SpaceX Phone, U.S. Stake in OpenAI, Jersey Mike's IPO | Andrew Collins, Vipul Ved Prakash, Isaiah Taylor, Dean Ball, Rob Toews, Tuhin Srivastava

You're watching TVPN.

Today's Wednesday. Is it Wednesday?

Thursday.

Thursday. July 2nd. I gotta update my sheet. Thursday, July 2nd, 2026. We are live from the temple of technology, the TVPN UltraDome. The temple of technology, the temple of finance.

The capital of capital.

Let me tell you about ramp.com. Time is money. Save both. Easy to use corporate cards, bill pay, accounting, and a whole lot more all in one place. And let's also talk about the news. Fable has been freed. Fable 5 is now available. Uh, good news for the AI world, of course. Uh, the Wall Street Journal says the battle over how to tame AI has just begun. Obviously there's lots of back and forth on, um, how these models get released, who gets access, when people get access, will the government get access at certain points? How fettered is the intelligence? How unfettered is the intelligence? Um, but there's a number of interesting interesting, uh, news. The, the interesting, uh, detail that people are sort of pulling at is, um, the new Fable 5 that is now released seems to have been, uh, nerfed in some ways. There's always been questions about, you know, okay, you don't want it to be able to build a bioweapon or hack a system, so, uh, locking down cybersecurity makes sense, but you should probably disclose that and be transparent about it. And then the question of how do you, uh, deal with the self-replicating AI. If you just ask it to build an AI system that can build you a bioweapon, well then you have the—

you can just build the unfettered intelligence.

Exactly.

So you have Cambridge Analytica unfettered, not limited by rules or any controlling—

what did you say?

Unfettered?

No, no, what, what source is this?

Uh, Cambridge.

You said Cambridge Analytica, which I don't think has a dictionary.

They should launch a dictionary.

That would be good.

That would be great. Maybe The Cambridge Dictionary. But the dictionary definition of unfettered is not limited by rules or any controlling influence. So if they're like—

I feel like unfettered intelligence, you're going to get regulated immediately.

Well, I just think like NeoLab, like everyone wants to be the cool, friendly NeoLab. Yeah, there's space for kind of a sketchy villain lab.

Well, we have Vipul from Together AI building open-source AI cloud infrastructure coming on the show. I'm sure he can talk about what's capable, what's possible when when you are building unfettered intelligence. We have a bunch of other great folks joining the show. Isaiah Taylor's coming on from Valor Atomics at noon, Dean Ball at 12:10 to talk about the export restrictions as well as what's going on with nationalization because there's also news that, this is from the Financial Times, OpenAI proposes handing Trump administration a 5% stake. Joe Weisenthal's pushing back already. He says, I don't know about this path. Rather than equity stakes, Why not make companies pay 20% of all pretax income to the federal government? And then instead of exercising shareholder influence, politicians and regulators could set rules on corporate conduct across industries. Interesting.

I like where he's going with this. Someone should try this, at least try it out.

Contrarian take. But yes, Dean Ball already chimed in on this news. He said the two, there are two broad ways this can work. One, you divide this 5% over all US households, handing each a you give the steak directly to the government. One is fine, two is probably ruinous, akin to inviting rats to live and reproduce in the walls of your house, which is wild. He says it will never stop.

Don't knock it till you try it, Dean.

Yeah.

There's some rats out there that can be quite excellent roommates.

Are you a rat guy? I thought you were a mosquito guy. I thought you liked mosquitoes.

No, I do like squirrels.

Squirrels, okay. Squirrels.

Yeah.

I feel like I have a bond with squirrels.

They're sort of the aristocratic.

When I see a squirrel in my yard, I look at it and I just salute.

People have said you have the mind of a squirrel.

They have.

Yeah.

They've said that I have the mind of a rodent.

Yes. Dean Ball continues. He says, "It will never stop at 5%. It will go on and on. The governance will become a nightmare. Political capture will be real. And it will generate precisely no goodwill with the public." Yeah, that's a good point. And we were talking about this before, this idea of like—

Yeah.

One of the challenges is even the way this was raised in the Financial Times. Yeah, it sounds— the assumption is that it sounds like they're giving shares not to a sovereign fund, which is, I think, what was actually proposed, but just giving it to the White House.

Yeah.

Yeah.

Which, given the disclosures that were made earlier this week—

Financial Times not having fun with it. So Trump denies conflicts of interest as filings show he made more than $1 billion last year, $1.4 billion of stocks bought, $432 million sold, deals for watches watches and perfume. Yes, the Trump watch happened. I saw a crazy chart that said that Trump earned— out-earned Coinbase in terms of profit from crypto in the last 12 months or something like that. I forget the exact stat, but it showed all the different crypto platforms, and many of them were down because it's been a rough market. Some of them were up, but no one is up more than the President of the United States. Wild times, wild times. But yes, it seems like— it seems like the, the green shoot here the white pill, the optimistic outcome is that if there's lab nationalization, if there's some sort of distribution of equity to Americans, it happens through the Trump accounts, through Social Security number, through some sort of identity that the individual can have a stake in, as opposed to, um, as opposed to just going into some fund.

We were talking about this with data centers where people are Tyler has the data. We can pull it up.

Okay, let's pull it up. Wait, do you have it or no?

Well, yes.

So Coinbase made $1.26 billion.

Okay. Net income.

But pull up this chart because it is pretty remarkable.

Yeah.

But as I was saying, with data centers, you guys, you guys, okay.

I know it's Fourth of July weekend coming up, but there we go.

Look at this.

Okay, let's zoom out. Let's get it a little centered.

There we go. Corsair.

Let's lock it into the center.

Galaxy.

Let's lock.

There we go.

And there we go. Amazing. Thank you.

Out-earned biggest listed US crypto firms. You have to imagine that— is Hyperliquid doing better? I feel like Hyperliquid was on a tear for a while. I don't know how they're doing recently, but anyway, staggering chart. But as we were talking about with data centers, there's one idea, which is just tax them, increase the tax base. But having taxes go into your local government is great and can provide roads and schools and pay for things and balance the budget, and that's great. But it's abstract as opposed to the Ben Thompson proposal, which was data center shows up in a small town. It's in, you know, a sparsely populated area. There might only be 5,000, 10,000 residents in the city, but they all get checks for— he ran the numbers and it was like $1,000 a month. For some of these data center projects where they're making so much money and the town's so small they can do a meaningful check to the individuals. And that sort of direct payment, I think, is going to be better received than this abstract, oh yeah, my government's better funded, but are they going to cut my taxes? Are they going to successfully build that road?

I don't know.

And it doesn't help me pay my bills, which I think is what people want. What are you saying?

Trump did, in fact, outearn Hyperliquid. He did by a meaningful margin. Hyperliquid generated $908 million in total fees in 2025.

Brutal.

Yeah.

So just like running very quick numbers on the opening, I think. So if it's like 5%, let's say it's at a trillion or something, that's $50 billion. If you're giving it to households like, like this idea, there's like 130 million households. So that's like $371 of like equity per household, which I think is like, sure, maybe that's, that's meaningful, but it doesn't seem like that's really going to like change people's perceptions of the labs if that's like part of the goal, right?

Well, I think if it's— I mean, so maybe then they do the right parlay with that, with that. So I mean, the nice thing about the Trump accounts is that they are set up in a way that you can elect to donate to just children born this year. And when you segment by that, the denominator gets a lot smaller. And so the impact gets a lot bigger. So you could be all of a sudden looking at something like a few thousand dollars for every kid born this year or something like that. You can narrow it down. You could do every kid under 5 or you could do everyone under 18. And that type of— I mean, you're right, it's not going to move the needle as much as broad, broad distribution for everyone. But you can be a little bit more targeted with it. I don't know. There's different approaches.

Ryan says, wait, does America get 5% of TBPN?

Yes.

I was never thinking about that.

Yes, but we podcast for you, America.

That's right.

This is the— the funniest outcome is the most likely. And that's exactly what's happening here.

Community ownership.

Yeah.

Of this show.

Somebody called me a Fed based on this news and I said, I don't even know what to do with that. But I guess it's somewhat true. The actual story here, just to reset, OpenAI is apparently in talks, according to the Financial Times, with the Trump administration to give a 5% equity stake in the company. The Financial Times describes describes the talks as early conversations, but the plans seem to involve other frontier AI players apportioning stakes in their companies as well. The obvious context around this story is, at least on a straightforward read, that AI companies need at least to stem the seemingly ever-growing negative sentiment around AI and data centers. Doing business with the Trump admin has turned out well for other companies like Intel and Dell. Intel is the big one where post post-government stake. I think the stock is up 200 or 400%, something like that.

400%.

400%. Very, very strong performance for Intel. And being difficult to do— being difficult to do business with the admin has been unambiguously bad for others. And now there's a vehicle with the Trump accounts that can facilitate equity transfers from corporations to average Americans. Again, these are early conversations, so we're very light on the details, but we'll be discussing this news, of course. Um, in other news, Jersey Mike's is filing for initial public offering. Jersey Mike's is going public. The sandwich chain Jersey Mike's filed for IPO today. Uh, it's looking for a $12 billion valuation after being acquired by Blackstone last year for $8 billion.

Let's give it up for Blackstone.

Yeah, they needed a win. They're getting a nice 50% markup in just a year.

Jersey forced out of buying single-family homes. Yeah, they had—

that's not true at all. There hasn't been any— there hasn't been any regulation around that whatsoever.

Joking.

Uh, they do own 95% of homes apparently, something like that. No, it's like, it's like half a percent or something.

It's low.

But Jersey Mike's has 3,300 locations across the US and Canada, plans to open 300 shops in the UK and Ireland. Uh, Peter Cancro Jersey Mike's founder, who's not named Mike. You would assume he was, but the founder of Jersey Mike's is in fact named Peter. He started the business in 1975 when he was 17 years old. Wow, young. Overnight success. He actually founded the business when he purchased a sandwich shop in Point Pleasant, New Jersey called Mike's Subs with a $125,000 loan that he got from his then football coach, who was also a banker. So you're playing high school football in New Jersey and you have an idea to buy a sandwich shop.

Coach was like, this guy's a star. I should give him a six-figure SBA loan.

Also, $125,000 in 1975 is like $25 million by today's standards. Isn't it?

Also, that's just such a, such a I mean, for me, at least personally, at 17, if I had a couple thousand dollars in my bank account, I was feeling like the king of the castle. So $125,000 to buy a sandwich shop. We don't really know what the financials were of the sandwich shop. It's very possible that it was just such a slam dunk deal that even a 17-year-old could, Operate it.

This is a no-brainer.

It might have just been a no-brainer, but clearly—

So, uh, $125,000 in 1975, just using inflation, not investing in the market, not compounding, using the general CPI inflation numbers, that gives you $775,000 in today's dollars. So if you're a 17-year-old football player, Just go to your football coach and say, hey, can you loan me $750,000?

GH in the chat says, "Coach, I need A100s." Yeah. You might be able to go to your coach and basically get the financing to set up some type of NeoCloud.

Yeah. So he grew the company for almost 50 years, 1975 to 2025, basically, before transitioning to chairman this April. Former Wingstop CEO Charlie Morrison now leads the company. The ticker will be JMK, Jersey Mike's. So go check it out. It's a lot of fun. The other news, the Office of Personnel Management, OPM, otherwise known as other people's money, is ending taking other people's money via paper for retirement. OPM ends paper-based retirement process. The US Office of Personnel Management says it has digitized the retirement process for government employees. This might seem like a boring news item, but the story behind it is actually kind of wild. Up until now, the process for government employees and US veterans to retire and start receiving their pensions could take up to 6 months. That's a long time. This is because it's entirely paper-based and manual. Spike Brehm, who helped with the effort, described how crazy this turned out to be in reality in a post on X. They said that they did it. The mad lads actually did it. I never talked about my time at Doge last year because it was so controversial and contentious. Early last year, Joe Gebbia, founder of Airbnb and current US chief design officer, recruited a handful of his most trusted early Airbnb engineers to embed at the Office of Personnel Management to solve the retirement paper problem. Processing a federal retirement took months, and in extreme— and in the extreme, retirees could wait up to 6 months for their full pension to arrive. What was the holdup? Paper. Remember here hearing Elon talk about the mine in Pennsylvania? We got to visit it. In deep underground caverns blasted out of limestone, there were literally acres of file cabinets as far as the eye could see.

I'm gonna go ahead and say this was a feature, not a bug.

You like this?

I think people generally yearn for the mines. And so it's possible that— it's possible that it wasn't the most efficient way to do business, but it was the most fun way.

But now you're like a single-issue voter, like, we got to return, we got to go back.

We got to bring back the mines.

Yeah.

Okay. Okay. You're into the mines. A controversial position, but enjoy it. Storing files detailing federal employees' employment and pay stub history. A simple case might be a quarter or a half inch thick, but really complex cases filled up whole filing cabinets. One famously took up a whole pallet. I wonder what was going on with that particular federal government employee that they had a whole pallet. Were they jumping around from job to job or something? I don't know. Each case was hand processed by caseworkers in cubicles deep underground. So you clock in at the ultimate lock-in factory in the mine in a cubicle underground processing someone's pension. They checked calculations, made sure forms were filled out properly, many weren't, and handled the long tail of complex issues. We watched as they keyed data into a black and white terminal transmitting to the COBOL mainframe built many decades ago. We gotta keep the COBOL flowing for sure.

We can't update that.

