AI Is Coming for Your Memes, Axios NPM Package Compromised, Claude Code Source Code Leak | Alex Pruden, Qasar Younis, Sebastian Mallaby, Forrest Heath, Dino Mavrookas, Will Ahmed, Jannick Malling, Ryan Daniels, Chris Yu

You're watching TVPN.

Today is Tuesday, March 31st, end of Q1 2026. Boy, we're live at TVPN UltraDome, the temple of technology, the fortress 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. Yeah, I got a new sound cue. I know you were pulling for me.

Dangerous.

I got 3 now.

Dangerous.

I got 3 now. Watch out. I'm not going to go too crazy. I'm not going to overdo it. We've been there. We learned, we changed, we evolved, we became better as people, as did our Linear lineup. We have a banger show. Let's pull it up. Linear, of course, is the system for modern software development. 70% of enterprise workspaces on Linear are using agents.

Alex Pruden coming in from Project 11, going to be talking about Google's quantum news.

The crypto quantum crash. Say that 3 times fast.

Crypto quantum crash. Then we have Kaser from Applied Intuition.

Yes.

Great to catch up with him, an absolute dog. Then we have Sebastian Malaby releasing his new book, The Infinity Machine: An Insider Account of DeepMind, Race for Superintelligence.

I'm a huge fan of Sebastian Malaby. You might have read More Money Than God. You might have read The Power Law: The History of Silicon Valley. It's the definitive account of how venture capital became what it is today. Highly recommend that book. This is a very interesting departure from that because it focuses on a single person. It's a biography, not a history of an entire industry. Industry, but very excited to talk to Sebastian Alibi.

And then we have Forrest from Somos raising a $40 million round. Then Dino from Saronic, Will from Whoop, Janek from Public, Ryan from Crosby, and Chris Yu, who's working on a spinout from Rivian.

Very exciting.

Already has a billion-dollar valuation.

There we go. Well, friend of the show, our president here at TD PN, Dylan Abruscato, headed to the TBPn newsletter, which you can sign up for at tbn.com, and wrote a fantastic essay summarizing a trend that we've been discussing with him around how AI is changing meme-making. And I found it very interesting. I'm glad that he wrote this piece. And so we'll read through this and then discuss it, debate it, and see where we can take it further. And then obviously—

And Dylan's from Long Island, New York. So John is gonna be—

I'm gonna do it in a Dylan Abuscato impression voice. Memes are changing. That became abundantly clear during the Oscars a few weeks ago when Conan tried to create a new Leonardo DiCaprio meme.

I don't know what kind of weird accent that was.

That's just like UFC announcer. To go alongside the classic Leo memes. In doing so, especially by using TFW, that feeling when, and the blocky white font that defined early internet memes, he inadvertently demonstrated that the meme templates millennials grew up with have become increasingly stale, even cringe. It's a good point. Instead, AI-generated videos are the new meme template that every network and studio should be focusing their launches on. Look at what happened— look at what's happening with the Harry Potter reboot. When the trailer first dropped, the reaction to the new Snape played by Ghanaian— sorry, he's from Ghana— Ghanaian English actor Papa Isidu was predictably and unfortunately negative. According to the LA Times, he received death threats since being cast in the new role. But after a few incredibly viral and well-produced AI videos, one an original Snape versus Black Snape MMA match and another AI-generated rap video and another Dripwarts, the school of drip, the narrative has started to shift. Have you seen any of these? I think I've seen Dripwarts, but can we pull up the original the quote unquote original Snape versus Black Snape MMA match, because I have not seen this one. And I think it is illustrative of what Dylan is talking about here. Snape v. Snape in the UFC ring. While we pull that up, let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest, securely connecting bank accounts to move money, fight fraud, and improve lending now with AI. And let me also tell you about Restream. One live stream, 30+ destinations. If you want to multistream, go to restream.com. So let's take a look at Snape v. Snape in the UFC ring. Any luck?

Here we go. I just dropped it in. Cool.

And we have a few others here. The videos have amassed tens of millions of views, and on Dylan's timeline at least, sentiment around both the character and the reboot has done a complete 180. Here we go. Really photorealistic. Who burned it? You!

No!

Yeah, you like that! Are there any red flags here as a UFC enjoyer? Does this feel like a proper UFC fight?

The actual video quality is insane.

He's so bulky. Yeah, the video quality is insane.

Wait, but Old Snape won.

In the fight?

Yeah.

Okay.

But I think it just, okay, wait, how do you know that? I think it just sort of like makes the characters more entertaining, more fun, shows you that this is, This is just creativity at the end of the day. This is just like, you should not be so up in arms about something that's a movie. Like, it's entertainment and here's some more entertainment. And so you're adding entertainment to the discussion and people are enjoying that. There's another AI-generated rap video about the new Snape, which we can pull up a little bit of here.

AI meme videos are inherently viral and driving real awareness in a way traditional memes no longer can. Not just because they're novel and more entertaining, but because a single AI clip can travel further and compound harder than traditional meme formats and social feeds that now heavily favor video. This suggests, yeah, that's interesting. On X, it's still very easy for an image to go viral. But if you think about, you know, Instagram, YouTube, like a standalone image just can no longer actually get that escape velocity.

I mean, what about dripped out Pope? Remember that?

Yeah, a little bit. But people are just spending so much time in the short-form feeds. And it can go in there, but they're certainly— Yeah.

Yeah, I mean, I guess even some of the dripped-out Pope— or what was it? Was it Balenciaga Pope? I don't remember what the name was.

Dylan Pope says, this suggests a new playbook for marketers, especially in entertainment. If you're about to drop a trailer for a new movie or show, you need to be thinking about your— ragebait character, the one people will latch onto, remix with AI, and build around. Conan tried to force a Leo meme down our throats at the Oscars. Didn't see that because I was sleeping. But this might have worked 12 years ago. That playbook is over. Today, enraged fans and communities will, if you're successful, take your characters or moments and turn them into something much bigger, entire cinematic universes. Yeah. I'm just very impressed by the overall quality of those outputs.

The Oscar selfie, I remember this. This, I think, became the most liked image on Twitter at the time in 2014, briefly. That image, this is the canonical clout bomb. If you're a fan of Bradley Cooper, you like it. If you're a fan of Meryl Streep, you like it. If you're a fan of Brad Pitt, you like it. And so you're amplifying all of it, the ultimate collab post. And this has become a format that's been used time and time again. It's effective. We did a little bit of it at the Super Bowl. It's fun. It works. But now, the future is AI. Let's pull up the DripWart School of Drip video. I want to watch this one because I saw a clip of this. I didn't see the whole thing. Let's see if we can play this. That's Harry Potter.

Are you really Harry Potter, my G?

Shh.

Type shit.

Type shit.

Type shit. Type shit.

None of that. None of that, broski.

We're all here on the Maybach Express for one reason and one reason only, and that's to go to Dripwarts, the school of drip. The Maybach following the train is pretty good. Hey, time out, tell them I'm a spaz on this bitch, I'm gonna go crazy, check this out. Run! And there's the new Snape character. So yes, very effective. I was reflecting on this and thinking about how It's not just AI videos that are unlocked as the new meme format. Like 20 years ago, video editing was extremely difficult. Like you had to do it on a desktop. You had to have a piece of software that probably cost a lot of money. It was not widely accessible. And so these image makers, image memes, we were, I was talking to Brandon about this, like Good Guy Greg was one of these or like the Insanity Wolf. And it would just be like a picture, one image of a duck and the duck would be on sort of sort of like a solid colored background and that would be the template. And then somebody would put white text with black, like block text, impact font on the top and the bottom. And that was like the image meme. And that was accessible in the sense that it could be like generated on MS Paint. It was free to generate it basically. Then we got video editing, you know, CapCut, Instagram Reels has an editor called Edits. And all of a sudden it became, easy for someone to take a Vine reel and put different text over it. I send you a bunch of these where I'll find some crazy Vine reel and I'll just recontextualize it with a new, yeah, you're laughing, new caption basically. And so the classic one is like those 4 jets in the new Top Gun and it's like, when you and the boys all drive somewhere in separate cars or something like that, you know, as an example. But now you can generate, you know, full AI videos that can express the joke of the meme. I think the next version of this is like software as a meme, SAAM, something like that. And we've been experimenting with this with the simulators. There's TBPN simulator, Jeremy Giffen simulator. There are more simulators coming. And all of a sudden, we, you know, the idea of building a video game, becoming a video game studio was like an impossible challenge. It would be months and months of time, maybe millions of dollars to get anything reasonable. So you had to be commercial about it. You could not do it as a comedy bit, but now you can, or it's getting closer. Certainly our organization is set up to where we can turn Ben or Tyler loose for a few weeks and say, yeah, like, you know, work on this vibe coding project for a few days, a few weeks. Like, it's okay, you don't have a lot of other responsibilities that are gonna creep in. But increasingly it's going to be more and more just like a few prompts on your phone to get the piece of software that is, that meme. And you can think about the Gmail suite from Riley Walls as another software as a meme moment, where he's making a commentary on the Jeffrey Epstein saga and all of that, but he's instantiating the humor, the commentary in a piece of software that actually works. Although, of course, the feature set is a little bit boiled down from the full Google suite, but the UI is familiar, and the UI is part of the joke. And so But I think that's a little bit of where this goes. Well, let me tell you about Cognition. They're the makers of Devin the AI Software Engineer. Crush your backlog with your personal AI engineering team. And let me tell you about Labelbox. Oops, sorry. Labelbox, RL environments, voice robotics evals, and expert human data. Labelbox is the data factory behind the world's leading AI teams. So there is a whole bunch of hack news going on. We're in a very weird week in terms of the news cycle because it's spring break. And so a lot of executives of big tech companies are like, don't launch, while my kids are out of school and we're going on vacation. I think that, I actually think this is my real theory. So we're in a little bit of a slow news week and you can see that like the Journal is covering announcements that happened last week. They're talking about Sora, they're talking about the Disney, they're talking about, you know, things that are more like reflective. In "Stratechery," Ben Thompson has sort of a 50-year retrospective on Apple. It's not driven by a news item. Like it's not like Apple launched a new product this week. So Ben Thompson is taking a step back and reflecting. It's a great piece. But it's not exactly news-driven because there isn't that much news coming from big tech companies, coming from the labs, et cetera. But there are a ton of crazy hacks starting with Axios. There is an active supply chain attack on Axios, one of npm's most depended-on packages. So if you have been vibe coding, Axios is a package that helps with HTTP requests. So it gets sucked into all sorts of different projects.

And if you—

upgraded to the latest version, you basically got a virus with that. And if that's running in the cloud, it's building, and that's probably maybe bad because it could steal API keys or SSH keys. It could do a lot of things. It could wreak havoc on your system. Also, if you built this piece of software and you included the contaminated Axios installer or pip package locally, it could potentially weasel its way out of your local environment and get onto your desktop. It's a virus. So be careful out there, and I'm sure people will be responding. The recommendation from Feros, who sort of broke the news over at Socket Security, is that if you use Axios, pin your version immediately and audit your lockfiles. Do not upgrade. Socket Analysis confirmed that this was malware. PlainCryptoJS is an obfuscated dropper loader that deobfuscates embedded payloads and operational strings at runtime, dynamically loads fs, os, and execsync to to evade static analysis, executes decoded shell commands, stages and copies payload files into OS temp and Windows program data directories, deletes and renames artifacts post-execution to destroy forensic evidence. So very risky.

I would say if you have installed this, you should just freak out, basically.

And if you break your computer, that's the first thing you should do. Try to slam it over the head.

Yeah, take the computer, throw it in the lake, throw it in the ocean.

That's how you should start.

I concur.

I mean, practical, yeah. I mean, there is going to be some sort of power law response here where of the people that are victims of the attack, they will go after the most vulnerable with the highest ransomware potential. And I think we're seeing that with one company. I believe Mercore was targeted.

But I don't know if that's been addressed. But I don't believe— was that? Yeah, my understanding is that— Yeah, the crazy thing is you have this like Claude code leak that—

That was completely separate.

Nothing to do with this. Even though I do believe they use Axios in Claude code.

Okay.

I saw something on that.

Sure, sure, sure.

And you have the Mercure leak, which is—

Well, it's not a leak, it's ransomware.

It's ransomware.

Yeah, someone stole some data.

Yeah, they stole a bunch of data and now they're trying to, you know, get bids on it. We'll get to that in a little bit.

Okay.

And then there's this Axios supply chain attack.

Yeah.

Anish had a little bit more context. He said, a tiny piece of code called Axios runs inside almost every app on your phone and every website you visit. Developers download it 100 million times a week. A few hours ago, someone poisoned it with malware that hands an attacker full control of your computer. If you've never heard of Axios, that's normal. It does one boring but important job. It lets apps talk to the internet. When a website pulls up your feed or an online checkout processes your card, Axios is probably doing the work underneath. Over 173,000 other code packages plug into it. It's everywhere. The attacker stole a lead developer's login for npm. Think of it as an app store, but for code that programmers use. Once inside, they swapped the developer's email to an autonomous ProtonMail account and uploaded the poisoned version by hand. They— that jumped past every security check the project normally runs before new code goes live. And this was not a rush job. The stackers staged the malware at least 18 hours before pulling the trigger. They built separate versions for Windows, Mac, and Linux. They poisoned both the current version and an older one within 39 minutes of each other, casting the widest net possible. Once the malware ran on a machine, it deleted itself to cover its tracks. The trick was smart. They never touched a single line of code inside Axios itself. Instead, they tucked in a fake add-on called plain-crypto.js, built to pass as a well-known trusted library. It copied the real library's description and author info, so nothing looked off at a glance. When a developer installed Axios, this fake package quietly ran the malware on its own. When a smaller package called UA-parser.js got hijacked back in 2021 with about 8 million weekly downloads. The security world treated it like a 4-alarm fire. Axios has 100 million over 12x the exposure with 173,000 packages depending on it. Socket, the security firm that flagged this, caught it in about 6 minutes. That's fast, but 6 minutes is still plenty of time for automated systems at companies everywhere to pull and install the bad version before anyone can react. If you or your team run Axios, freak TF out.

No.

Lock your version to 1.14.0. Change every password, API key, and access token on any machine that installed the compromised update. And check your network logs for connections to sfrclak.com or the IP address 142.11.206.73. Carpathy had some context if you want to go through this, Sean.

I will, but first I'll tell everyone a very important message from CrowdStrike, which is super relevant today. Your business is AI, their business is securing it. CrowdStrike secures AI and stops breaches. And I'll also tell everyone about Cisco. Critical infrastructure for the AI era unlocks seamless real-time experiences and new value with Cisco. So Andhra Karpathy said, new supply chain attack, this time for npm-axios, the most popular HTTP client library with 300 million weekly downloads.

That's a lot.

Scanning my system, Andre Karpathy says he found a use imported from googleworkspace/cli from a few days ago when I was experimenting with Gmail gcal-cli. The installed version luckily resolved to the previous version, the unaffected 1.13.5, but the project dependency is not pinned, meaning that if he did this earlier today, the code would have resolved, everything would've updated, and he would've been pwned. It is possible to personally defend against these to some extent with local settings, e.g., release age constraints or containers or et cetera, but I think ultimately the defaults of package management projects, pip, npm, et cetera, have to change so that a single injection, usually luckily fairly temporary in nature due to security scanning, does not spread through users at random and at scale via unpinned dependencies. So very, very crazy, crazy story. Scott Wu said the Devon Review caught the Axio supply chain attack for multiple Cognition customers before the attack was publicly known. These attacks will be 10x more frequent in the age of AI. It is critical that repo maintainers start using AI for defense as well, showing one example below where Devon Review caught the attack within an hour of its release text minorly edited for anonymization. So I was debating this with Tyler earlier. The question is like, how does this update diffusion of coding agents, diffusion of vibe coding? Is this, I was sort of saying, is this bullish for Cursor, Windsurf, you know, code readers? Because you would see an organization that said, hey, we were having a great time by coding, but going forward, we have a standard in this organization that we're going to have more humans in the loop. Does this make people be more inclined to put humans deeper into the situation? Tyler's counterpoint, I'll let you explain how you were saying that maybe this is actually bullish for just more token generation, more code gen.

Yeah, I mean, clearly like there just needs to be more code review, right? Okay. The package was still seen within 7 minutes by an automated system, right?

That's true.

So, like, yeah, I think people will just— like, there's going to be much more of an emphasis, like, okay, use a coding agent to write the code. You also use a coding agent to review the code every time. Like, right now, that's kind of a thing you do. Maybe later, if you're in a big team, you have code review. But if you're just doing it solo, maybe you don't do as much code review, right?

Yeah.

But it just becomes more embedded within the agents, right? You talk to Codex, you talk to Cloud Code.

Yeah.

There's already, like, every single—

I just think it's bullish overall for cybersecurity. Like, I think every cybersecurity company will probably do well. People are on edge already. And even though this type of attack has happened for years, long before, like, the popularity of vibecoding, it just feels like there's a bunch of new solutions that are needed. The kind of incumbent cybersecurity players will do well. They're going to release a lot of new products. I think the question that I have is, like, why 7 minutes? Right? Like, if—

Why not check it before it's merged in in the first place?

Yeah, yeah. Or just like, you know, these are machines, so theoretically they can be constantly monitoring versus like—

Yeah, I don't know. And the question is, we're going to be digging into this story more over the next few days, but I'm interested to know, like, it's found in 7 minutes. When is it actually rolled back? If you look at 300 million weekly downloads, like, clearly there are people that were downloading it at that moment in time. At all 7 of those minutes, there's probably like thousands of downloads, if not, you know, tens of thousands. Just doing like a rough ballpark on what 7 minutes means over a week of 300 million per week. But the question is, like, how quickly was it rolled back? So is it only if you're in that 7 minutes, or was it— it was discovered in 7 minutes and then it took them another 20 minutes to roll it back and stop stop serving the contaminated package. Understanding the scope of this, because it's very clear that as Andrej Karpathy explained, like he was actively using it every single day and yet was not caught in that 7-minute window. And so he was cleaned. And understanding the scope and scale of the impact is very much determined by how many— just how broad and how many installs happened during the contamination. Anyway, Will Brown has a good take. He says, I hope someone at Axios is reporting on this. And I completely agree. It's going to be confusing when they do. Anyway, let me tell you about Gusto, the unified platform for payroll benefits and HR built to evolve with modern small and medium-sized businesses. And let me also tell you about ElevenLabs. Build intelligent real-time conversational agents. Reimagine human technology interaction with ElevenLabs. So moving on.

More hacks.

Last night.

More leaks. What's going on?

