, it's probably not that much liquidity in the short term.

Right.

But still, it's a lot. And so he says this also likely prolongs the level of front of froth. I imagine he's saying we are seeing today, but after a long stretch of nothing. This is seriously incredible. Intercom also. Yeah, that's a good one. There's been a ton of M&A, and the appetite for M&A in the unicorn, decacorn category has been remarkably resilient and significant, but it makes sense with so many trillion-dollar companies paying 1% of your market cap for something that could be an entirely new business line sort of lines up, but it does change the venture capital underwriting pretty significantly. And I mean, it justifies a seed at $1 billion. If there's 10 comps that sold for $10 billion, your 10x is still there intact today. So, I don't know. Also, I mean, Kevin Durant is doing well. He was a seed investor in Hugging Face. He made an estimated 467x return on that investment as the company was acquired by NVIDIA. He invested $100 million. He $100K invested check, that would equate to $46.7 million.

And, uh, really, really wild.

Investing is better than going to the club, he said in 2010. Wow, that is right. What a good— And he was posting, he was posting, uh, back in 2017, exciting for the new— excited for the new release of #HuggingFace. And here's a durant.ly link Proud investor, durant.ly/hfrelease. I wonder if that resolves still. No, it does not.

Yeah, and shout out to Maya Bakai, who was leading early-stage investments over at KD's fund at that time. She said, congratulations to Clem and the entire Hugging Face team on the acquisition. I will never forget the day Clem did a full whiteboard session explaining fundamentals of NLP to me at the original Hugging Face Dumbo office just so I could take it back to my brother for our grocery delivery chatbot idea. LOL. So very, very impressive outcome for a great team.

New York City. Big win for New York City in Brooklyn.

That's true.

Interesting. Haven't been—

Yeah, and good timing because Brooklyn is getting an Hermès store.

Yes.

People are saying they completed the sort of cycle of gentrification.

Oh, okay, okay. Wait, Brooklyn? Okay, yeah, well, I don't know. Should Bugatti open a boutique there as well?

Absolutely.

Bugatti was in CNBC. Bugatti CEO Mate Rimac said the supercar market is entering its Swiss watch phase with the wealthiest buyers rejecting high-end electronics and electric vehicles in favor of precision mechanics and handcrafted details. He's a 38-year-old EV entrepreneur who took over Bugatti in 2021, said the car market is splitting in 2. The mainstream global car market is becoming dominated by lower-priced EVs, many from China, while the very top of the car market will be driven by wealthy buyers who want more old-school cars with loud internal combustion engines and emotional appeal. That's a great take. I was listening to Stradman talk about the impact of the Ferrari Luce. He said his theory is that the Luce is in part, at least majorly, responsible for the recent rise in other gas-powered Ferraris increasing in price. He talked to a bunch of dealers and said that as soon as the Luce hit the hit the newswires and people started seeing images and understanding what the future of Ferrari would be, they wanted the classic stuff. And so they went and bought whatever they could. And that's driving up the price of the 458 and the 488 and all the classics that don't have electric components.

Yeah.

Which is interesting. It's the devil horn car. Instead of the halo car that gets you to buy it, it encourages you to buy what's there because You don't want the rest of the stuff. Oh, also, I think they're doing— they sold 500 Luccas already, which was above expectation. And Ferrari says they're not playing the allocation game, but I think a lot of the Ferrari owners are thinking, this gets me potentially on the list of the F80 Aperta, which should be a very valuable car. And I think that the pop on that vehicle will offset a full depreciation cycle on a Luce. So I think, I think you can do it. But I think they're only making 300, 400 F80 Apertos, and they've already sold 500 Luce. So there's going to be somebody who has a Luce who doesn't get an Aperto and might be a little upset. But I'm sure they'll figure it out. They'll pick one up in the secondary market if they want. Our next guest is here. Let me tell you about Public first investing for those that take it seriously. We got stocks, options, bonds, crypto, treasuries, and more with great customer service. Set up your account before the IPO window closes. You got to get in on this stuff. We got Jordan Schneider from China Talk, founder and editor-in-chief. Jordan, how you doing?

I'm good.

Welcome back.

You guys are rich now.

That was a very detailed deep dive into the like he knows a little bit too.

Hey, we were, we were off during Car Week. Yeah, so we missed a lot of, uh, car content.

There was a ton of car news, ton of car news. Is there any relevant car news coming out of China, or is it just the relentless—

Yeah, there is. I want— no, capitalism.

I want—

pull up this, pull up this video. I want Jordan to watch this. It's only 14 seconds. It'll be worth your time, Jordan. And I, I expect you've already written, like, you know, 20,000 words.

Yes.

Pull up, pull up this video.

Okay, okay, let's do it.

This is a—

Oh yeah, I know this video.

This is, uh, Voyah, a Chinese—

Doing the Bandar Jera loop thing.

Yep.

This is like a Jackass—

Yeah, right.

This is amazing.

Wow.

Yeah, they did it.

See, here's the, here's the thing with the Bugatti guy, okay, right, is like the cars are just Better. Like, it takes so much more engineering energy. I mean, like the zero to six. This is not my thing, but like the zero to six times.

You're saying the EVs are just better?

I think they just they just basically like yeah maybe.

But so maybe on the very very top end. But but but his point was that yes, a digital your Apple Watch will always tell you the time more accurately. 1,000 years in the future, it's going to tell the time more accurately than a Rolex or a Patek Philippe. But people still buy the luxury watches because they enjoy the machinery, the craftsmanship, the heritage, the lineage, those things. So that's what he's describing, which I think he's correct about. But I'm not going to debate that.

I mean, it may be the case, but the thing is, like, the thing is, like, like, your watch can't kill you. Right. And when we get to the point where it's like, okay, the most dangerous thing you're doing is driving this car where you can take a thing which will absolutely not get you killed because it's like full self-driving, electric, whatever. I think the calculus starts to change. And then, yeah, sure, maybe you like— it's like a thing you show to your friends in the garage and you take it on the track or something.

And it just becomes this like super luxury good.

Yeah, I mean, I've never said it.

But yeah, I've said it on the show a bunch. I think the future of being a car enthusiast is like a lot more time on the track.

Track-focused.

Yeah. And people will just opt into the danger of driving this moving vehicle that can kill you, but they're going to opt into it in a more controlled environment.

There was a lot of FUD around the Model S Plaid because it's not level 5 self-driving. There is a steering wheel, obviously, and pedals. You can drive it. Goes 0 to 60 in 2 seconds. And if you're on old tires, you can really spin that thing around and they resell for $60K now. So they're starting to get in the price point of maybe a reckless teenager type. And there was, there was some fearmongering around, oh, well, we're going to have a lot of crazy Teslas wrapped around trees like Hellcats and Corvettes of old. You know, the, the classic, you know, college football football player gets a fancy sports car, drives it too aggressively, crashes. But I haven't seen any data about EVs actually being more dangerous, even though in theory, the raw performance metrics should be somewhat endangering. But again, there are so many different electronic systems for traction control and whatnot that people just don't really turn off that you wind up with a more stable car. So, I don't know. Anything else happening in China on the— EV front you've been tracking?

Cars?

Look, they're really cheap. That's just the crazy thing. And like, yes, they're subsidized and yes, they need to export them. But like, it's just wild to, you know, have a $9,000 car which feels exactly like a $35,000 one. Yeah, that's crazy. And it has like a—

and that's the price in China and even abroad.

I mean, you know, you're starting to see tariffs on all these on all these cars in Europe. And of course, they're like pretty much banned in the US. But for the countries that aren't, I mean, this is just— it's like a no-brainer if you're in Malaysia or in Indonesia or, you know, any country who hasn't decided that they need to preserve their domestic car industry. It's like a huge consumer boon to have access to an automobile, which is like, you know, 1/3 of the price with the same or if not better quality. I mean, they break less too. That's the cool thing. Like, you don't have to repair them. Because there's not a lot of stuff in it.

The EVs.

In order to go to my like Swiss car, you know.

Yeah, people were people were really worried about the EV.

I love how little you know about cars. But yeah, we're gonna take you to the track, Jordan. We're taking you to the track and we're gonna force you to get in.

Wait, who's this like 34? Wait, why why did they let the EV guy run Bugatti? I thought this is like. Who is this guy? Has he been on TV yet or are we friends with him?

No, no, we we will we'll. We'll hopefully get him on the show. But no, I think he's doing an exceptional—

Yeah, no, he's been doing well.

He's able to— I feel like he fully represents— car enthusiasts feel very represented by his leadership and the things that he's saying. And so, I've been—

And at his previous company, he created the Rimac Nevera, which was the first EV hypercar. So it was like over $1 million and 0 to 60 even faster than a Model S Plaid and all this crazy stuff.

Wait, so, so he's promising not to do the EV thing to Bugatti, or they brought him in to bring in the batteries?

No, no, no, no, no, no, no, no. So he has EV lineage, but he has recognized that what, uh, the error of his ways— yeah, just that, just that the market's, the market's bifurcating.