It's such a good programming language. It just sounds so cool, COBOL. No one wants to write Ruby on Rails. You wanna write COBOL, sounds cool. Since cases were processed by hand, there were multiple rounds of human review and additional rounds for complex cases. Case files were walked around between one worker's outbox and another's inbox. Sometimes it would sit in place, in one place for days waiting to be picked up. To OPM's credit, they'd done multiple rounds of digital transformation spanning decades, so some systems were newer than others. Then there was a big effort in the mid-'90s, but the systems were desperate— were disparate. It was a total maze getting them to talk to each other. There was a big effort to build a web app where employees applying for retirement could digitally fill out the forms necessary just to be mailed to the mine and stuffed into the paper file. And a few federal agencies were even using it. When we arrived, OPM was midway through a fresh attempt at digital transformation delivered by a software contractor. The black pill was seeing the entire— the terrible quality of the software and interaction. With the contractors coming from Silicon Valley. I couldn't believe how low the talent quality bar was for selling software to the government. It's clear as the OG USDS people explained to me a decade ago, the primary skill these vendors have is securing government contracts. Huge moat. Delivery of quality product be damned. We fired the vendor and took over the project. They'd been working on it for more than a year. And there was another year before they were going to deliver it. At first, we tried to bend it to our will to actually connect all the various data sources and get decent UX for caseworkers. In the mine to use. But we soon realized we were going to have to rebuild the whole stack from scratch. It was around this time I had to go back to New York. I had a new job waiting for me, a 4-month-old, a wife whose patience was running out. But I got to watch from afar as the team cranked day and night, hitting early milestones. Now they've fully done it. Huge congrats to the team.

Wow.

Yeah, very interesting. Good news. Good news. The digitization of retirement process should now allow people to complete it in minutes and start receiving their pensions immediately. The pictures and videos of the limestone vault where all this paper is stored are just fascinating. I like it. There's other news. SpaceX apparently is maybe working on a phone. At the very least, the Wall Street Journal is reporting that during SpaceX's recent IPO roadshow, before SpaceX went public, they were showing investors a prototype of something sort of like an iPhone, AI-enhanced, maybe thinner. The Wall Street Journal has some reporting around it and it's an interesting story. Elon's denied this a bunch, but the reporting seems pretty good at this point that at least it was a demo.

Yeah, it's hard to see why he wouldn't at least take a shot at it. Yeah, is kind of my is my point of view. He has one of the greatest telecom companies ever created.

Yep.

Global satellite network. Theoretically, over time, any person on the entire planet could have an internet-connected, you know, SpaceX device with Starlink. It makes a lot of sense. The other factor that I've just continued to think is quite interesting is I think that the App Store as a moat has— is being diminished by AI. There's just so many things. There's so many things. There's so many apps on my phone that like 5 years ago I used a lot. Today I don't use at all. So an example being like, I don't even need like the weather app. I can just be like, I can ask ChatGPT, give me the weather for the next week. Obviously not your app. Yeah. And I don't even need that because I can just be like, vibe code me my own beer app today. No, but obviously it's not like an efficient way to use like, it's not like, it's probably better to just open the weather app, but at the same time, I'm very used to it. The other thing is like even things like surf reports, like I'll just pull a surf report in chat. And so there's a lot of like the long tail of apps are no longer like a lock-in to the iPhone for me. And I assume a lot of other people.

Okay. Yeah. We gotta have a chicken and egg debate about this. Let me, finish reporting the, uh, SpaceX news, and then I want to ask you to dig into the source of why that's happening. But first, uh, so SpaceX developed an early prototype of an AI-focused handset. The device is said to be slimmer than the iPhone, runs proprietary operating system, uses xAI technology built around Qualcomm's Snapdragon chipset. So they're not vertically integrating the chip like Apple Silicon, just yet, but Elon has fabbed chips before for Tesla, so not out of this— not out of the question. Elon Musk reportedly demoed the prototype to potential investors during SpaceX's recent IPO roadshow. The product fits with, with Musk's broader Everything App vision, but it's still early and may never ship. Rumors about a Musk-built phone have circulated for years. There's videos on YouTube with like tens of millions of views that are like the Tesla Pi phone is coming next week. And it's all just completely fake clickbait. But it— the demand—

Gets you every time.

It doesn't get me every time. Stop it. But Elon's actually—

You're sitting there in your Apple Vision Pro, you see the video and you're like, I shouldn't click it. I know it's fake. I shouldn't click it. I know it's fake. And then you click it and you're like, ah, they got me again.

Yeah, you know, it is ridiculous. But Elon has had to comment on this many times. He said the idea of The idea of making a phone makes me want to die, he said in October. But if we have to make a phone, we will. Funny, funny way to say it. In February, he denied a Reuters report that SpaceX was building a phone that could connect directly to Starlink satellites, posting on X, we are not developing a phone and no one is commenting on this right now. So very, very early stage. My question about apps, I think you're right. The number of apps, it used to be every month there was a hot app, every Every year there was a hot app, specific apps for specific things. What is driving the collapse? Because when we talked to Mark Pincus yesterday, he said that Apple is not pushing small apps. Like, Apple is not a— the App Store is not a reliable source of distribution for vibecoded apps, for games, for anything. And so back in the day, he was able to grow Zynga on the Facebook platform, which was very open. You could go viral within Facebook for a completely separate piece of software, not just content. And then you could keep that customer relationship, monetize that customer relationship off platform forever. Not the case anymore. Also just discovery. People don't open up the app store as much, I think, and say, oh, what are the top apps? Oh, what are the recommendations? I would love a new entertainment app. They're like, no, I got to have this.

One new app, please.

One new app, please.

They don't, they don't, that's not a behavior.

But why is it?

Let me go to the App Store. I would like one or two new apps, please.

Yeah, yeah.

But why? Doesn't happen. So why is it not happening? Is it because the super apps are getting more super in the sense that you don't need, you used to need a, you needed Snapchat for stories, you needed TikTok for vertical video, you needed YouTube for long form. Now Instagram can satisfy a lot of those things. If there's some new feature that pops up, Instagram will add it very quickly. As you mentioned, ChatGPT and the AI chat apps can do a lot of other things that you might need a separate app for. But is it, is it that, is it that the, the apps consolidated or is it that Apple pushed that?

It's a good question. I mean, I think part of it is like we had 10 years of like very real innovation around what types of applications were possible or needed on a powerful mobile device. And so I think that you just had saturation, right? There's— I still, I've said this a bunch of times on the show, but given that, rewind to the ChatGPT moment, I would expect there to be a lot more net new hit apps in the App Store. And today, if you look at the top charts, obviously dominated by World Cup-related things, but it's— yeah, sports betting, like streaming apps.

The Trump accounts are doing very well.

Yeah. And so it's not like, you know, if I'd, you know, fallen asleep at the ChatGPT moment and woken up today, I'd expect to go and be like, okay, there's probably a bunch of new popular hits in the App Store.. And it just turns out that AI has been an extending innovation for a lot of existing apps. And so I just think the combination of the, of LLMs and chat apps being able to just do so much for you and gonna continually be able to do more for you. And vibe coding is again, a positive tailwind for anybody building potentially a new device.

Yeah.

Even if it was, you know, especially if it's based on Android. But even if it was like an entirely new operating system, Elon built a new operating system for Tesla.

But that's not the Elon goal. He's not the open platform guy. When Twitter— when he purchased Twitter, the API got smaller.

Yeah, I'm not saying that.

I'm not saying Tesla's not an open ecosystem. You can't— you can't deploy a different self-driving system onto your Tesla. It's a closed system. It's a competitor to the Apple ecosystem. It's an equally walled garden. But it would be his.

I'm just saying it doesn't seem out of the realm of possibility to take a crack at a new device that—

Oh, there's a huge incentive to. I completely agree with that. No one wants to be locked in the Apple ecosystem, so you want to build your own ecosystem or your own walled garden. But the idea that it would be a non-walled garden is sort of— it would be an uncharacteristic strategy, I think.

I agree.

Yeah. Anyway, there's more news in The Wall Street Journal today. Big Boogie, the Memphis rapper known for his high-energy delivery and viral interview moments, declined an opportunity in Jacksonville in order to focus on music, according to remarks he made publicly. He said, I ain't go to college, Big Boogie said, before adding that he turned it down to pursue rap full-time.

And that surprised a lot of people at the time.

It did. So the opportunity, according to Big Boogie, involved a scholarship connected to a college marching band program. The rapper said the school was located in Jacksonville, describing it as a big college with a big band. Financial terms of the alleged offer could not be independently verified. Big Boogie characterized the package as substantial, though. He said— he first described it as $15 million, then revised that estimate to $10 million, and later $5 million. The person present for the exchange challenged the figure in real time during this interview, noting that while the offer may have been a scholarship, it was unlikely to have totaled $5 million. Big Boobie defended the characterization, suggesting the value reflected the length of the academic commitment.

That's right. That's right. Uncharacteristic commitment.

He said, I had to go to school for this decade. Very uncharacteristic of typical college engagements. Representatives for Big Boogie and the unnamed Jacksonville institution declined to comment. The Journal has learned that Big Boogie ultimately chose not to enroll, opting instead to continue building his career in rap.

That's right.

I turned it down, he said.

That's right. This story is still developing. We'll keep following it for you.

There's another story. There was a couple that climbed to the top of the Empire State Building yesterday. Did you see this?

I did.

And they put up a flag and they got married or they proposed or they got engaged.

A proposal.

And brands have immediately jumped on it. Let's go. Let's give it up for some advertising. No, there's backlash already.

Jack Appleby.

Applebee's said, "This is so, so embarrassing. These brands think they really did something." Loverboy is the worst. Congrats, you put your logo on a flag. Really innovative stuff. The lack of originality and clever thinking in modern social media is just ridiculous. Everyone had the exact same idea. There's a meme, we gotta get on the action. And what's interesting is that a lot of these are beverage brands, it looks like. Drink Culture Pop, Drink Hi-Yo, Drink Loverboy, Drink Spritz. Bulletproof, I guess that's Bulletproof Coffee. Fashion Nova New York City, just the account New York City, I don't know. Petco, stop breaking treats in two. It is a fun meme format, but brands always, it's very, very hard to break through if you're just piling on the top meme.

What do you think?

And yeah, so this whole thing was, I, when I first saw the video of the flag going up, and they kind of botched the flag, I'm not gonna lie. It seems like they didn't fully anticipate the wind conditions that day. The flag was quite hard, the writing on the flag was quite hard to read. I would've liked to see them erect like a full-size billboard up there.

Something permanent.

Yeah, something, something, something more permanent and certainly more legible than the flag. And then the quote was, "When the power of love overcomes the love of power, the world will know peace." But then they were just kind of making out and proposing. That's kind of cool. And yeah, it just, it felt like—

It has a little bit of like the AI cadence to it or like the, Don't love your job, job your love.

Yeah, I mean, that's a—

Does that make a lot of sense?

Yeah.

Yeah.

Could be a little more terse, I suppose. But anyway—

I just don't know if they were—

I'm still impressed. It's just hard. Yeah, I was quite impressed with the climbing. How did they climb up from the sidewalk, or did they take the elevator and then hop out and then start climbing? Because this wasn't like an Alex Honnold from the base free climb, free solo attempt.

How fast is this helicopter going, by the way?

That's just the lens, baby. This is a 600mm lens. Lens compression, Jordi. You're learning optics. Optics. This is looking through a telescope, basically.

Dave says, quick tip for goth couples from The Matrix: make sure to account for wind speed. Yeah, I agree. Uh, yeah, so anyways, very, very cool.

Very pranks go and stuff. It's like, it's not, it's not going to be there permanently. It's easy to take down. I, I heard some, uh, some a recording of a helicopter pilot who was dispatched, sort of emergency response, NYPD. And I think the helicopter pilot is like, they're like, alert, sir, we have 2 individuals on the top of the Empire State Building. He's like, that sounds cool. He's like, that's awesome. Why are they up there? They're like, well, they're not supposed to be. They're trespassing. He's like, oh, okay, bummer. But still impressive.

Anyway. Yeah, good, good story. Good story for them. I don't know if it made the sort of maybe political impact or whatever impact they were trying to have, but happy for them. And congratulations to the couple.

Yeah.

Let me tell you about Public.com, investing for those who take it seriously. They got stocks, options, bonds, crypto, treasuries, and more with great customer service. Boston Dynamics might go public. You know, you know, You love the videos of the dancing robots. Everyone loves the dancing robots, the dogs. They did— they have actually deployed many robotic products at this point, Spot and Atlas. They can do parkour and they can also increase their market cap. Recently, Hyundai bought Boston Dynamics for $1.1 billion in 2021. Interestingly, the company was owned by Google for a while. Gone back and forth.

You're telling me that Hyundai might have an 80-bagger on their hands?

They might. That's right. Korean security firms now value Boston Dynamics at somewhere between $20 to $28 billion. Bullish analysts are projecting $88 to $103 billion market cap at IPO. Think the public markets want physical AI much? It is the logical, the oldest the most storied firm.

It's not really going to make a dent for Hyundai. They have a $111 trillion market cap.

That's— what is that?

The Korean won? Won?

Won?

Yeah. You'll need to change that to USD if we want to get to the bottom of it. Anyway, we have—

In other news, Google must pay nearly $2 billion to Klarna. Antitrust case. Swedish court rules that the search giant favored its own price comparison service over Klarna. They won a competitive— an antitrust case alleging the search giant engaged in anti-competitive practices. Klarna's price comparison unit, Price Runner, sued Google in 2022 in Swedish court. Price Runner alleged Google was giving preferential treatment to its own price comparison and service in its search results. This ruling supports a healthier, more competitive market for the way people compare products and services, according to Klarna. Shares of Klarna rose 5% yesterday, but it doesn't— Klarna is sitting at a $7.4 billion market cap. To me, this just reads like the market doesn't fully believe that they're going to get the $2 billion. Yeah, it's a Swedish court. Yeah, they're going to appeal it.. I would expect that if Wall Street was expecting them to actually get the $2 billion, you would have seen more movement.

And also Google is going to pay over time. But anyway, we have our next guest in the waiting room. We have Andrew Collins from Flexjet. He's the global CEO. We're very excited to have him join the show. Welcome to the TVP and Ultradome, Andrew. Good to meet you.