Last night, pod code source code was leaked via map file in the npm registry. There's just a link to—

wait, someone just actually— do not click a link. If somebody ever says, hey, I got some really great source code here, just click this link, probably don't click it. Let other people screenshot it. There's plenty of meta-analysis over here. Seems messy, seems unfortunate. Heart goes out to the folks who who are dealing with the situation. At the same time, Codex is open source. It's not the end of the world, but it did reveal a bunch of things about the roadmap and also some of the internal— April Fools. April Fools. That is the worst part. We love a secret surprise April Fools joke. I love a good joke and nothing spoils a joke like hearing about it a day early. But much more importantly, there are lots of other critiques of the way Claude code is implemented. What are the bad parts?

What are the good parts? I don't think this hurts their business at all.

No.

Because people are using Claude code to make other products.

Yeah.

And then also having to take basically a fork of Claude code, maintain that, try to be shipping features against it, which is, again, I think it's not, seems to not be legal at all to just fork the code base just because it's out there.

Oh yeah, you can't just start stealing it if you're a business.

People are converting it into other languages, and maybe there's some argument there, but still, I don't think this hurts their business at all.

It's a little like you should read the attack script, understand some of the secrets, what's special. But at the end of the day, all of these tools, especially something like Cloud Code that's so new, like it's more of like the process.

It's more bad for the overall brand of vibe coding.

Totally, totally. Yeah, yeah, it's rough.

And the irony here is that every time Anthropic has released any feature related to cybersecurity, all the big cyber companies have been selling off, you know, tens of billions of dollars.

Yeah, yeah, yeah. The question of like, yeah, does this build trust in like using Viya code?

Yeah, so overall, overall it hurts some trust, but again, you know, very obviously gonna get through this.

Yeah, so the how it started, how it's going is of course landing like a ton of bricks. In the last 30 days, 100% of the contributions to Cloud Code were written by Cloud Code. And the how it's going is that it leaked the source code, not what you want to have happen. Yes, Angel says, "Mythos is so good at security that Claude code source code got leaked." Okay, let's, I don't know, should I, this is like, you know, you didn't get to watch the Super Bowl, you have it DVR'd at home, do you want spoilers? Should we review the April Fools joke or should we leave it unspoiled so that we can enjoy it tomorrow? What do you think?

I mean, it's not— it's not— it's cool. It's very cool.

You've already read it?

I read through it, but it's not— it's not to my— were you belly laughing? I don't think we're getting a knee slapper out of it, but it's very cool. Okay. I think it'll be cute.

Okay, well, then we can move on. What else— what else did we learn? Tuki summed it up here. Do you understand what just happened to Anthropic? Someone on their team ran a production build of Claude code. The compiler generated a.map file, which is literally a blueprint that reverses the entire code base back to its original source. And then they published it straight to npm for the whole world to download. And it really does show you how fast the npm downloads. Like there are people that are downloading it every single minute. And so if even if it's only up there for a minute, someone's gonna get it. And then all they need to do is send it to somebody, zip it, and post a link on X and it goes viral. It's like locking every door in your house, installing cameras, hiring armed guards, then accidentally uploading your floor plans to Google Maps. Does that?

Matter?

No, that's a bad analogy. I don't like that analogy because floor plans are not why I lock every door in my house. I install cameras, I hire armed guards.

Aren't floor plans public on like Zillow?

Oftentimes.

I would say we can probably skip over this, John. If you scroll down this account, just kind of posts like the same format every single time. So we can skip this. Let's go over to Lisanne.

If there's a red alert emoji, you got my attention.

Let's go over to Lisanne. I'll get you.

Yes, yes, yes.

A few takeaways from the Claude code leak. Anthropic is actively using Mythos for development.

Okay.

They are already at Capybara V8. We learned last week that capybaras are extremely deadly, but can be deadly in the right context. Capybara still has issues.

Foreshadowing is crazy. The foreshadowing is crazy. We were talking about how the Faustian bargain that is getting a capybara as a pet. It seems so cute, but it can bite you. And it seems like that might be what happened.

Cappybars has 1 million token context window and fast mode. That's really cool.

Numbat is another interesting code name tagged with @model launch. Remove this section when we launch Numbat. Fennec seems to be— the fennec fox. Fennec fox is very cute, but also not a domesticated animal. How about we get some golden retriever code names? How about big fluffy poodle? That's a good code name for your animal-themed AI model. Anyway, let me tell you about Consul. Consul builds 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 Lambda. Lambda is the superintelligence cloud, building AI supercomputers for training and inference that scale from 1 GPU to hundreds of thousands.

Jordi Hays says, hot take, Anthropic leaked Claude code intentionally to get a nerdosphere code review it would've never gotten. If they had just open sourced it.

Oh, that's actually true.

Way more attention.

You don't leak your entire feature roadmap and you don't do, I mean, it's funny and I'm sure they'll make the most of this.

This is 4D chess right here.

But I'm not seeing the 4D chess.

I'm seeing the 4D chess now. I'm convinced. This is, I mean, we're in completely uncharted territory for marketing stunts and pre-releases and sneaky footage that goes viral and maybe was planted and you don't know. And it's like some leaked account account. Like, I don't know. I think everything's— I think the gloves are off. Everything's on the table. This could be an April Fool's joke. This could be a stunt to drive attention to an open source move. Although, Tyler, you said that Dario is not a fan of open source at all, right? He's like against it unilaterally.

Yeah, Dario says he doesn't want to do open source. I feel like—

Isn't there some steel man there where if you open source, like, I don't know, like Opus 2 or something that's like really old, it's entirely commoditized in the research community. So all of those secrets that went into like making Opus 2 good, those have been commoditized. They've been discussed at the house parties in SF. The researchers have moved from one place to another. So everyone knows these, they've implemented it, they're available as open source. But by open sourcing your model, you can share with more of like the up-and-coming academic community. Like if I'm a computer scientist—

Yeah, but if all the research is already commoditized, then—

Yeah, I guess you could just use the other ones. It doesn't really have a benefit, maybe.

Yeah, so has anyone at Anthropic commented on this at all?

I haven't seen anything.

I haven't seen anyone. Yeah.

What is undercover mode?

That is a way to contribute to projects without letting people know that you're using Cloud Code.

Oh, interesting.

Um, uh, that's a hack. Gergely over at Pragmatic Engineer says this— Gergely, sorry. This is either brilliant or scary. Anthropic accidentally leaked the source code of Cloud Code, which is closed source. Repos sharing the source are taken down with DMCA, but this repo rewrote the code using Python, and so it violates no copyright and cannot be taken down.

Okay. And there's a warning, "Do not store the code." Even though it has leaked, do not store it because you might get DMCA'd according to Primeagen. The last time Anthropic, in their infinite PhD-level wisdom, leaked their own source code, February 25th, that happened? I missed that entirely. They DMCA'd all repos that had their code. Careful storing the code because Anthropic will have no mercy. 40,000 users forked it, so maybe unfork it if you did that because it sounds like you might get a legal letter. Again, a DMCA is not like an actual lawsuit.

It's more just a— Unfork!

Unfork! Unfork it. And people are of course making a joke that the Codex source code has been leaked in full here, and they're linking to the GitHub because Codex is open source, which is cool. I don't know. It's interesting. And more and more people are building their own harnesses. There was some interesting data that Opus performs extremely well, better than in Clogcode in Cursor on some benchmarks. And so there is this new paradigm of like, you know, how can you add different value when you're building a harness?

So what else?

Yeah, certainly there's other, plenty of other companies that are building harnesses and they're gonna be able to dig through this and get some benefit, be able to improve their products. That's not great, but at the same time, Codex has already— the source code is already open source. And so that hasn't been hurting Codex's progress and growth. So end of the day, ultimately, I would say, I'm assuming very embarrassing for the individual that ultimately contributed to this, but they will get past it.

Well, We will put the blame squarely on the AI model so they can take it. Dax is getting line of code mogged, LOC mogged, because Clog Code Source is 512 lines of code, whereas open code, his project is only 118 lines. And so he's got to get those numbers up.

And GT moggs everyone with over 50 billion lines of code.

He doesn't have 50 billion lines of code. GT, Gary Tan will be coming on the show hopefully this week, and we will get the full scoop on how he's using G-Stack and other models, which should be fun. Before we move on, let me tell you about Turbo Puffer, serverless vector and full-text search built from first principles and object storage, fast, 10x cheaper, and extremely scalable. And let me also tell you about vibe.co, where D2C brands, B2B startups, and AI companies pick channels, advertise on streaming TV, pick channels, target audiences, and measure sales just like on Meta.

Zach says NDAs are a great way to keep your corporate secrets safe from one or two beers. —but not 3 beers.

Why is there a community note on this? Oh, this joke was posted before on Instagram. It's a little joke theft. Interesting. Got him. Interesting. But it's a good joke. And I'm glad that he brought it over to X where we could enjoy it along with 36 other jokes.

The original post was, an NDA is a lock and 3 beers is a key.

Okay. Well, yeah, he toned it down for the— for the timeline. Anyway, there is news out of Google. A Google paper warns crypto on quantum risk ahead of 2029 timeline. So we've heard about the risk of quantum computing affecting the cryptocurrency industry, crypto projects broadly. There is some new research out of Google that provides some more perspective. So Google researchers have warned that future quantum computers may be able to break some of the cryptography protecting Bitcoin and other digital assets with fewer resources than previously thought, adding urgency to the debate over how the industry should prepare. The researchers did not indicate such a machine exists today, but said new work suggests the computing power needed to carry out that kind of attack may be lower than earlier estimates had suggested. In a Google Research blog post, this is from Bloomberg. The researchers said that a future quantum computer could break elliptic curve cryptography, a form of public key encryption used across much of the market. Their latest estimate points to a 20-fold reduction in the quantum computing hardware needed to break what's known as ECDLP-256, a mathematical problem that helps secure crypto wallets and transactions. That does not mean Bitcoin and Ethereum are suddenly exposed, but the researchers in a white paper dated Monday said the clearest defense is a shift towards post-quantum cryptography, or PQC. I'm sure this will be a hot topic over the next few months. A newer form of security designed to withstand attacks from powerful machines. They also urge the crypto industry to cut avoidable risks. In the meantime, we urge all vulnerable cryptocurrency communities to join the migration to PQC without delay. Google cast the paper as a warning meant to give the industry time to time to act, not as a prediction of imminent collapse. Last week, the tech giant introduced a timeline to fully migrate its own security systems to post-quantum cryptography by 2029. Fears around quantum computing as a realistic threat to crypto have swirled for years. In January, Coinbase established an independent advisory board to study what quantum computing could mean for the blockchain. That same month, Christopher Wood, global head of equity strategy at Jefferies, removed a 10% allocation to Bitcoin from his model portfolio, citing fears that the advent of quantum computing could undermine the token. On Tuesday, Bitcoin shrugged off the news of the Google paper making the rounds, rising as much as 2.6% to $68,300. I'm not sure where it is today, but Jordi, I'm sure you can pull that up. Even so, the researchers said the time left before such machines arrive still appears longer than the time needed to move public blockchains to post-quantum cryptography. However— BTC is currently at 67. 67. So slightly off of yesterday. A lot of this stuff has been discussed ad nauseam in the crypto community for years. I remember hearing about quantum potentially breaking Bitcoin as far back as 2016.

So you're saying you were already in that kind of like post-quantum mindset?

Yes, 100%. I was locked in. I don't know.

But I was aware of it. [Speaker:JASON] One concern that people in the community have had that I've seen talked about is this idea that if you did have a computer powerful enough to crack these encryptions, you would, unless you were like Google and you already had billions and billions and billions of dollars of cash flow, you wouldn't exactly stand up and say, like, hey, I have cracked Bitcoin, because the incentive for a certain team would just be to go around and find these wallets that were maybe— didn't have any activity for a long time and just start cracking those individually. Because if you just stood up and said, hey, I have a quantum computer that destroys Bitcoin, the price would go down and then the hacker wouldn't get any benefit from it. Yeah.

It's interesting. What are quantum stocks doing on this?

Quantum, probably ripping. They rip on everything.

PsiQuantum, is that one of them?

Yeah, Rigetti's up 8%.

Okay, there we go. Oh, PsiQuantum's privately held. There's another one, D-Wave, right? D-Wave, are they public? Yeah, they're up 10% today, but they're down 12% over the past 5 days, and 25% of the last month. Month and 42% over the last 6 months, but they're up 88% over the past year. Let's go. D-Wave is a $5 billion company.

Yeah, there's apparently a bull market in Nick on our team's email inbox. Oh yeah, quantum companies. Really? Right now? Okay, talk about—

well, we do have someone coming on, right? Uh, we have Alex Pruden from Project 11, uh, coming on to break it down for us at noon. Uh, so Nick Carter, uh, was talking about this. He said many are wondering what Google saw that caused them to revise their post-quantum cryptography transition deadline to 2029 this week? It was this, and it's from research.google, research.google, which we will go through. Max the VC says Google is basically saying we've cut the quantum resources needed to break Bitcoin's encryption by 20x. We can now break it. We can prove it. We're just not going to tell you how. We've slowed down research to give crypto a chance. "'You have until 2029 to figure out a solution. Good luck.'" Elon chimed in and said, "'On the plus side, if you forgot your password, the password to your wallet, it will be accessible in the future.'" Also to everyone else. Yeah. Yeah, I don't know. I mean, how do property rights, if somebody does have a quantum computer and they crack your Bitcoin wallet that you forgot the password to, you can prove that you owned and then they get busted for stealing your Bitcoin, you could potentially get it back.

Do you ever really own code?

I don't know. Nick also said, "And the craziest thing is that the quantum AI, the Google quantum AI paper is maybe not even the most concerning quantum paper released today from Project Eleven," who's coming on. Shor's algorithm is possible with as few as 10,000 reconfigurable atomic qubits. So this will be interesting to dig into further. The, within minutes with the 500,000 physical qubits, Google is now more confident on a 2029 post-quantum transition. Well, speaking of Google, let me tell you about Gemini 3.1 Pro. With a more capable baseline, it's great for super complex tasks like visualizing difficult concepts, synthesizing data into a single view, or bringing creative projects to life. And let me also tell you about Graphite. Code review for the age of AI, more important than ever. Graphite helps teams on GitHub ship higher quality software faster. So there's a lot of news about this quantum story.

Nick said, good morning. Now is not the time to panic. The time to panic is if Bitcoin devs read these two papers and double down on their chosen solution of hoping it goes away. Then panic.

It's ridiculous. Dan Shipper says, first thing I've seen that could make Bitcoin go to zero or allow competitive coins to catch up. So other coins clearly have at least like a very clear marketing story to tell if they are like the quantum proof, or the first to be quantum proof, or the most seriously regarded in the quantum proofing race. Will be interesting. Ariana Simpson. Chimed in and said, "Except all or most other coins have this problem too." But that is the opportunity that someone can maybe change something. So the chance that NASA lands on the moon, we were tracking this yesterday, the missions are starting to happen before 2028 on Cal-SHE is now at 14%, before 2027 is at 4.7%. So they are racing. Of course, this Artemis II mission is not boots on the ground on the moon. It is rocketing around the moon. We'll have more about this tomorrow.

They're just going to check it out.

They're gonna be gone for 10 days. They're gonna be in space for 10 days. And will be very interesting. Brendan Gurrell was doing some deep dives on the technology, the streaming technology, what we really care about here, that will be on board. Something like 20 cameras, 4K live streams, laser beams to make sure it's low latency. Super chats would be a lot of fun. Super chats would be good. We got to get a chat going. I'm sure there might actually be, because they usually stream on YouTube, and so I wouldn't be surprised if—

Is it going to be a 24/7, like, perpetual stream that's always on? Yes. Even when the astronauts are taking a sleep? Yeah. Taking a little nap? Yeah. Yeah. Okay. Yeah. Okay. Yeah, it's gonna be funny. There's so many— all the conspiracy theorists are going to be sitting there watching it very closely and then pausing. And there was a glitch there.

Did you see that glitch? That was, that was VFX. That was AI. No, this is my mark. I will believe that it's real if I see an astronaut put 3 fingers in front of their face.

Yep.

Because this is the one thing that the AI can't do right now. If you're ever on a Zoom call with someone who you suspect of being fake, a scammer who said, hey, "Let's get on Zoom. Let's talk about some financial investment opportunity." And it looks like someone you think is the person, but you suspect that it might not be. And they will be able to show you, "Look, look at the fingers. The fingers are perfect. It's fine, it's fine." That's because this part is not AI. Just the face is AI. This is the deepfake stuff that's happening. So what you have to do is you have to ask them to hold up 3 fingers. They'll be like, "Yeah." 3 fingers. This is fine, right? I satisfy the task. You got to say no. Put the 3 fingers in front of your face because if you put the 3 fingers in front of your face, the AI gets confused and it breaks the deepfake that's happening underneath. So you got to go like this. Show some depth of field. 3 fingers in front of the face. This is the trick. This is the only way you'll survive in the future. Be careful out there. Mercore had a breach. This is crazy. The design language for the hacker is very hacker-coded. The hackers that are putting out a bounty created like an image that looks very aesthetic to me. The design of this, I didn't realize hackers did stuff like this. This is very interesting.

But, you know, it's like you think about the hackers, you don't think of them as playing as hackers. They are. But they have—

They adopted the green text in the black terminal. It's like they're living the— It's like life imitates art, that type of thing. Anyway, it seems like a very rough leak, very unclear what's actually happening. There's a whole bunch of different questions in here. There's a database of candidate profiles, source code, video, all sorts of stuff, Tailscale VPN data. Unclear how much of this is real. They could be faking it. I don't know where the comments are, but, um, the risk is that, uh, if there is some sort of Equifax-style payout, that could be extremely costly, because if they have millions of people in their database and they've got to pay everyone $400, like Equifax did, uh, because they have sensitive information, that could be very, very expensive. Well, uh, we have had the Mercure folks on the show, uh, many times, and are hoping hoping that they get through this smoothly and everyone's— Yeah, absolutely brutal.