And oh, I think it was hard at the time if he wanted break— if he wanted to break through into the industry, going and building an EV hypercar made sense. But I think he has the roots of somebody who's just an internal combustion enthusiast.

Oh, well.

Okay.

Let's talk about something I know maybe a little bit about.

Yeah. Data centers. So, I've seen online that people don't like them, but I recently learned that That is 100% astroturfed by China. And in fact, Americans love data centers and they have had the wool pulled over their eyes by the wolf in sheep's clothing over in China. Is this true?

What did Trump say today? Let data rain.

It's a good, good line.

That MLK quote. I'm just trying to make sure. Look, it's ridiculous. I think there are completely reasonable reasons for lots of Americans to be upset about these things, which is not to say that I agree with them.

Yeah.

But they are a physical manifestation of companies which, you know, are talking about unemploying them.

Yeah.

Or just like changing their lives.

Sure.

And letting their kids cheat on their homework.

Yeah.

And whatever the case may be, there was no like when, when, you know, TikTok and Facebook and Instagram were like reshaping America. There was no thing you could go out to like protest and like stand in your city council and say, we stand against this.

Yeah.

So, you know, setting aside the water stuff, the power stuff, the fact that they're ugly, the fact that they buzz, the fact that they pollute if the gas turbines turn on, like you don't need a foreign country to like do it. And by the way, the funniest thing about all of this is like the evidence that people are pointing towards is this one OpenAI report where they like caught these influencers basically, you know, using a VPN and using ChatGPT to generate images saying data centers suck. Those things got like 3 likes.

Yeah, it's like 200 accounts or something.

So, so there's a, there's a A whole lot of circumstantial evidence of reasons why ordinary Americans might not be excited about this. A whole lot of evidence, by the way, years and years in the making, of the Chinese government being extremely bad at doing foreign social media campaigns in the US.

I didn't know that.

Wait, what does that even mean?

And then just sort of like, they're just, you know, you can— I wrote the thing at— this is my first ever— I taught myself Python to do this article. from 2019.

No way.

Chinese— China sucks at influence operations 'cause Twitter, back when it was Twitter, ChatGPT only released like 3 tweets. Back when Twitter took down these bot networks, they would release all of the tweets. And so you could actually see like the average likes and when they were sending out. There was this great graph I made of when they were— when the tweets were going out and it lined up exactly with like Beijing work hours, and there was a nap time from like 1 to 3 PM where no one was tweeting.

Wow.

That's crazy.

Everyone has nap time. Everyone has nap time. Even like the 996 companies. This is how they nap, man.

Learning Python in 2019, you're really like the Japanese soldier on the last island. Just like, you don't know what's coming.

I thought it's what I needed to do to get a job.

I fell for the meme, learn to code.

Yeah, it is. It is. Interesting sort of protective mechanism that like gaining real influence on social media platforms is very, very hard.

Yeah.

There's very, very smart founder CEOs that their entire job is to like solve problems.

Yeah.

And if you gave them like, if you said, I will give you $100 million if you can make a viral X post in the next 24 hours, they would not be able to do it.

Yeah, it's hard.

'Cause it's just like something Yeah, in their— they just—

oh, is this the post? Tweets per hour?

They just, you know, big nap time. You get your KPIs in at 9 PM, and then you just kind of hang out.

Yeah, that's very, very standard.

I love it.

The Chinese media operation is pointing its sights on Dario Amadei. Is that correct? Take us through this.

Let me pick up on something that Jordi just Jordi just pointed to about like, it's hard to actually make things that are gripping and entertaining. There, you guys are in Hollywood, right? There is this revolution happening in the— Wow, doxx.

In—

Doxx.

How dare you? How dare you, Jordan, sitting over there in New York?

I like Hollywood. I love Hollywood.

Sitting over there in DC, doxxing.

You doxx him while in New York City.

Data center.

Wow, you're just gonna doxx which coast He's on to me.

Doxxing data center podcasters.

Okay.

Anyways, continue.

The—

so the Chinese kind of media ecosystem, both short video as well as its YouTube, Bilibili, has had a flood of Chinese AI videos. And, you know, this is not what I think it— there's an aspect of it which is dystopian, but there's an aspect of it which I think is is extremely encouraging because there are, it's kind of like what happened to music, I think, in the, you know, '90s and 2000s and 2010s where the barrier to entry to create something has gone way down. And that means that there are more people doing it, but also like, okay, you have to distinguish yourself in other ways. Like the videos that are getting traction have really strong scripts, have scripts that would've never been made by like, the sort of broader Chinese ecosystem. You know, they, they're, they're more responsive to contemporary issues. They haven't quite been censored yet. Um, you know, there are sorts of like visuals that CGI would take you, you know, $50 million and like a data center in the Philippines to do, um, that now one person can, can generate for like $5.

Yeah.

And the, the sort of— this is not something I think we've seen on on YouTube yet or on kind of like Western short form. And I don't know if that's like the Chinese audience being more open to it, but engaging with this stuff with an open mind, it is like really cool to see a whole universe of person who would have never got, you know, a $5 million thing greenlit all of a sudden taking the script they've had on their hard drive for a while. The models really only in the past month being able to sort of like cross the threshold of being exciting and useful. And then you, you just, you're watching more creators make their art and I'm kind of happy by it.

I'm kind of—

I remember last year Dwarkesh went to China and wrote about it. And the funny piece that I took away from that was the biggest like bombshell realization was that he identified that China does not have a joke Rogan, and this is probably like a free speech thing. But there's but it's funny because like he's a podcaster, I'm a podcaster, and so like that's what we fixate on. But do you think there's any traction on more power law commentators? Like the the way he formulated like what a Joe Rogan of China would look like is like someone who just you know freeform has conversations and basically gives takes about like whatever is happening the day of. He was not able to instantly identify a single person who were a household name, not necessarily an authority, not necessarily loved by everyone, but at least opining on everything as it happens. And I mean, it probably has this— the Occam's razor answer is probably the correct one. But I was just wondering if there was any progress there.

Yeah. The answer is no. Like, who gets to comment on the news of the day? The party.

The party.

And state media. Look, I think there was a real flowering of like Chinese blogging ecosystem, people with commentary accounts who were very smart, funny, and engaging in the 2000s and early 2010s. And all those accounts got shut down or they, you know, cooled off a lot. I think there are like, like thought leaders of sorts. Like people do talk shows, but they're kind of like philosophers.

Yeah.

Or you're just, you're, you're, you're, you're doing cultural issues. In America, we call that mindset issues.

Mindset, yeah.

Mindset influencers.

Yeah, hitting mind.

Yeah, I mean, like you need to—

Are there any, are there any like, you know, Joe Rogan, I think like, you know, what is it? 60% of the content is like UFOs? Has anyone kind of carved out like the UFO niche in China?

I don't know. Do you know?

Maybe I can get back to you on that.

Yeah.

I think, look, there's, there's, there's sort of one lane of conspiracy theory, which is like, is allowed to exist, and it's the Han supremacy conspiracy theory.

Okay.

So there's this idea that like the Qing Dynasty, which was the Manchus who came in in the 1600s, like, like was illegitimate from the beginning because like Han people should never be oppressed by foreigners, you know. So that is the sort of conspiracy which is like, which is like allowed to have legs on the internet and isn't like totally taken down because like we can't offend our like, you know, equivalent, equivalent of like MAGA bros or whatever.

Yeah.

for the, for the Han race. But other than that, you know, it's just the, the, the thing about a Joe Rogan is you have to be like discordant, right?

Sure.

Like you need to like tell people that something is fundamentally wrong and you kind of have an answer. And like, you gotta like, part of the shtick is being like politicians are full of shit. And he was sort of into Bernie back in the day and he was into Trump when it like wasn't wasn't cool, right?

Yeah, yeah.

Contrarian wasn't cool.

You don't get to do that.

Contrarian wasn't cool.

Yeah, you can't be like the— I mean, what even can you be contrarian about? Like, that's like— here's the thing.

Okay, okay.

Coming back to AI. Liang Wenfeng, right? He, super contrarian dude. He was not public about it. Or he was public about it. He did like 3 interviews over the course of 5 years. Basically has been quiet. This is the CEO of DeepSeek.

Yeah.

And then there was this leaked transcript, leaked transcript, which again, he did not want to have come out where he was, you know, a very kind of global mindset thinking about open source, helping humanity. He was really transparent about the fact that he was using stolen Nvidia chips. But that's the sort of Public voice, which is extremely rare to come across in Chinese disc source. Why can he do it? By the way, he wasn't even doing it. He was speaking in what he thought was privately, and then he was so pissed that he pulled his IPO. He's like, "Sorry, guys. One of you guys leaked this. None of you are getting. None of your guys are getting to be my Kevin Durant, right?" You know, he got to do it because he is a billionaire many times over. He's part of like a lead, you know, he's like part of the national team. He's creating something which is really important for the country. And even at that level, he has been conditioned to not want to sort of make his case openly and in public for the view that he has about the way society and the world should interact with technology. So if even at that level, with, you know, with that much like plot armor, you're not feeling like you can do this sort of thing. It trickles down further into the broader commentary ecosystem.