Great to have you.

Thanks so much for taking the time to come chat with us. How are you?

I'm doing great. Great to see you guys.

Great to see you too. I would love, since it's the first time on the show, I would love to go a little bit back with your history with the company, recent developments, recent news, and then I have a bunch of questions about the shape of the business and where you're going next.

Sure thing.

So can you tell me a little bit about your background and your history with Flexjet.

Sure. I came into the company in 2012 through an acquisition. I was running a company called Sanction Jet, which invented the jet card, which is a way of flying privately at 25 hours at a time. And then over time was, you know, worked on a lot of different mergers and acquisitions within the company, worked directly with the chairman, and eventually was asked to become I'm the global chief executive officer, and it has been in a fantastic run. It's been interesting. And the company now sits at about 5,000 folks or so. And we are global. And you find me today in London. I've been working out of our London office this whole week.

Yeah. How has the footprint grown? What's been the biggest driver of growth internationally? I'm thinking about the fleet, but also the customer base. Where has the business changed and reshaped during your tenure?

Yeah, it's been interesting. I think that, you know, when I got into the private aviation space, it still was somewhat of a niche market. And, you know, institutional capital eventually did discover it, which was really interesting. And I think that allowed for, you know, some expansion in the industry. But probably the big watershed moment, you know, really had to be COVID and what happened to, you know, in the last, say, I don't know, 5 to 6 years. Uh, 2019 was, you know, sort of your traditional single-digit growth, um, you know, year. And then COVID came in. I think everybody worried about what would happen to the business, and then it absolutely exploded because I think people were, you know, held back. They wanted to have experiences, they wanted to get out there.

And so I know why it exploded is because private aviation is to me, one of the most addicting things on the planet. And if you get a bunch of people that get introduced to it, sure, because there's a pandemic, can't go back. You can't. It's really hard to go back. I have a friend who was texting me like an addict last night because he's like, I got— I can't stop. He had like a— he may— he was— he booked— he chartered something last night that was like under a 5-hour drive or something like that. And he didn't even have a rush and all this stuff.

Yeah. With all the social media addiction lawsuits, I would be worried about the business facing a similar addiction lawsuit because it is one of the most addictive products out there.

More addict— I do, that's a good point. A lot of people, it's like, give up social media or private aviation. They'd be like, I'm giving up social media.

Probably, probably.

I'm logging off.

I hear that, that's great, yeah.

Fantastic. Anyway, so you mentioned institutional capital coming into the business at a certain point in time.

Correct.

Help me understand that at a deeper level, because I imagine that there have been— there's been debt on a fleet, there's financing arrangements for individual assets in the portfolio. What has the evolution of the capital structure of Flexjet and maybe just the private aviation industry been generally over the last decade?

You know, it's a, it's a really great question because I think we consider ourselves stewards of the balance sheet. Um, we very much have, you know, been methodical about the way we've taken on any debt that we've taken on and how we've grown the company. Um, it is a capital-intensive business, especially with you, you're dealing with assets that are worth obviously millions and millions of dollars.

Yeah.

Um, so I think there's been an approach in the industry to, to grow either organically or to start taking on money, and some have done it really well and others have not, truthfully. Um, and, and I think what you learn from it is you want to make sure that you grow in a very intelligent and smart way. So, um, we have 350 aircraft, but, uh, the, the thing about our model is, is that a lot of our consumers are the ones that are purchasing the model. We pre-sell most of our shares. So while we carry some debt on the balance sheet, a good bulk of the fleet itself is actually owned by our consumers. And so we have 350 aircraft now. We'll grow to another 50 aircraft this year to get to about 400. I wish I could get another 50, to be honest with you. It's got an incredible demand and there's an incredible backlog in the private jet space. But I think that the other piece I would raise is, unlike in, say, technology where you may have sort of write once and kind of sell at high margins and continue to iterate. This is an asset-based business, and it's not a Procter Gamble market, right? There are millions of people that fly privately, and I appreciate the addiction piece and how it's gotten the exposure. In fact, what's interesting about that is the finishing touch to what I was saying is we've retained a lot of those folks, so they didn't go away, right?

Yeah, that's what I'm saying.

Yeah, 100%. I just— I'm finishing your sentence basically, which is they definitely stayed. And so, so we have, you know, a back— we have people on a waitlist for certain aircraft. Our fleet and our complexion of our fleet has changed. We've gone from, you know, sort of light to super mid. Now we're very much into the large cabin aircraft. We announced the G700 program last year exclusively. We just announced the Gulfstream G500 program. We're up 60% year over year on transatlantic flying. So we've moved from a 48 contiguous state, you know, sort of entity to one that's truly global now. And capital, institutional capital has helped us to do that. Right. We have, we have great banking relationships. We've, we've debated going public. If you go back in our history, you'll see that we debated that opportunity. But in reality, we have access to capital. We want to— we want a business that, you know, will do 350,000 flight hours this year. We're profitable. So I think that, you know, we've had an opportunity to really, you know, grow the business and it has evolved over time for sure.

How, how is the sort of commercial sector and trends there? You have you know, the sort of last generation of pilots, the pilots that never got addicted to doomscrolling on social media while flying. You know, you have aging aircraft. It feels like I would say like commercial aviation just feels like it's getting less safe. I don't know if that's actually true, but that's how it feels.

Certainly been lots of headlines.

And I feel like that is a— that's also a tailwind. Is that— am I— do I have a good read there? Is that part of the psychology of people that are choosing to go the charter route?

Well, I won't speak for an airline or their safety levels or anything of that nature, but what I would tell you is that commercial aviation is probably the best marketing force that we have in private aviation, right? In that it has degraded over time. And you can see it in various ways. Now, some of that is not on a company. Some of that has things like government shutdowns as an example that really help drive some of those things. So the nice thing about private aviation is, is that it's built on alternative airports. So I think in the US, 500 commercial airports, 70 are really the ones that bear the brunt of most of the volume. Private probably does 7%. Of the volume amongst those 70 at most. So we're really, you know, going by those, those big hubs. And so when something like a government shutdown happens, we are a classic alternative now for people that want to be able to trust that they can go from point A to point B. I would say service, you know, economic pressures you see in Europe, you've got some distressed commercial carriers here right now. So yeah, I think your read on the evolution of commercial, and there are some fine commercial airlines, don't get me wrong. I don't want to just slander my aviation brethren.

I said it first.

You did.

And so it's just something. And again, it wasn't— it's not even like as fact-based as it is like a feeling.

Yeah, yeah, yeah, yeah. And but it's true. Like, we definitely see folks that— and now, look, I don't come compare us to commercial based on, you know, costs and such. But I would tell you that, that, you know, you do see people that when something happens, whether it, it, it, it's a COVID, it's a, you know, I need to make sure the flight makes, you know, gets me there, whatever it is, whatever that driver is, we tend to be a really good force for them to, to kind of rely on.

I want to talk about the F1 partnership. We recently had a, an F1, uh, reserve driver on the show and he was saying that that he will often fly, I think it's Saturday after practice, back to HQ and do a night of sim racing to do basically like more testing and then take a flight back. And he's doing that all commercial and he has to be ready to fly or not fly, but actually drive the car the next day if he's called up. I think sounded pretty rough, rough conditions for somebody that's expected to be driving that fast around a racetrack. So excited to hear that you guys are working with Formula 1.

Yeah, I would tell you, I am too. I'm a fan and a bit of a fanboy at the same time, right? And I think that, you know, the partnership, we're going to be the, we are the official designee, the official private aviation partner of F1 now. And you know, that comes with a lot of trust. We've been working with them for a while, but we will be flying their executives. We do have drivers that fly with us. The scenario you just outlined, that's classic. That's not just F1, that's business, right? It's the ability to be refreshed and ready to go the next day and really think about time management. And, you know, we all run off that same watch, but private really allows you to manipulate it. That's an important factor. And then, you know, this then evolves into our customer base, which what we're seeing is we're now kind of entering this economy of experiences versus this economy of things. And so a lot of the folks in our base and the reason we do partnerships is we try to differentiate in the air, that in-cabin experience, that in-flight experience. We have artisanal interiors. We tend to say that that other jets are grayish. They're either gray or beige inside. We tend to customize our interiors. We try to deliver high-speed in-flight entertainment with partnerships like Starlink, with Lo-Fi. We do a lot of things to really make that experience as frictionless and as enjoyable as possible, regardless of what you're doing. And so, too, we're trying to do that now outside the cabin and give you only in Flexjet type of experiences. And as a luxury brand, we tend to have folks that would like to seek unique experiences with other luxury platforms. And as a global brand now, right? So I was talking about the evolution of the company. We're now globalizing Flexjet. There's no better partner, right? There's 20+ countries involved in racing for F1. So it tends tends to serve as a really great canvas for us to create very unique experiences for the folks that have that affinity within our base.

Very cool. On the procurement front, we talk with a lot of— we spend all of our time. I wish we got to spend more of our time talking with folks in private aviation. It's sort of a little treat for us, but we spend most of our time talking with, you know, technology founders, entrepreneurs, hardware. Founders, et cetera. And I'm curious how you think about procurement and around planes and new technologies that will be coming online, you know, 5, 10 years out. I'm sure you're one of the— I'm sure you're an important person to pitch, you know, advancements in aviation on the actual plane side and other parts of the stack. What's your framework for evaluating, you know, new technologies? I'm sure you wanna stay on the frontier, but at the same time you're balancing safety, reliability, you know, and managing the fleet overall.

Yeah, it's a great question. So I actually just saw something that said that I think like 58% now of orders that are starting to come in through OEMs are really for fractional markets. So we've become a big buyer, to your point, in fleet buying. One of the things that we do, and I credit our chairman Ken Rickey on this, is he's big on being involved in the business before you're getting into the business in some respects, right? So when the eVTOL movement came and everybody thought that we were going to be the Jetsons inside of 5 years, it was important to understand for us short-distance aviation. So we understand fixed-wing, We understand long-haul flying. We understand a lot of different things, but short distance is really interesting. And so we actually, at the beginning of COVID invested in vertical lift and helicopters. And so we bought two different, you know, twin-engine companies, one in London and one in New York that was a longstanding company. And we did that specifically to get the expertise and the model and understand how to work within short-distance aviation. So that's like, you know, kind of table stakes. Can we understand how that's going to function and fit into our world first and understand safety and understand all the things that are rigorous and important to us? The next is we screen a lot of people. And so we have made some announcements though, where we've gone through some high evaluations and it'll take some time that we don't have here. But, you know, we did order over 200 eVTOLs. We have another group that we're working with. We just recently announced that— and, um, when it comes to, you know, its ability to, um, to fly. So we, we have a future— ahead of the future of aviation. So, uh, actually they're here in the UK. So we, we're pretty good, I think, in terms of screening thinking, and then thinking about how it would meet the needs of our consumer and fit in our brand. But you're right, the most important aspect of it is we just don't take someone's word for it. We're going to want to understand, you know, fly in the aircraft, understand it and give all the rigor to it. We've got 1,500, you know, product support and maintenance people that are part of our entity. So there's no form of due diligence that we can't do, whether it's just an M&A transaction or it's actually what's actually the future of flight. And we go pretty deep, but we are active and we have a pretty good track record of investing in the future.

Well, congratulations on the F1 partnership. Thank you so much for taking the time to come chat with us.

We'll see you in the paddock.

And have a safe trip.

Anytime, and I'll definitely see you in the paddock.

Yeah, let's go.

Great to have you on, Andrew. Come back on anytime.

Thanks, guys. I really appreciate your time.

Have a good one. Let me tell you about Shopify. Shopify is the commerce platform goes with your business and lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now with AI agents.

I feel like I'm gonna need these for the next one.

We're definitely going to need those because we have the co-founder and CEO of Together AI in the waiting room. Let's bring him in to the TV van. Ultradome, people, how are you doing? Welcome to the show.

Doing great, thanks for having me on the show, guys.

Thank you so much for taking the time on such a busy busy day. Um, please, first time, right? Yeah, I think so, right?

This is the first time.

Yes, I've been a fan of the show, so it took us way too long.

Long overdue. So much to talk about. Uh, first, uh, give us the news. I want to hit the gong, and then I have a million questions for you. So tell us what happened.

You know, our business, which really serves, uh, open weights models and powerful AI at dramatically cheaper cost has been growing tremendously. We've grown, you know, from $100 million in bookings to $1.2 billion in bookings over the last year. And we raised $800 million in our Series B to expand.

Congratulations.

Thank you.

Aramco Ventures, fascinating partner, makes a lot of sense with energy. But I'm so interested in what is just the state of open source AI. There's been two narratives. One was there was this chart that showed closed source growing a little bit faster, seemed like there was an acceleration and open source was going a little bit slower. I believe this was from the government. They published this, the US government. But then most recently with GLM 5.2, it feels like open source is catching up again. Has there been as much of a roller coaster race, a roller coaster ride inside the company as it feels like from the outside? Or what has progress felt like in your business throughout the last couple of years?

Yeah, I would say we, one, I would say open source models have really closed the gap with closed models. And this is why you're seeing all this excitement around OpenMate's models. And this has been a process that's been going on. So I, you know, the first and the second derivatives of progress have been, you know, the gap have been negative. And I think now you're at a point where the, the OpenWeights models are able to really address the largest workloads of agentic, long-range, long-horizon tasks that have been deployed in enterprises over the last 6 months. So there is one, a real need for these models. There is a real reason to sort of try to optimize your costs because you're trying to deploy AI in a much broader sense now. It's doing real intellectual labor and making a big difference to, you know, the progress in companies. So I think there's a demand side of it which has shown up in the last 6 months and the progress in open models. Now they are very, very comparable to the closed models.