I mean, it sucks first and foremost for all the individuals whose PII is now potentially floating out there. Totally. It's probably quite bad for their customers who paid for some amount of this data and now it's just floating out there. And then obviously, unfortunate for the company, but it's still unclear. I was looking up Lapsus$, which is styled as L-A-P-S-U-S$ with a money sign, classified by Microsoft as Strawberry Tempest, and more recently identified as or a part of Shiny Hunters, is an international extortion-focused hacker group known for its various cyberattacks attacks against companies and government agencies. The group was active in several countries and has had its members arrested in Brazil and the UK in 2022. Wow. According to City of London Police, at least 2 of the members were teenagers. Lapsus$ uses a variety of attack vectors, including social engineering, MFA fatigue, SIM swapping, and targeting suppliers. Once the group has gained the credentials to a privileged employee within the target organization, the group then attempts to obtain sensitive data through a variety of means, including using remote desktop tools Attempts at extortion follow. Initially, the messaging app Telegram has been used for communications to the public, including recruitment and posting sensitive data from their victims. The first major cyberattack attributed to Lapsus$ was against the Brazilian Health Ministry's computer systems in 2021. Lapsus$ gained notoriety for a series of cyberattacks against large tech companies, including Microsoft, Nvidia, and Samsung. Following these attacks, City of London Police announced that it had made 7 arrests in connection to a police investigation into Lapsus$. Although the group had been considered inactive by April 2022, it is believed to have reemerged in September 2022 with a series of data breaches against various large companies through a similar attack vector, including Uber and Rockstar Games, with subsequent arrests again by City of London Police and Brazilian police. The group appears to have become inactive after September 2022, with members perhaps dispersing to other groups. And the conviction of 2 British members.

It's also interesting because they don't enforce brand intellectual property around hacker collectives. And so anyone can pick up the brand and use that whether or not they're in the organization. It seems very fluid. But good luck to everyone who's working on the response and hopefully a good resolution that is resulting quickly. Let's move on to something I have some good news. We will be having Sebastian Malaby join the show at 12:30 today. But Colossus Magazine published an exclusive chapter from the book, which Tyler has there, the biography of Demis Hassabis from Google DeepMind. And he secretly built a hedge fund inside of DeepMind trying to beat Jim Simons. Google shut it down. So there's this interesting There's this interesting screenshot that Colossus shared. Hassabis, for his part, assembled a secretive hedge fund operation within DeepMind. He recruited a team of 20 researchers to train high-frequency trading algorithms and explored a collaboration with the Wall Street behemoth BlackRock. It was not a project of which Google approved. But Hassabis, a 5-time world games champion at the International Mind Sports Olympiad— sick— and hoped he'd found another game that he could win. One day, I asked about the story of this trading project. I was told that Hassabis wanted to beat Jim Simons, the mathematician who founded the wildly successful algorithmic hedge fund Renaissance Technologies. RenTech operated in secret, which Demis loved, my acquaintance explained to me. Did the secret DeepMind trading team make money, I wondered? No, came the answer. Because of Google's wariness, it was quietly disbanded. I heard about about something, maybe it wasn't this DeepMind team, but—

Dave says, but did they rip CIGS? Oh, yeah. Could have been the missing ingredient.

Jerry Simons, he also never wears socks and always speeds and just pays the tickets because his risk-adjusted value in terms of his opportunity cost is that he should never drive the speed limit, which is sort of a wild move. True, true wild man. I heard about the potential of a Google hedge fund. Years ago, I don't know if it was related to DeepMind though, but just the amount of cash they have on the balance sheet, like they need a trading desk basically to move that money around. Even if they're just buying treasuries, they need a strategy, forex. There's so many different operations. And there was a pitch I heard about years ago that they were thinking about like, should we be more active? We have a lot of information. We have a bunch of great engineers. We could build a hedge fund here. But they decided that it was not compatible with the don't be evil philosophy. It was not core to the mission, and that at some point, there is risk associated with active trading, and so you could potentially blow up. There are certainly plenty of examples of hedge funds that had fantastic teams, but could not stick the landing and wound up zeroed.

Sophie says, Google shutting down a DeepMind hedge fund. Right before they were about to strike it big. It really is this meme.

They probably would have printed, although it's not like the high frequency trading firms are not using AI or not using— I mean, Jane Street invested in a custom server company or custom silicon company, something along those lines, specifically for high frequency trading. They have a lot of AI researchers there. And you see this with a lot of the labs saying, hey, does anyone from the high-frequency trading industry or quant finance wanna come work over here? We can maybe start matching your salary, maybe give you a more interesting project that you can actually talk about and people will be potentially excited about.

I don't know. Anyway. BoneGPT, the rapper Eater shared, I don't want this part of my brain to grow, which is a quote. From this. So in the weeks after the presentation, the two sides finally converged on a fleshed-out version of the Pichai plan. Suleyman would lead DeepMind's applied side from within Google, while Haseebis would run research as an independent global interest company. For Suleyman, this was a triumph. Google had finally signed a complex term sheet granting most of what he wanted. Haseebis was equally pleased. The plan guaranteed him an astronomical $15 billion in Google funding to sustain AGI research over the next decade, and it would put an end to the meetings on corporate structure, which he found screamingly boring. After 2 years of negotiations, he had hit his limits. I don't want this part of my brain to grow, he often said when asked to get his mind around another legal document. That's hilarious.

That's, uh, that's a great, uh, that's a great saying. I don't want this part of my brain to grow. Uh, it's so funny that you're going through 2 years of negotiation and they're like, okay, you're gonna have so much funding to build AGI, $15 billion. And you're like, oh, so like a seed round for like a Neo Lab, like great. Like, oh, so like one data center from a Neo Cloud or something. It's like the numbers have gotten so, so big that $15 billion does not feel like anywhere near enough at this point.

Colossus, hit me, John. Let me tell you about Railway.

Railway is the all-in-one intelligent cloud provider. Use your favorite agents to deploy web apps, servers, databases, and more while Railway automatically takes care of scaling, monitoring, and security.

Tell me about the— Elon— Colossus shares Elon has spent a decade trying to control an AI lab. He tried to absorb DeepMind into Tesla in 2014, then OpenAI in 2018. When that failed. An intern spoke up. It did not end well. Interesting. Okay, let's read through this. He also tried to control xAI to some degree.

Well, doesn't he control xAI?

Well, he controls it, but at what cost, right? All 7 co-founders gone. Oh, true, true, true.

That's what you're referring to.

Got it. Anyways, from the book, pushing back against Musk's obsession with the race against Google and DeepMind, Brockman added, it doesn't matter who wins if everyone dies. Musk responded the next morning at 3:52 a.m. He confronted Brockman with a proposal that recalled Pichai's pitch. OpenAI should spin into Tesla. Initially, OpenAI's team could accelerate Tesla's development of autonomous vehicles. Next, it could use the profits from self-driving cars to fund its AGI moonshot. Tesla is the only path that could even hope to hold a candle to Google, Musk declared. Even then, the probability of being a counterweight to Google is small. It just isn't zero. Back in 2014, Musk had Skyped Hasebas from a closet in LA.

What a funny— what a funny—

proposing that Tesla or SpaceX should absorb DeepMind. Almost exactly 4 years later, the new version of this proposal played into Altman's hands. It proved Musk's power hunger. With little difficulty, Altman now persuaded Brockman and Sutskever to take his side. Together, the 3 told Musk that OpenAI would not attach itself to Tesla. At an all-hands meeting on the top floor of a converted truck factory that housed OpenAI, Musk announced to the employees that he was quitting the lab scornfully, adding that I need Raptors.

I need, I need a new Ford Raptor potentially every day. We got to put this lab inside of a truck factory.

This is me scornfully adding that OpenAI would have to sprint faster to stay relevant. Yeah, I guess they did. I guess hoping to lure away some researchers, he declared there was a much better chance of building AGI at a strong business like Tesla. Yeah. Showing courage, or perhaps just youthful innocence, an intern asked Musk if speed might be reckless from a safety perspective. Besides, wasn't developing AI at a for-profit company like Tesla the same as creating it at a for-profit company like Google? Isn't this going back to what you said you didn't want to do? The intern demanded. You're a jackass, Musk retorted.

Then he stormed out of the meeting. That intern.

Tyler Cosgrove.

No, that intern was Steve Jobs. Just kidding. That intern was Taylor Swift. It is, my interesting read on this is like, it's crazy that Elon was interested in basically buying all of DeepMind, absorbing all of DeepMind, and then 4 years go by and he's like, I still want a lab, I want to absorb all of DeepMind. All of OpenAI instead of just incrementally adding to an internal lab at Tesla, just one researcher at a time. Like he was able to assemble—

Yeah, they were already working on self-driving.

Yeah, and he was able to assemble 8 co-founders at xAI. Of course, like they wound up leaving. But if you just think about it as like, okay, there's gonna be some churn, maybe the churn will be higher even. But if you start the process in 2014 and you're hiring researchers researchers continuously and using cash flow from Tesla to fund that. And then yes, researchers might leave, but then you get new ones and you're just building that capability. It's like the Supercharger network or Starlink, like you have to build a team and you have to continually add. But instead, Elon's been in this world where it's always like all or nothing, which is a very odd strategy to me, instead of just like home growing it. I don't know. It is just like an interesting— well, it's an interesting strategy, I suppose. Let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. And let me also tell you about Figma. Agents meet the canvas. Your AI agents now create and modify your Figma files with design system context in beta starting today. And without further ado, we have our next guest in the Restream waiting room. Let's bring in Alex from Project 11 to the TV Alex, how are you doing?

I'm doing great. It's good to be here, guys.

Is it over, or are we back? What's going on? How bad is it? Tell it to me. Give it to me straight. How long do I have?

Uh, well, according to Google, you have until 2029. Okay, that's like forever.

That's forever in my mind.

Yeah, uh, although maybe not forever if you think about blockchains that take a long time to change. Bitcoin's last, uh, upgrade took 4 years, and that's 2029 is less than 4 years away.

So yeah, that seems really risky. Uh, take us through like what actually changed, because I—

before we get into that, a little context. We'd love some background on, on you and Project 11, and then we'll get into all the papers. Yeah, all good.

Uh, yeah, so me, I'm a former Army Green Beret. Uh, got out— or I got really interested in Bitcoin working in the Middle East. Uh, got out, went to tech. Thanks. Uh, nice, nice sound effect. Uh, got out, uh, went to Stanford, and then entered the blockchain space at a company called Aleo, where I was at for 5 years before getting really excited about solving this quantum problem that blockchains face. And so that's what led me to found Project 11. So Project 11, I kind of gave it away. It's all about securing digital assets on blockchains into the post-quantum future, right? So the way I like to frame it is that what quantum computers threaten is the underlying foundation of cryptography that all blockchains are built on, right? So it's that level that we have to fix. And then ultimately everything on top of that, as you guys know from tech, like there's all kinds of dependencies at each layer of the stack and we have to rebuild the whole stack. And that's what Project 11 is all about.

Just to put it into perspective, when did you found Project 11? You said it was 5 years. You spent 5 years at the last company. When did, when did you actually get started on this?

October 2024. So just over, I guess almost a year and a half ago.

And what was, what was the general dialogue around around quantum and the risk to crypto at that time?

That it wasn't real, that quantum computers were always going to be 20 years away, that, uh, you know, that no one had to pay attention, there were bigger things to worry about. Uh, honestly, I feel like that's slowly changed. And I think today's not just— so there's a paper from Google, there's another paper out of Caltech, both dropped on the same day, both effectively lowered the bar massively that a quantum computer had to clear. To be considered cryptographically relevant to threaten Bitcoin. So that was the breakthrough. And I think this is a watershed moment where really, at this point, when Google, the head of the Ethereum Foundation, and a Stanford cryptography professor all pound the table and say, we cannot wait to migrate anymore, then people are going to start paying attention.

OK. What actually changed? Because it doesn't seem like the number of logical qubits or physical qubits, that seems to be growing exponentially. But even when you trace out the curve, you're 5, 10 years away from what we thought we needed. So is this a new algorithm, a new stack of code, or is it new math? Like what changed that we got this 20x increase in efficiency in terms of cryptography breaking via quantum computers?

Yeah, a couple things. So first off, these two papers are not necessarily about a quantum computer bigger or more capable. So this is— they're about what is— what it takes to break cryptography. Right. And so what changed? So one of the things that changed was that, interestingly, physicists and kind of quantum cryptographers that looked at this problem for a long time studied an algorithm called RSA. It's not worth defining, but it's kind of an older cryptographic algorithm. But that's not what really any blockchains use. Right. Because, you know, RSA keys are very large. Right. So it turns out this was kind one of the key upshots of the Google paper. It turns out that if you actually focus on the cryptography used by Bitcoin, Ethereum, and other networks, it's actually way easier to break than they thought it was compared to RSA. And so that is one of the major things. The other big breakthrough, and this is from the other paper from Caltech, is that quantum computers, as you guys may or may not be aware, as your audience may or may not be aware, are kind of— they're very fragile, generally. So to be useful they need to have what's called error correction applied. And that error correction can kind of result in a lot of overhead. You need to have tons of physical qubits to get to— you mentioned logical qubits— to get one logical qubit. Well, this Caltech paper basically showed, hey, we have some new ideas to do error correction. And it turns out if we apply those, we don't need hundreds or thousands of physical qubits. Maybe we just need a handful to make one logical qubit. So that— the title of their paper, actually, the headline is You may only need 10,000 physical qubits to break Shor, to run Shor's algorithm. And by the way, they demonstrated last year 6,000 qubits.

Okay, so we're close, yeah.

That's, you know, no one can put a timeline on it, but how fast do you think you can close that gap is the question.

Okay. Is it possible, Jordi was throwing out the idea of someone having a secret quantum computer going around the blockchain siphoning Bitcoin from cold wallets that haven't been—

I wasn't implying that it exists yet, just implying the incentive of if somebody were to create one of these. But the way you're reacting, I'm imagining it's like if somebody does it, it'll be Google first, which is maybe a good thing.

I don't know. It's really hard to know how it's going to play out. I wrote a whole blog post on our blog on projecteleven.com people can check out called Quantum War Games, and it was really fun because it's exactly— it's like the what-if scenarios, right? Um, you know, because why do people want quantum computers generally? Uh, well, they're great for science. Two, like, you can imagine governments that want to do espionage might want the ability to break cryptography too. They probably don't want to reveal what they have, certainly not if it's China or Russia, right? And, um, you know, but— and private companies, maybe not Google, but maybe some of these pure-play quantum companies, like, how are they going to make money? Well, one way would be to recover Satoshi's Bitcoin as if it were buried treasure, right? Like, oh, buried treasure. Satoshi's not here. It's mine now. Right. So, I mean, that could be another scenario. So, look, I think that there is just a whole bunch of uncertainty about how this is going to play out, about who's going to execute the attack, about how long a quantum computer will take. And again, because blockchains like Bitcoin fundamentally rely on this cryptography, like it's existential for them. That's one of the reasons, like, I founded Project Eleven and we pursue this, you know, solving this problem very vigorously is because everything's on the line here. And we have to solve it for these chains like Bitcoin to have a future.

Yeah. Is there generally low optimism right now that Bitcoin developers will be able to react quickly enough?

What's the statement about— like, I think Churchill said about democracies, or the Americans maybe, where it's like, they'll do the right thing when every option is exhausted. I think this is— look, I think this is true of Decentralized networks like Bitcoin, their greatest strength is the fact that there's no single party that says how it works or how it should work. This is encoded into how it was built by Satoshi as a reaction to the Great Financial Crisis. That's a great philosophical strength in the face of a crisis like this that demands a massive technical effort to overhaul. It's a daunting challenge, because unlike like, say, Google, which Google has said they're going to upgrade all their systems by 2029. That's just someone at Google can make that decision snap, right? In Bitcoin, because it's a distributed community, everyone's kind of first has to agree there's even a problem, then everyone has to agree on the solution. But I think there's examples of places where blockchains, like I'll take Ethereum, have done amazing things, right? So one is they transitioned from an old system of consensus called proof of work to a new system called proof of stake. It took 4 years, to be sure, but it involved thousands of people all over the world. And they did it. They did it. The blockchain's been running. Ethereum is the second-largest blockchain by market cap. So I don't think it's impossible, right? But I do think, especially in light of these two papers, these two breakthroughs, you just can't stop, or you just can't wait anymore before starting that process.

What about the rest of the digital world? Because if Bitcoin is having problems, then and so many other kind of core institutions and companies, organizations, I imagine would have issues as well. Maybe because they are centralized, there's easier to react, easier to kind of lock things down, but still need to upgrade overall encryption.

Yeah, there's no doubt that other institutions need to upgrade, but in my mind, there's also no doubt that, blockchains and digital assets are just the most vulnerable? One reason is obvious. Satoshi's Bitcoin, so the founder of Bitcoin, who we think has gone away or died or something, they have a bunch of their early Bitcoin that hasn't moved. There's a bunch of lost coins. All in all, it's about maybe 15% of all of Bitcoin supply is estimated to be lost. That's hundreds of billions of dollars, potentially, in market terms. That's just a huge incentive that— let's take the counterexample of if someone wanted to hack into a bank or something. And as you pointed out, banks are centralized, they can kind of react. Also the cryptography, the way that banks implement this cryptography is just kind of one of many layers of security, right? So it's kind of, this breaks, like theoretically, if someone tried to wire all the money out of my account, my bank would call me.

[Speaker] Yeah, no, they literally have tape drives where, you know, they have cold storage, they print things out and they have a ledger and they can potentially roll back, which is crazy to think about, but like they could, if there was like, like a catastrophic hack, they could be like, look, everyone's just going back to yesterday's accounts. And that's better than the chaos that we were in.

That's it. And that's not true for Bitcoin. All I need is one signature and all of Satoshi's or Coinbase's or Binance's Bitcoin is mine. There's no fallback. There's no anything. And that's how it was designed, right? That's the point. That was the point. Permissionless finance. That was, so that's the challenge.

So what is the state of the more faster moving coins, faster moving chains? Are you consulting or do you think you'll plan on launching something yourself that is quantum secure, quantum proof? How do you think this plays out? Because it does feel like, you know, I'm optimistic that I'm rooting for Bitcoin. I hope the devs figure it out quickly. Hopefully that happens, but it does just feel in terms of like the marketing of a new project, there is a bit of a white space to say we're the ones that are taking this particular feature most seriously.

Yeah, look, I mean, it's kind of hard to know how things are gonna play out, but the white space that we're occupying is we want to be the bridge for digital assets to the post-quantum future, right? Now that doesn't necessarily rule out potentially having a platform to issue on top of at some point. But I think for now, our priority is more or less, you know, people have already decided that things like Bitcoin and Ethereum and Solana and stablecoins have value. Yeah. And I think overwhelmingly they would like to keep the things that they already value and just make them secure. So that's what we focus on. Right. And there's no shortage of things for us to do because the protocols all have to get fixed. All the smart contracts have to get fixed. All the apps have to get fixed. And then all of the user wallets have to get fixed. And so again, going back to the fact that this is a stack and you're breaking the bottom part of it. So I mean, we really focus all the way across. So we've done work with Solana, the Solana Foundation. We did the first post-quantum testnet for them. We've worked with a few other protocols as well. We designed actually a new novel post-quantum algorithm designed for blockchains with the founder of Zcash. We've done that too. We collaborate with the EF and we're doing— we're getting ready to launch our own post-quantum wallet as well.