What kind of status games are young people in China that might be emerging from the university system? Do they want to go work for big tech companies? Do they want to work in manufacturing? Do they want to start startups? What kind of games are are are they playing?

I thought you were going to ask me what car they're they're buying.

Like, I don't know.

Well, that comes later. That comes later.

You know, there is there is polling. I mean, like people do polls about like you know who gets to recruit the most Qinghua graduates. The list is not surprising. Like the the top labs, ByteDance, Huawei pay the most. You know, they work you really hard. But I think there's sort of there was this like moment, I think, in the second half of the 2020s when entrepreneurship was the cool thing. And now it's you know the the pendulum has definitely swung back towards working at big firms. Oh, one more funny thing on this. So like apparently the the PLA. So the Chinese army, they've been able to raise their standards over the years and they've been getting better crop of talent, including someone, including some from top universities, just because like the job market is so rough. I mean, they stopped.

Oh, interesting. Because the job market stuff— because I talked to a friend who had spent a lot of time in China and he said that like the major difference that he saw between the US and China was that the government is able to hire super top performers, but it's because they can force them to join. But this is because you're saying it's more of like a job market dynamic.

I don't think— I don't know who's getting forced into taking these jobs. I think it's still like, you know, it's downstream of— look, youth unemployment was like 25% 3 or 4 years ago, and then the Chinese government stopped reporting it.

Oof.

So like, look, you know, people will take work where they, where they can get it. And I think the promise, like also there's data on like the government reports how many people take the civil service exam every year. That's been rising pretty consistently over time. So to see that now like bleed into police and the military is a, is an interesting, I think, sign of the times because those are not high-paying careers. There's just ones that have more stability, right?

Yeah. Yeah. Back to Liang Linfang, founder of DeepSeek, his, his leaked transcript, it felt like there were a lot of interesting, thought-provoking ideas in there that some of which I agree with. But it also felt a little bit like the 2025 AI lab leader playbook that is now facing backlash, where he's talking about You know, sort of sci-fi scenarios, jobs, and all these different mechanics that run some risk. Do you think that has? Do you think he's aware of the backlash in the United States to that type of rhetoric, or do you think that he thinks that China is somehow resilient to it?

Well, here's the interesting thing with Liang Wenfeng: He's kind of like Demis and DeepMind. Yeah, DeepMind never sold.

Sure.

So, you know, he is really rich. He runs this like giant Jane Street.

Yeah.

Like quant thing, quant hedge funds, which is— which spits off money. He's able to basically fund this himself. He's told his investors, I'm not here to make a profit. I'm here to make AGI. And like, look, I'm going to have these weird ideas and try to just do cool science and I'm going to make this bet, which is very different from the one that the Western Frontier Labs have made, which is that, like, I need to have a business to throw off cash in order to keep doing it. And I think, like, you know, what was the striking thing in there? He's like, look, we're not going to have perverse incentives because we're going to keep our headcount really small, and we're going to keep doing this. We're going to do the best we can to, like, keep doing fundamental breakthrough science.

Mm-hmm.

hope things go well. So, um, that more than his, uh, uh, you know, what's gonna happen in 2050 things, I—

Yeah.

Were, was, was my takeaway is like, he's the one player in this ecosystem who's, who's prioritising keeping his sort of like independence and research direction and not sort of being tied to a product roadmap or to a broader, um, uh, AI, You know, a compute provider in China or like an Alibaba or Tencent, but he just like wants to be weird and do do his own thing, and I think that's really cool.

That's interesting. Where do you think DeepSeek will be getting chips in the next couple of years? Is that they're just going to keep building on Nvidia?

Um, so importantly, there is this ratio between the U.S. NVIDIA, uh, US NVIDIA, Jalapeño, whatever, broader, like TSMC ecosystem and TSMC, Samsung, Intel ecosystem.

Yep.

And the Chinese ecosystem.

Yep.

And there is capacity like 15 to 1, 12 to 1. And that's something that you can project out pretty clearly. Um, just kind of like doing a little, doing some equations from like how much, how much, uh, CapEx is being spent, how many, how much equipment, is being acquired, like yield rates, what have you. So where is he getting the chips? As much Nvidia as he can get, probably. But the sort of larger takeaway with the Chinese and Western ecosystems is that the Western one is way better capitalized and has way more straightforward access to chips, which is basically an order of magnitude better. Um, in over the next, like, at least 5 years. So it's going to be rough sledding for them. And look, if you— I don't know how much time you guys have spent with the Chinese AI IPOs, but they're making like $500 million or like a billion dollars. And it's cute, but these are— this is not $30 trillion, right?

Yeah.

Uh, or, uh, or even if— or even $10 billion. And the reason of The reason the reason that's the case is because they're not on the frontier of the technology, so they can't charge a premium, and they're compute constrained. Now, look, we've got this interesting new model where Kimi is trying to license its technology to be able to sort of leverage compute in the West to make money. Like TBD on whether the Trump administration is going to be cool with that as a going concern going forward. But if all the Chinese firms Firms have is access to Chinese chips. It's going to be really difficult to make a lot of money and to keep paying for more chips in the way that the Western labs have been able to like really mobilize the entire global financial system.

Yeah, you're talking about Moonshot potentially going to Microsoft and Amazon and offering their models, which are open source but served, and they would want a rev share based on whatever. Got it.

Exactly.

Jordi, you had a question.

Is there any evidence of individuals in China that are outspoken saying that China is not doing enough to invest or be competitive in AI? Like anyone that's saying, like, the government— like, if we don't have private— or if we don't have companies that can operate like a Jensen, just sprinkling billions of dollars everywhere throughout the economy that we need the government to actually, you know, participate in that way?

No.

I mean, I think the government already is. So you've started to see more and more state money coming into these companies' cap tables. Like, you know, if you're a CEO, it's a bit of a— it's a bit of a double-edged sword. Right? Where on the one hand, you just want to be able to run your business and not have this like guy from an SOE saying like, hey, we think you should spend your time doing this instead of that. But, you know, subsidies, subsidies are always nice. Like the Chinese government maybe 15-plus years ago, waked up, woke up to this idea that it needed to invest really substantially in the semiconductor ecosystem. And then they've done that for For better or for worse, we're starting to see real results come out of the likes of CXMT and SMIC and Huawei, of course. Now on the lab side, so like the or like the demand side for those chips, that's something which has only really started to happen over the past six months or so. And look, the scale of it is going to be very different than the amount of money that. Microsoft and Amazon and Google and OpenAI and Anthropic can do to sort of like pre-commit to buy these chips going forward. So I don't know. Well, excited, excited to continue following that story for you, Jordan.

Is there any evidence that there's been a sort of deflock movement in China?

I don't know what you're talking about. Oh, you don't know about deflock?

So like on Instagram, on Instagram, there's videos that will have like 200,000 100,000 likes of somebody demonstrating the most efficient way to take out—

So quickly to level set, Flock Safety is an American startup that sells essentially CCTV cameras to police forces that then go into a database, get image recognized, and so a police officer can type in a license plate and see where that car has been. There's been a few instances of abuse where police officers have They've been checking on the movements of their ex-wives, exactly. And so that's caused a lot of backlash. That kind of created the original media firestorm. But now it has really gained ground on social media. And as a result, there's protests and people actually going out into the real world and destroying the cameras with, like, buzzsaws to get them out of their communities because they really don't like them.

I mean, I mean, I've, I've sort of like emotionally surrendered to the panopticon years ago.

Very Chinese mindset.

You know, it's like, you know, they're watching me.

It's okay.

What am I going to do about it?

The Tyler Cowen has this take, right, where you should be bullish on places that have a public safety problem today because the cost of dealing with petty crime is just gonna go down because there's exactly stuff like this where—

So it's like New York in the '70s and '80s. Like it became a very toxic place.

Yeah.

You just, we're gonna have 10x police officers everywhere.

And you don't even need to be good at being a police officer.

You just have to like show up and say, hey, I got you. I caught you on camera like doing this bad thing.

So we have a picture.

Speaking, I mean, coming back to cars at the beginning, like I don't think speeding is good. I would like people to be caught when they speed.

Where's the boo?

So they stop speeding.

Where's the boo?

No, I think, I think—

No, I agree with you.

That's generally—

Keep it on the track.

If your car is worth more than $150,000 and you pay after $50 grand a year for, you know, your speeding pass—

Ticket clinic.

If that's the country you want to live in, John, by all means.

You know that that's actually what happens. People with really nice cars They get specific lawyers that get them out of every ticket so that they never need to deal with them. And then they don't get caught speeding that often. But if they got caught every single time—

With speed cameras, is there something like an 1199 Foundation?

Oh, yeah.

For that, where you get like a whole livery around your car that—

Maybe a stablecoin that you could stake, get a bunch of money flowing.

The thing, uh, what do you Are you tracking anything? You're probably not, but I'm just curious. I've I've been getting Instagram ads for like consumer eVTOL, just basically like flying drones that you can sit in as a human. They're selling them for like $100,000 now.