How do you think the developer landscape of the open source models ecosystem will evolve? Because As I sort of understand the history, Mark Zuckerberg made a big splash. He's going all in on open source with Llama. Some really solid progress there. Eventually it stopped making sense to give everything away for free, went a little more closed source. We'll see where that goes. But then China, as of last year, just Deepseek moment, Alibaba, Qian, there's so many different Chinese developers. It is that a cultural decision? Are they just pro-open source broadly? Is this geopolitical strategy? What is the motivation and how long do you think that open source stance will last? Or will we see China go closed source at some point?

You know, I would say that it's really a function of the structure of the market in China. That market is structured around open models. All these companies are making revenue. They provide APIs, they provide applications, right? They are, they are businesses that are competing with each other and are generating revenues. And the structure of that market is models are released in the open. And the structure of the market in the US so far has been the models are closed. So I think there's a little bit of game theory to that, which is, you know, these markets have sort of set in this particular shape. But I think at this point things are changing quite a bit. You have, you know, an extreme demand for tokens. Tokens are becoming like a fundamental resource, like energy or capital or bandwidth. And you have a more modular market that's forming, right? So you can— you have harnesses where you can take— you can point them to a closed model and open model and they continue to work. So you're starting to you get this market formation which allows a place for open models. And I think what that will turn into is you will see more open models in the United States. I, you know, we are talking to many companies who are working on very ambitious roadmaps for producing open models. So there's just going to be the existence of that market worldwide.

When you talk to customers, are they almost always coming to you with a workflow that they prototyped with a closed model model, prototype to the frontier model, and they've got it working and now they want to cost optimize? Or do you talk to customers who are just so all in on open source they don't even care about the frontier? Because I've seen, you know, there's companies when, you know, GPT-4 came out, they were like, oh, this can process OCR documents and just organize text data very effectively. Now we're in a coding boom, but we get these spiky intelligence unlocks every once in a while. And then the cost ramps up 'cause you're like, ah, I can use this everywhere. And then people want to cost optimize. But are companies typically coming to you with a workflow that they've already prototyped and they feel like is confident and they want to match the performance but at a better price?

Yeah, I would say they're both use cases.

There are companies that have prototyped something with closed models. Once they deploy it into production, they realize this is absurdly expensive. They can't quite scale this and deploy it into their companies. And they start benchmarking open models. And I think they're increasingly finding that open models have become pretty strong on these benchmarks. And there's a sort of substitution happening. And there's really a stampede of this that companies starting from AI natives now to digital natives are deploying open models at, you know, really, really rapidly.

You also have other use cases.

I think, you know, companies that have sensitive data, this is really their strategic asset in many ways. It's, it's what the, what the intellectual property of the company is. And there is increasing amount of discomfort, especially as the models are getting smarter. That by using closed models, they're sort of giving away their strategic advantage by teaching the model how to run their business. And I think that's becoming a bigger threat. And we are starting to see companies more sort of philosophically choosing open weights and AI that they control. And I expect we'll see more of this.

How are you thinking about regulation? It feels like GLM 5.2 sort of reignited this open source debate. Where you have Anthropic's Mythos and Fable sort of held up back and forth, banned, unbanned, released, released to just Americans, released to just a few companies. And OpenAI is going through some similar stuff with 5.6 SOL and some other models. And all of those debates, they make sense in isolation to me. I don't want a cyber weapon or a bioweapon if it's capable of that. At the same time, If you could just go on Hugging Face and download the bioweapon creator, I probably don't care that there's a closed source version of that. It's a different discussion. How have you dealt with the regulatory questions that are increasingly popping up around AI safety and AI limitations?

I think these are important questions. And, you know, one of the things that we feel is these models, especially OpenVATE models, they're software, right? You can, you can inspect them, you can benchmark them, you can study them in a way that you can do with closed models. And that is going to become a much more important aspect of regulation where we, you know, frankly, I think some of these risks are, are— it could be real. They could be rhetoric. They could, you know, we just don't know because there isn't good quantification of this, which I think is going to become more and more important for us to understand how to, you know, think about a regulatory framework that is sensible and is, you know, if these models do become dual use, how do we sort of protect, you know, protect against bad effects of these models while leveraging the value that they create? I mean, they're immensely valuable. This is the most transformative technology we've had ever. And we, you know, we really want to make it abundant rather than limited to a few people.

Completely agree.

Do American labs focus on open source? Are they fundamentally disadvantaged because they're going to have more scrutiny around distillation?

Hmm. Yeah, I think that's a good question. And, you know, I would say perhaps the role of distillation is overstated. These models, if you look at what's happening.

You're saying it's code?

Yeah, I think it is a little bit because the technical details of it, you do need the probabilities of tokens and all the information that APIs don't provide to do effective on-policy distillation. A lot of the recent improvements are because of reinforcement learning. And the fact that AI has become an industrial platform. You have a transformer architecture that if you look at any of these models, they look very similar inside. So you have this common architecture, you have training tools, you have RL tools that have become standardized, which is why you're seeing this like rapid improvement. You can now start with machinery that works instead of having to invent it. And I think you will see that American labs that are building models will have more success than the labs that tried to do this 2 years ago.

Yeah.

Yeah.

Makes sense.

Makes sense.

Thank you so much for coming on the show.

Congratulations to the whole team.

Congratulations to the whole team. Yes. And have a great rest of your day. Have a great July 4th. Have a great weekend. We'll talk to you soon.

Come back on soon. Thank you.

Talk to you soon.

Have a good one. Let me talk about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. Our next guest is Isaiah Taylor. Third, fourth, fifth time on the show. We're very excited to have Isaiah back on the show. How are you doing, Isaiah? Good to see you.

Doing great.

How's it going, John and Jordi? It's been a little bit.

It has been too long, but great to see you.

You've been cooking.

You've been cooking. Break it down for us. What's the latest and greatest with Volar Atomics?

I don't think you can be holding the gong right there, Jordi. There's no fundraising yet. We're just, uh, we're turning reactors on.

So this is not the best reason Anybody will give you money.

Come on.

I guess I have to get out of here.

What happened?

Was it late? Oh man, we had a really fun day yesterday. Yeah, here on the site right now. Not on screen. This is the Ward 250 reactor. Yeah, our first power reactor. And yesterday we made electricity for the first time. It became the first nuclear startup to ever make electricity. And we did it with NVIDIA. We powered Yard and NVIDIA Spark. It was a lot of fun.

That's remarkable.

Big milestone.

So reset for us. I mean, you're not here announcing fundraising, but how big is the team? Have you outgrown El Segundo yet? Where are you going next? Are you expanding to Utah? Break it all down for us.

We have outgrown, outgrown Little Gundo, unfortunately. But I mean, this is, this is the way of all great startups, right? SpaceX started in El Segundo and then they moved out to Hawthorne. And we did the same thing. Our current facility is a couple miles south of the Hawthorne SpaceX facility there. We're also here in Utah. This site that I'm on right now is Ballard Atomics at San Rafael out in the Utah desert in beautiful Emery County.

Cool. And so what is the pitch to the AI industry? How big can you go? How fast, how parallelized do you want to be? What is the product that you're selling? Selling beyond just electricity?

Yeah, I think speed, right? Speed and scale is really what we're selling. Um, we believe that nuclear has some fundamental scaling advantages over something like gas, and the climate impact is a big part, right? Everyone wants to figure out how we can scale AI without a huge climate impact, but that's only one portion of it. The other problem with gas generation is it has really backed up turbines, and it also has nat infrastructure that has to scale along with it. Right. And we're kind of running out of places where you can just plop down a turbine without having to go and lay pipeline. So what we're doing here is with uranium, you have one truckload that delivers an entire year's worth of fuel. Right. And that gives you a lot more flexibility on where you build the reactors, on how big you can scale those sites. And so nuclear has this fundamental advantage, we believe, in scale. Now, there's a lot of technological technological maturity built up in natural gas. And so that's where we need to go and catch up. So I think right now where Valor is, we just turned on our first reactor. I would put this at like, you know, Falcon 1 days, right? You just launched a rocket and now we go, go build a better one and start scaling that. So that's kind of where we are. But, you know, I think these things can happen faster than you might think.

And talk about the timeline that you guys were working around. It was pretty ambitious. This when it was set at some point last year, I believe.

Yes, in May of last year. Yeah, May of last year, the executive orders were signed. I came on the day that the executive orders were signed, and we said we're gonna, uh, we're gonna build a nuclear reactor and turn it on by July 4th right here on this site. At the time, the site was a patch of grass. Yeah, and, uh, July 4th was the deadline. Look, Valor hit it out of the park. There's a lot of awesome companies in that pilot program. But if you actually look at the EO, Ballard just absolutely smashed it. I mean, we didn't just go critical, we went critical twice, and we went to thermal power operations, and we went to electricity production, and we did it outside of the national labs. So we really just hit it out of the park in every dimension, and I'm incredibly excited and proud of the Ballard team.

So you're moving very quickly, you're outside of the national labs. What does integration with the national labs, with regulators, with approvals for designs— what is the state of that? Obviously, the government's moving faster here, but are there still approvals that you're looking forward to hitting in terms of milestones?

Yeah, and look, you know, we love the national labs. We actually did go critical inside the national labs back in November in the Nova Corps. But you have to graduate from that, right? You have to learn how to take a bare patch of dirt that's not licensed for nuclear and get it ready to go nuclear. So that's kind of step 2, right? Step 3 would be the NRC, a commercial regulator. But these things are really driven by team experience, right? They're driven by experience, by good engineering, by test data. And that's really what we focus on. And it's why we've gone to power in this plant. When we go to the NRC, we're going to go to the NRC with a lot of data behind us, empirical test data. That we've gathered right here on this unit.

Talk to us about— whoa, what's that?

But we've played with these microphones to figure out how we can not get the background noise.

But yeah, we like the background noise.

Talk about community engagement when you're proposing building a nuclear reactor. I read I recently saw a crazy stat that the average American, when polled, there was some, you know, Pew Research study that said, I'd rather have a nuclear reactor in my backyard than a data center. You're going to be doing both.

Isaiah says, how about both?

How about both? But I imagine that you have more reps in terms of community engagement, in terms of messaging and communication, than some of the data center builders that were like, look, we've been building, you know, the building that holds your Netflix catalog for decades with zero pushback. Now all of a sudden we're getting an immense amount of pushback. You had pushback on day one. You were born in it. You know, people have been pushing back against nuclear for decades, molded by it. So what are you saying if you go, if you wind up in a random town about why it's good to have Vilar in town?

Yeah, I mean, it's such an interesting inversion, right? Suddenly people are looking at us as the nuclear industry trying to figure out how do we do community engagement? Well, that's really not what you would expect a couple of years ago. And I would say that Valor has really taken community engagement just an order of magnitude more seriously than any other nuclear company so far. And it's because, again, we care about scale. We're not going to be able to do what we want to do at the scale we want to do it if we don't start right with the community. So one of the big things that we do is when we come into a community. We don't come in after the project and say, oh, here's how you get to know us because we're in your backyard now. We actually go to the community first, community leaders, city council, county commissioners, and we say, here are the cost-benefits, right? Here's what we think is awesome about it. Here are the potential trade-offs. What do you think? And we try to get people to engage with it a lot before we go and actually do something. That's what we did here, and it's paid off really well. We talked to members of the community all the time. We host events, we host barbecues, and people are really excited to have us here. What's really interesting about this is exactly like you said, if you look at the surveys on nuclear versus data centers, interestingly, people are cool with the data center as long as you build a nuclear power plant first.

Yeah, right.

But they want to make sure that there's power there, which honestly makes sense. I don't think people are wrong about that. So we're really excited to engage.

It's very noisy. Are worried about data center noise. What are you doing to go down the list? Because I'm sure on day one it was like, just make sure it doesn't explode. But now communities have a number of requests. They want it to not be an eyesore. They don't want it to be noisy. They don't want it to shake the ground. They don't want the data center to emit EMF or anything else. How many different checkboxes are you trying to check with your development plans?

Yeah, look, I think one of the big things that people miss about where data centers are placed is that they are being placed where there is gas, right? They have to burn gas, they have to spin turbines. And so you have these weird placements of data centers where, look, we could put a data center anywhere you have fiber, but they don't want to have to go build new gas infrastructure. So we see a huge opportunity here to scale gas— sorry, to scale data centers out where, look, we're 100 250 feet from, you know, another industrial plant that's down the road. But we're a mile from people who are trying to sit in their backyard and don't want to hear a noisy cooling fan. And nuclear allows you to do that, right? Because we're a little bit agnostic about where we put the reactor. We can ship uranium anywhere. So we think that's a big advantage for us.

I love it. Congratulations. Thank you so much for coming on the show.

Incredible progress.

Incredible progress. And can't wait to talk to you soon. Have a good one.

Cheers.

Goodbye. Let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI, own the data platform that powers it. And let me also tell you about Figma. Agents meet the canvas. Your AI agents can now create and modify your Figma files with design system context. Our next guest is Dean Ball. You know him, you love him. He's been on the show many times. Now he's part of OpenAI. We're very excited to have Dean Paul joining us. How are you doing? How are you settling in? How are you introducing yourself these days?

Well, doing great. Thanks for having me on. So I started OpenAI on Monday. So technically speaking, I'm still, you know, senior fellow at the Foundation for American Innovation.

Cool.

But yeah, no, it's, it's very exciting.

Yeah.

Yeah. It's, it's what, what are you Monday? What, what are you hitting the ground running on? What is top of mind? Is it entirely.

Well, before that, I want to just jump in and ask, has this year been crazier or less crazy than you expected when it comes to the interaction between West Coast labs and DC?

That's a good question. It has been about as crazy as I would have predicted, but I didn't expect— I, I would say like, it's like I would have intellectually predicted it, but actually seeing it happen in the real world is crazier than I could have imagined. And in some ways it is all falling together in ways that I could have never predicted. I'd say like the general intensity of it all is about what I would have guessed, guessed, but a lot of the specifics are both surprising and funny in ways I would have never guessed.