Yeah. Uh, talk about the information flow. How much of of the work that you do, the work that will be done by the Bitcoin Foundation, Ethereum Foundation, all the different developers, how much of that is open source by default or licensable or just can be understood by other parties and implemented very quickly? Like, how should we expect diffusion once this problem is solved to actually roll out? Will it just be like, oh yeah, like we're just following the, Salana standard. And so we're just going to mirror that over onto whatever chain we're working on.

Yeah, I think there will be diffusion. I think there will be consensus, if you will, around a certain subset of post-quantum algorithms. But I don't think it's just one and done, because Salana is a very different system than Bitcoin. Bitcoin's digital gold. It's sort of meant to be slow. There's no apps on it. Salana is meant to be fast, right? Yep. And so the cryptography that works for Bitcoin might not be this cryptography that works for Solana. And this is actually— it was kind of the results of some of the experiments we ran with Solana. And look, this is one of the challenges, right? And this is again why we keep saying this is like time to start is now because we don't know how long it's going to take to migrate because these new algorithms, there's trade-offs that come with them. And by the way, even if you choose to implement one, you need to test it and you make sure it's secure, all this stuff.

So have you tried to quantify or guess, estimate what the quantum discount rate is on Bitcoin right now? Because it's an interesting thing where like, if you own a lot of Bitcoin, there's a bunch of people on the timeline today talking about this. They don't have an incentive to like really freak out and spread the narrative, but they have some incentive to say like, hey, we need to have a conversation, we need to make progress on this. But do you think that's factoring into price at all? Totally.

I mean, the way I would put it is I think if this risk didn't exist, Bitcoin would be priced significantly higher. So I think exactly what you said is right. You're like, oh, I'm not going to sell my Bitcoin because, you know, it's maybe not right around the corner, and I'm hoping people fix it. But I also think there's people that would maybe enter, and they're like, you know, Chamath has said this exact thing. He's like, hey, is this really digital gold? Can we— without— with this quantum threat hanging over everyone's head? So I think if that threat was removed, then, you know, you remove this cloud over that ecosystem potentially, you know, you have a lot more people coming in and therefore price would be up.

Yeah. And it's easy to imagine as you approach that 2029 mark, more selling pressure, more concerns if meaningful progress isn't made in the next 2 years. What countries have the most kind of advanced quantum projects outside of the US? I can guess, you know, China's China is, you know, investing heavily here. Do they have their own retail quantum companies that are trading like crazy? What's going on over there?

Yeah, first off, I think it definitively the leaders, both companies and research, is American. So I think we should be proud to be an American here. But look, I think one interesting thing about the way China has chosen to attack this is they've made quantum computing a priority. And what that means in China is, you know, it's like there used to be at a lab at Tencent and at Baidu and a few other places. And at some point, a Chinese Communist Party official came in and said, guess what? You guys all work for us now. And guess what? You're all working together now. And guess what? You're not allowed to talk about it anymore. And that's the state of things. It's kind of a— I don't want to say Manhattan Project, but it's like that level of secrecy in China. And there's a legitimate question around how far back they are. So the best estimates that we have from Quantum Fit, like we have quantum physicist who's an advisor to Project 11 that tracks generally resource estimates across the world, and their view is that China may be 6 to 12 months behind at most. Yeah, exactly. That's not that far. Yeah, that's really quick. Can we expect a quantum computer in the hands of the Chinese Communist Party that maybe is more willing to crush dissent in places like Hong Kong to care as much about the philosophical principles of Bitcoin and decentralization if it serves their purposes to do otherwise, I don't think we can. And so I think, again, back to the fundamental problem, uncertainty, right? And it's better to be safe than sorry. So we need to basically prepare today to prevent the crisis tomorrow, to keep the trust in these systems.

Well, thank you for everything that you're doing. Thank you for your service, both here and before. Project11.com is the website, correct? Yeah, yeah, everyone should go check it out. All spelled out, yep. Fantastic. Thanks so much for taking the time to come chat with us. We'll talk to you soon. Have a good one. Great to be here, guys. Thanks. Cheers. Goodbye. Let me tell you about Vanta. Automate compliance and security. Vanta is the leading AI trust management platform. And without further ado, we have Kaser Younis in the Restream waiting room from Applied Intuition. He's the founder and CEO of the company. We'll bring him in in just a second.

We are running a little bit of time, a little bit ahead of schedule, a little bit ahead of schedule. We need a little bit of time.

And it's hard because there's not all that much short-term news. There is one news item that we can go through quickly. Allbirds just sold for $39 million. The company was once worth over $4 billion, was D2C darling. Follow this company closely because I was building a D2C company at the same time. I was like, wow, they are really getting big. But it seemed like it did not particularly scale. It was more of a niche product potentially. And of course, margins and cost of sales creep in, and then everything collapses to private equity multiples. Did you ever wear a pair of Allbirds? Would they ever pull you away from Bottega, get you in some Allbirds? You know, it's Australia. Australian wool, you know, it's kind of like Italian leather, Australian wool.

Yeah, there's something like that. No, I never, I never owned a pair.

Birds. Not even if you're visiting San Francisco, it's a great, it's a great, great sign of respect.

Did you ever, did you ever have a pair?

I had, probably did.

I think I had one pair at some point. Um, they were okay. They didn't, I don't know, they, they, they sort of like look okay and are comfortable on day one, but then they deteriorate a little quickly.

This would be probably a $2 to $3 trillion company if the shoes had aura. Yeah. But they had not. They should have released a lot of aura. It seems like they potentially had negative aura.

Yeah. And they got— and I mean, the stock suffered.

They had to pay the aura tax. The aura discount.

Massive aura loss. Yeah. I mean, still a very interesting— launch, very interesting go-to-market, telling the story of where the materials are from. That was certainly a playbook that was adopted by a lot of companies, putting the supply chain on display, basically. It was the right fit at the time. But people are not into it. The chat is not happy about— Let's ask our dear friend— Kaser Yunus from Applied Intuition, because he's here. He's in the TV PM. UltraDome now. Kaser, how you doing?

I'm doing great. How about you guys? We're doing great.

We have to ask you, do you own a pair of Allbirds? What's your preferred shoe when you're walking around a factory like that?

I, uh, don't own any Allbirds. When you're in a factory, uh, you're gonna have to wear— you're— yeah, exactly. You have to have steel. There's no Allbirds.

Maybe they should— maybe that's the comeback story for them. So they just— they were worth $4 billion, now they're worth 40, maybe the steel-toe Allbirds are what gets it done. A steel-toe Allbird would look fantastic. We seem to be having a video delay. I think the team will work it out, but we can't hear you. Can you hear us? Okay, great.

Yeah, I can hear you and I can see you. Okay.

Okay, fantastic. Exactly what's going on. Well, great to have you back on the show. I'd love for you to just reset with us for the shape of the business, where the company is today. How big are you? Give us, you know, the broad strokes. And then we'll go into the partnership today.

Yeah, thank you. Thanks again for having me. The company Applied Intuition, we're a $15 billion company still doing what we were doing before, which is taking intelligence and putting into physical machines. Today we have our first ever physical AI day where we're bringing lots of investors together, bringing industry analysts, you know, bringing everybody who's kind of relevant in the field to talk about up all the things that are happening in physical AI. We're, we're pretty strong believers that the future, you know, the next kind of big thing is AI going out of screens and going into the real world.

Yeah, I couldn't agree more. Uh, talk about the, the most recent partnership, LG.

Yeah, LG and Intech, we just, uh, we just announced this a couple of days ago. The, uh, I don't know, uh, how many of your viewers know, um, but LG provides—

you're putting AI in TVs, that's what you're doing. Yeah, the AI is going in the TV and I'm going to be able to Last question.

That is not what we're doing.

Much more serious. I mean, what's happening in the self-driving space is there is— now the models are basically working and they're figuring out. So really there's an aggressive downward pricing pressure of how to make self-driving cheaper. The research kind of question is done and now it's just an engineering question. And that's just another way of saying it's a cost question. So companies like LG who are doing, you know, sensors at really, really large scales and really, really cheaply. Yeah. You know, they're entering the space as well. We're working together with them on self-driving. Yeah.

So, yeah, take me through when people think self-driving, they always think Waymo, Tesla, but the market map of like products that need autonomy that would be defined as vehicles. Give me some examples. I mean, you're standing in front of something. I know that it's very broad. What's in this partnership? And then what else are you focused on? What's adjacent and what's, you know, on the roadmap?

Yeah. So I think what's different about us versus, let's say, vertical players like a Waymo or a Tesla is we provide this, you know, AI across all types of machines. So you see ag machines behind me. If you were— you guys were here for Physical AI Day. We do. We take the same models and we put them in defense. We put them in commercial trucks. We're running driverless trucks in Japan right now that are going into commercial operations in the next quarter. We are running in mines. So both all the way from Arizona to Australia. So our hypothesis basically is is these technologies, whether it's self-driving or the underlying operating system, they're so expensive and they're so complex to build and maintain. The only way that you really make this a viable business is that you actually spread this across lots of manufacturers and lots of industries and lots of use cases. I mean, our kind of crazy claim to fame is, you know, our company's almost 10 years old and we've preserved basically all the capital we've ever raised. It's amazing. It almost sounds like BS, right? It's crazy. The whole mantra is, you know, raise a lot of capital and deploy. We're a real AI company. We have real AI bills and we've figured a commercial model which has allowed us to scale. We have over 1,000 engineers. And so we're one of the, if not the biggest physical AI companies on the planet. That's obviously also commercially viable. So, but it all goes back to that simple thing is like you want to distribute all this cost across lots and lots of companies, lots and lots of verticals. What about shared learnings?

Like, is a team that's working on mining, are they able to find a breakthrough or discover something you can apply to trucking in Japan? Like, is there a— Bingo. Absolutely.

That is the heart of the company. And so there's all of the— what you described as like shared learning kind of broadly, but there's also technical advantages. What we've seen is taking data, which is just obviously also not obvious, but taking really, really diverse data from a mine actually makes our self-driving car system better. And taking, you know, data that we have from our software and car in Germany, you know, makes our defense work better. And so it really is, it really is. Yeah. Yeah.

I've heard so many stories about that where like there will be like exactly one instance of a chicken being chased by a woman on a tricycle. In the training set. And so it's very hard for the machine learning system to actually understand that if you see that exact scenario, you gotta slow down. But that's the nature of big data and machine learning and these scaled systems. And it's not, it sounds crazy, but it's not that crazy to imagine some weird scenario that you see in a mine actually teaching you something that you could use just on a normal street.

Yeah, maybe getting a level lower, just, just so, because I was— me being an engineer always bothers me to talk in pure generality. Yeah, I tend to mix and miss things. Uh, just getting to a level— or what you're really talking about is anomaly detection. Sure. And it's not necessarily like, you know, you need to see the chicken running across the road in Thailand and that's going to make them— but what's really happening is models are getting a better understanding of the physical world around them. Yeah. And the kind of parameters around them. If you look at, uh, you know, kind the last kind of generation. I'm crazy to say last generation, but really large language models. Large language models really improve with diversity of data. That is really like, you know, a kind of a big breakthrough. And of course, scaling laws, all of that stuff is being brought into the physical world. Yeah. And we're powering that. Yeah.

I mean, truly no one would have predicted or I mean, of course some people did predict, but I would have never predicted that, like, including poetry would help a model get to like solving math. Like I would just see those as different things. I'd say, put the poetry team over there, put the math team over there, but actually bringing all these things together worked really well. Play out the counterfactual for me. You haven't been a high burn company, you haven't been super capital intensive. If you'd done vertical integration and built the tractor behind you, that would have been extremely capital intensive, correct? Is that like impossible?

Well, nothing is impossible. There were jobs. My undergrad was at Stanford, at this obscure school called the General Motors Institute. And as the name implies, it's really about automotive. It's like the West Point for automotive. And when you spend a lot of years in factories, as I have, there are some deep lessons that get imparted into you. And one of those lessons is, holy crap, these factories are extremely cost-intensive, capital-intensive, and they're extremely complex. And the strengths of Silicon Valley are actually don't quite overlap with the strengths of building a large factory. Now, in terms of the core question, we had Marc Andreessen here today. We, you know, we talked about this. Marc was one of our first investors and has kind of been along with us with the entire ride. I mean, all the way to the presentation today. And we asked him this question about vertical, horizontal. What do you see happening in AI? What do you see happening, specifically physical AI? And the punchline is, you know, we all have our values that Applied Intuition can be be reduced down to two words: radical pragmatism. And if there are verticals that we think that we should be a bit more vertical in, we'll, we'll do that. And I think it's kind of a false trade-off to say what we do in, you know, trucking is what we're going to do in construction, which is what we do in agriculture is what we're going to do in mining. What we're really trying to do is bring intelligence out into the real world. And each of these verticals are facing really, really different problems. You take ride, you know, with a tractor behind me. The average American farmer is 58 years old. There's nobody coming to replace that person. And so what is going to happen? Because, you know, if you take that person, their kids have left and they're often not coming and taking over the farm like maybe in previous generations. So that farmer needs, you know, we don't need to teach them how to use Claude code. That's not what's going to change the farmer's trajectory. What's going to change a farmer's trajectory is the machines are intelligent and they're working harder and smarter on their behalf. And so he can run an entire farm with a swarm of machines. And that's not too far into sci-fi. One of the key components here that we're doing and we believe is you need to abstract that hardware and software away. We as technologists, you look at like your laptop and your phone and you kind of take for granted the miracle that exists Android runs on thousands of hardware devices flawlessly. So that's also something that Applied does. We're just abstracting the hardware and software. Once you do that, you can make every machine intelligent.

Have you tried to estimate the economic impact assuming you guys stay at the current kind of improvement rate or accelerate as the technology kind of starts to diffuse in some of these industries like trucking in mining and agriculture, like what are the downstream impacts? I mean, there's such a debate right now around what impact will AI have on the economy? So much of the economy is like moving physical things around, producing things, shipping them.

Yeah, yeah, so let's separate a little, 'cause economy is such a generalization. So when you're talking about like, you know, CodeComplete and white-collar work is very different than, you know, trucking. Where there is a huge labor shortage. It's very different than in mining where, you know, people don't want to go live in kind of remote areas doing 12-hour shifts. I mean, literally labor shortages are preventing construction companies from, you know, collecting billions and billions in revenue. So these are industries where AI can't get there fast enough. Yeah, it's a very different calculus than a kind of, you know, I think what the normal narrative is. And, uh, and then we're super obviously excited about that. Let's take defense as a particular example. It's a very salient example. We don't need more warfighters in harm's way. We need less warfighters in harm's way. And no warfighter wants to go out in, in, into that ecosystem where autonomy is really, uh, becoming the dominant thing. And so, so I think the way to think about this impact in the physical world is it's a lot less resistance, there's a lot more pull. Now, the first question you asked is the size of impact I don't want to sound like I'm pitching my own book here with you.

I'm asking you. I'm asking you. I want— I want the biggest number.

I want the biggest number. The numbers are absurd and ridiculous. But I can tell you this much. If you think about, you know, the way I think about— I used to be a Y Combinator before I was a CEO and, you know, ran the firm and funded lots of interesting companies. And one of the analogies I used to use to help founders understand market potential, market size is, yeah, I grew up in Detroit. You're sitting in the Detroit Metro Airport and you're sitting in a gate and you look around. How many of those people are like really deeply using Claude code? I mean, frankly speaking, not many. Now, a lot will probably be using something like ChatGPT, some variant of that, maybe Gemini. But how many of those people drive? How many of those people work at construction sites? How many of those people ride in buses? How many of those people serve in our armed forces? The point is a much, much larger group. And I feel a little— again, the engineer in me feels a little awkward saying these kind of, you know, pitching these things. But I think the market for physical AI is way, way bigger purely because the surface area is much bigger. And it's compounded by the way that the way technology diffuses with phones and laptops creates this like rabid competition that you see in, you know, you're seeing in all these kind of subspaces, right? Yeah. In physical AI, you got to kind of know what's going on in the car business. And I'm not saying, you know, I'm not gatekeeping and saying, yeah, you got to go to the General Motors Institute to build technology for the car business, but you bet your bottom dollar it helps. And we're doing that across a bunch of industries. I think it's, you know, I'm as confident about the company as ever before. You know, the question we always get asked ask this question, why the hell did you raise all this money? You know, almost $1 billion, you're just going to keep plowing it away in the bank account? We're doing it for a simple, simple reason, because we can, if we need to, we can invest very aggressively to take opportunities that we think we can accelerate, you know, beyond just traditional organic growth. And so far that's worked. It's not to, you know, promise the future that we won't, uh, but that's— those are kind of debates we have every single day.

Yeah, that makes a ton of sense. Well, thank you so much.

I just want to say I can see the path to $100 and then a trillion $100 billion in run rate.

I agree. Well, I mean, Waymo, you know, a company that we love them. They're a local, you know, we're also in Mountain View, now Sunnyvale. That company, you know, is a great company, but is burning a lot of capital and is a smaller revenue base than us and just raised that $126 billion. Oh, okay. I mean, I love those guys. I mean, we have so many friends there. I'm not trying to— talk poorly about this, but no, we love Waymo too.

It's very impressive what they're doing.

You know, 15 that we're at and 120— I think we got room to run. Massive, massive. Yeah, we got room to grow.

Tell, tell Mark, tell Mark you're ready, you're ready for the big one.

Believe, believe me, everybody wants to, you know, it's— I feel like it's foie gras where they want to get— keep putting, you know, money into the company. We, we don't need—

that is the best analogy for a venture capitalist. Yeah, yeah, they are the farmers stuffing the goose.

We, we have our own, we have our own, uh, farm, you know, and we're making our own money. So that, that's really great. And frankly speaking, I mean, like I said, Waymo is great, but it's just robotaxis. Yeah, yeah. And, and, and that's a small— it's so— go, go to, go to, go to Warren, Michigan. Yeah. And you just go, go, go to the party shop in the corner and say, hey, are you excited about Waymo? I don't think, you know, it's not hit the masses yet, which just shows obviously Waymo's growth potential, but also shows, I think, how big physical AI is going to be.

Yeah.

Well, thank you so much for taking the time to come chat with us.

Thanks for having me again. Fantastic. I'll talk to you guys. Thanks. Yeah. See you.

Let me tell you all about Public.com. Investing for those that take it seriously. We got stocks, options, bonds, crypto, treasuries, and more with great customer service. And we're talking to them later today. But first, we have Sebastian Mallaby. He is the author of The Infinity Machine. Sebastian, how are you doing?