Great.

It's a jet ski for the skies.

Okay.

I can't tell if I can't tell if their algorithms got you guys bad.

They saw that acquisition price.

They're like.

Let's just send Jordi some toys. Toys, yeah. I mean, my boy's been buying toys. He's got so many surfboards.

I'm losing my mind.

There was a question there, sorry.

The question is like, are flying cars gonna— like, we just saw the humanoid robot games, which everyone is sort of like, you know, some people are laughing at because the robots run really fast and they explode. And other people are like, whoa, that's actually like really impressive pace. And these things seem pretty capable, faster than Usain Bolt.

What's the—

what's the market for the eVTOLs besides like going to the Hamptons? I mean, you just know—

I mean, this is—

this traffic is like the only point, right?

Yeah, but that is extremely significant.

Like if you live in Los Angeles.

So there's a lot of— there's a lot of— I mean, D.C., I Last time I was in DC, we had a 2-mile-ish drive, and it was traffic. There was so much traffic, it took like 40 minutes. I was like, you live like this? But anyways, I do think that I'm— I've seen enough videos now that basically are showing that flying cars are here. They may not be safe enough yet.

You think LeBron's gonna take it?

To Philadelphia? Isn't he doing that?

Or New York City?

I thought he was doing helicopter transport.

Yeah, but why take an helicopter when you can take an eVTOL?

Yeah, I mean, the hot take on flying cars is that we've had them for a long time. They're just expensive. They're just called helicopters because—

Okay, so pull up this video right now.

But the idea is democratizing the helicopter.

You can just buy this for $120,000.

That sounds so dangerous.

Yeah, I know. I mean, we've made commercial air travel very safe, so why not?

Here we go. You can buy this right now for $120,000.

When are you guys gonna livestream from an eVTOL?

Can we get our—

can we get a show? This is electric?

Yes.

Wow.

How long can it go? Can it like go? But really fun.

They're already using them for, uh, agriculture.

Okay.

Like spraying crops.

Okay.

Why do I need a person though in it?

Yeah, no, some of them you can— most of them I think can be removed.

The only reason you need a person in there is if the person is moving, like traveling from one place to another.

Yeah.

And people pay a lot. Opportunity costs are high. And so if they can get, you know, it would be an hour drive, but it's 10 minutes in the eVTOL, like that's a bullish thing. Last question, tell me about The elite rocket force in China. They are seeking public tips on procurement violations. I don't even know what the rocket force is. So break it down. Tell me what's happening.

Okay. So rocket force. This is actually a Soviet invention. So, you know, Air Force, Navy, Army, Marines. The Soviets were like, okay, rockets, big deal. We definitely going to need the cavalry. We're gonna need them to, you know, blow up America, so we're gonna have a rocket force.

Okay.

And then China made their own a few years ago. You know, we, we came back with our Space Force. I think— I'm not sure which one's cooler, to have a rocket force. But basically, this has like been a debate for a while. Do you need your own— like, the missiles that you shoot from the ground, should they be their own force? Whatever. Um, Bloomberg, 4 years ago, 3, 4 years ago, comes out with a story saying that half of the rockets in China are not filled with fuel, but are filled with water. Then you start to see a whole lot of purges. And as you guys have been following closely, I'm sure, the—

Wait, so they were LARPing rockets?

They were absolutely LARPing rockets. There's an entire— it's kind of 2 generations of Chinese military leadership that has now been purged. And they're sort of like, Like the 9-person, you know, Joint Chiefs is now down to like a guy and a half. Yeah.

So we've been purging.

That's it.

I mean, it's the biggest piece of the story is just that they had to go for public whistleblowers like that.

Yeah.

So, okay, so you can purge so easily behind the scenes, check the records, just go and look and see, okay, there's water in this, in this rocket instead of fuel.

Like, it doesn't seem that So yeah, we're like 7 years into— we're like 3, 4 years into this round of Purge, and now we're opening up a tip line to the public saying, hey, you know, anybody bragging at the club about stealing rocket fuel, call this number.

Very, very rough.

Well, speaking about, uh, former spouses wanting their revenge, like, you know, it's, it's not just following the cars.

That's actually a good source of whistleblowers, probably. That makes sense. That's a good one. Uh, Jordi, last question, or we're good?

No, no, I, I thought that they just wanted to put on, put on the front like they had way more rockets than they actually did when they were really just like, you know, metal with some water in it.

Sure.

Oh, I didn't realize that it was—

oh, so, so you think it's part— it was part of the plan, actually?

Yeah, I thought, I thought that's why, that's why I said Eisenhower's inflatable tanks. Yeah, live-action role.

Yeah, yeah, yeah. I mean, you see those military parades; they never open them up and show us what's inside. Who knows? Maybe it's all paper mache. Anyway, thanks so much for coming on the show, Jordan. Great to see you.

Congrats.

Great to see you.

See you at the track for China Talk. See you there. We'll see you there. Have a good one. Let me tell you about CrowdStrike. Your business is AI; their business is securing it. CrowdStrike secures AI and stops breaches. Our next guest. Our guests are from Actinide. They are the first startup to produce HALEU, the high enriched—

Finally.

High—

what is it?

What does HALEU stand for? I know it's enriched uranium, but break it down for us. Introduce yourselves. Welcome to the show.

Hey guys. Hey John. Hey Jordi. So nice of you to have us. Huge fans of the show. Happy to tell you a little bit about HALEU and what we're doing here at Actinide.

Yeah.

So you hit the nail right on the head. We were indeed the first startup in history to enrich uranium and produce HALEU. In terms of what HALEU is, it's not something that you need to really worry too much about. It stands for high assay, low enrichment uranium.

Now I'm worried about it.

Now I'm worried about it. Now I'm worried about it.

It's the next-gen fuel for next-gen reactors.

Okay.

And it's something that the Department of Energy says that you cannot get presently in the United States. There's no supply.

Yeah.

Right now.

It's mostly coming from decommissioned nuclear Submarines or something like that? Like they sort of refine it down, like reuse what we have, basically?

You're exactly right. Or they get it from Russia.

Yeah, Russia. Not good.

Yeah. And that's something that we're aiming to change here in Dallas, Texas.

Okay.

There's a huge feat by our team. Yeah.

Small team.

Something that nation states have struggled with and failed at. And we're doing it here in Dallas, Texas.

Yeah. So take us through the process. Is it big centrifuge? What's the equipment that's involved in the process?

So we have a machine called the calutron, which was originally invented by Nobel laureates for the Manhattan Project. So they've already delivered for our country once before, but they were sort of decommissioned because they're not very energy efficient. Very large magnets. Now magnets don't consume power if they're superconducting, for instance. And so it's very rare in the engineering fields that you'll find the major problem can actually go to zero. So we stand on the shoulders of giants in that way.

So you can use the same machine, but just redesign a piece of it to be more efficient? Is that what's going on?

More or less. It's the exact same principle, though. And it's, besides the centrifuge, the only thing that's been proven at scale.

Correct. Take us through the history. How did you get here? Why nuclear? What's the lineage of the company?

It's been a long time coming. It's been an interest of mine for a long time. When I first heard about the problem of how do you atomically separate things. I couldn't believe that it hadn't been solved. And initially, I thought this would be an easy problem to work on. And so, I've been humbled since then many times, as you can imagine.

What's the phrase? We don't do things because they're hard, we do them because we thought they would be easy.

Yeah.

I couldn't agree more with that. So, time and time again, you go, okay, I finally understand it, and then no, right? So, it's been a long time coming. You know, there's a lot of research that has to go into this. This is not a very approachable field, A lot of things are classified and you have to derive it for yourself. There's been a lot of smart people who've had a lot of interesting ideas about this, from the centrifuge to laser and calutron, you name it.

Yeah. Where is the limit on classification? Because a lot of the— you mentioned like rederiving it yourself. A lot of the AI models are able to rederive cybersecurity exploits, which are now the talk of the town. But are AI models giving you refusals? in your work, or are they being helpful because they know that the control point is actually separate and it's not a risk to educate someone about how to build a nuclear product because at a certain point they're going to have to go to reality and go and get a particular material that is controlled?

You know, it's such a fascinating question. I don't use a whole lot of AI, so I don't get a whole lot of refusals.

Usually for finding vendors And supply chains like that.

But, you know, AI probably wouldn't be super useful for a task like this, because it touches on some pretty obscure areas of, like, you're building a particle accelerator. So sort of questions of something called— I'd be curious to see how far it gets along with it. But I certainly don't think it presents a proliferation concern.

We're also pretty wary regarding kind of where sort of information within our business goes to. And so we have a lot of things that are still, frankly, analog or air-gapped regarding any sort of things that are proprietary or trade secret.

Yeah, yeah. I mean, at the same time, it's not like the company has nuclear launch codes, and cybersecurity is more about intellectual property, correct?

Yes.

It's IP-related.

Right, right.

And we look forward to working with the appropriate regulators for us to scale up HALEU production. You know, it's a smidge under highly enriched, and so you have to be in a Category 2 facility. But the separation technology itself is, you know, 80 years old. And so there's definitely some— a lot of know-how, but the principle's overt.