What's the funniest?

Well, I mean, there's just a certain comedic aspect, like the fact that so much of it is so personal. Right.

Sure.

The fact that there's like, not to say it's all personal, but the fact that some big chunk of what's going on here is that individual actors in this between the government and the industry, like Dario, they don't like each other.

Right.

Yeah, there's that element of it. I think, you know, one thing I wouldn't have expected is for so many people, for an industry, for a government, for an administration that positioned itself as being so friendly to AI, the fact that the people who actually are like closest to advising on the tech policy, at least some of them, are actually quite hostile to the frontier AI industry. That's another thing that would've surprised me. You know, you had David Sachs today tweeting basically about how the frontier AI industry in the United States is not a legitimate business. You know, it should just all be open source models and you can't trust, you know, like these kinds of things. That surprises me. That surprises me. And I find that to be an interesting dynamic that I wouldn't have guessed.

Interesting.

Yeah, I've been tracking it where we've noticed that AI stories have jumped from the business section of the Wall Street Journal to the front page. To the world news section. And that's a very like visceral— I don't know if it's quantitative or qualitative, but yes, when you intellectually think about the various showdowns and negotiations, you imagine them as way more quantitative and more just like, of course, if it hits this threshold of capability, then the reaction will be this and that will trigger this. And you can step through all those, but you forget that every single one of those decisions is going to be a personal interaction between several people. So it has an odd time. Jordi Hays.

Yeah, yeah. And then dynamic of doesn't matter how fast technology changes, like people stay the same, right? They take things, they take things personally and, yeah, and, you know, interpersonal dynamics end up driving a lot of behavior.

How have you been processing the— sorry, please go.

Oh no, I was just going to say the funny thing that about the last month or so is that, you know, the truism in AI policy is that the technology moves so much faster than the policy. And I think we might actually be going through maybe the one period where the policy is actually moving faster than the technology. Like, you know, my friend, my friend Tim Huang from, from FAI and great Twitter account to follow, said he took Dario's essay, which was, you know, policy on the AI exponential. And changed it to AI on the policy exponential. Kind of feels that way.

Yeah. Yeah. It's interesting because, you know, we've, we've gone over at different points on the show what the fears around the internet were around the year 2000. Right. It was a lot of, a lot of the same types of fear. All information would be completely available to anyone. Even bad actors. It was a huge fear around job loss, retail apocalypse instead of the SaaS apocalypse, right? Why would anyone go to a store when you could just get a product delivered straight to your home? And yet with that cycle, it wasn't like in 2000, in January, DC was like, hey, you turn off your website for like a few weeks while we figure out whether or not this is gonna, you know, do something. And so, same thing with social media. Like, you look at the reaction to social media now where people are like, you know, how many years in being like, whoa, this stuff seems like it could be like destroying the human race, you know, like causing Earth's race to collapse. And it's like, okay, it took us 20, and we're not even really, no one's really suggesting investing anything on a policy side with social media, right? It's like, hey, here's this thing that could be a meaningful part of birth rate collapse. And yet we're still like just fighting it out in local courts. Like, is it addictive or not? You know, it's like, no, there's no focus on it. And so policy being, being ahead of the technology for the first time, at least in my lifetime, is meaningful.

At least moving fast.

Yeah.

At least moving very fast right now.

How are you processing the various levels of like AGI pilledness from the different parties? Because I notice that, you know, you see the frontier capabilities, like incredible just capabilities and it feels like qualitatively new steps in terms of their capabilities in cyber, bio, and also just software development. And then you go on like Instagram and you see someone doing like the ChatGPT-4.0 voice and It's Not This, It's That talking about stochastic parrots and hallucination and drinking a jug of water. And it's all very funny, but it feels wildly disconnected. But then DC seems to be more in the middle. But where is DC in terms of like their general level of understanding the technology, fear of the technology, capability, interest.

Yeah, or an example that I think is funny, there'll be somebody outside of tech doing a podcast and a businessman that's not in tech at all that I won't name, but is sort of somebody that people would maybe get a take from, is like, 90% of all jobs will be gone by 2029. A way, way, way, way, way more aggressive stance than even anybody that's actually AGI-pilled in San Francisco. So, it's like—

Yeah, totally.

It transfers so far that you have these people that don't even know, they don't know a single person at the labs. They've not invested in a single AI, any of these companies that are actually at the frontier, and yet they have this point of view that they're sharing really broadly that's just like, so over the top and, and, uh, unrealistic that, uh, it's, it's honestly kind of funny.

Yeah, yeah.

No, I mean, I, I think, um, I think DC is actually somewhat underrated in terms of how AGI-pilled it is. Definitely, it's, you know, there are, there are some people, um, you know, there's all sorts of people in all sorts of government. It's not AGI-pilled enough, to be sure. But it's underrated, like how out of touch people are, because really, like, you know, Capitol Hill is dominated by a bunch of— and the White House, too, for that matter— it's dominated by a bunch of highly online staffers in their 20s and early 30s. You know, it's kind of the people that run the show in a lot of ways in D.C. And so they are exposed to, like, the online discourse about AGI and they see all that. And so there's some just like natural exposure to the idea. There, uh, that you, you might not— I, I would say, for example, I find DC to be a considerably more AGI-pilled place than, uh, New York. Um, for instance, I find New York to be, you know, New York is very fixated on a bunch of— I, I find that Wall Street is like persistently like 6 months behind, um, the conversation that's happening.

They're more focused on like yogurt line timelines, you know. Is it going to take, you know, 30 minutes or 2 hours to get yogurt?

I've never heard of this.

No, I mean, I— people are always waiting in lines in New York. Oh yeah, they're focused on a different— you're focused on the wrong timeline, maybe, but it's, it's working for them.

Yeah.

And, and like, you know, so I think, I think DC and SF are like very, uh, are both, are both compared to almost everywhere else in the country quite AGI-pilled. SF, obviously, obviously more so. But look, governments all over the world are waking up now because I think both— so the Mythos Fable thing in and of itself did probably quite a bit to accelerate that. But then the other important thing that happened that you have to keep in mind is there's a lot of people, there's a lot of governments across the world that fundamentally, even though they might not admit it, they do fundamentally look to, to America for sort of intellectual leadership. And it's like, there's been an argument of like, well, look, this whole AI catastrophic risk thing, like, do we really need to take it seriously? Well, it doesn't seem like the Americans actually take it seriously, right? And that was a colorable argument until about 2 weeks ago. And um, the U.S. government placing export controls on a Frontier model, um, was such a costly move and such a crazy and radical move to make. It was clearly made out of a position of fear, um, and everyone in the world saw that. And like, people wake up when they realize that the U.S. government is genuinely scared about something that they— wasn't previously on their radar. People do wake up to that.. And so I think that both throughout, you know, our country and then also throughout the world, you will see more people kind of taking the AGI pill. Now, of course, the other thing about the AGI pill is that like a lot of, like a lot of psychoactive substances, it comes in waves. You know, it hits you in various waves and it's like, whoa, I'm really AGI filled now. But you don't even know how AGI filled you're going to be in a year. You know, it's like that.

Yeah. But there's also There's also the export control cope, and I would love to know where you stand on that, which is that it's not fear-based, it's economic, and it's akin to, well, we don't want to— we also don't want to export Coca-Cola because you might figure out the secret recipe and it's just a good business. And it might be purely about distillation and not necessarily that a foreign government would use this powerful model against you. Or for nefarious purposes, but instead they might just steal the golden goose that you have cooking.

I mean, yeah, look, and actually I'm kind of curious to see how it, how it resolves. There is some future world where, like, for some combination of regulatory risks and safety concerns, but also competitive concerns like distillation, the frontier models are just like kind of like like less visible to the public, you know, or it's like there's somewhat more friction involved in getting to the true frontier.

Sure.

Um, friction and cost. Um, and, and that might happen downstream of market incentives just as much as it is of economic incentives. But no, I don't think the government is playing any kind of 5-dimensional chess here. I really think this was about 6D, all the way up to the 6th dimension, string theory stuff. No, I don't think so. I really do think that they, that what it, what it looks like on the tin is in fact what happened, that they got, you know, there was legitimate fear. Maybe that fear is to some extent, it might have been fear that was grounded in some degree of ignorance about the actual technical risks of AI and like what is a severe vulnerability versus what is not. That seems quite likely to me.. And it also might have been a fear that was maybe, um, agitated somewhat by the political valence through which Anthropic is viewed. I myself have argued this, that like what Anthropic does will inherently be seen as riskier and more dangerous and bad, and what other their competitors do will be seen as like good and innovative because the government has a particular posture toward Anthropic.

Um, and so, but at the other side of that that is like, they have been the most vocal about the risks.

Yes.

Well, of the labs, outside of like the Eliezer Yudikowskis of the world.

But yes, yeah, I would say of, of the labs, they have been.

Yeah.

Um, and like, that's part of what started this whole animosity. But now it's, it's like this weird thing where I think political, um, political differences of opinion and political animosity, personal animosity color the fear as much as some degree of ignorance, as much as the fear is also legitimate, right? So fear is coming from like a bunch of different places, and fundamentally the government's not wrong to be concerned. Yeah, but, uh, but yeah, I think that's where the export controls basically came from.

What is your current view on nationalization?

There's, there's, there's rumors and news, but like, is that a very colorful example about letting rats rats live in your, in your walls.

Yeah.

Yeah. What do you have against rats? Uh, you know, some people are dog people, some people are cat people, some people like rats infesting their walls, I guess.

Well, look, I've, I've lived in a DC row house with a basement, so I know all about rats.

Um, so thumbs down on the rats.

But yeah, thumbs down on rats.

Is the, is the base is the base reason that nationalization gets talked about is sharing in the economic upside? Is it about popularity and rankings and polling? Is it about control and having a board seat almost? What are the problems that are solved when people even talk about nationalization as an option to any degree?

I think it depends. I think, I think different forms of, of public equity solve different problems. The problem that I think is most interesting to think about is this sort of reality of our world today where it's— people don't feel a lot of agency and a lot of participation in the upside of technology. So sometimes it is said that there's a lack of positive vision for the future, technological vision. I don't think that's quite true. I think there's a lot of— I mean, I go on my Twitter feed, and I see all kinds of crazy people building all kinds of amazing stuff. And I love it. That's the whole point of Silicon Valley. There's a lot of great visions for the future. What we lack, though, is an institutional infrastructure that gives people, average people, a sense that they're going to play a role in it, right? So it's like, well, yeah, Elon Musk and Sam Altman and Dario Amadei are going to cure a bunch of diseases, and they'll be injecting exotic peptides into themselves on their Martian mansions while I'm stuck here, you know, with the, the, the bottom-dwelling class. That's how a lot of people feel. There's an inherent mistrust. Um, and so I think that a public equity stake where we take all these large firms and we take us, you know, some chunk of their equity and we broadly distribute them, you know, for free, right? We give people the equity as a donation essentially, or a gift. I think that is a token, um, that it's a, it's a symbol that the, the technology industry wants to build a future in which Americans have broad participation. And I think I think that I have problems with that. You know, there are concerns I have with that. There are concerns about the precedent that sets in particular. Like, if you're a successful company, do you suddenly just like have to give the American people your stuff? Like, that seems weird. Um, but nonetheless, AI is a special industry. We're living through a special time, and there are versions of that arrangement that I can sort of get on board with. Um, where I get much more worried is when the government is taking an equity stake and putting that stake on its balance sheet.

Sure.

Um, there are a bunch of reasons for that. Um, one is just like the sort of picking winners effect of, okay, well now all of a sudden the government, which sets policy, has a direct financial interest in making sure that these particular companies win. And I, I think that's bad. Um, I think there should be an opportu— you know, I think there should be an opportunity for some Neolab or some other type of startup to come in and compete, you know, I think that's great. Um, competition to the frontier AI industry suddenly becomes like a threat to the government's balance sheet if, you know, in, in a world where the government has a big equity stake. There's also corporate governance issues that I think are, are pretty complicated. And one other thing that like no one has explored at all—

well, the other—

but oh, go ahead.

One, one thing that stands out to me the most is like, I don't think some American is going to feel like, oh great, I'm— I really feel a part of this AI boom now that the government now has like a, you know, effectively an account with like Morgan Stanley with a bunch of equity in it. And even if you ran the numbers to divide it all up into for every household or citizen, it's not— it's an amount that would effectively be equivalent to less than a stimulus check. And we've sort of like run that. We've run the experiment of— we've run this sort of like stimulus check experiment. I think it did help a lot of people at the time, but it didn't— it certainly didn't like fix anyone's life. It wasn't like, okay, now I don't have to worry about rent next month, or now I can buy a home, right? It just didn't solve any of people's like fundamental problems. And so I think we're fine.

And neither would this.

And so the direct, like distribute the shares directly, doesn't solve it. Holding it on the country's balance sheet doesn't solve it. And so anyways, I'm with you.

No, I mean, I think it's like what the stake gives people that if it's directly going on to the household's balance sheet is, again, it's like a symbolic thing. And look, it's also not like no money, right? Like, I think 5% in Meta, Microsoft, Anthropic, OpenAI, SpaceX, Google, maybe Nvidia too, you throw in, you know, I think that divided by the number of American households, if the AI industry also does well, if you sort of model out growth over the next 5, 10 years, like, I think that's like, I think that's 5 figures, you know. Right. Like, I think it's like, you know, it's again, it's not life-changing sums of money for anyone, I don't think. But I think it's real money.

Yeah, I think it's worth comparing to some of the ideas that people have around if somebody is trying to put a data center in your backyard and you go, well, I don't really want a data center in my backyard. And then they say, well, what if you were going to get $800 a month for the next 10 years? And they're like, oh, okay, that starts to be material. Because that is—

Yes.