I'm doing great. Thank you.

Thank you so much for coming on.

So great to have you. This has been— We're very much supportive. You've been on John's radar for a long time. Dream guest list for a long time.

We met maybe 4 years ago at a talk you gave around the power law, and it was very fascinating. I love that book. This book goes in a different direction. And after that conversation, I asked you probably the worst question you could ask to an author. I asked you, what's your next book gonna be about? Because you had selected venture capital, venture capital had done very well, and I presumed that whatever you would pick would be a great investment category because you seem to be a good picker. You told me that you were thinking about biotechnology, biotech in investing. You went a different direction with AI. Is there anything I should read into that?

Well, I always kick the tires on a few ideas before I settle on one. And I think it took me from the power law coming out in February of 2022 to somewhere around the summer, maybe August, when I really settled on AI. And then it took me another, I don't know, 3 months to get the courage up to go and pitch Demis Hassabis on the idea of giving me a ton of access so I could write this book. And then I got lucky because I pitched him, and one week later, guess what? ChatGPT came out. So what I thought was maybe a fringe subject— yeah— went mainstream super fast.

Super fast. Uh, what was the response with the team? What was the process? Obviously, he's extremely busy. He also sleeps at very random times. We can get into that, but, uh, What was your actual interaction? How much time did you spend?

What was the research process like? So once he agreed to be in, he was really in. It took about 6 meetings, 2 with him, 4 with his team, to get them to agree. And, you know, my pitch was, hey, if you say in every speech you give that artificial intelligence is the greatest invention in history, yeah, that means 'You're way too important not to have a book about you, and it's going to happen, so you better get used to it.' Yeah. And also, if you're going to upend our lives, you know, change the way we think about ourselves as humans because it's a rival form of machine intelligence, you know, you better explain your motives to people, otherwise they're not going to accept it. So that was the pitch. He agreed. And then once he agreed, we would meet like for 2 hours at a time. We would go to a pub near his home in North London. And there was a kind of secret staircase at the back, go upstairs, kind of dusty little room with nobody else there. And we would sit there for 2 hours usually. And he would just riff, you know, talk about philosophy, movies, computer science, neuroscience. I mean, he's such a range of a person, by far the most fascinating person I've ever written about.

Wow, that's high praise. How do you think about balancing the biographical timeline, the history, the financial impacts, which you've covered in the past? When I think about the stories that you've told in The Power Law about venture capitalists, there's a little bit of their philosophy, but it's a lot of how the deals came together, fly on Man in the Wall. I love that type of storytelling, but this goes a little bit of a different direction. So how did you, how did you think about balancing all the different perspectives that you could bring to his story?

Yeah, I mean, I always want to do the personality, the figure, but then the landscape behind as well. So it's always a mixture of, you know, you need a character who drives the story, but the story is boring if it doesn't mean anything. So you have to link it to larger stuff that's going on in capitalism and how society is going to change and all that. And I mean, in this case, because Demis is who he is, um, and he would just riff in these extraordinary paragraphs of like storytelling and, and theoretical stuff, and you know, it's just so fascinating that I did give him the microphone more than I have in any other book. I mean, I basically quote quote him at some length, you know, and it's broken up with me asking him questions. And so I'm kind of the reader's lens through which to see Demis talking. And I'd never used the first person really before in my other books, but in this case, those 30-plus hours I had talking in a pub with this extraordinary man, that was the gold dust I had. So to really make the most of it, I did have these passages of us talking together, which kind of intersperse the more analytical or narrative bits of the book. You know, how after the Transformer model dropped, what was Ilya Satskeva's reaction and why did OpenAI get ahead? You know, I cover all that stuff as well. But I do have these passages where you see events through Demis's eyes because, and I think it's worth doing because he's so unusual.

What was your understanding of AI or view on AI in 2022 before you meet Demis for the first time? Were you aware of the doom arguments? Were you a believer in the technology? Did you think it was 20 years away, 100 years away, 2 years away? Like, where were you before this book? Because I feel I think it probably changed you. Yeah.

You know, I had met Demis before at tech conferences because of the power law and writing about venture capital. I would go to tech conferences in Europe and he would sometimes be there. And I actually cheekily, you know, raised the issue of hedge funds and especially, you know, the one Renaissance Technologies, the main CEO, Peter Brown. Had done a PhD with Geoff Hinton about AI back in the day. And of course, I knew that Dennis would know that. And so I said, do you know these guys who use deep learning and applied it to markets? And that got his attention. So I talked to him a bit. So I knew that he was amazing. I knew that the technology was ripe in the sense that he'd produced this string of breakthroughs, AlphaGo, defeating the human Go champion, then AlphaZero, which was even stronger. Then there was AlphaFold, which won him the Nobel Prize for predicting all the shapes of protein in nature. So there was a series of good models, and what they all had in common was they dealt with insane amounts of data, crazy combinatorial spaces. Like in Go, there's 361 first moves you can make. Okay, That's way more than chess. And so unscrambling Go and the strategies in Go was much harder than chess. And I knew that these breakthroughs were not just cool in themselves, but they represented the coming of a time when machines could make sense of an almost infinite amount of data and extract meaning. And hence the term, the Infinity Machine, the title of the book, and hence my interest enthusiasm for writing about it. And, you know, so I knew it was breaking out. I knew Demis was amazing. What I didn't know is it was going to break out literally the week after I met him and pitched him on the idea.

What has surprised you about how the industry overall has evolved since you started meeting with Demis? Because in some ways, I have to imagine you kind of— maybe it hasn't been that surprising at all, even though a lot of the growth is impressive. But do you feel like you had a view into the future from those first conversations in the pub?

Yeah, I mean, I think, you know, I was lucky that the meetings were bookended by, you know, going to see Demis sort of maybe the third meeting or something. ChatGPT by this point had gone viral, and him saying to me, okay, this is war, you know, and you could see his competitive side come out, This is war. These guys, he said, have parked the tanks in my front yard. I'm fighting back. So you could see that. And then after that comes the merger between Google Brain in Mountain View and the DeepMind team in London. So the two kind of halves of Google's AI talent base are united. And then I think you get what a business school professor historians in the future are going to write about. It's kind of like a textbook case in how you make a merger successful. Because everyone knows that mergers are hard. And when you do it with 8 time zones between the 2 teams, one in London, one in California, and you're doing it in the middle of this knockdown, drag-out capitalist fight over building LLMs, this is going to be— most people said this is going to fail. And I would come to Silicon Valley while I was doing the book, checking with my friends at different venture capital shops, and they'd always say, game over, OpenAI has won. And the surprise is that within 2.5 years, Demis Hassabis's Google DeepMind model Gemini was doing better on the rankings than the OpenAI models. So that was an incredibly fast comeback. Yeah.

How do you think about the importance of being like a business person or a deal guy in AI. It feels like Demis has this quote, he doesn't want to grow that part of his brain, referring to some legal negotiations that took a number of years.

Suleiman.

Oh, oh, was that Moussaieff?

No, actually that was Demis saying, yeah, I don't want part of this, this part of my brain to grow, right? Take away these legal briefs.

I completely get that. And it seems incredible for him to stay in the research mode, build the research organization. And yet because of scaling laws, we're in this weird regime in this paradigm where sometimes doing a deal to marshal an extra $10 billion of compute actually does unlock a new capability and is almost in the research track. And I'm wondering how you perceive the relative importance or Demis' perception of the relative importance of these sort of like business dealings that might be more critical path to AGI than the Demis of 2021 might think?

You know, I would say that the single most important business relationship in the world today is between Sundar Pichai, you know, CEO of Google, and Demis Hassabis, who's running the AI brain trust. Because, you know, Sundar has Demis's back. He deals with providing the resources, you know, supporting the notion that you're going to spend all these tens and maybe hundreds of billions on compute. You know, that's Sundar's, that's what he delivers. And he gives Demis the oxygen to then just go do the science. And sure, he has to build products, but I think, you know, he said to me several times, Demis, we've got to a point where building a product like Gemini is in fact advancing the progress towards AGI. There's no tension between the two. If you had stopped your AGI research 10 years ago and taken a sidetrack to build some widget, yeah, that would have been a waste of time. But now that it's so mature and you're actually to build the next LLM, you've got to kind of figure out, you know, a reasoning model, then it's going to be agentic and then you're going to be scaling it even more. And all this stuff that we've seen, this is genuine scientific progress as well as product advance.

Yeah. Do the products also help sort of shift the culture in Google? I'm interested in understanding this concept of like AGI AI pilling, becoming a believer in the Demis worldview of the impact of AGI, what artificial intelligence will do. That mindset has to diffuse through Google. The ChatGPT viral moment clearly had an impact. I'm sure agent-to-coding has had a similar impact. But what has Demis's role been in being the culture carrier of that belief in AI progress internally?

Well, I think he's used all his sort of visionary communication skills to unite the Mountain View team and the London team. And I think the one sort of organizational contribution he's made, which is sort of super powerful, is that from a long time ago, DeepMind had always two different cultures going on at once. There was a kind of blue sky research for scientists. You get a lot of freedom, you could publish papers and all that, you know, go find what you want to find. And then there were moments when Demis decided that if you pushed really hard on a particular product or a particular project, you could get a breakthrough achievement that would really shock the world. And so he did this repeatedly. With AlphaGo and AlphaFold. And this was sort of like his judgment, scientific taste being applied to knowing when the moment was ripe to really go for it. And once he decided that, there was the blue sky research kind of bottoms-up stuff that went out the window and it became a top-down strike team, they called it. And in a strike team, there was a lot of top-down direction and kind of everybody had to work on on the same code base. You couldn't just go off and code your own experiment on the side. You had to be contributing to the main one. And that drives towards— a team is driving in a united way towards an outcome. And I don't think Google Brain had that. Google Brain had much more of the bottom-up stuff, and there was no strike team component. And so Demis brought this strike team idea, and it came came from video game design. You know, earlier in his life, he'd been a builder of video games, and he in fact started a company doing that. Yeah. And so this was like how you ship product, and, and that's been a key insight for Gemini.

Yeah. Google and Gemini have every advantage just due to the massive cash flows that they have from their other businesses. How do you think that has impacted the culture of DeepMind, given that they have something very real to lose, right? It's not just about the, you know, maybe Demis's personal desire to be at the forefront of this research. But if you're not successful, then you lose one of the greatest business, you have the potential to lose or have your kind of core business threatened. Google Search. Yeah, Google Search in such a big way. Yeah.

Yeah. I think Google Search stands for a more general point that Google's whole reputation stands on providing reliable information. And so the penalty for screwing up is very high. You've got this very valuable company, you don't want to spoil its reputation. And so I think they were more worried about releasing a chatbot fast. And so they had prototypes of a chatbot. In the fall of 2022, and they didn't want to release, and then OpenAI went ahead and did it. And so that kind of forced their hand. But their first instinct was, this is going to hallucinate, this is going to do bad stuff, we can't afford that hit to our reputation. And, you know, Demis was quite honest with me in saying, well, you know, the surprise was actually the public's quite happy to play with the tool that hallucinates. You know, they're It still went viral. So we should be less inhibited. But that was an example of how being at Google could be a kind of inhibition in moving ahead.

How does Demetson— he's always told a very optimistic story about AI. I love him as a science communicator, as an optimist. How has he interfaced with the effective altruist community more of the AI doom crowd. Does he, 'cause he doesn't talk about it that often, but does he think about it often? He does think about it.

In fact, you know, he met his co-founder Shane Legg at an AI safety lecture, right? They bonded in a safety lecture. Yeah. And so right from the beginning, safety has been a big part of the agenda. And when Demis sold his company to Google in early 2014, 2014. Part of the deal was you're not going to use this technology for weapons ever, and you're going to have a special independent oversight kind of committee, you know, which will decide on AGI deployment, because we don't want that to be just up to the corporate board of Google. Now, you know, he's, he's slipped on some of these things, particularly the military stuff, but he has been thinking about safety. And, um, you The question is, what has he got to show for it? It's all very well to think about something, but what can you deliver? And I think this is why towards the end of my book, he's talking to me quite honestly about how it's a sort of paradoxical moment. He's doing great as an AI inventor. He's doing terribly badly as a sort of AI steward, as making it safe. It turns out that there's a race dynamic. The race dynamic includes China. How do you control this technology when everybody is racing to jam it out the door?

Yeah, it was interesting last week we were talking about how Bernie Sanders has come out with a push to pause data center development. And he quotes Demis and some other lab leads talking about how they would agree to a pause if other countries were kind of in agreement. Disagreement on it. And Bernie obviously left out the fact that I don't think any lab lead could see China pausing on development. So how has your personal definition of AGI evolved over the last few years? Because everyone has their own definition. Half the people that come on the show says it was 6 weeks ago exactly. And then yet we're still in so many I've been in some of these kind of more high-level conversations, lab leads talk about race, you know, AGI is 6 months away, you know.

You're completely right. Everybody has their own definition. And so my solution is not to have that discussion. I mean, who cares what, you know, this, you could say right now it's Gemini is artificial, it's general, it's intelligent, game over. But just a definition thing. I think the other one, which is sort of usually fruitless is AI conscious? Could it become conscious? What is consciousness? Nobody has a good idea. So let's just sidestep those things.

I heard one good idea, which was something around training a model specifically that would hold back all data and all training data related to consciousness and the idea of consciousness. And so it cannot just pull from the archive and reference or sample simulate consciousness, and if it develops consciousness from that and can talk cogently about consciousness without having any training data ever seen the idea of consciousness, then maybe that's conscious. I don't know. That's interesting. It was just an interesting thought experiment. I don't know that anyone's actually run the test, and I don't know that I would even accept it if they did, but it is something to think about.

Did you feel an acceleration in your personal writing process due to AI?

I felt an acceleration in my learning process, which is sort of what I do before I go interview people. So because all of the computer science papers are basically on Archive and recently there's been less publication, but certainly up to about 2022, you could go see a scientist either at DeepMind or one of the rival labs and just have a conversation with the model about, okay, this person has done 4 papers, what was the connecting thread, why is this person different to the one I interviewed last week? That was a super efficient way of getting up to speed. And I didn't worry that it might be wrong because I was going to speak to the human and cross-check it. But that was helpful.

Do you think Demis's having a home base in the UK is— in what ways do you think it would have helped or hurt the company so far? Far? Like, has it been beneficial from the talent war standpoint? I'm sure, I'm sure lab US, some of the other lab leads have taken a trip out to the UK.

Mark Zuckerberg's hoping hosting nightly dinners for AI leads at his house, which is just a couple blocks from all the other labs. Can't do that if you're in, if the researchers are in the UK, right?

Yeah, I mean, there is movement across the Atlantic, but I I think the deal is if you're in London, it's a little harder to recruit people, but once you've got them, they're probably stickier than they would be if you're in Mountain View or somewhere. I think Demis has stayed in London because actually he is weirdly patriotic. He comes from this mixed up sort of Greek heritage father, a Singaporean Chinese mother. In a way that makes him a typical Londoner. London is really a melting pot. But he stays there because he feels he wants, he believes in sort of British social democratic values. It's maybe ironic to an American audience, but actually he thinks it's more egalitarian in Britain than it is in the US. The US, if you go to the DeepMind office in London, you go past the sort of public spaces like this kind of fountain with toddlers playing in it. Kind of free movies in the summer by the canal. And then you get to this green space where the kids from the local housing project are playing soccer. And then you get to the DeepMind office, and it's hard to imagine you'd find that on the way to the, you know, the Apple headquarters or something. You know, it's just a different vibe.

Could you ever see Demis as CEO of Google, or would he have to do too much paperwork?

You know, this gets to the heart of the dichotomy about Demis. It's so difficult because he's so many different things at once. I mean, you know, he is this Nobel Prize-winning scientist who would love to just do pure research, and he often would sort of fantasize to me about, hey, I want to retire to the Princeton Advanced Institute of Advanced Studies and do what Einstein and Oppenheimer did before me and go there. And that, you know, I think he really means it when he says that. At the same time, he also wants to be in command of an AI lab and he likes the capitalist competition. So if he had the opportunity to be chief executive of the whole of Google, I suspect he's too competitive to say no. But who knows? I mean, there is that science side, so it's genuinely unpredictable. He probably doesn't know himself.

Were you left personally optimistic after this process about just our AI future broadly, net positive impact from AI?

I mean, clearly there's a lot of upside, especially in medicine and pharmaceutical discovery. I think you have to be honest and say there's also downside, significant downside. And as I continued to do the research for this project, you know, I, I became more worried about the downside because, you know, people like Geoffrey Hinton— I went up and spent 2 hours in his kitchen in Toronto and sat there debating not necessarily whether machines would be more intelligent than humans, obviously that they will be, um, but whether they— whether machines have a motivation to harm humans. And he's very persuasive in arguing that they will. I mean, my point was no, look, humans evolved over centuries to survive, to pass on their DNA. We're hardwired to survive. That's why we fight each other. Machines aren't like that. So why would they attack us? Why are you so worried, Geoff? And Geoff is like, well, you know, imagine you've got this super powerful AI and it's going to be attacked by the enemy AI and you have to defend it. So you tell the AI, if you see a cyber attack coming, you've got to fight back, you've got to defend yourself. And now all of a sudden you've given your AI a survival instinct. And so don't tell me that evolution has to happen as it happened to humans. The evolution can happen in a machine way, but these systems are going to want to survive and they're going to be more intelligent than us. So we're in trouble. And I think you can't dismiss that. So I'm kind of both worried and excited at the same time. And I think that's how humans Humans generally respond to all technology. And if we didn't take that trade and move forward with both the excitement and the scariness of technology, if we didn't take that, we'd be still living in caves. So in some sense, the story of Demis is like the story of all of us, but magnified 10x.

There's a documentary that was just released or might be releasing right now that features an interview with Demis. The AI Doc, and it spends more time talking to all the different lab leads and voices in the industry, more focused on the doom question. Can we be apocalyptic? Should we be optimistic? But what I found most interesting was that the creator of the documentary, you know, summed up his full takeaway and said that AI is a Ponzi scheme. And I'm wondering if you got Any vibes that everything will collapse, that this is just not financially viable?