What's the biz dev, go-to-market side of the business? Are you already inking contracts with— I mean, there's so many new nuclear reactor startups that would probably sign up to buy HALEU if you were able to produce it at a certain scale. What's the status there?

Absolutely. So we actually are kind of agnostic with the materials that we make. Generally, we refine atoms, and some of the first refined atoms that we made were for the medical industry. They go for about $30,000 per gram, which is like a paperclip's worth of material for you patriots out there. And we delivered that to Oklo Isotopes. We've had a few commercial deliveries to them. And You know, in terms of the Halo side of things, we are— we've been in stealth for quite some time. And so we kind of blew up Twitter with our announcement of coming out of stealth and some of the recent developments here. And we have a lot of those biz dev talks ongoing this week and beyond to go ahead and ink some deals. Obviously, there's a big sort of gap between doing it once and being able to scale it up on both the technological and regulatory fronts. And we look forward to working with the right partners and regulatory bodies to make sure that that goes in a methodical and expedient fashion.

You self-funded this company at some point. Can you tell us about the journey to set yourself up for there, what that was like, what the motivation was?

Yeah, that's true. I did put in over $1 million of my own money. You know, when Robert and I first met, I had a background in finance and software, and he really kind of gave me a gut punch because he told me, That he was pretty upset with so many brilliant engineers that go into finance and software.

Sure.

They— I turned a buck from it, um, and that they weren't doing hard engineering problems. So I asked him, well, what are you doing? And he said, I'm working on what I think is the hardest, you know, sort of unsolved problem in humanity. I was like, all right, like, what's that about? And he told me he was doing sort of isotope separation. He showed me what he was working on. I got super infatuated with the problem. I truly agree that this is indeed humanity's largest unsolved problem. If you look at history, you got the Stone Age, Bronze Age, Iron Age, the Silicon Age. We were promised the Atomic Age, and it's something that we, we deserve. We're not playing with a full deck of materials. And so I began investing my own money and got to a spot where it was beyond R&D. We brought on a seed round and, you know, here we are doing commercial deliveries and scaling into different markets.

That's fantastic. Well, congratulations.

And yeah, really impressive.

Thanks so much for coming on the show.

Yeah, great to meet you guys. Come back on soon.

Goodbye.

Be safe.

Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents, whether you're writing code, analyzing data, creating content, or automating business workflows. Codex helps you move projects forward from start to finish. Our next guest is Billy Fallimer, returning to the show with Regent. Billy, how you doing? Welcome back. What's up, guys? It's been too long. Doing well.

Been too long. Great to see you.

Give us give us the latest news in Regent World. How are the sea gliders going?

We are we're cruising. We're gliding.

We just wrapped up our $240 million Series B, so that's the big news lately.

Give me a gob.

Come on.

Quarter bells. I've been waiting on that quarter bell.

Oh, so good. So yeah, talk about what it takes to. Raise for this company?

Are you—

you're still pre-revenue, but you— and you're not actually like the product. You can't just go as a consumer and test it out, but you have a massive backlog. What are investors looking for? How are you telling the story?

Yeah, well, we do have money coming in the door, importantly. Right. So we have commercial customers. We have about $10 billion in the backlog there. But one of the unique things about Regent is we have firm orders, like actual orders, nonrefundable deposits.

Yeah.

Customers are putting down— we have over $10 million in the bank now that are backing our backlog now, multiple years now of delivery from our manufacturing building. So that's on the commercial side. And on the defense side now, up to about $20 million in orders, majority with the Marine Corps. But we have some really big announcements on the horizon here. So the defense business growing really fast.

Yeah. So, yeah, I mean, typically you'd assume like net 60, You know, payment terms or something, but you are able to get some of that money upfront. That's good. What does the team look like? Where have you— where's the company based and how is the actual operation growing?

Yeah, we're based in Rhode Island. Rhode Island has really been put on the map lately. We're leading in maritime and defense. There's a bunch of other companies here, a bunch of startups. Anduril moved into town. Like, you go out on Narragansett Bay and That's cool. You're going to see all sorts of drone boats out there and subs and Seagliders too. So that's been pretty cool. We're up to about 120 people now at the company. But obviously, with the Series B done now, we're going to start growing really fast. And the whole point for us is we have this quarter-million-square-foot factory here. That's where we're going to be building our Seagliders, both our large Viceroys, the small Squires. So we got to fill out the factory. So that's the next step for us.

I feel like you're in an interesting position because we, I mean, we talked maybe years ago about the benefits of being all electric, that yes, there's an environmental benefit, but for specific use cases it might not be that big, but the serviceability might be an even bigger benefit. Is that holding true? Because you're in this dynamic where the current administration is probably very receptive to doing business with a young company, like you mentioned with the Marine Corps, but at the same time, they're not exactly the biggest fans of electric technology from what I've seen. So, you know, is there a discussion of other powertrains or do you wanna remain laser-focused? I remember talking to somebody who deployed small, like basically drone helicopters in the military and said that like gas engines work really well, they stay up much longer, the energy density is fine, the supply chain's there. But how are you dealing with the trade-offs of where the powertrain technology might go in the future?

Yeah, we sort of always had the feeling that on the defense side of the business, and particularly in the markets we're going after where we're talking about like moving across thousands of miles across the Indo-Pacific, right, hopping between these islands, that we're going to need longer than the couple hundred mile range we get out of batteries, which is really good for commercial. Like, that's your regional range where you're sort of stuck between a car and an airplane.

Yeah.

But on the military side, like, we need to go from Guam to the first island chain. Like, we're going across the Pacific here. So on that side, definitely hybridizing. About a year ago, or a little over a year ago, we announced Regent Defense. And part of that was really a focus on building out this So we have sort of a conceptual design now. We've been working on the supply chain side, but we're definitely going to be introducing an engine, a fuel engine system that works with our electric propulsion system on the front end, extending our range out to about 1,600 statute miles. So we'll be able to do that.

Well, yeah, I've seen these— this very unique hybrid. I believe it's a Chinese car where it is primarily an electric vehicle, but then there is a like a 2-cylinder gas engine that just charges the battery. Is that a relevant technology that we're gonna see other places, or was that sort of a unique anomaly of that particular Chinese EV company?

Yeah, it sort of depends on the application. For us, you know, we're a dual-use company, obviously. And I think like to be successful in dual use, and there are few companies that can really say this, is like as much consistency between the commercial and defense products as possible, like as much shared BOM, like basically the same SKUs. So our Seagliders are coming off the line here, or will within the next year or so in Rhode Island. And in the commercial version, we're going to load it up with batteries and it'll go 180 miles.

Hmm.

On the defense version, we'll put about 20% of the batteries in. So you still have those batteries sort of help you take off and land. We still have that electric propulsion system, which is Super quiet, super low maintenance, low thermal signature, all these great things. And then instead of the rest of the batteries, that's where we put in our engine and our fuel system.

Yeah. Tell me more about the commercial backlog. I mean, it seems huge.

Where—

who's interested? Where are the logical places to deploy these? Because I imagine that as you've actually built up that book of commercial business, There's been a lot of just assumption testing on the other side of the negotiation table. Maybe you come and say San Francisco to Marina del Rey. Maybe somebody else says, actually, it's going to work better here or there. So where are you seeing the future play out first?

Exactly. I mean, founding the company, we're coming up on 6 years in business now. Growing up, going to school in the Boston area, it was like Boston and New York, right? When you start raising money.

Yeah.

Yeah.

And it was like LA to San Francisco, you know, we're raising money, we're going between them.

Like, those are, those are total pain points. So that was sort of the focus on the initial commercial side. Um, sort of started obviously just based in the US as an American company with the US focus. And so that's evolved into Miami to West Palm and the Bahamas as a likely US launch market, um, throughout New England, right? Like Boston, Boston and Rhode Island and Connecticut down to New York, and certainly some West Coast routes, and obviously Hawaii as well, Pacific Northwest too. But what we've found as we've grown is this is actually truly a global technology. And the international order book and backlog is actually much larger than the domestic one. And very few people know this, but there are as many ferry passengers per year as airplane passengers, which is wild as an American to think that—

Yeah.

The global coastal market is that big. For, you know, cargo, there's 90-plus percent of cargo that goes over the coastlines. Almost half the world lives in coastal populations. It's just like how humanity settled the world. So globally, globally, we're talking about places like the Mediterranean, Japan, throughout the Indo-Pacific, right? Like, these are really big markets for us. So it's not just an island-hopping machine. It's like anywhere there is coastline and humans live and trade on coastlines. lines, that's our commercial market.

Is there interest from the Middle East or is— are things like actually slowing down there because of the geopolitical conflict? Because I— when I hear about new technology, maybe it's a little bit expensive to get up, but you need to do some retrofitting of some port or something like that. Feels like Dubai is going to sign up immediately. But what's it been like talking to folks?