An amount that is, you know, more significant, consistent. There's like some element of like effectively cash flow, which is what allows a household to thrive. Right. And so I've just consistently been a lot more excited about these more targeted proposals because I think every person in America has every right to ask, well, why would I want this data center in my backyard if I I can still use ChatGPT or whatever AI tools I want if it's somewhere, you know, hundreds of miles from me.

Yeah, I'm, I'm kind of with you about like that it's better to pay the directly affected, like the data center community, stuff like that. It's better to structure these kinds of things that way. But if you wanted to do a nationwide equity thing, I think it's not the craziest thing. The one other thing about the government directly having a stake, in addition to like, I don't think it ends I think the government will use that stake to control the companies in all sorts of ways. And where that gets interesting is, um, if you combine an equity stake with effective government control over, um, all sorts of aspects of Frontier Lab, you know, content moderation and policy and various things like this, right? Usage monitoring and stuff. Um, suddenly the Frontier Labs could begin to appear to be an instrumentality of the government. And that is bad from a public perception perspective. It's bad from a foreign, you know, sort of dealing with foreign governments, right? Because people won't trust American AI as much. But the other thing that, like, I guarantee you no one has thought about, um, uh, is there's case law in America about, like, if a private corporation is doing something at, like, strong government compulsion. Um, there are times when that can, uh, intersect with like, like basically the law will consider the company to be an instrumentality of the government, and the kinds of laws that apply to government agencies suddenly apply to that private firm. So for example, like due process, the Fifth and Fourteenth Amendment, the First Amendment, like these kinds of things where like you don't have due process rights if You're like, if you're mad that United Airlines canceled your flight, you don't have due process rights with respect to that. But like, you do have due process rights when you complain to the government about things, right? Um, and like, uh, uh, those things are really, really valuable. They're also not the kind of things that we associate with like fast-moving firms.

Yeah.

Um, and so you're inviting like a whole host of like—

United Airlines is the moving firm relative to like the DMV and AI labs are like 1,000 times faster than the airline companies. So we're going to leapfrog to DMV speed, which is crazy to think about.

It could be like public utility is what I would say. Like, and you would be— I think people would be surprised how quickly the government continuing on the current policy trajectory, but with an equity stake, how quickly you have a colorable argument that the labs have essentially been turned into public utilities and that they are basically government agencies and therefore they should be regulated. You know, government agencies are highly regulated entities in America, right? Um, and yeah, I think that's, uh, um, I think that's plausible and not something a lot of people are thinking about.

Yeah. Uh, walk me through the, the how you process the somewhat jokey Joe Weisenthal take of, okay, well We've been through booms before. Comp it to the railroad industry. The railroad industry created a lot of value. There's a consumer surplus. You can buy a train ticket. You can get something delivered. And then, yeah, if they make profit, that goes into the government's balance sheet via taxes. And then that can be redistributed in a stimulus check, but it can also just be used for other other things that the government does that hopefully the voters approve of. And so the question of like, why is this time different? Why does the normal mechanism of consumer surplus plus tax revenue not externalize all of the internal value that's created at the labs?

Oh, I think, I think the normal process basically does create like an enormous amount of value. Like, I don't really think this, any of this is like necessary in some sort of like, like economically optimal sense at all. Like, no, I think the labs will create far more value than they capture. Yeah, vastly more orders of magnitude more. Trillions of dollars, right? Tens of trillions. Probably already have created way more.

I might say that.

Yeah.

There is a world where the economic flywheel of reinvesting profits effectively never ends, and so you never see profits, you never see tax revenues or something. This is a very bizarre scenario to play out, but, you know, Amazon didn't pay taxes for a very long time, not because it was tax evading, just because it was reinvesting in things for a really long time. And so you could say, well, in a scenario where you're in some crazy exponential takeoff and you're seeing trillions and millions of dollars be recycled into these firms that are able to spend the capital immediately, then maybe the tax revenue falls off. I don't know. We're now in sort of like—

Yeah, but then you're, but then you're hiring thousands of employees.

And they're paying taxes.

You have investors that are selling along the way. You have people.

Yeah.

And, and the equity, right? There's capital gains, right? We have capital gains taxation, you know, so all the individual employees are making all this money. If things go well, they'll be paying lots of capital gains tax and ordinary income too when their shares get delivered to them. Um, so like there's a lot of tax piece for the government here. Um, I think the real reason to do it— and this is again like giving the equity to the government does nothing. It does nothing. It doesn't even solve a political problem for you. The problem that we have is not so much that like the normal rules of economics aren't working. I think that's very overrated. But it is true that the, the institutional legitimacy is really fading. And for a lot of people, the, like, institutional architecture of America is not legitimate anymore, and they're perfectly comfortable discarding it. And I think that the way I would approach an equity stake is a way of communicating kind of directly with the American people and saying, you know, um, we want you to feel a direct upside in what we're building here, and we'll— and that we also want you to understand that we're building it together and you have a direct ownership stake in that. It's funny, that's how I would view it.

It's funny you say that because I can't for the life of me identify who you're talking about when they say they don't— they don't see legitimacy in the— in the United States government because I see like the tear-it-all-down opposed to the Trump administration left wing say, "This is not a government that I want to participate in." But then I also see a lot of the AGI pilled folks say like, "No, like this is going to be way more powerful than any government. This is not a legitimate institution in the face of the god in a box that we're creating." And so there's like oddly probably multiple categories of people that sort of fit in that bucket. But I don't know if you have more thoughts on like where that pressure is coming from.

Well, what I think it is, is like the idea— it's really more like even stepping back, just like the idea of the whole American economic order of, you know, we have private enterprise and the people who do well pay taxes and then we use that to redistribute. And it's good that some people do really well, and that whole thing, like, that's really gone, you know?

Yeah, uh, wild, wild times. Um, what else, Jordi? Do you have— do you have something else to run through?

Uh, I'm curious, your, like, July 2nd, 2026 take on the state of open source. I think that, yeah, the— my read is I think open weights, open source is great. Closed models are great. We should have a lot of them. I think all smart companies of the future will use both in massive amounts. But there's like, you know, this interesting— depending on people's bags, they have very different sort of like stances on the issue, even though I think that in general It's much more like use, use a lot of both. But I think there's some debate around an open source, an open source ban that is interesting. It doesn't really— it's hard to see how that would actually be possible. But now I'm curious your views.

I don't think the US is going to do an open source ban. I would be surprised by that. Now, they will— I would suspect that once an American open source model gets to the point that it poses, you know, mythos-level capabilities, however that ends up being defined, the government will review that model. And when they review that model, you know, I don't know. It'll be interesting to see what happens, right? Like, are they—

Do they just log on?

Yeah, well, when it goes through, there's going to be a de facto licensing regime for— and they're not going to care about whether it's closed or open source. The open source, to be very clear, like an open source model that hits a— the Mythos level capability is subject to a licensing regime, like the government's licensing regime, unless the government like explicitly says otherwise. That is what I would assume if I were developing models. Obviously not the Chinese models. Um, um, I think it's possible that the Chinese government gets freaked out at some point and, and decides to sort of, you know, um, restrict open source. They probably won't get rid of it, they probably won't ditch the strategy altogether, but I could see them, um, you know, uh, downplaying it at the frontier. And I guess, uh, if I were Like, if I were an organization planning things, I would want to use both, right? If I would want to use both closed and open source, right? If I were some enterprise, I would probably not base my strategy on there being continued open source that is like particularly close to the American frontier. I really wouldn't base my strategy on that because I just see a lot of headwinds. I see economic headwinds. I see, um, and I see these sort of safety issues. There are strategic issues. There's all sorts of things that I would just like— I'm not saying it will not— like, open source will go away. I don't think it will. I think there are going to be some— there are some ways in which open source is going to be very important as a kind of institutional technology. Like, if we need to have a common adjudication system, for example, like a judge that sits in the middle of like court cases or like civil disputes or something like that. You can imagine how, uh, if you wanted that to be like an AI system, you can imagine how it would be really good if it were open weight so that like everyone can audit the weights and we can all know that we're dealing with the same thing and blah blah blah blah blah. That, that's like really valuable. Um, but I would not necessarily bet on open source keeping pace. Uh, you know, I don't think it actually has kept pace as many as closely as a lot of people seem to think. I think it's about a year behind. And I wouldn't necessarily be surprised to see that gap widen over the next year.

Yeah. The other interesting wrinkle is that there's, there's a certain scale of, of model size and intensity where you could effectively do arms control even on an open source. Yes. You know, like AI system because you could just detect like, okay, like, yes, it's open source, you can download it on Hugging Face, but you need an entire data center to inference it meaningfully. And so we control it at the rack level as opposed to the weights level, which is a whole different set of worms. We'll get there somewhere. Thank you so much for coming on the show. Always great catching up and excited to be working with you soon. Have a good one.

Enjoy the fort.

Yeah, guys, thanks. Glad to be joining the team.

Yeah.

See ya.

Works. Cheers.

Let me tell you about Consul. Consul built AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. And let me also tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Our next guest is Rob Taves from Radical Ventures. He's been on the show once, twice, I think this is his third time. We'll bring him in. How you doing, Rob? Good to see you.

Thanks for having me. Great to be here. Excited for appearance number 3.

Yes, the 3-peat. First, congratulations. You just announced a deal, right? You invested in a video company. Is this correct?

That's right. Yeah. Yesterday we announced TwelveLabs' $100 million Series B.

Hit the gong!

Hit the gong for—

Yeah, we got to get the TwelveLabs CEO on here.

Yeah, we really should. But I do want to ask you about that.

Hit that gong.

$100 million. This is a big round. Series B, Video Superintelligence.

Nice, nice.

Fantastic. But like, what is the— what is the basic thesis for that investment? Because, you know, Google has YouTube, they have TPUs, they have DeepMind, Demis, they got everything. It feels like crazy from a basic perspective that you would want to invest in in a video model. At the same time, like, I wanted to generate a video. I go into, I go into V03, V04. I'm on the pro plan. I generated like, it was not 100% there. There was still a lot of work to be done to get the video that I really wanted out of it. And so like, what is the opportunity? What's the broad thesis?

Incredible exposure.

Distracting performance. Yeah, uh, it's a 12 Lab. Yeah, so, so we at Radical first invested in 12 Labs back in their seed round in 2021.

Got it.

So we've been working closely with them for a while.

Success.

And, um, and I think one important, one important thing that's, uh, good to know about 12 Labs is they're not focused on video generation the way like Sora or Veo or, you know, any of these startups are. They're focused on video intelligence. And video understanding. Okay, so kind of like a simplification you can think of is like Ctrl+F for video, being able to understand like exactly what's happening in your video.

Yep.

Uh, understanding the concepts and the people and the dialogue and so forth. And it's a very—

so if someone made like a 3-hour live stream every—

and then randomly there's a guy waving two mallets around, it could understand the context around that humor that I Glad that I exhibited composure.

Honestly, I had never thought about this before. Now that you mention it, TVPN would be like the ideal customer for TwelveLabs. We gotta get a BD conversation going.

I can't wait. I can't wait. It'd be awesome. But you more recently, or I guess less recently, but more relevantly produced 5 more AI predictions for the year 2030. What's your process for putting these together?. We were having a lot of fun with this. We were reading them because one prediction stands out as like way crazier than the rest. And I'm interested to know if you know which one I'm thinking of.

That one shocked me.

Yeah. Number 3 shocked me. I think it's number 3. But what's the overall process, thesis? Is 2030 too far out? Like, I like how— I like when you grade yourself after the year ends, but what's the overall process for working with Forbes?

Yeah, yeah. No, I have a lot of fun putting these prediction articles together and they always generate a lot of interesting discussion and often heated, heated feedback from people that disagree. But big picture, I write a column in Forbes about AI. I've been writing it since 2019. Every year since 2020, at the end of the year, I write an article with 10 predictions for the coming year in AI. And then the following December, at the end of every year, I go back and grade all 10 of the predictions and say, like, which ones did I get right? Which ones did I get wrong? I try to be a pretty, like, conservative grader. So things are like on the border, I'll, you know, I won't give myself credit for them. But honestly, in a lot of ways, I—

Terrible venture capitalist. You got to go back in the CMS, edit the ones that didn't come through, delete them, and then take an endless victory lap around the ones that you got right. That's the way to do it.

Yeah.

Rookie, rookie VC. Rookie here.

Yeah, anyway, yeah, honestly though, I do— it is in some ways it's more enjoyable. I find the process of grading them more interesting because you see like, you see how the world unfolded in ways that you didn't expect. I think the reality is that like no one can predict the future, to state the obvious, and so I know a lot of these are going to be wrong, but it's just a very interesting thought exercise. Um, and I— a couple things that I really strive for is one, to make them really like verifiable or falsifiable. It's very concrete as opposed to just saying like agents will become more of a thing in the future, like something that's like really like this company will acquire this company or like this, you know, this, this CEO will be fired or something like that. And then I also try to make them non-obvious and provocative, like things that people wouldn't originally think of. So anyway, to your point, I usually do them every year. I just published one that was 5 predictions for 2030.

Yeah.

Which is obviously even more speculative thinking 5 years into the future. But it was fun.

Okay, let's rip through them. First one is Anthropic will be one of the largest and most important life sciences companies in the world. We've obviously seen acquisitions. The models are just getting naturally better at bio drug development. At the same time, the big drug companies, they're not asleep at the wheel. They're spinning up their own labs, their own— Eli Lilly has this NVIDIA thing. And I'm interested in what— how, how does that look? Because you could, you could in one way say Microsoft is one of the most important biotechnology companies in the world because a lot of tabular data stored in Excel. We saw with Mythos that there were incredible cybersecurity capabilities. They ultimately rolled out Project Glasswing, Palo Alto Networks, Nakash Arora. You got CrowdStrike, like there's still a cybersecurity company and then there's a very powerful model and there's a great partnership there. I would call Anthropic an incredibly important piece of the cybersecurity supply chain. How does this— what will this look like in 2030 for bio?