No, my view is that there's no AI bubble, but there's only just an open AI bubble, you know, in the sense that the technology clearly is getting better and better pretty fast, right? You know, we had the first chatbot in 2022 and it hallucinated, to kill the hallucination. Then they had longer context windows. Then they had multimodal systems that could handle video and pictures. And then they went to reasoning models. And now we're getting agentic models. And now next there's going to be world models that will be built into these things. This is a lot of progress in just 3.5 years. So I think it's accelerating in progress, and therefore it's not a bubble. But what is true is that it's super expensive to develop. And if you're not attached to a really deep pocket, like, you know, Demis is attached to Sundar, you're in trouble. Because I don't think OpenAI can raise enough money to bridge from today when they have a very popular chatbot, but almost none of those customers pay for it, to some future where the product is stickier somehow and they can charge much And so I think, you know, OpenAI has been running two simultaneous experiments. One is with a new frontier technology, and the second is how deep are global capital markets? And they already raised, you know, $41 billion last year, which was a record for any private fundraising, bigger than any IPO, by the way, as well. And so, you know, kudos to Sam Altman for raising that much money, but can you pull that trade break on a bigger scale every year until 2030 when they hope to break even? No. And that's why they're cutting products like—

Except for right now, because they just raised the $120 billion.

But if you look at the $100, it's kind of smoke and mirrors. A lot of that is contingent on you get this money if you go public, you get this money in the future, you get this money in kind in terms of compute or something. The 100 wasn't really 100.

Isn't there a little bit of a dynamic where you could wind up with like an anti-Google alliance? Maybe there's, you know, there's like tension between the industry and Google. This is like the foundational, like, myth of the AI industry and the AI labs broadly, although they of course have fractured many, many times at this point.

Yeah, I mean, you know, of course Of course, there's always rivalries. One might say people will gang up on NVIDIA. That's the occupational hazard of being the leader, right? I think they'll deal with it. I mean, I don't think it's a winner-takes-all market, by the way. I think that there'll be space for others.

Yeah, that makes sense.

Last question from my side, then we'll let you go. Did you have any takeaways or ideas around the diffusion of physical AI and robotics. Did anything stand out? Do you have a strong opinion there? Right now, it feels like we've entered the sort of software singularity. But we just had Kaser from Applied Intuition on, and we're talking with him about how autonomy and AI is diffusing through the physical world. But I'm curious if you had any takeaways.

Yeah, I mean, look, I think that I think that one thing sometimes people don't quite understand is that large language models and the transformer architecture that underpins them is super consequential for lots of applications, not just chatbots. And so robotics is being improved by this technology. And I fully expect to see a huge breakthrough over the next 2 or 3 years with the capability of robotics. And so I think that's going to be the big story. I kind of agree with the guy from KSR you had before that the movement of atoms is going to be affected as much as anything else. So that's part of why I don't think this is a bubble. I think the potential in this super powerful AI is enormous.

Yeah. Well, thank you so much for coming on. Thank you so much for coming and joining the show. It was a pleasure talking to you. The book is "The Infinity Machine." It's available everywhere books are sold. Thank you so much.

Fantastic cover too.

Yeah, beautiful cover. Yeah, beautiful book. Thank you for having me. We'll talk to you soon. Have a great rest of your week. All right. Goodbye. Take care, guys. Let me tell you about Finn.ai, the number one AI agent for customer service. If you want AI to handle your customer support, go to Finn.ai. And we will kick off the Lambda Lightning Round. Let's see. Oh, look at this. We got everything. This is amazing. Let's bring in Forest. Forest from Sumos. Welcome to the stream. How are you doing?

Hello, guys. How are you doing?

We're doing great. Greetings from Medellín. First time on the show. Please introduce yourself.

What a setup here. This is a great setup.

This looks fantastic. A little PowerPoint too.

A bunch of nerdy stuff from the team, all the things we're building.

No, hey guys, I'm Forest. Yeah, break it down.

Amazing. Yeah, kick it off.

Let's see, where do I start? I'm Forest. I'm a fancy plumber building infrastructure from scratch down here in Colombia to make internet way better and make compute and everything kind of work well in Latin America and beyond. So very weird journey, but making the fastest internet in the world at the lowest cost is kind of the summary in a very short nutshell.

Amazing. You have a lot of fans. I got a bunch of texts from investors that were excited for you to come on. What were you doing before, before we get into Somos? What were you doing before this?

Dude, I came to Medellín when I was like 18 years old, basically just dropped out of high school and started building stuff and ostensibly came to visit a friend for a couple of weeks. And that was 8 years ago. And here I am. So I made websites. I kind of took a very nontraditional path. And it's led me to doing a bunch of weird stuff here.

That's actually a non-traditional path. Normally when somebody comes on and says they had a non-traditional path, it's like, you know, Stanford to a Meta internship, a Google internship, a VC internship, but very, very cool. What, yeah, break down, like, you know, Somos, you're raising a Series B today. Like, how long have you been at it? What initially, I can kind of guess what the initial inspiration might have been. You're building digital products, it sounds like, and were probably frustrated with internet speeds, but what's the backstory?

No, I came to Colombia in 2018. I got roped into helping people build this blockchain incentivized blah, blah, blah back when everything was gonna be blockchain in urban slums here in Medellín, and as those blockchain projects happened, everyone got bored really quickly, so they just left me with all the equipment, And I just kind of became fascinated by how do you build connectivity in slums and didn't speak any Spanish and know anything about telecom and just started tinkering. And first couple of years were me like literally living in a slum here in Medellín, building internet, stringing stuff up in the middle of the street and kind of bit by bit, I just became fascinated by the concept of like internet is the most basic thing for the modern world. And we just sort of assume that it's been solved and really we haven't done engineering in the last like 30 years. It's the same underlying architecture that like John Malone was building in cable cowboy days. Yeah.

Well, yeah, every time I go travel to another country, I always look up like who the biggest industrialists are and who the richest people in the world are. It's always the people that built like the railroads, the mines, the telecom infrastructure, the Wi-Fi, the internet, the power lines. It's always like the stuff that once you install it, it provides value.

So yeah, it sounds like you're doing a lot What is in your strike zone of things that you want Somos to take on versus where do you work with external partners? Like what is the kind of core path?

This is kind of the insanity of Somos. So we literally do everything from like, we interface with the submarine cables, we built like a nationwide backbone crisscrossing the country. We like make the Wi-Fi routers. So like the things that go inside of the thing here is all PCB.

Oh, that's a Wi-Fi router. I thought that was just a pretty lamp. That's crazy.

So this is literally the router that goes in customers' homes. So instead of being some like ugly blinky box, it's like a beautiful lamp.

That's super cool.

You're like, wait, does it have to be an ugly box?

No one thought of this? Dude, it's incredible.

Like the internet is like literally the most incredible machine humanity's ever built and we just relegated it to being this like ugly piece of infrastructure and kind of our idea is like, what if it was dope? That's amazing.

So yeah, like how do you, How does the coverage map spread out? Everyone's familiar with those AT&T and Verizon coverage charts of the map of America. Do you have to go city by city, block by block? Are there ways that you can, how do you actually scale out to reach an entire coverage area?

So I literally have 1,000+ employees cabling the streets. We have linemen and installers, and they all work directly for Somos. And this is kind of the insanity of what we're doing. And it's been a bit of a, like, slow flywheel to get ramping. But then at the end of the day, you're building this actual moat because it turns out it's really hard to build infrastructure from scratch. So we literally are cabling the entire city from scratch with a new type of fiber to connect people to faster, cheaper internet. Yeah.

Well, in what ways do you think it's easier to do in a place like Colombia than the United States and and vice versa?

Yeah, I think this is one of the interesting things is like the US is kind of looks archaic in comparison to some of the stuff we're building here. So like my base plan for customers, like $12 a month and is a gig and we're giving people 10 gig connections and soon 100 gig connections in their home. So the kind of weird thing that has happened in the US is we sort of believe Comcast is sacred, so we don't let people go build new infrastructure. And we're almost— there's a world where it's like 10 years from now, we're like, shit, the US has like third world internet infrastructure. And what we thought of as the third world has infrastructure that lets us do all these crazy things that AI is enabling that you couldn't do before. So I think there's this inversion happening kind of in the same way as there were no landlines in Africa. We're kind of leapfrogging all the crappy old cable and just building internet as it was intended back in the ARPA days. That's amazing.

How do you get people to move to Colombia? To work on this, what's your pitch? Not saying anything wrong with Columbia, it's just, it's a long way away. It's a long flight. The time zone's not too bad, but yeah.

It's 75 degrees the other year round. We have awesome engineering. It's like San Francisco without the fog and we get to just build whatever the hell you want. It's kind of like cowboy Western technology, Galt's Gulch in the middle of the jungle.

That's amazing. Last question for me. I'm sure you got asked this in all the VC pitches. How is Starlink rolling out? How does satellite internet fit into this? Is that a competitor? Is that a complement? A lot of people in America sort of have both. But how, how, how does that fit into the picture?

I think the big takeaway here is I would sell your Comcast or Telefónica or Telemex stock because I think what's happening is Starlink is attacking lower density areas, moving stuff, mobile. Somos is basically building, think of Starlink for cities. So we're building a pure play, just the best internet for dense urban environments. And I think there's like a density threshold where below that Starlink will win. Above that, a thing like Somos will win.

But realistically, if I'm on level, if I'm on the 3rd floor of a 15-story building in the heart of the city, like you're gonna win on reliability, connectivity, speed, cost, all of that, right?

Yeah, just base physics will be cheaper and way better and more reliable. Like even Elon will say this, he's like, Starlink isn't really for cities, it's for everywhere outside of that, right? Yeah, okay, yeah.

So like basically unaffected by everything there because of physics.

Always good news. Mutual friend Zach asked me to, told me I should ask about how kind of AI is impacting bandwidth with needs and how that factors into the opportunity for you guys.

Yeah, I think we're going to look back and say like, damn, the infrastructure we have currently is far underbuilt for the future of applications with AI. And one of the things that we are thinking a lot about in Somos is what if you extend the data center to everyone's end home, all the crazy applications that you can build on top of that, not only just speed, but reliability, redundancy, low latency. I think there's a world where compute lives in data centers and we're streaming your OS, think of like super Chromebooks to everybody and that that's a world where you make compute way cheaper and way better in a way that historically wouldn't really be doable in traditional telecoms. And I think this is back to like, there's a world where LATAM has orders of magnitude better compute than parts of the US just simply because we rebuilt the telecom infrastructure from scratch.

Wow. What other, what other markets are on the roadmap?

Yeah, well, we're heading to Mexico in the very near future, and I don't know, there's some interesting neighbors of Colombia that are becoming open again that might be a new place to expand to from Colombia. So I think Somos Caracas might be a thing in the future.

Before we hit the gong, another mutual friend, Aaron, asked me to ask you about the High Frontier. What's up with that?

Yeah, I mean, I'm obsessed with, with this vision of the future as it used to be. And I think one of the things that is stoked for me right now is like, it feels like we're building awesome things like new ship factories and new infrastructure in the world that are like building a future as it used to be. Like we used to think the future was gonna be awesome and we kind of got okay with a very boring, muddling kind of version of it. And it feels like we're now turning this wave of like, let's go build orbital space stations and the Lagrange points, let's build fleets of autonomous vehicles. Let's build all these amazing things from scratch. So it's like build infrastructure, rebuild the world, make awesome things happen again. It's fantastic. How much did you raise? We just raised $40 million in this round. There we go.

Who came in?

So Ribbit led this. We had Bracket Capital and then USV, Kasek, and Y Combinator all kind of doubled down from the past. So I love it.

What a great YC story too. Yeah, fascinating company, fascinating industry. Just, yeah, I love it.

Did you, did you pop up to the West Coast for the raise or did you make everybody visit you?

Oh, it's a little bit of both. Like, we get people to come down to Medellín and they're definitely not regretting when they come visit. So we have some fun adventures driving to the countryside.

I love it. Well, thank you so much for taking Thank you for taking the time.

Great to meet you, Forrest. I'm sure you'll be back on soon. Have a great day.

Cheers, guys. Goodbye.

Let me tell you about Phantom Cash. Fund your wallet without exchanges or middlemen and spend with the Phantom card. Our next guest is Dino from Saronic. He's in the waiting room. Let's bring him in to the TVP and Altium.

Dino, how are you doing? Good, guys. How are you doing? Thanks for having me on.

Thanks so much for joining.

Great to finally have you on. Yeah.

Give us the State of the Union. What's going on with Saronic? Yeah, likewise.

I mean, you may have seen today we announced a $1.75 billion financing round.

There we go. So do we need to build some ships?

What's going on? We got to build some ships. We're super excited for this. I mean, it is a true byproduct of the execution that the team has delivered on over the last not just 12 months, but the last really 36 months since we started the company. I mean, our team is truly A+. If you look at where we were just, just a year ago, we came off a $600 million financing round. What that let us go do is we opened our first shipyard. We launched Marauder, which is a 180-foot autonomous and unmanned ship. We then announced a multi-hundred million dollar project into that shipyard to scale production of Marauder. And then Corsair, which you see behind me right here, our small USV platforms, we've already taken production of those well into the thousands. So now as we look forward, what we're going to do over the next 12, 24, 36 months with this capital raise is we're going to accelerate that. We're going to accelerate the production, accelerate the deliveries of our vessels to the US and our allies around the world. We're going to launch new products, we're going to build new ships, and then we're going to go and build new shipyards. Yeah, right. We're going to invest in the shipbuilding industrial base in this country to the tune of billions of dollars. We're going to create thousands of jobs. And ultimately, we're going to, we're going to unlock production rates that we haven't seen in this country since World War II. And we're going to do it in a very technology-first, software-first approach.

You mentioned USV. Is that unmanned surface vehicle?

What's the acronym? Unmanned surface vessel.

Vessel, got it. And then, so walk me through the use of, you know, boats used to transport people. Now we put equipment on them. How versatile are these vehicles? What are the different use cases? Are some weapons platforms or some just ISR capabilities? Like what is the range of utilities that the armed forces will get out of these? Out of these USVs? Extremely versatile.

The whole point of the platforms we're building is actually for them to be modular by nature, right? And actually, we actually try to change the acronym around a little bit. USV is like unmanned surface vessel. You know, when you look in the past, it's really like a remote control platform, very similar to a Predator drone. We use ASV, autonomous surface vessel, because what we're building at Saronic is not just one-to-one control, but it's true maritime autonomy to then go and deliver these platforms at scale and be able to control them at scale, meaning fleets of hundreds or thousands of vessels through the most advanced software on the planet for the maritime domain. And then when you're looking at the missions, the use cases that you mentioned, it really all just boils down to scale. Persistence and risk reduction. Yeah, right. How do you operate large numbers of vessels? How do you do, how do you do that continuously in what's becoming an increasingly dangerous maritime environment? And then how do you offer like real capability to commanders while keeping sailors out of harm's way? Keeping people safe is very, very critical and a key point to what we're building here.

I don't want to diminish the work, but I'm curious about how much of a challenge is it actually to create an autonomous surface vessel? Because it feels like when you're driving on the road, there's so many random conditions and the car can flip over. But boats, it's a little bit safer, I would feel like. But am I missing something there?

Like, what is the big factor there? Well, the other issue is you have people, other boats that are trying to kill you.

OK, OK. Maybe that's it. But I'm just thinking like a plane, you know, if it doesn't land perfectly, it'll crash. Like boats, you know, they just kind of rock through the water, but there's obviously more to it. So what went into making it fully autonomous?

There's a lot of— yeah. So the ocean's just a completely different environment altogether. So we deal with a lot of different challenges.

Sure.

One of the challenges that's really different from self-driving cars is, yes, there's a lot of complexities on the road, But that singular car really only cares about how it gets to its end destination. It doesn't care about how the other 100 cars get to its end destination as well. Sure. And how they're all working together collaboratively on a mission. Got it. Oh, and then you start throwing in 6, 8, 10-foot seas, high winds, enemy environments, some of the things that we're seeing now. And like whether it's the Black Sea, the Middle East, and we're anticipating in the Indo-Pacific, Those are very, very complex challenges that we're solving at Seronic.

Yeah. What goes into setting up a new shipyard? Do you have to kind of co-locate around existing shipyards? Can you kind of stand something up, you know, totally independently? How does that work?

Yeah, I mean, when you look at shipyards and the shipbuilding industrial base in this country, it's really how do you bring on net new capacity. You're not really co-locating next to anything because a lot of that capacity has really atrophied over the last 30, 40, 50 years. So what we're focused on is building new shipyards and then building the ecosystem and the infrastructure to support that as well through partnerships and vendor relationships. But one of our, one of our main projects and one of the things a large part of this capital is going to go towards is Port Alpha. Right. We have a shipyard in Franklin, Louisiana. I mentioned that we're investing hundreds of millions of dollars in that yard, but we're looking at a brand new shipyard, building this from the ground up, completely greenfield, investing billions of dollars to 10x the size of our existing yard. Right. To bring on new scale, new capacity and rebuild the shipbuilding industrial base from the ground up. That's what's needed, because if you go around the country right now, you go to places which used to be shipyards and you'll see apartments and condominiums that are called naval yards. That's not just a name they came up with. It actually used to be a shipyard. So what we're doing now is we're investing in the shipyards of the future again to produce at a scale that we haven't seen since World War II.

What's the best way to get a job at Saronic?

You can apply on our website. I mean, we are hiring, we are We are growing. I mentioned how amazing our team is. What we're doing is absolutely critical for the country. The team comes in every single day. The work they're doing is changing the world. And so if you're a top engineer or looking to get into the defense tech space, please, please apply. Everything we're doing is absolutely critical. We grew the team from 200 to 1,300 people over the last 12 months. And that's, that's only the beginning, guys.

That's amazing. Well, thank you so much for coming and breaking it down. Have a great rest of your day.

Incredible progress. The chat, the chat says just put the S-1 in the, in the SEC mailbox. We'll talk to you soon.

Good to see you. Thanks, guys.

Very much. Looking forward to following up.

Let me tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk, secure every Secure any agent with Okta. And our next guest is already in the Restream waiting room. We got Will Ahmed from Whoop. Whoop is back. How are you doing? Good to see you again. Welcome back, guys.

Thanks for having me. Give us the update.

What happened? How are you doing?

Things are good, thank you. We, we announced a round of financing today, our Series G, $575 million round.

So Yes, there we go.

Is this, is this the capital you finally need to make a bust-down whoop? Diamonds from the factory, that's what I'm looking for.

Can you do anything? You know, we do, we do have some, we have some premium offerings in the mix that are going to be just right for you, actually. Perfect. Very happy with the product development.

Jordi wants a bust-down. No, no, uh, seriously, give me, give me updates. What What's changed? Where do you want the product to go? And yeah, I am interested to know, are more collaborations in the works? Do you see that as an important point or are you just focused on more and more channel partners, more and more distribution, more growing because the product's pretty dialed at this point?