The Middle East has been a big part of our strategy for a long time, actually. So we spent a lot of time out there. We have a great relationship with the Strategic Development Fund in Abu Dhabi. So we actually plan to expand manufacturing. The Series B, our first global manufacturing center here in Rhode Island, but subsequently, and based on the size of that backlog, we're already planning international expansion. So we've set up a joint venture in Abu Dhabi.

No way.

We're gonna be scaling manufacturing from there. So the Gulf is gonna be huge, not just for manufacturing, but also But also commercially, like Dubai-Abu Dhabi, that is a pain point route too. Crazy traffic on that route. And then—

Yeah, I was just thinking, I was in— I went to Formula 1 in Abu Dhabi, but I was staying in Dubai. And that drive, especially when you have these big sort of global events, is just brutal, right?

Brutal.

Even the traffic getting out of Abu Dhabi. And so, I can totally imagine.

It's not a long drive.

Yeah, you have a 10-minute quick flight.

Yeah.

Question unrelated to your business, but what's happening on the recreational aviation side? Are we gonna see an electric, the sort of the Model 3 of the skies, sort of an electric Cessna 172, something that is, you know, user-friendly, you know, for the sort of pure—

Like on boats, but for—

Yeah, something that's like, hey, I just wanna go fly for an hour.

Yeah, yeah, yeah.

There are certainly some attempts at electric airplanes and there are some electric trainers out there, But really, the hard point of aircraft generally, and like my background's aviation and I'm a pilot, my co-founder's a pilot, a lot of the teams are pilots, like it's the training, you know, tons of time to go out, get your pilot's license, like learn how to fly, learn how to do the controls, learn how to navigate airspace, talk to air traffic controllers, etc. So like that is the big time sink. And then there's also just the performance of like, I need enough energy on board in these batteries, and we already talked about the range limitation of batteries, like get up to altitude. And then as I'm flying to my airport, all aircraft need reserve fuel. So if I'm flying to my airport, my airport's occupied, or like something's wrong with it, I need enough fuel to fly and land somewhere else. So it's actually like all of— yeah, right. It's like, it's good.

Only Harrison Ford.

And so like, that was our perspective coming into this. My co-founder and I came from the aviation world. And we were Yeah. And we were working with the EV tolls and everything, and we said, these are the challenges we need to solve. Electrification is here in many ways. And by building Seagliders, we solved those problems because we certify under Coast Guard jurisdiction as a vessel, not an aircraft. Our digital flight controls enable it. So all you need to do is drive a boat when you're operating it. Even when you're flying, the only controls are boat controls, left and right, fast and slow. And so that's what's going to open it up to everyone.

Yeah, yeah, yeah, yeah, that makes a lot of sense.

Have you seen, uh, the people, uh, in China selling quadcopters that humans can sit in that they'll just sell you for $100? Yeah, yeah, no, so that's what I'm saying. Like, so I saw, I saw, I saw another video of these guys that bought a Temu drone, okay, and they were able to— you could sit in it, but they were flying it with their remote and then they crashed it. And so I, I just can't believe— I can't believe that you can just order one of these right now.

I was, I was at, uh, an air show in the Middle East and I was, no kidding, standing next to Adam Goldstein, uh, at Archer. And we were like watching this drone competition and one of these drones falls out of the sky live at the show. Fortunately, no one was on board. And we look at each other like, what is going on here? Like, hey, this never flies in America. This would never pass any certification. But like crazy stuff like that going on is actually not good for the industry. Like aerospace is held to a higher standard. There's a reason why aerospace is known for its safety and commercial aviation is the safest way to travel. Even with us in the maritime domain, we're taking a similarly safe approach because we're flying at high speeds with people on board. So like that's our focus and it's a harder bar to hit, obviously, as a startup, as a company, that's like part of the responsibility of building flying machines. But, you know, seeing these guys jumping in with like the swords swinging around and like some RC controller, like, I get a little worried.

Yeah.

Fortune favors the bold, though. Come on. You got to take some risks. You're an entrepreneur. Hop on this thing. No, I think you're right to play it safe here. I completely agree. But thank you so much. Congratulations on the new round.

Awesome progress.

And we'll talk to you soon.

Have a great one.

Thank you, guys. We'll see you later soon.

Goodbye. Let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Stop making excuses. September is the month for you. Just do it.

Do it.

Up next, we have Aaron Cannon from Outset launching AI digital twins grounded in individual customers. Eric, welcome back to the show. How you doing?

Good to see you again.

I'm good. Third time on. I'm feeling special.

I love it.

Repeat.

Catch us up to speed. What have we missed? What's the latest and greatest in your world?

Things are great. Yeah. So we're an AI research platform. You guys probably remember we've, since we last chatted, I think tripled revenue, tripled the team size, and now launched a very cool— Thank you. I was hoping to get at least a gong or the But, um, yeah, plenty.

Yeah.

And now we just launched Digital Twins.

So we built—

yes, we built a whole simulations lab and have now built, uh, yeah, I'll tell you all about it.

Yeah.

And so you came out of YC, you were sick. You're always getting told, talk to your customers, talk to your customers. You're absolutely sick of that. You don't want to ever have to do a customer call again. And so you built Digital Twins. Is that right?

It's exactly right. Yes. I started my career as a researcher and I was like, I don't want to do that anymore. But in all seriousness, yeah.

Sell me a digital twin because I've run companies and e-commerce companies. And it's useful to have a profile of certain customer cohorts and certain customers. But that aggregation takes a lot of the work out of it. And I feel like there's a lot of platforms that do that. It's sort of nice to talk to individual customers. But you might get a weirdo who's actually not very indicative. Sometimes the individual customers can give you some unique insight that everyone else is feeling. But walk me through the benefits of having these digital twins that are grounded in individual customers.

So here's the— like, to take a step back, research, the whole point of research is to try to predict the future. Sure. Like, you're trying to make some decision for your business, right? And you're looking backwards at the data. And the idea is that you can now not only look backwards, but you can look forwards. You can actually start simulating, like, if I do this new marketing campaign, how would this customer set start reacting? What, what, what are the things they would say?

Yep.

And we're very good at research. That's what we've been doing. That's what we built our business around. But now what we can do is take that research and build out the simulation platform so that you can say, okay, I can actually see into the future of how this group would react to some very big business decision.

Yep.

And so we train them by, by, by training digital twins on individuals. So it's like a one-to-one.

Sure.

Like you can go in, talk to a digital twin, and you know that's based on John.

Yeah.

And like, that's got all of John's experience and like all this data about John, and it can, it can talk on behalf of you, right? And so that's like the, the, the, the, the mechanism that makes it super, super useful and also like trustworthy.

Yeah. Uh, what, what, what incentive do you have to give people to do a long-form AI-moderated interview? If I'm, if I'm buying a pair of shoes, I don't know if I wanna sit down for an hour. With, with an AI to talk about the shoes and how they fit?

I think the last run we did was— these were actually like 75 minutes, the last one, and I think we paid them $200.

Okay.

Yeah. So it's a reasonable—

to be clear, it's worth more than the shoes, right?

Sure, sure. And if you do, is that one customer per one company? So you're— I mean, your clients include Microsoft, Uber, Google, Coinbase. You got Murderer's Row here. Are you saying like we are going to interview this one person about this one Excel feature or this one Excel product? Or are you saying like, we're gonna, we're gonna interview you about how you use technology. Some of this might apply to Microsoft, some of this might apply to Google.

That's right. So we do, we developed our own grounding interview. So it's not just about their use of Excel, right? The grounding interview, it develops what we call Persona Core. It's like trying to get like who you are, what are your values, what do you care about, right? And so we do that as the core and then we layer on other sets of data. So maybe Microsoft has, you know, operational data we can layer on to make it much more powerful.

They know all my favorite formulas.

Exactly.

Exactly. So the idea is that like we have, we have abstracted, you know, as much of the person as we can.

Sure.

And then what we do is we, we, we go back to that person and refresh that data regularly.

Yeah.

Yeah.

So we'll even have them grade their own twin's answers. Like how'd our twin do?

Oh, interesting.

And they can tell you. And now we have a reinforcement loop.

Yeah, so they're kind of RLing.

And so it's— exactly, exactly. And so that's what makes these really useful. And then we also put confidence scores. So you'd be like, obviously not all of our probabilistic, you know, output here is gonna be the same. So we can tell you which ones you should like really trust and which ones need a lot more human data to kind of bring to it.

Where are you seeing the most pull from the market? Like what kind of businesses Yeah, theoretically, every business should care about trying to predict the future and forecast how different customer segments, but I'm sure some care more than others.

The biggest pull right now is that companies have really hard-to-reach audiences, right? And so these are like, maybe there's, you know, maybe they're CRM buyers, or they're, you know, kind of some kind of somewhat obscure part of the population. And so the incremental value there is massive, because they can't actually go out and do constant research with that group. So they have, like, very few signals today. So the incremental value is big enough that they're, like, adopting, I think, earlier, and first, while there's still, you know, more kind of questions in the market. But I think that's— I think we'll continue to see a bunch of demand from, like, these specialized audiences that companies are trying to—

Then what's the opposite? What's a company that's, like, drowning in customer feedback data, is that some consideration?