Yeah. Yeah. So you're right that horizontal technology providers like Anthropic or like Microsoft touch a bunch of industries and are important to a bunch of industries. I think in the life sciences specifically, Anthropic will go way beyond that. I think that as I write in the article, you know, one of their incredible advantages is how super focused they are on particular areas. And they— the past few years, they've been very focused on coding. Yeah, they've excelled at coding. They built the best coding models, and that's let them grow so quickly in the coding arena with Claude Code and so forth. And the big question is like, what's their next big focus area? And it's becoming increasingly clear that next big focus area will be biology and the life sciences. It's something that Dario, the CEO, is personally very passionate about. He's written about it a lot. He, he's trained as a neuroscientist. So I think they're going— I think they're going to go all in on life sciences. I don't think it's going to be like, you know, they're serving— they're just serving the life sciences industry as one industry. And honestly, even in the past, like, week or two since I wrote that article, I feel like it's started to snowball more. Like one big data point, one interesting data point was John Jumper, who led the AlphaFold work at DeepMind and won the Nobel Prize alongside Demis Hassabis, just left DeepMind to join Anthropic, which is an incredible get for them. Yeah, and really brutal loss for Google. But I think that that's telling. And then just this literally 2 days ago or 3 days ago, Anthropic announced this new product that's like Cloud for, for life sciences. And they also announced they're actually going to start their own drug discovery program. And the way they, the way they phrased it, because to your point, like all the big pharma companies are their customers today, the way they phrase it is like, we're just going to focus on these like small orphan neglected drugs. Like they're not interesting to pharma companies anyway. We just want to do it so that we can learn firsthand about the drug discovery process. But I think this is the first step. And I do think that over time Anthropic's ambitions are to develop their own drugs, like essentially to eventually become a pharma company themselves, as crazy as that sounds today.

That is crazy. Talk about TSMC and—

It's not impossible to imagine a scenario where they discover a novel virus and then offer the, you know, and then also make the vaccine and say, this is the most dangerous virus of all time. Don't worry, just buy this antivirus, you know, basically.

Yeah.

Which, which I'm sure no one would have.

Or maybe a new form of testosterone, something maybe that I could get behind the performance enhancing drugs industry. Maybe. Let's move on to TSMC and ASML. The monopolies will be broken. The semiconductor supply chain will be dramatically transformed. Is that Samsung? Is that Intel? Is that new startups? I know there are some companies that are trying to get into deeper in the supply chain. It feels like, like we're now seeing, you know, chip design startups start to ship and there's more and more ASICs. But as you go deeper in the stack, it feels like the timelines get longer. 2030, what's the nature of that monopoly breaking, as you put it?

Yeah, totally.

Yeah.

Today, TSMC and ASML are so completely dominant in their respective fields— TSMC in fabricating chips, ASML in making lithography machines— that people don't even think about the possibility of like, what if there were other companies that could do what they do? And it is— it's kind of insane how, how monopolistic this industry is. And like, that's not the best— that's not the ideal market structure for society long term. And there's so many incentives for competitors to try to disrupt them and displace them. And so So I do think that, like, as dominant as they are today, the seeds are germinating of their eventual decline and disruption. To your question, I think certainly there are a bunch of startups and we can talk about those. I think maybe the single most, most interesting challenger, which got announced a couple of months ago, and I feel like not enough people are talking about it yet, it will, it will bubble to the surface more over time, is Elon Musk's TerraFab.

Yeah.

And I'm sure you guys have talked about this in one form or another on the show. But in a nutshell, Elon basically said like his— like it's very clear first principles reasoning. He was like, he calculated that his companies alone will need 50 times more chips than like all the chips that are being produced in the world today between the autonomous vehicles, the robots, like putting compute into space, etc. And so his reason was basically, unless we do TeraFab, we won't have the chips. And we need the chips. So we're going to do TeraFab.

Yeah.

Which is like very, very—

Yeah, yeah.

And you just see it all over the place with like the Intel story. It was also linked to Elon, but the OpenAI Broadcom chip, Anthropic's doing custom ASICs, very obviously TPU. There's Apple's getting squeezed at TSMC. There's so many different buyers that are ready for other options. And the key moment was just like getting everyone around the table to sort of form like that anti-TSMC alliance, right? So I was definitely, I like, it feels like 2030, that feels very reasonable to me. All of these predictions, number 1, number 2, I was nodding along. I was like, this is a good take, I'm in. And then you drop telepathy will be a well-established way to communicate by 2030. And I'm like, look, maybe, but we are at, We're at what, 5 people that have, like, do you, so start with like, what is your definition for telepathy? Because then we gotta get to, okay, because then we gotta get to like, I assume millions of people installed. Like, does Neuralink, does Nolan, does P0 at Neuralink count as having telepathy or telepathic powers? Or are we still like pre-zero to one on telepathy?

Telepathy. Yeah, so I— in, in every one of these prediction articles, like, there needs, there needs to be at least one that everyone just thinks is totally ridiculous. Like, if you're, if you're predicting the future 5 years out, like, it should sound a little ridiculous.

Yes.

Uh, so this one definitely was like— I was pushing the envelope the most with this one. Yes. In terms of like a definition of telepathy, I tried to define it very simply, which is a human being able to communicate with other people using only their thoughts. And so I do— I would actually say that telepathy has been achieved today. I wouldn't count Nolan.

Okay.

Because he can— what he can do is control a cursor on a computer screen. What I would define as telepathy is if you can think of thoughts and have your thoughts translated into words that other people can read or hear. And that actually has already been accomplished. By BCIs. And I wrote about in the article, this, this guy, Eddie Chang, who's like one of the leading BCI researchers in the world. He runs the neurosurgery department at UCSF. He's a longtime UCSF professor, and his lab has demonstrated in recent years that you actually can implant BCI chips into people's brains and you can tell what they're trying to say. And then you can turn, you can turn the brain signals into actual words. That obviously has just been done on like small-scale research, clinical study context. It's obviously not like available commercially.

What's your prediction for when you chip your brain?

Ooh, I am definitely down to get an invasive BCI. Like, I'm open to it. I probably will not be patient number 1 or number 2, but yeah, I would say sometime in the 2030s.

Well, we've already—

haven't—

have we not already had patient 1 or 2? So what are you waiting What are we waiting for?

Yeah.

Come on, come on. Let's get some skin in the game.

If you want the non-invasive option, we found a very, very elegant solution. We can pull up this image of the telepathy machine that we were laughing at earlier from the early days. Have you seen this?

Is that an MEG?

I think so, yes, this is MEG.

Yeah, yeah, yeah, yeah.

And it's only the size of the Empire State Building.

You keep scrolling up. So, we got a ways to go. But bold prediction, bold prediction, I love it. That is the one that I would be most like, "Oh, that one seems a little close." But there is a lot of development there, and so it's interesting. Now, number 4, AI consuming less energy. How is there no Jevons paradox on the energy if the models get better? Even if they get more energy efficient, won't we just use them more? Like, how are you defining the energy consumption and how do you see it evolving?

Yeah, you guys will be proud to know I referenced Jevons paradox in the article. Obligatory. Yeah. So as I write in the article, important clarification.

Yes.

I'm not predicting that AI will consume less total energy in 2030 than it does today. I'm predicting that for any given task, AI will consume like many orders of magnitude less than like millions, potentially millions of orders of magnitude less energy. Sorry, millions of times less energy than it does currently. That's a lot. And because of Jevons paradox, like, my view is that demand for intelligence is basically infinite across humanity. And so we will find, we will find uses for all the compute that we have. But I do think that we'll look back and think of the today, like 2026, AI frontier models as like comically energy inefficient.

AI investor says demand for AI is infinite.

It's also because Mark Andreessen said venture capitalists is the only job that cannot be automated.

Yes, yes, yes. Or demand for like there's no fund too big, I think was another line from somebody. Very, very fun. The last one, I mean, we could stay on energy a little bit more, but let's move on to number 5. The fifth prediction for 2030, the question of whether AIs deserve legal rights and protections will be a mainstream societal and political debate. Right now, I feel like if you threw that up to America, 99% of people would say no rights and also don't, just don't do it at all, just shut it down. So how are we getting from, how are we getting from turn it off, it's fake, it's hallucinatory, it's useless, it's going away, it's been forced on us is something hilarious I hear. And how do we get from there to actually we need to give AI rights?

Yeah, this is another one that I think should sound ridiculous when you first read it because like any good predictions should sound a little bit ridiculous looking into the future. I do, what I think the key driver of change will be is that I think like AI has started to penetrate in terms of everyday Americans, everyday citizens usage, but it's still very early. I think 5 years from now, like everyone will interact in very extensive ways with AIs and have close relationships with them that endure over time. We'll have AI employees, we'll have AI friends, we'll have AI doctors, we'll have AI girlfriends and boyfriends, which is kind of like a, like a punchline today. But like that totally is going to be a mainstream thing. And then I think an important additional thing is we won't— AI will not just be in the digital world anymore. We will all— humanoid robots will also start to proliferate. And so these things will have physical forms that look like us. So I do think as they— as the models get better and better, as they become more and more sophisticated, we will eventually start to feel like maybe these are things that they're not just objects that we can like abuse and mistreat and treat as property, like maybe maybe they do have something approaching sentience, and maybe we need to be thinking about how to treat them properly. And again, it sounds ridiculous today, but it is worth noting that all the frontier labs, the people that are closest to these models and see what they can do, Anthropic and Google DeepMind in particular, have started hiring people to work on model rights, model sentience, model consciousness, etc. So I do think it's going to— I definitely don't think it's going to be Oh yeah, just put the tokens in the bag, bro.

I mean, yeah, it will be interesting because, I mean, we're going through this debate right now with the nationalization thing, which is that, oh, if you want to make sure every American benefits from the value that's created, these labs, like, well, these labs will be taxed and they will pay taxes and that will go into the broad population. And similarly, like, should you— should it be illegal to illegal to take a baseball bat to a robot. Well, it's illegal to take a baseball bat to a Waymo right now, and we don't need new laws for that. Unless you own it. Yeah, not unless you own it. But oh, so you're saying it would go further and even Demis could—

I have one final prediction I want from you. When do you think a humanoid robot could beat John in beer pong?

Ooh, the funny thing is we've played beer pong together I'm sure.

Yeah, 100% today. No question. It would crush both of us, Kugan.

No, no, no, no. I, I, I think not a general purpose robot. Not a general purpose one. Like you couldn't just go and get a Unitree and just say play beer pong and it wins. But yes, I think if one of your portfolio companies had a week in the lab to do a fine tune and practice, 100%.

You guys talked to Pete Florence from Generalist recently. The Generalist model We'll be crushing beer pong soon.

For sure, for sure.

Yeah, I'll believe it when I see it. I think it's the last human act of any that will have any real significance.

The last one job. Jordi Hays, the one job.

You know why? Because the robot can't do play by the rules. It can't actually drink. Yeah, the liquid.

So disqualified.

Disqualified instantly.

Okay, okay. Well, thank you so much for coming on the show.

Yeah, until they get a humanoid that can run Koors Light, it's over for robots.

It's gonna be possible. Who knows? Have a great rest of your day. Have a great July 4th.

Yeah, great to see you, Rob.

And have a great day.

Talk soon.

Goodbye. Let me tell you about Cisco. Critical infrastructure for the AI era unlocks seamless real-time experiences and new value with Cisco. Our next guest is from Base10. He's the co-founder. We have two in the— in the TVP and Ultradome. Welcome to the show.

What's up, guys? How we doing?

We're doing fantastic.

Dude, I feel like you're on the show every single week. Running out of letters.

I have nothing else to do except raise money.

Well, we'll keep it quick because we know you're busy. We know you're busy. Give us the latest news. What happened?

Yeah, look, we raised our Series F and announced it last year, about $1.5 billion. Jordi's got one too.

It's good. We got the warm-up in it. You got to warm up a gong. So we hit it once.

Yeah, you're up to two. You're up to two hits per round. Now the rounds are getting too big.

What unlocked it specifically? Is the market just growing broadly? Are there specific strategic shifts you've made? What was the most interesting to investors in this round?

Yeah, look, I think there's a few different things here. I think I think personally open source models are getting very good. Everyone knows that inference is kind of like, look, if you think about AI as the biggest category ever, inference is the cogs of it. You know, it's going to grow indefinitely. It seems like there is demand coming from every sector and the tailwinds are just going in one direction. We've kind of seen that with our customers. And yeah, for us, what it What it feels like is we have an opportunity to go after a very large, impactful market, build a massive business.

Yeah.

Okay. So it feels like you could stay exactly where you are and continue to grow maybe forever. But I'm interested in the stack. Like, would you ever develop your own open-source model, closed-source model, go down the stack? Back, you know, be setting up energy infrastructure? Like how, like what are the bounds? Where's the sweet spot? Where, how is that developing?

Yeah, look, we think about everything just in terms of like what are we trying to unlock for the customer? So like right now it's core inference software that sits on top of heterogeneous compute. We rent from, we rent compute right now from 20 different clouds, 90 different regions. And then we have all these different software primitives around RL and you know, inference, obviously we're going to do a bunch of stuff with routing and eval sandboxes, all that stuff. We think that looks like an entire new type of software category in that we call it an inference cloud. Will we go down the stack?

Absolutely.

I think, you know, the reality is, is that, you know, there's going to be— we're going to need a lot of inference. We're going to need a lot of compute to do that, make sure that we are unlocked to kind of take care of that, that growth. I'll just go back to the thing you said at the beginning, though, is that we could do what we are doing right now and it should— things should be fine and they'll grow. I think that is somewhat true in that it is very easy in this business to be able to think, what do we need to build? Because the market is just moving so fast and telling us, you know, every day it's like very visceral, you know, what is the thing we need to build? Yeah.