Well, look, it's been an extraordinary 12 months for the business. You know, we ended 2025 with over 100% year-over-year growth, you know, $1.1 billion in bookings run rate to end the year. Our membership is growing around the world. We're operating in 60 markets. We've got members in over 200 countries. We've launched medical-grade technology now, coming out with blood tests around the world. And so Whoop's really become this broad-based health platform. And, and you can see that in the financing announcement. Today. You know, we've got the history of Whoop with the world-class athletes we've added. Now LeBron James is a new Whoop investor. We've got Cristiano Ronaldo on the cap table, Virgil van Dijk, Matthew van der Poel. Some of the world's very best athletes, really from every sport, are now on the Whoop cap table. And then in addition to that, we've got a bunch of great existing investors continuing to support the company, Collaborative Fund, IVP, Foundry Group, and more. And then we've brought in some sovereign wealth funds from the GCC, proud to have Mubadala and QIA and 2.0. So really some of the biggest funds in the world that are, I think, phenomenal long-term partners. And then lastly, to the point about health, you know, we've added the Mayo Clinic is an investor in Whoop, one of the premier health institutions. What are they called?

That's a niche, that's a hard to, rare Pokémon in the venture world. I haven't seen, I haven't heard of them on a cap table.

They're not ripping a lot of seed checks at YC Demo Day.

Yeah, they haven't done a lot of investing, but we see a ton of synergies from a research and medical capability standpoint. And then we also have added Abbott as an investor, and you know, really one of the best medical device makers also in the world. So it's a phenomenal mix of investors, I think we've been able to achieve this because of, of the remarkable growth that we're seeing as a company. And I think also just the, the tailwinds around health and longevity.

How do you think your marketing mix will change over the next few years? Because I'm, I'm seeing $575 million LeBron James, Cristiano Ronaldo that has like crazy Super Bowl ad written all over it. At the same time, you're tech native. I could imagine going way further into AI-generated personalized ads, pushing the performance marketing further. Like, what appeals to you? How do you think you'll change? What will, what will stay the same? What will change over the next few years from a marketing perspective?

Well, we never wanna lose sight of the fact that we started in sports and we've built this aspirational performance lifestyle brand. Yeah. And so that'll really be at the core still of of a lot of what we do for Whoop. But we also now have a product that, you know, can detect whether you have AFib and tell you your blood pressure every morning and help you do blood tests. So it's just a much broader health platform than it's ever been before. And our marketing needs to reflect that. You know, one of the areas I would say of maybe not weakness but opportunity for Whoop is, is brand awareness. You know, we don't have massive brand awareness around around the world. And so this capital's gonna give us the gunpowder to really grow broadly internationally. And so you're gonna be seeing a lot of Whoop wherever you consume content.

Super Bowl ad incoming.

Yeah, I mean, if you wanna reach me specifically, maybe a partnership with like an athlete like Johnny Knoxville might work. Yeah, it just might make sense.

Whoop Live heart rate on some of the stunts.

I think that would do the trick.

I'm curious, like, how, uh, Whoop, uh, like, how do you guys think about improving accuracy at this point? Like, I'm assuming, like, a lot of, like, so, so much progress has been made over the last however many years, but there's still always incremental progress. Like, you can always be more accurate. How do you think about that? Is that still like a top priority or are there other, is it accurate enough at this point that there's better things you can focus, you know, the core energy of the team on?

I mean, I think big picture, we want the product to be getting constantly smaller and smarter. You know, we want it to be an aspirational product in the sense that it's something something cool that you wear on your wrist, or we want it to be something that disappears throughout your body and can essentially be invisible. And so however you can gather this data super accurately, have the datasets grow in nature, more sensing, more capabilities, more medical approvals, the better. And I think that's going to continue to expand our TAM. I think it's going to continue to deliver deeper insights for our members. So we're going to be leaning in pretty heavily on research and development. You're going to see a lot of very powerful sensing coming from Whoop in the years to come.

How is the peptide boom affecting Whoop?

Well, I think the underlying reason for the peptide boom is that people want to take more control of their own health. Yeah. And they're sort of generally frustrated with the tools that they've had to improve their health. And so that leads in different directions. Ads being part of it, supplements being part of it, concierge doctors being part of it, AI health coaching being part of it. But broadly speaking, it's, I think, good for Whoop that people want to take more control of their health. And 10 years ago, I would talk about health monitoring and people would say, well, that sounds like something niche for athletes. And now everywhere I go, people want to talk about how they can improve their sleep. Or improve their VO2 max, or what is heart rate variability. So there's just clearly been a cultural shift to care a lot about your health. And, you know, longevity has become one of the most common reasons that people use the product. Our Healthspan score, the Whoop age score, has become the most screenshotted page in the Whoop app. So clearly you've got people who want to show off, uh, how old they are, or, you know, who want, uh, some counseling for how old they might be.

Yeah, yeah, that makes a lot of sense.

Yeah, there's an interesting dynamic with the various health platforms where there's like kind of an incentive, there's like a weird incentive to like, you know, measure, like say somebody's, you know, their chronological age versus their biological age, make it like lower, so people are more likely to share. It's like, I've seen some people have, you know, come in and say like, well, my biological age is 19.

I did one test that said I had the mind of a 5-year-old. It said, it was testing my brain health and it said that I had the—

Is that good or bad? It's not.

It's extremely young. I'm way over 5. So I assume it's good.

Look, I think we've built the most credible biological age metric because we did, first of all, we did it in partnership with the leading longevity institute out of California, and then the Buck Institute. And then we show you in great granularity each of your biometrics and what's improving it and what's not and by how much. So here's a trivia question for you. What percentage of people on Whoop do you think have a younger Whoop age? Oh, interesting.

I mean, if it was perfect, it would be 50/50. I would think 70% because people that use Whoop are much more likely to be healthier. Oh, true. Okay, so let's go with 70. What is it?

It's 55%. 55%. But what that means is 45% are older on Whoop. You know, that was—

They're early in their Whoop journey. You guys have done a lot, but this whole show is really dialed in.

No, so 45% are—

Yeah, sorry, it's idiot.

Yes. So, you know, but that shows that like, it's not just telling you what you want to hear. Yeah. It's, you know, it's gonna push you.

Yeah, yeah, yeah. And I only brought that up originally because I've seen some of these come out and I'm like, okay, there's zero shot this person's biological age is lower than their chronological age. Based on the life they live. And not every platform, you know, platform are just kind of— every platform is going to run their own kind of algorithm to determine that, and not necessarily working with the Buck Institute. Yeah, exactly. Yeah. Um, well, very cool. Fantastic. Great to get the update. Congrats to the whole team. I feel like just in the last year, like, the world has woken up to this kind of category. And your guys' progress is a testament. And I still think it's— I still think it's very early. It must be fun to walk down the street for you and 'Cause I mean, someplace like LA— Oh yeah, you probably see 'em everywhere. And I'm sure it's like every 10th person has a Whoop band on, but that means there's 9 or so. Yeah, it is a trip seeing Whoop in the world.

And for people who decide to take the hard path of building hardware, it's an amazingly rewarding feeling when you see a physical product that your team's built in the world. So I will say that's a very gratifying thing. Thing. But you guys, you guys owe me a question here you haven't asked me.

What is it? What's the question?

Aren't you gonna ask me if the job's done?

Oh yeah, is the job finished?

Job's not done, guys. We gotta keep going.

Thank you, thank you for that.

Great stuff. Congratulations.

Well, all right, I appreciate you guys. Keep it up. Thank you.

We'll see you soon, I'm sure.

Yeah. Yeah, get ready. The chat's gonna be your strongest supporters. Thank you so much for taking the time to come chat with us. We'll talk to you soon, Will. Have a great one. 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. Continuing our lightning round, you know who we got? The CEO of Public.com. Yannick, how are you doing? Hey guys, I'm doing well. How are you? We're doing fantastic. Look at that stuff in the background there.

I know. Do you recognize any of this?

Oh yeah, yeah, you got the whole head to toe, head to toe.

I don't recognize what Jordi is wearing. I don't think he's got one.

This is the new polo. You're right there, top left. You got lead left.

There's plenty of room on the wall.

There's plenty of room on the wall. Anyway, uh, Enough about our merch. Let's talk about your business. Walk us through the launch today.

Yeah. So today we launched AI agents for investing. Yes. The easiest, safest way to put agents to work directly inside your portfolio. Yeah. And the way it works is pretty simple. There's now an agents tab directly within the public app. You chat with an AI to set up agents that monitor markets, move money around, and even execute trades for you. All within the app, right? So there's nothing to install from a security standpoint. Obviously, everything stays in a safe, controlled environment because it's within the brokerage. Install Axios.

No, do not do that. That is such a niche joke. Everyone's going to think you're talking about the publication. Yeah. No, but I mean, it's been really fun.

I have a bunch of agents running now on my account. It's been really awesome to see how it's changed my behavior as an investor, right?

So Relax. I like the sound of that.

I knew you were going to do that. We got Lil Jon in the studio.

No, so makes total sense. What— if people are already signed up, which they should be, what should they— what's the first thing that they should try to set up or experiment with? What was your first few agents that you set up and continue to keep running? Because I'm assuming you you can effectively run them, you can retire them at different points.

Exactly. Exactly. The first one I set up was one that checks the oil prices before the market opens and buys protective put options every day as a hedge if there's a spike in those. I call it the I'm tired of seeing red due to war agent. And so today, that didn't fire, thank God. I have another one that just looks at my bank accounts and automatically sweeps any cash in excess of a certain amount into my bond portfolio. So I'm always yield maxing. You want to be yield maxing always, especially while rates are still high.

Yeah, that makes sense.

And then I got some more advanced stuff, like I got one that scans the markets for opportunities to write covered calls across my top positions. So if there's a low-risk opportunity to make like $20,000 a month selling options premiums, I instructed it to just go ahead and place those orders. And so that just rolls every time, all the time.

And I don't think that exact strategy was the first thing a private wealth manager ever pitched me in my career, like a decade ago. They were like, I— and they had like a guy that did it and you had to have a lot of money to like access that. And there were minimums and stuff. And now I check on that guy after today's launch. But so, so there are obviously like incredibly advanced things that you can do with these agents in in the market. I'm also interested in just driving behavior change because a lot of folks that I know are earlier in their career, they don't want to necessarily take a ton of risk. The biggest lever on their financial future will just be seamlessly funneling money from their paychecks into something as simple as VTI, and then they can do something more, more advanced down the road. But what does it look like in terms of the best practice or best functionality for just creating a set it and forget it? I want to make sure that every time I get paid, money's flowing into the market. How easy is that these days?

Well, I think with agents that becomes really, really simple, right? I think this is sort of the whole point. You know, the stock market's always been about manually entering orders, right? Like you do all this work and eventually you end up manually being like, buy 200 shares of Apple this price. Yeah, I think that user interface is now shifting to something like increase my position if valuations compress 15% from here. Yeah. You know, and it stays within my defined risk tolerances and so forth. And so I think it changes how people think about and manage their portfolios in a pretty profound way. And we do see it as a user interface shift. Like, you know, the interesting thing about this industry is every technology kind of had their model of brokerage, right? Like we started on the horn and then the dawn of the internet gave us the discount broker. With mobile came the neo broker. And I think now with AI, it's the era of the agentic brokerage. But what's uniquely interesting about this shift is every previous shift was about streamlining the process and reducing the responsibility set of the broker, you know, to basically just trade execution ultimately. But it actually used to be much more more full service. To Jordi's point, there used to be a guy used to call you to pitch all these kind of ideas, risk, you know, trade ideas, etc. I think with the identity brokerage model, you're reversing back to that. And it's much more full service than obviously any human service could ever be because this thing can like write an algorithmic trading script for you in 10 seconds. It can do tax loss harvesting. It can instantly analyze risk. Sure. And so it's a shift back to a world where the brokerage plays plays a much, much larger role than just trade execution. It sort of goes into the realm of maybe a quant and a financial advisor. And that's what we're excited about, the brokerage playing a much bigger role through essentially a GenTech.

Jordi, can it, can, can it pull in external data sources yet? I'm thinking like fear and greed index. Like, should I, could I set something like unemployment if you have like MACD? CPI, that you want to buy on days when the fear index is high, but that might not be an actual instrument that you can buy and sell directly.

Yep, 100%. Unemployment data, CPI, Fed cuts. Like one that people have been kicking around today is, you know, whenever there's a Fed cut, move money obviously out of my high-yield cash account on Public, put it to work into high-growth tech. We like the sound of that. All right.

That makes sense.

There we go.

And so there's a lot of those CPI, like the fear and greed one was requested today. I think that's coming in the next couple of days. And so really it's about getting all that into this natural language interface and just letting people kind of instruct AI to do this on their behalf.

Last question for me about AI on the platform. What are the capabilities? What have you learned about users educating themselves about various financial instruments within the public ecosystem? You know, okay, I see a company, you're gonna surface price-to-earnings ratio, market cap, the usual stuff. But there's so much more that you can ask an LLM these days about what does a company actually do? What is their strategy? What's the history of this company? What's the founder like? These things are perfect for LLMs and you can vend those in, but what are users actually using? What's adoption been like? What have the learnings been?

Yeah, I mean, I think one of the core tasks in application layer AI is to sort of figure out obviously what do users want to achieve and then which model and kind of harness is best suited to achieve that purpose. But then also focus on like, what are we uniquely able to deliver, right? So we know what you own. We know what you used to own. We know what your risk tolerance is. By the way, we also know the difference between what that actually is and what you said it was when you signed up. And we have real-time data feeds of everything. Right. And so I think as a product builder, those are like some of the situations where you can really create a magic moment that general purpose LLMs can't. And I think a lot of that comes because we just have a lot of that kind of history about how people people like to invest, what questions they've asked, to your point, in the past about the P/E ratio or what the founder is like, etc., because we've been basically running a research assistant since 2023. Yeah. And we're only a 6-year-old company. So it's already for sort of like half the time that we've been live. And so we've been able to gather a lot of data for the last 3 years that we can now kind of repurpose into this.

What's your theory right now? It's obviously day day one, but do you think in the future we'll get, you know, more volatility because you have like financial institutions that are effectively using like agents or algorithms to do trading and then you also have retail. So when you get like a new CPI print, you know, you have, you know, even, even, you know, additional trading activity off of these single events. Like, do you think this is something in the future that, that everyone will effectively have like a handful of agents running just naturally? In the product, and then some people will be like maybe prosumers, people that are more into it, will have many? Or at some point is everyone— at what point is it old-fashioned to be just buying a stock yourself, even with a button?

Totally. I actually think we will look back at tables that buy buttons and feel that's a little antique, maybe already 12 months from now. But I think I think the effect is things will get priced in faster for sure. On the institutional side, they've, you know, they've used some version of AI for the longest time. Right. But then at the same time, retail have gone from being like 5% to 25% of the market. And on the retail side, folks haven't been as fast to react always. Right. They haven't been that disciplined. They're not necessarily glued to the screen 24/7. And so, you know, they can't always react as quickly as they want to. And agents obviously change that, right? And so I do think that there might be— I mean, it's a little bit like whether it's crypto prediction markets, there's always a little bit more of a, of sort of an alpha opportunity or an opportunity in the early, early days. And then over time it becomes more mainstream and that kind of fades away. And I would suspect that this follows something like a similar pattern.

Yeah, I've just been thinking about it because because the content on X is primarily user-generated, at least the big accounts, it means that an event happens in the real world, it gets reported on, or it pops up on a website, or you get a newswire, and then a human takes that and puts it on X. And then this trade, even the majority of retail volume is like flowing off of like that human seeing the news, posting it, and then you get this sort of trading activity. But in a perfect world, you see news, and then you go to your broker and the right trade has already been made on your behalf. Exactly. And anybody that's not adopting this will just be like, well, I missed kind of— I missed the opportunity. Unless you want to get out entirely.

Totally. Speaking of X, a fun one is that was from the US. If DJT says buy, just buy. That's actually really smart.

I think this probably backtests very well. It backtests extremely well.

What a crazy timeline we are in. Well, thank you so much for coming on and breaking it down for us.

Great to see you. I miss you. Let's, let's— I miss you too. Be sure to hang out. Let's hang soon. We will too. Yeah. All right.

Bye guys. Cheers. See ya. Ciao. Let me tell you about Sentry. Sentry shows developers what's broken and helps them fix it fast. That's why 150,000 organizations use it to keep their apps working. And without further ado, we have Ryan from Crosby. How you doing, Ryan? Welcome back, guys.

Good to see you again. Dude, you're on, you're on here like every week.

It's getting ridiculous.

Yeah, it's getting a bit much. Let me guess, you should just bundle all the fundraisers together into like, uh, series alphabet.

Then you just do it.

I never get to see you guys.

It is, it is more strategic to break them up. But tell us what happened. How much did you raise?

What's going on? We have two announcements. The first announcement is we've raised $60 million led by Lux Series B. And our second announcement— thank you. The second announcement is we did some math last month and we have now closed contracts worth over $1 billion for our clients.

Whoa. It's a big milestone for us.

Okay. Okay. $1 billion for the client. That's good. Yeah. Power on there? Was there like one sneaky $950 million?

Did you get 1% of this new OpenAI round in there? Somebody was like, I'm going to review one clause and you're like, I'm getting— No, it sounds like, it sounds like there's a lot of lawyers using it. That's right.

I mean, these are small deals, so it's a lot of velocity. I think definitely my corporate law friends are like, that's like one deal for me, so that's not interesting.

But for us, that's a big number. That makes sense.

So it's a really good milestone.

Yeah. So yeah, yeah. Take us through I mean, it sounds like, like the shape of the work that is being augmented by Crosby these days.

Yeah. So, you know, we're about a year and a half into it. We announced our seed around 230 days ago. We do commercial agreements. These are the sales agreements, MSAs, NDAs, BPAs for like fast-growing AI companies. Now we're branching out, but since the beginning we've been a law firm. So we have about 30 lawyers here. Who I'll give a shout out to, are just the last day of the quarter. They're working so hard for our clients getting the deals closed and we close deals fast, like in a couple of hours. And so this idea has just taken off with a lot of tech companies now and now even bigger clients who just want to close faster.

How, how have you been processing your kind of, I would say, very tapped into how well the models work in different roles. I'm curious your view on on how application layer legal AI companies will do compared to just the labs themselves, right? I feel like every other day on X somebody says, wait, this LLM seems to be doing just as much as, you know, this application layer company. You guys are using all the models internally for your own internal tools, but like how are you processing of what feels like, well, ultimately in the same way we saw with CodeGen, where you have application layer companies and foundation model companies, and then you have foundation models with their own applications. I'm assuming we'll see that in legal, but how have you been kind of processing it?