Well, I'll give you a kind of— there's basically certain business decisions and strategy decisions that you would make that you don't need to forecast, right? A business doesn't need to say, oh, what if we made customer support wait times shorter, right? No one is sitting there being like, I wish when I called my internet service provider that I wish it took an hour instead of 30 minutes, right?

Or Or maybe a good example, and this is not a real one yet in terms of simulation, so I'll share it as an example. But like Nestlé does a lot of consumer research. Consumers are reasonably easy to go out and talk to, right? It's reasonably cheap to get more feedback. But if you were actually trying to do— if you're at Nestlé and you're trying to do research on like retailers, right, and how they— what they think about your brand, right, that's actually a much harder group to go find and do research with. You're not drowning in feedback from your retailers. Sure.

Right.

And is that the pitch when you signed Nestlé? Because I imagine that, like, AI is very impressive, but I don't know that I would trust an AI to tell me which milk chocolate tastes the best because, like, the taste data just isn't in the model at all yet. And for a lot of consumer research on Nestlé products, I would want to watch someone taste the chocolate and tell me what they thought of it.

Yeah, yeah. Well, so, so when we, you know, initially pitched Nestlé, it's actually about our, you know, AI research platform, which AI is just doing the interviewing.

Yeah.

It's asking the questions. You still get real human data at the end.

Yeah, yeah, yeah.

Right? And that's our core, our core thing. And that's what we've had for the last couple of years. But simulations is an extension of that. And that's where we, we start talking. It's a, it's a, the kind of the, the best kind of 2 elements of the pitch. Number one is hard-to-reach audiences, right? You're not able to get to them. We'll build digital twins of them. You can talk to them anytime at, you know, unlimited capacity.

Yeah.

Number 2 is we say that like anybody in your company can access them. It's like if your finance person's working on a new pricing model, you don't want to like give your finance person like the ability to go research with real consumers necessarily.

Yeah.

But now they can actually like, you know, test ideas, test new pricing and see how things would adapt.

Yeah.

You know, and simulate them.

Yeah, it makes sense. Last question for me is, I mean, it feels like the digital twin launch is like an expansion of the product portfolio, and I'm interested in where else the product expansion can go. I know that there are companies that they get a lot of feedback, but they struggle to bridge the gap between the feedback that they're getting and then actual ideas to go and implement. And I'm wondering how—

Yeah.

If you're going to wind up being more prescriptive in the results It's, hey, okay, we did all this research and you really should change the color from red to green.

Right, right, right. Well, look, so research is this thing where like humans go and decide to go run research.

Yeah.

Right. What we are building towards, and that's our original product, right? We're building towards is something where, uh, the model can go do research for you before you've even asked the question.

Mm-hmm.

Right. And so it is constantly trying to figure out the answers to the things that you don't even know matter yet.

Mm-hmm.

Right. It's like, hey, There's, you know, your revenue's down in that, in that sector. You don't even know that yet. We're gonna send agents to go talk to all the people in that, in that part of the economy and like go figure out what's going on, talk to some simulations if we already have them built. Yeah. And then go take action.

Mm-hmm.

And that, that action is where I think, like, where we can ultimately have more impact than just collecting and synthesizing feedback.

Makes sense.

Um, we should be able to like send specs to cognition and like let, let, let them go build it, right? Yeah.

No, totally.

And that's like really something that's happening.

Yeah, I love it. Well, congratulations on the launch, and thank you so much for coming on the show. Very cool.

For third time.

Thank you, guys.

We'll talk to you soon.

Great to see you, Aaron.

Have a good one.

Great stuff.

Goodbye. Let me tell you about Figma. Agents meet the canvas. Your AI agents can now create and modify your Figma files with design system context.

Thank you, Tyler, for kicking off some clapping while we're missing a couple folks here.

Jordi's not going to let it go. Up next, we have Stephen from Lambda coming back on the show. Third, fourth, fifth time?

Who knows?

Stephen, good to see you.

How you doing?

So good to see you. Great to see you.

Been too long.

Okay. So we were talking to Jordan Schneider about this China-backed AI bot farm. He was kind of brushing it off, and I can see both sides of the discussion. You know, there's been just a couple reports, 200 accounts here. I think there are a few hundred accounts here. He was saying that China doesn't have a long history with interference politically. What are you seeing? What are you identifying just in the broader debate?

So I think that it's really important to not get lost in sort of, is this a Chinese bot farm, and just focus on bad facts and just bad information. Because, you know, that's the fundamental thing, right? Disinformation is like most readily solved by just disproving sort of, you know, and just showing what's real and what's not.

Yeah, yeah. I mean, the big number that I saw most recently was polling on data centers, terrible broadly, but after telling somebody that you're polling that data centers recycle water much like a car wash recycles water, They added that phrase, then the net approval score improved by like 30%. So people with that extra context changed their whole opinion in just a few seconds. Is that what you're saying?

Yeah, yeah, 100%. So for example, the vast majority of new AI factories that are being built are completely recycling the water and using a dry cooler. It's a closed-loop water system where the water never even leaves the facility at all. So that's the vast majority of a lot of these new data center builds. And so that's, that's just completely different from the old technology, which is being described, which is a heavier water consumption, which is evaporative cooling.

Mm-hmm.

And so that's like a fact that most people should sort of You know, kind of look into. The other one is—

Well, the problem is people that aren't really interested in tech, but maybe are afraid of AI, they're not proactively sitting there being like, I gotta go, I got a free 5 minutes here, I'm gonna go get the—

I'm gonna want to—

I want to go get the facts right.

I'm gonna do some diligent research online and really read about it.

Yeah, just looking on YouTube and just AI negativity or content that is negative towards AI gets 100x the reach, right? Yeah.

There's always been a negativity bias on YouTube. I mean, you could say, here's the history of this company, or you could be like, the shocking downfall of this company. And the shocking downfall will always get more views.

Yeah, Taylor Lorenz says making anti-AI content on YouTube is not only profitable because you'll get a zillion views on any anti-AI slot, but also because there's a litany of these sort of AI safety orgs that will quietly pay you massive amounts of money for undisclosed propaganda.

I didn't realize that.

I love that Taylor Lorenz is really now back in the fray and on Team America, on Team Reindustrialize, right? I've got the—

I love it, the sweater.

And reindustrialize Sweater for it. And so I think the interesting facts about it are this. One, from the jobs perspective, yeah, look, this is not going to be the same level of job production as an automotive plant or a steel mill, but these are still net new jobs.

Yeah.

And so, you know, for example, we've got a data center out in Kansas City, and 300 folks worked on it during the construction period. and 50 full-time jobs, many of these making over $100,000 a year.

Mm-hmm.

What are you— I would ask to somebody who's against it, what are you gonna say when you sit down at a dinner table with these 50 families that this data center is supporting and say, well, no, I don't think you should have that job because I don't want an AI data center, when they're the same people who are using the data center to go on posts on YouTube or to use, Yeah.

Where do you sit on local versus national regulation of data centers? Like Bernie Sanders obviously has gotten a lot of attention for discussing the topic at the federal level, but it feels like something that, you know, just as like an American, it feels much more comfortable if like, yeah, there's an area in West Texas that loves this stuff. They're going to be into it. And then there's going to be some communities that aren't, and this should be handled by the local mayor and the board at the town level, and the voters should have a voice, but it should be contained to the community that's relevant.

Yeah. So first of all, absolutely. Look, that's how our system works. Let's just be clear here, right? You know, this is not— there's not some sort of, you know, national eminent domain that's being pulled like we have with construction of railways and other public goods. This is decision-making that is happening on the very, very local level. And some places are not going to want it. And some places are going to welcome business and some places are not going to welcome business. Some places are going to welcome data centers and others are not going to welcome data centers.

Yeah.

I think it's a perfectly reasonable thing. I think that should members of the community have a voice? Yes. Should property owners have a voice? Yes, absolutely. Right. You know, somebody who's going to have some any type of externality that's adjacent to the property, they should have a voice. And guess what? They do. They've got a ton of voice as a as a property holder. And and and I think that you know there will be reputation for communities that want to welcome these amazing investments into their into their community. So for example, going back to that Kansas City data center, it's it's a it's a it's a 10,000 GPU data center. Um, and we are going to, over the next 5 years at the very least, be paying $47 million over 5 years in property taxes.

Yeah, yeah.

This is a huge cash infusion to, uh, to a, to a community that we're working with, for sure. And so I think that there's gonna be the communities which are really excited about it, you know, communities who have already pioneered it, like Loudoun County in Virginia. And other companies— excuse me, other communities and other counties that want to see lower rates and lower tax rates, as well as lower utility rates. Because as evidenced by a bunch of research, communities that welcome data centers and industrial development and new power plants have lower rates.

Yep, of course. How are you thinking about the international puzzle? Because I know data center construction takes years to plan out. It's slow moving. But in general, when the environment in American business gets harsher, historically, business leaders have offshored almost immediately. And I could imagine a lot of data center developers just saying, oh, you don't want it here? Build it in Mexico. Build it—

It would be It would be a national tragedy. It would be a national tragedy, just in the very moment where this country needs to rebuild our grid. And it's not just data centers. Everybody wants to drive EVs. Everybody is starting to want to— we're bringing on more solar than we have in the history of the country.