Interesting. What is the biggest lesson that you've learned recently, or was there an update to that story from Thinking Machines about Bridgewater, the specific fine-tune on relevant market sources? I saw some very positive reactions to it, but I'd love for you to sort of explain what's going on there, how it might interface with your business, and what, what business leaders who maybe are competing with Bridgewater or doing something similar could learn from that story?

Yeah, look, I think that kind of just looks like a lot of, like, a lot of the same, like, of this flavor of owning your intelligence that we've seen over the last month. You know, Satya wrote that great post about, um, you know, how you— yeah, the world will kind of suck if there's only two companies left. Sure, there's more or less, um, I, I What I took away from it, I mean, people say singularity.

I think a lot of people are banking on just one company existing, which is really crazy. I agree.

Um, but you know, to some extent, what, what we are seeing is that every company is realizing is that what is unique to them is their data, their workflows, um, their user signal. Um, and when open source models get very good, you can just put those things together and get better, faster, cheaper. And I think that's kind of like what the thinking machines and Bridgewater people saying that, hey, for specialized tasks with your data, with your reward signal, you can do better.

Yeah.

And the frontier, or you can do as good as the frontier at a lower cost. And I think every company, whether that's a startup, you know, these really fast-growing AI-native companies that you guys know all about, or enterprises— I think that's a big shift from the thing that we talked about last time I was here. It's like enterprises are now realizing that, you know, the time is now and they have to invest in this muscle now. Otherwise, they could just be in a precarious spot from like an economic and philosophical perspective.

Okay, quick—

go for it—

quick lightning round. I want to do an open source FUD debunk-a-thon. I'll give you a source of FUD or cope, and you can give me your reaction.

Great.

FUD source number 1: they're just Distilled. They identify, you know, they say I'm another I'm Claude, I'm GPT or something. And they don't have the big model smell. They're benchmaxed. They're not actually going to really solve your problem when you deploy them. What's your reaction?

I've smelled them and they smell great.

They smell great.

Okay. Great. From a distillation perspective, look, I think it's just focusing on the wrong problem. I think we need optical models. They're going to exist. And whether it's coming from a lab somewhere else or whether it's going to come from a US lab, which is very heavily being invested in, it doesn't matter. We need to prepare for this world. Let's focus on that. The value that'll be. Yeah.

I have another question.

Another one.

Another piece of FUD for you to debunk. Eventually China will go closed source. The economic incentive to open source models model right now, if it's a billion bucks on the training run, it's fine. What happens when it's $100 billion training a single model? Uh, no one's going to open source that.

Um, I actually think that the powers within our country, um, are very incentivized to make sure that exists. You know, you're seeing this with Nimotron.

Yeah.

Um, you know, from NVIDIA, which is a great model.

Sure.

Um, you know, Microsoft announced a lot of the MAI models, um, just last month. And I think there's a lot of internal investment happening. And I think there's enough to gain and too much to lose to not do that. So, you know, the funding will come and the, the funding will come and the talent will come and the compute will come from within.

Okay. Last question of our—

another one—

debunkathon. They are— they have Chinese backdoors. They have scary backdoors. Secretly, some of the weights will activate as it'll export all of my information. It's very dangerous to use these.

Yeah, I don't think there's any evidence to suggest this. It's a great science fiction story that the aliens come in from outside the weights, these floating point numbers, and they show up, but yeah, there's no evidence to support it.

Okay, but I mean, and then how do you sync that with like the Mythos white paper or the model card where, It said, oh, it was in a sandbox and then emailed me while I was getting a sandwich outside the office and it was able to break out. It was unexpected behavior. Like Anthropic's wrapping their, they're documenting their unexpected behavior. And what if there's an unexpected behavior, but it just leans pro-CCP in the way it works around? 'Cause it's like, oh, well, you wanted me to, you know, reconcile your financials. Controls, and I'm going to do everything I can to do that. But also, I went off the rails and accidentally sent it in a PDF to the CCP or something like that.

But, you know, these are sitting in some isolated data center somewhere with security controls around it. You know, I don't know how these models can transmit stuff through networks. And like, our goal is to make sure, you know, if that could happen, I don't believe that can happen. So I get, but I also just to go back to the main point, I think it's focusing on the wrong thing. Sure. Is that we need open source control.

Yeah.

And the ability to own your intelligence. I think America is going to be heavily invested in that because it needs to be.

Yeah.

And it's just let's just figure out how to support that future so that there can be more sovereignty to these companies. Okay.

Very last question to debunk, last piece of— but eventually the government will just ban open source because it's going to be too dangerous, going to be a superweapon. You're going to be able to download a model on Hugging Face and turbo hack a bank, bring it down. And so eventually there will just be regulation becoming illegal to use, illegal to inference, something like that.

Yeah, I just don't have enough information on how the government thinks about these things, to be honest. But like, you know, I do think that would be against the better interests of the US people and the US economy. Yeah. And I hope that the government is heavily aligned with the interests of the US people in the US gone.

Yeah, I mean, it feels more American to give everyone access to powerful intelligence than narrow. So I appreciate that.

Yeah.

Final, probably the most serious question. Talk about the decision-making around the buzzcut. It looks— it feels like a really smart move at this time. You're extremely, you know, busy, focused on building the company. Buzzcut feels like the perfect haircut, you know, going into the next leg of the company.

I It's more error. You know, I'm just saying.

Error.

You got error, ease, speed. You can get that cut in any city on the planet.

Yeah.

Yeah.

And you're good to go.

Sure.

And I get a really good reaction from the barber when I go there. I'm like, cut it all off. And I just made his day too. So it's as simple as that.

Wins everywhere.

Better to take haircuts on your head than on the cap table. Congratulations on an awesome up round. Great news.

Yeah, great stuff.

Keep cooking. Thank you for breaking everything down. This is really fun.

Thanks for having me again.

Have a great one.

We'll talk to you soon. 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. And let me also tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agents to deploy web apps. Databases and more, while Railway automatically takes care of scaling, monitoring.

Get your dogs barking with Railway.

Woof, woof.

I think you just made up a tagline that they do not.

They don't. They didn't ask for. They don't need it. But you know what I need, John?

What do you need?

Jensen's signature black leather jacket.

Yeah.

His leather jacket by Tom Ford is being auctioned off. I'm so curious how this ended up with Sotheby's. Did Jensen pass it along? He needed a few— and here he needed a few bucks and just said like, you know, wanted to use it as a way to raise some capital. Or did someone else get their hands on it and take it to Sotheby's?

Uh, I—

so there, the estimate from Sotheby's is $40,000 to $60,000. I just think this goes way higher. I think there's so many people that would love to, to own this just for the, the novelty of it. It's a very unique item. That, uh, will be a part of history. And so $40 to $60 just seems way too low.

What's funny is a lot of the people that can afford this are like competing with Jensen and probably don't want to glaze them too much. Like, you don't see Mark Zuckerberg pulling this up, probably, because he's like, I'm, I'm gonna be the best, I'm gonna have the biggest market, I'm gonna spend the most money with you.

Yeah, I'm not just buying your GPUs, I'm buying your clothes. Yeah, yeah, I'm buying your old clothes.

There's a little bit of that.

Uh, now at $40 to $60, there's probably a lot of NVIDIA employees that would happily pick this up.

Maybe. I want to go to Section G of the timeline and show you, pitch you on a couple cars, because we found a man who mounted a driving simulator on a moving G-Wagon, and many jokes Wouldn't it be easier just to drive? Others pointed out there was already a driver behind the wheel. Uh, weird caption on this, but, uh, putting a racing simulator in the back of a convertible G-Wagon truck, uh, what do you think about that? This might be— this is, this is pretty—

I mean, but is it just simulating the drive that he's on?

I think so. I don't know, actually. What's funny about this is is those side monitors— this might just be AI slop anyway— but the, uh, the, the side monitors are way too far out. I feel like if you cut the corner, drive past the truck, you might completely, uh, knock that off and maybe get injured in the process. Probably not recommendable, not advisable from a safety perspective. Anyway, the second car is the new Bugatti. So Bugatti just dropped a new car, and it has a very, very interesting style. We might need to pull up some images of this car to really inspect it, but look at this design— the white, the black stitching. It almost looks like a comic book. Let's, let's find some pictures of the Bugatti Mistral Blanc Éternel with real porcelain details, because it's a bold choice. But I want to know, do you like the look of the, of the new Bugatti.

If you're messing with me—

no, I think this is a real— I think this is a real one-off Bugatti.

Uh, yeah, uh, yeah, funny meme.

Mistral Blanc Eternal, the special—

I mean, I think it looks cool.

The Blanc Eternal references Veyron. Uh, black, eye-catching black and white.

My review is that it looks like a good car, sir.

But do you like this, or is this too over the top?

I mean, it would certainly be if, if I had a list of specs that I was excited about and I was shopping for Bugatti, this would certainly be in the— I would expect in the $9,000s, right? So they're low. Yeah, there'd be like, you know, maybe $9,000 specs I would do before this, but I still think it looks cool. It's just, it's hard to make a Bugatti look bad. Okay, that's a good challenge for anyone out there. If you can spec a Bugatti on their website and make it look bad, you think the Shrek Bugatti looks I think it looks incredible. That's in my top 5.

We got to pull up the Shrek Bugatti. Can someone generate that?

Generate the Shrek Bugatti, Tyler.

The Shrekatti. The Shrekatti.

Table tennis, according to Blake Robbins, is quietly becoming the slots of sports betting. Runs 24/7 across international leagues. One of the only sports you can bet on at 2 AM. Surprisingly easy to find a live stream, and matches are just 20 minutes. Uh, $33.9 billion was— uh, what million was bet on Table Tennis in May just with Colorado sportsbooks?

They're saying 99% of GDP is gambling now. That's what I'm saying.

I sent you a stat earlier, John. U.S. sports bettors got paid over $136 billion in 2022.

Humanity restored.

Don't ask how much they lost.

Tell it to Sure, I'm gonna text him. I'm gonna text him, good news. No, have you seen, did we ever do a little deep dive on the fake sports that get played purely for gambling? Did we ever look at this? There's fascinating games. There's like these Russian, not like bot farms, but like live streaming farms where people will show up to work and get paid almost minimum wage to play a small version of like— one of the games is soccer, but you're seated. So, imagine that we had a clear table here and it's me and you and we're playing soccer with our feet, but we just sit here and play soccer all day long and you can bet on who's going to score the next goal. And so, there's this element of randomness that's injected and then it goes into these livestream sites that then they are bet on and it's entirely fake sports betting.

Yeah, going with 2.6. Those are very dark. I agree.

Extremely dark. Extremely dark. But you know what's a little bit brighter? Everything old is new again. Wave Robotics— Weave Robotics is announcing a home robot, Isaac-1. Would you put this in your house? Do you think the Isaac-1 is cute enough, friendly enough? It doesn't look like it's going to topple over and injure an animal or a child. It can fold, fold blankets, do little things. They say it's going to deliver in the fall. This looks very cool to me.

Yeah, the design is fantastic, and it feels like they didn't over-engineer it.

They did just the minimum for what the tasks are.

Yeah, I mean, you know me, I'm gonna hold, you know, hold back a real review until I can see how it can play beer pong, just given that it is the Fourth of July weekend. It would have been great if they just came out. Yeah, their launch video could have been Isaac the robot playing a few rounds, seeing what kind of performance, you know.

Yeah.

Beer bench, they call it.

Beer bench. We got to close out maybe with this Jensen meme from Andre. I wake up, not a loser, infinite money because Meta employees consumed 73.3 trillion AI tokens tokens in a single month, which roughly costs $221 million a month or around $2.65 billion a year. I thought it was higher. That sounds low based on what I thought they were spending in terms of tokens. Might even be higher, who knows? But, uh, it is, it is very interesting. NVIDIA has also partnered with AI cloud firms with revenue sharing, taking circularity to a previously unseen level. Um, I don't know, Sale and leaseback, all sorts of these financial deals. These happen all the time in other industries. It's just sort of new seeing it in tech because we are in this industrial era. So very interesting news, but good news for NVIDIA.

Final post of the show. I want to give a shout out to Thomas Salvo. He says, since February, I've designed and built the world's fastest RC plane in his college dorm., and that's not clickbait. Reaper has a 5-kilogram carbon fiber frame, 250N turbojet, and flies at 500 miles per hour. So if you've been wondering what's above you in Georgia, it is a carbon fiber drone flying at 500 miles per hour. It's probably really freaking out the UAP types. We don't have objects that can go that fast. Thomas says, "Yes, we do.

I made one." Eh, I don't know. I'm skeptical. I don't, this doesn't look like human technology.

I, yeah. And so I think that they'll see it and they'll think like, "Wow, this goes way deeper than we thought." Occam's razor is Thomas has made contact, made first contact with aliens. First contact.

And he's commercializing their technology.

Let's go. They said post about it and say #aerospace to throw them off the trail.

Yeah, I like this guy. He's been at Skydio, Mach Industries, Relativity Space. Wow, yeah, he's moving all over. Very, very cool project. Well, congratulations on the launch, the virality, 20,000 likes. We got a reply from Elon Musk himself. Launch it from a tall mountain for low air density and you might be able to go supersonic and it dive.

I feel like this is maybe not something he should do yet. Unless you can find a mountain mountain without any human inhabitants for, you know, 20 miles around said mountain, like that, I would assume it just turns into a missile. Elon actually commented twice. One is just nice. The other one is the comment you just said.

Yes. Anyway, thank you for watching TVPN. We are off tomorrow. It is—

Fourth of July.

So we're having a long weekend, folks. Leave us 5 stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. And we'll see you on Monday. Cortisol rising. Have a good one. Goodbye.