So, I mean, that is the question we have to ask ourselves every day. We think that cogeneration more or less is kind of like a year and a half ahead of the sort of non-self-verifiable domains. So anything that's not like math or code, and law is one of those, but it's a huge service area. And our sort of like insight a couple of years ago was, let's, you know, not think about these sort of like AI copilots that are kind of like, you know, the equivalent of what Cursor was a year and a half ago when you kind of hit tab to autocomplete, but these long-form agents with bigger contexts that could do a full job end to end. And if you have agents that can do entire swaths of legal work, work, then the best thing you should do is start a law firm because you're selling work to clients, not, you know, fractions of work or kind of helping them along. And in truth, we were ahead of the models, and so we were selling something that, like, we weren't able to fully automate. And as the models are progressing, we're seeing more and more of a compounding advantage as, you know, we have more and more contracts that we're processing, we have more and more lawyers that were able to help us, you know, tune judges and, and, you know, like, create better agents. And so we're able to just do end-to-end work in a way that like if you were just selling a legal, you know, copilot, I think you're going to face a lot of competition just from the models with no customization. Yep.

Sorry. Wild. John's back. Yeah. Wild, wild moment. What I'm assuming you'll also face competition from clients that are just like, hey, we can— should, should we, should we have an in like Should we have an in-house lawyer that we can speed up? But everybody's competing with everyone. But yeah. How are you tracking the legal education market?

I've seen it, it seems very hard to predict for me. Like there was this weird spike, I wanna say like it was maybe post-ChatGPT where there were like more people signing up for law school. And that was like sort of contrarian based on the model capabilities capabilities like the AI/SF discourse, but maybe it makes more sense. Like, are you tracking that data? And then are you tracking like how legal education is changing? I imagine that using AI tools already happening in middle school for a lot of people, high school, definitely college, definitely law school. How will that, how will that all trace through and how closely are you following it?

I mean, I think every industry is asking themselves like, how do people get the entry-level jobs to learn those skills and then become really good and senior and get leverage by agents. I went to law school at Stanford. I'm talking to a lot of professors there who are struggling with this question. I think our insight for now, like the stat we found recently, is that the top 100 law firms last year made a little under $70 billion in just profit just in 2025. That's just paid out to their partners. That's just salary. And that's for lawyers.

And that's not a good year. It's so good to hear.

And that was, that was more money they than Google spent on all their R&D. And so like our insight was— Logs. Which is great. So like if we could just put some fraction of the profits law firms are making into building better tools and experiences for lawyers and for their clients, I actually think the legal industry gets a lot bigger. And so it's like for a person in law school today, it's a good time to be thinking like, how can I just build better stuff? And that's just a new way of lawyers thinking.

Okay, that's one way to put the profits to work. Let me pitch you another way. If I'm a partner I'm running a law firm, and I see that, yes, agents can do the work of the associates that I would be hiring. Maybe, you know, contrarian in me wants to still hire associates just for the mentorship and building, like, the pipeline of partners that will do more human work, more deals work, more interpersonal relationship work. But I know that if I don't I don't start buying and paying for that service right now, even if I'm getting less margin on it loosely, 'cause I'm paying an associate a bunch of money and it's work that an AI agent like sort of could do and maybe they're a little bit more free. I'm actually incentivized to figure out how to accelerate them faster in their career, have them start working on larger, stickier deals that AI can't necessarily navigate just yet.

Yeah, I mean, I buy the argument. I think that there's two jobs for lawyers really to focus on right now. One is just doing client-facing work and being really good at being like, you know, talking to people and understanding what their points of view are and not being buried in the sort of paperwork like a typical associate. And the other is being able to explain reasonably well to an engineer or a researcher what it is you're doing and what you're thinking about and all the subtleties of context. And those two things, I think, are both things that if you're not hiring enough lawyers, you can't do well and you can't build better legal technology. Experiences. And so I think we're feeling this and every law firm is feeling like you just need people to be really thoughtful about doing both of those jobs.

Yeah. Yeah. Are you guys fine-tuning any models, you know, based on fine-tuning like open-source models or is that not even, you know, I've seen like Fin and Notion have had some success with this. Is that even a good use of time right now? Because I'm assuming your guys' like actual like inference costs are not that high relative to what you can charge clients, even if you're using the frontier models. But how are you thinking about that?

I think, again, if we just look at cogeneration as like the blueprint for the future, you'll see like a lot of the cogent companies got a lot of lift from just like, you know, the main, you know, 3 big models. And over time, you have to start fine-tuning your models as you get scale, as you get data, and as you need a more competitive edge. So we're not there yet. We have a lot of lift from just getting the right context to models, building the right agent flows. Like just doing some reinforcement learning on like basically, you know, we work with really, you know, OpenAI, Anthropic, and Google's models. But yeah, in a year and a half, as you get really specialized in use cases of law, I'm sure like we're going that direction. And part of the reason for this funding and doing it so quickly is to start investing in a research team that can kind of push the boundaries there.

That's very exciting. Well, congratulations on the funding round. I'm sure we'll see you. We'll just book it now. You just tell us. Us. And same time next month.

Thank you, gentlemen.

We'll talk to you soon. Have a great day, dude. Thanks, guys. Let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents. And without further ado, we have Chris Yu from Allsow. How are you doing, Chris? I'm doing great. Taking the time.

Yeah, thanks for having me. Chris, welcome to the show.

Since this is your first time, please introduce yourself.

Yeah, my name is Chris Siu and I'm the co-founder and president of ALSO.

Okay, break it down for us. What is ALSO? Give us some of the corporate lineage, the strategy, the product, just sort of everything.

Yeah, so actually before 2022, RJ and I met and we immediately hit it off on this one topic. And so that turned into me joining Rivian at the time with the explicit mission to create a startup within a startup. The entire thesis that we had, and that's— this has turned into also with the spinout last year, is that, you know, if you look at the vast majority of trips that happen for the movement of people and goods around the world, they happen in smaller-than-car things. Yeah, but nearly none of them have been electrified yet. And so it's really taken the Rivian or Tesla playbook and applying it to these smaller form factors.

Okay, uh, smaller form factors, that means everything from, uh, hover board to a horse and carriage? What are we thinking? We're focusing on wheels first, but yeah. Okay. But yeah, yeah. Narrow down the product for me, the go-to-market, the quad and the pedal-assisted electric bike. Are there timelines, shapes, sizes, ranges? Like, how do you think about narrowing down the product set? Because it is a really wide and diverse category. Totally. Yeah.

So we think of it as, in a way, two phases of the business. Phase one of the business is how do we create a vertically integrated software-defined EV platform but optimized for small form factors? And we've applied that to our first products. We call them EVs that you can pedal. And we announced those back in October. So one is a consumer e-bike and the other is a pedal quad, which we partnered with Amazon to deploy soon. That's really exciting. But if you look globally again today, things move around in things like two-wheelers, like scooters, boda-bodas, tuk-tuks, microcars, kei trucks. There's just these rich, diverse set of form factors. And again, none of them have been electrified and all of them are ripe for a really kind of tech-forward platform, which is what we're building. But importantly, today we announced a partnership with DoorDash, and that kind of underpins phase 2 of the business, which is Very good.

Thank you. Yeah.

So if you look at the world becoming more and more autonomous and even as that happens, some fundamental constraints don't change, meaning these trips are all happening in dense urban suburban environments. Congestion is always going to be an issue. Cost per mile is always going to be a factor. And so we believe really strongly that even in a fully autonomous, autonomous world, small form factors make sense for a lot of these trips. And that's really what this partnership is about.

Talk to me about, I mean, I love RJ. Rivian's an incredible company, obviously a younger company in many ways than other EV makers that might be more vertically integrated. So I'm wondering like how much is it that you're taking the supply chain knowledge, the expertise, the best practices, the connections, and setting up sort of an entirely new supply chain that's distinct versus you're just going to be able to buy stuff from Rivian or license it, or there's going to be more of a business relationship other than just funding? It's all of the above.

We, RJ and I, talk about us as kind of like sibling companies, if you will. So I think, I mean, there's a few aspects. One is we share the latest and greatest from a technical architecture standpoint. So if you look at how also vehicles are built, they're very, very similar in terms of how Rivian is built. There are some commodities that are shared. So battery cells, we are— our first products actually use the same cell that are in Rivian R1s, and that helps a lot from a scale standpoint. But there are other areas where we are taking a decidedly different path because our products are different from a car, truck, or SUV. So supply chain that you mentioned, that's actually one of them. We are fully engineered in-house, but we partner with contract manufacturers across across the world to be able to do assembly, and that's right for smaller-scale products.

Yeah, and with cars, you almost always want to manufacture them where they're going. Like, that's why even, even like the Japanese car makers have facilities in America or Mexico, because you just would drive the car as opposed to putting it—

That's exactly right. I mean, if you look at a car, I mean, the size of tools necessary, the custom they are, tariffs, like they all add to like, you have to have your own factory in region. Yep. But if you look at any product south of a car, almost all of them are built with this contract manufacturing model. Okay.

Talk about the name and the brand. Rivian has those delightful headlights, a lot of different interesting brand decisions around Rivian. What are you taking? What are you thinking? And where does the name come from?

That's so great. I love it. We— naming something is so hard. And so RJ and I battled quite a bit with this, but when we landed on this one, we knew it is the one, because if you look at transportation, it's always been so singular narrative. It's like it's just cars or it's just not cars. Yeah. And for us, the approach is like whether it's a commercial enterprise or a consumer, it's all of the above.

Yeah, most likely.

Meaning that I want to use my R1 to go on a long weekend trip, but my school drop-off with my kid, it's a pain in the butt to sit in the car line, um, and something smaller probably makes more And so the transformation and electrification of transportation is also— it requires all of the above in a way. And on kind of like how we present ourselves from a brand, you know, one analogy that RJ and I really love and use often is it's kind of like we're two characters in the Marvel Universe, if you will. So it's like we have the same mission, but we can have very different personalities. And so also has an opportunity be maybe in a way really expressive and take a little bit more liberties, which you're starting to see in some of our products than a more grown-up vehicle brand may need to be.

Okay.

How do you think about competition with Chinese manufacturers? You know, Rivian's had the benefit of not having to compete with all the Chinese manufacturers in the U.S. Yeah, I imagine micromobility, you know, is not going to be having the same kind of export restrictions, you know, how do you, how do you think about that threat? You know, assuming that, you know, there's Chinese companies out there that for one reason or another will be able to like sell at a loss for some amount of time.

Yeah, the DJI story basically. Yeah, that's a great question.

I think there's a couple of ways to think about it. Increasingly, as you get to the larger form factors in our portfolio and certainly as we get into autonomy, I think a lot of the similar factors that we're seeing in the automotive world in terms of the natural firewalls that are happening will exist in our space to some extent as well. But just to back up, I think one of the things that gets lost is there are a tremendous number of products in this kind of like small mobility or micro-mobility space that are coming out of China for sure. But I think it's without debate that the vast majority of these products are commodity, like relatively low-quality white label type products. That being said, there are a small handful that are really, really great products and using the latest and greatest tech. And if you look at— take apart one of those products and you take apart one of our products architecturally and from a technology capability standpoint, they're actually more the same than not. And I would say also is probably one of the only brands outside of China that you could say that. Yeah. Of within this space. Space. And so just purely from a product feature quality and technical capability standpoint, we feel like we're very, very competitive.

Okay, product pitch. The Rivian R1T has a gear tunnel. It fits a snowboard. Electric longboard with a handle that flips up like a giant Razor scooter that fits perfectly in the gear tunnel. Am I on to something?

I love it. That, that's not the first time we've heard that one. Oh, really?

Okay, the gear tunnel, it just, it just does feel like such a unique feature, and it just, it just demands some bespoke thing that fits in there. You know, you want like a big speaker, Bluetooth speaker that fits in there, like barbecue or something. It just— I want an ecosystem around the gear tunnel, even if it's, you know, who knows how viable that is. Uh, anyway, very fun.

Jordi, anything else? Very cool. Thank you so much. I'm on the website right now. I'm interested.

You're shopping? I'm shopping. I'm shopping.

We'll hook you up, just let us know.

It'll be very excited to ride these around. We've been doing, uh, we've been doing some, uh, office chair racing in the studio.

This is apparently a whole— that's what I want. I want an electric office chair.

Well, there you go. We have, uh, well, we have in-house vertically integrated motors and burners. We can power them, we can soup those up. There you go.

Adjust me to the left 1 inch.

Yeah, just a little joystick.

Uh, thanks for hanging out, Chris. Pilot me You know, if you're not in the right shot, you're a little bit to the left, production can just move you.

That's actually awesome. I love that.

You will have one or two customers for this if you build this. Autonomous chair, I love it. Autonomous office chair. Hey, you have to leave the meeting, go back to your desk. Drive you around. I think we're on to something. Well, thank you so much for taking the time to come by.

Great to meet you, Chris. Yeah, thanks for having us on the show. Thank you.

Thank you. We'll talk to you soon. Goodbye. Let me tell you about app-loving profitable advertising made easy with Axon.ai. Get access to over 1 billion daily active users and grow your business today. What's up? Brad Adcock. Yes.

On the show yesterday. Yes. You had some interesting comments about the state of AI. Okay. I disagreed strongly with many of them. Okay. But we have to cover this video. Yes. From the Sean Ryan podcast. Yes. It's a new gate. Okay, it's a— they're calling it a telegate. Ad gate? Ad— well, maybe that too. Who knows? So, so he is hanging out with a figure robot. Okay, on the Sean Ryan podcast.

Oh, okay, you went outside for it. Yeah, I was wondering because like Sean Ryan normally shoots in that like very cinematic whiskey bar, but he's outside and there's like a 2-minute video where they're hanging out with the robot.

Yes, let's pull this up. Up, and I want to get your take. All right, turn around. So this is the first time he tells it to turn around.

The thing too is like we, uh, we basically, uh, the robot—

end of the video it starts turning around and then he says turn around.

And Nima here says the video is the smoking gun that figures robots are teleopped. Again, I love teleop, not a problem, but Brad has consistently said he's not doing teleop.

He'd never do teleop. I don't know.

This is not autonomous. Notice how the robot starts turning around before Brett says, "All right, turn around." Yeah, you can skip forward a little bit.

And there's, pull up this other video that I'm actually on.

There's another, yeah, there's another video quoting this that shows it on repeat.

Let's see, now.

The other thing too is like we basically, the robots almost all fully software. It's by Vic. Vic Courtreyan said, yeah, okay, yeah. It's definitely not waiting for the command.

Okay, it's very— This one. All right, it's very subtle, but you can see it's turning around and then he says, "All right, turn around." All right, turn around.

Yes, but the steel man here, premonition. All right, turn around. The robot knew what was gonna happen because personalized superintelligence understands that a turnaround command is coming before Brett even says it, starts turning around before. So that would be one possible solution, but yes.

Who knows? Also— The moment from yesterday that stands out to me is I said, why build a separate AI lab focused on personal superintelligence outside of your company that is trying to sell some combination of intelligence in the physical world? And he said, I really value focus, which I thought was fascinating given that he is diverting his personal focus.

Yeah, it's like focus within an organization, like a specific, like the leaders that join that company can focus just on that problem. It was an odd comment to sort of process. Yes, I mean, I don't know. I haven't watched the full interview with Sean Ryan. I wonder if he talks about Web3 whether or not this particular robot is teleop because it's totally reasonable that a company would have some teleop robots, some autonomous robots, and sort of mix and match them based on the particular demo. Obviously, if you're doing some sort of pre-scripted stunt, dance, or parkour scenario, you might pre-script that. And then ideally you would say, hey, you know, this one's teleop. Here's a demo of what we're capable of when we're using teleop. Here's a demo of what we're capable of of when we're fully autonomous, when we're partially remote controlled or something like that, somewhere in between. I don't know, we'll see. People will continue to dig in. I mean, all of this, the rubber meets the road when the robots are out in the wild, when people get them and they start shipping and people can see, unless you buy one and it's secretly teleop, that would be wild. You're like, wow, this is remarkable. I can give it the most common commands complex. I can give it the most complex vague instructions, it just does exactly what I do.

If the Figure robot can simply open a Diet Coke for John, we're a buyer.

That's, that is, that's the goal.

I don't— we don't need it to do everything. That's the goalpost.

Just need to crack open a 6-pack of Diet Coke. Uh, Tyler, what do you think about, uh, the Figure, Figuregate?

Uh, I mean, it's hard to say just from that video, but I think broadly, like, People are probably like too against teleoperation generally. I agree. Because like, you know, the lesson from Waymo is that like actually it's goaded, you know, part— maybe if it's totally like 100% teleop, like, okay, that's not great. But if it's like partly like there's someone overseeing it, yeah, and maybe they're like pretty involved sometimes, it can be like extremely valuable. Like Waymo is a great product, whatever. Yeah, like even just like deploying robots in dangerous locations, if it's fully teleopped, that's still like a great thing, hugely valuable. And like clearly I think the way we get fully autonomous robots is by starting out with partially teleopped ones, so you get the data. There's like a very clear loop there. So I think broadly people are too against.

I completely agree. Yeah. Anyway, anything else we need to talk about? No. We talked about all of this.

We'll be back.

Oh, we do have one follow-up to yesterday. So we did a little deep dive from the Wall Street Journal on Mark Lanier, the lawyer who's successfully argued that Meta and YouTube are addictive in the LA court last week. And we posted the clip. A lot of people enjoyed learning about him and in particular the fact that he has a menagerie that contains lemurs and llamas as well as a 120-person train. We love the way he's living his life. We're huge fans of Mark Lanier, although do have some disagreement around the legal findings. But a lot of people chimed in. @xlraider said, "By the way, this is the Lanier Theological Library in Houston, which is open to the public for touring. Incredible guy." Shares two amazing images of, you know, what an amazing contribution to the community. And Eric Sufert quote tweeted our post and said, "This is true. I grew up down the street from his property and he hosted a high school graduation party for one of my friends. He recently bought an adjoining horse ranch and built a seminary on it. So a lot of people coming out in support of Mark Lanier. And yeah, I mean, just seems like a fantastic lifestyle.

We'll get him on the show.

Fantastic menagerie. And you really reset, you know, everyone focuses on, oh, are you flying private? Are you post-economic? Like menagerie is clearly just a different tier, different ladder. That's where you wanna go in life if you're successful. And he's done it. It. So congrats to him. Anyway, thank you so much for tuning in to TBP today. Leave us 5 stars on Apple Podcasts and Spotify. We sign up for our newsletter at tbn.com.

A wonderful last few hours of your quarter.

Yes, it's been an honor. See you tomorrow. Goodbye.

Bye.