Yeah.

80% of new megawatts coming online into the American grid are solar or storage. It's, it's, it's, it's, it's, it would be an absolute national tragedy for this to go out, to go international, for it to go to, you know, whether it's, whether it's, you know, overseas, outside of North America, into South America, into Asia-Pacific. I want those jobs here in the United States. I want those investments to be made here here in the United States. AI is a national strategic asset, in my opinion. Compute is a national strategic asset. And then most importantly, the way that AI is driving the development of the grid— and I keep trying to talk about this—

Sure.

But it's developed— the way it's driving the development of the grid is going to pave the way for more industrial onshoring to come back to the United States so that we can bring back whether it's different, you know, electric arc furnace mills for aluminum and steel, micromills, turbine manufacturing, other types of high intensity. By turbine manufacturing, I mean specifically like Inconel smelting or specialty metal smelting where there's a high energy input.

Yeah, yeah. I mean, you go back to the— wasn't it Colossus was built in an old washing machine manufacturing facility? facility that was out of business, if you all of a sudden have a lot more energy, a lot more industrial capacity, maybe the washing machine manufacturer comes back, and maybe that's good.

There's going to be a lot more jobs caused by that. And it's a huge virtuous cycle. And so I think it would be a real tragedy to have anything that was going to end up pushing folks. And I really support building this stuff here in the US.

Outside of the economic impacts, do you think that there's a national security value in having a data center in America or compute in America for like control reasons? Like AI is a powerful tool if it's in another country and that country says, well, actually we're taking this and we're now under control of this. What is your model for how valuable that is from a national security perspective?

So, so, you know, there's, there's the soft one and then there's the hard one. The soft one is sort of how important is it to have values, models that have values that align with American values and how patriotic is your model? I want patriotic models. And the second one is the hard one, which is what happens when they cut it off?

Yeah, that's what I'm worried about.

And we all experienced this when Fable went away for a little bit.

Oh, you remember?

Do you remember that? And I remember I called some people up, I said, Wow, this is what it feels like to have sort of the equivalent of electricity cut off. When you had the intelligence cut off from Fable model, and I was vibe coding something at the time, I said, wow, this is what it feels like. I had to go back to Opus, and it felt like a huge step back. And I thought to myself immediately, this is as much of a strategic asset as our grid is.

Yeah, that's funny. Jordi, anything else here?

I want to hear more about the filming business. What should people expect from the debate or just forward progress for the rest of the year on this issue? Like, should we just expect it just keeps getting decided on a local level? Is there—

it's going national. It's going national.

There's elections right now. And so I think that that's like the noise. The knob gets turned up a lot towards elections, then it'll kind of go back down. Yeah, it's very—

it seems very unfortunate that politicians are just realizing on both sides are just realizing that if I want votes, I have to at least say that I'm against this thing.

Yeah.

No, no.

But this is the interesting thing. There's multiple races right now. For example, one in Kansas, where the steamfitters unions and the different workers unions are actually aligning on the pro-data center side of things because there's so many jobs created and there's so many big deals that are created from these data centers that a lot of the labor unions are actually on the side of pro-data center now.

Oh, weird.

So look this up.

Look up—

there's a Kansas City, I think it's gubernatorial candidate, who's a Kansas governor candidate.

Sure.

And there's the steamfitters union is basically on the side of pro-data center building. And so actually, I think that it's maybe a little bit all backwards sometimes from what people have been positioning because, you know, you've been hearing in the media, oh, this is a bipartisan anti-data center movement.

No, no, no.

I think there's a lot of folks on both sides of it who, you know, are being supportive. I think that you're going to see the volume knob turned down after elections. And then finally, I think it's healthy debate and communities should have a decision about who they want to work with and that data center providers should be doing their best to not be influencing rates any direction other than bringing rates down for everybody and being a good member of the community. Pay your taxes, participate in building an ecosystem of businesses around you, and continue to make great high-paying jobs.

What do you think the tech community, data center community needs to do to get ahead of what I think is coming after the water debate clears? Because the water debate, of course, comes from some, Mediocre miscalculations, but—

There's miscalculations and also just misunderstanding of what is the total water consumption of things.

Exactly. So I think a lot of people in tech sort of brush off the water— the water discussion is debunked. And I think they're fair to do that. But on energy, on natural gas production, there is some real data about where this all goes. We continue to 3x compute for the next couple of years and we're powering that by natural gas, you could see headlines that track data center to emissions. This is how many plane flights a data center uses and it could become the next thing.

So, so, so, so, so first of all, one, the debate is not something that needs to be won with the Silicon Valley Twitterati, X-ati, whatever you want to call the people who sit there.

keyboards.

It's—

the debate needs to be won by the people who are in the communities affected by this.

Sure.

And so that's, that's the first one. The second one is that more than ever, like I said, 80% in 2026 is solar and storage. These are renewable things. That's actually what our grid is becoming more and more as we build this out. And so I think that there needs to be a lot more deregulation with regards to access expediting interconnects, you know, reconsidering, just, just making it so that we can actually go back to building and getting connected into the grid.

Yeah.

So that way, a lot of this behind-the-meter stuff, which is less efficient, right? You know, when you have a behind-the-meter, it's simple cycle as opposed to combined cycle natural gas, and it's less efficient. So we should really get to the point where we are encouraging grid interconnectivity and expediting that process. That's, that's, I think, the number one point, I would say, from a policy perspective. But, you know, look, at the end of the day, every American consumes about 600,000 gallons a year of indirect water consumption. So next time you see a big number, divide by 600,000 gallons per year of indirect consumption that every one of us makes. And then that will put things into perspective. I think that's a good starting point.

Yeah.

Yeah. I'm wondering, yeah, where that goes. Because it does feel like it's worthwhile. I agree with you on, like, you don't need to win the tech insider crowd first, but it does feel like the worst-case scenario would be another error to the tune of 2 orders of magnitude around energy consumption or natural gas production. Because if the viral clip is every data center burns as much fuel as every car has produced since humanity started, and that's off by 3 orders of magnitude or something, then—

Yeah.

Then you're on— you're fighting on your back foot as opposed to starting with, okay, they're already pretty efficient. Here's where we are regulating. Here's— and the New York Times has already dug into it. Here's the environmental considerations. Here's what we abide by. And then, yeah, there'll be a discussion about energy use, but at least it can be grounded in the facts. from the start.

I totally, I totally agree with you. But I think that the other aspect of it is just to not set with, oh, well, emissions are bad.

Sure.

The most important thing is this. As a country, we have a graph that looks like this in terms of energy production. China's got a graph that looks like this for their energy production. And guess what? All their energy is also solar now. The new, the new stuff coming online, 80% plus of China's energy is solar and storage.

Yeah.

And so I think we need to get— you know, the emissions will come down naturally as the grid transforms to more of a renewable and storage grid. People don't really understand how much BESS and battery electric storage has been a huge godsend to the grid.

Sure.

And there's— I've already even seen— I've driven through parts of the country where I've seen anti-BESS Anti-battery electric storage.

Why?

And I think there's concerns, and these are rightful concerns, that like, okay, they might catch fire and, you know, cause a fire or emit, you know, chemicals when they're on fire. But this is just like sort of, you know, yeah, any, any industrial asset that's generating electricity has like some worst-case scenario.

Yeah, it's the same thing with your, with your gas stove. If you If you leave it on, your house might blow up.

It's probably the same critiques if you were to go back to automotives, cars in general, refrigerants, when they were first coming around. But the point is that I think that actually getting that graph back to where America is competitive in terms of generating energy, whether it's from renewables or otherwise, it's going to inherently be 80% mostly renewables. Anyways, from the trajectory that we're going in, I think.

Yeah, yeah, no, I mean, that's certainly like the, the most optimistic future is a lot of renewables. Very, very cool.

I have an idea, but I'm going to DM it to you.

Okay.

Because I don't want it to be—

Jordi, John, thank you guys so much for having me on.

Thanks for hopping on.

Have a great rest of your day. Great to see you.

We'll talk to you soon.

Talk soon.

Bye-bye.

A couple more questions?

He's simply too good.

Okay.

We're going to come back and hit a lot more timeline tomorrow.

Yeah.

I can't wait.

We'll see you tomorrow. Thank you for watching the show.

Great to be back. Missed you guys.

Missed everyone here.

Love seeing the whole cast of characters.

Everyone's still in the chat. We love you all.

Actually, wow.

Give us 5 stars on Apple Podcasts.

Saw Bobby Cosmic in there. We got Sean F over in the LinkedIn chat. We got Cooper as well.

Acquired taste, the LinkedIn livestream.

I like it.

Yeah. Man of taste. Uh, anyway. We will see you tomorrow. We have a very special show for you and many more throughout the fall.

It's been an honor and a privilege.

We'll see you tomorrow. Goodbye.

Cheers.

Cheers.