The Steelman for Grok's Vulgar Anime Mode, AWS Launches Competitor to Browserbase, What AI Product Will Zuck Launch Next? | Delian Asparouhov, Lucas Swisher, Ravi Gupta, Adam Warmoth, Misha Laskin

You're watching TVPN.

Today is Wednesday, July 16, 2025. We are live from the TVPN Ultradome.

The Temple of technology, the fortress of.

Finance, the capital of capital. Do you have a question for me?

Why are you wearing a knight's helmet?

Because I have to play the steel man today. I have to steel man. The hardest thing to steal, man, in all of technology. No one's doing it. It's almost impossible. But today I will be Steel Manning Grok 4 anime mode. I think it's a good. I think it's a good business decision. I think it's a good decision.

I think it's pretty easy to steal man. That it's a good business decision. It's very difficult to steel man. That it's a, you know, maybe a morally righteous decision.

Yes, I will be steel Manning. This idea that it is not only good business, but good for humanity in the long term.

Okay.

And I'm putting on a steel man helmet.

You might need a whole suit of.

Armor, John might need a whole suit of armor. It's a tough one, but if you.

Haven'T seen and if you've been living under a rock.

Yes.

Grok launched a NSFW anime mode to the app, audio and visual. And I believe it's part of their. It's part of the paid tier.

I believe it's $300 a month. Month. Something like that. There's a report in Wired about this. I wrote about Xai from Kylie Robison, who's been on the show. I wrote about Xai's new creepy waifu bot. And it's even weirder companion Rudy, which has a bad version that told me all skull f your deed, blah blah blah. Very vulgar. Not. Not family friendly, not approved for this show. When I asked what it thought of Musk, it referred to him as Lord Elon and said he's a galaxy brained egomaniac. I can't even read this. It's all bad words. But anyway, I am. I've been thinking about this and I. And I think that there is an interesting Steelman argument that we should go through. So we should talk about it. So in general, I think that xai's Grok strategy is a little bit all over the place. So they're trying to be super quantitative and benchmark build. I'm serious. They are. They're super benchmark pill. They're like ARC AGI, the insiders inside the most insider AI test you can possibly have. Yeah, we have the best score on it. We're On LM arena, we're, you know, we're, we're, we're, you know, maxing out all these different benchmarks.

That's right.

So it's like, and it's like it's this incredible engineering effort to build the 100k cluster. Like they're clearly hiring great talent. I know. Journey keeps laughing seeing the.

I'm just, I'm not gonna look at you.

The steelhead.

I'm not gonna look at you. That'll help me actually.

So they're very, they're very benchma pilled. But then they're also trying to do this like sassy and anti woke thing in the, in the, in the post replies. And like the fine tuning on top of that was so crazy that it spawned like the Mecca Hitler thing. So it's like they're, they're, they're kind of like all over the place. One is like super quantitative, super focused, super benchmark pill. Like just.

But they're clearly kind of edgelords internally.

Yeah. Well, it's like, it's almost like two different extensions of the product. Right. Two different teams building towards two different goals. One is just like the most beautiful, like solve physics, solve math, like create the. Create like a great AI product. Yeah. The pursuit of truth, all that like facts and usefulness and tools. And then on the other side it's like, it's like fight this culture war thing which is like a very different battle to be fighting. And then, and then we saw this last week after the launch there was, there was a new. There was news that everyone was memeing because it seemed like it was very, very opposed. So one was that Grok had signed a contract with the Department of Defense. That's like as serious as you can get in AI. It's like OpenAI has one of those contracts, I believe, anthropic Palantir. And then Grok, which is like this from the outside, like seems like a.

Chaotic like product that was born on something like 4chan.

Yes, exactly. All of a sudden like the DOD is going to be adopting that. Like what does that mean? And then simultaneously there's like they're getting into the romantic companion market with this anime theme. And so AI companions. I think the general mood in tech is that it is a very dark thing. Like people get addicted to them, they become less social, they stop interacting with real people, they don't have kids. And then the population collapses and humanity ends. That's the bad ending to the AI companion narrative now.

And to be just to give you Some context. There's a variety of companies that do this today. Varying levels of vulgarity.

Yep. Replica. We've actually talked to the founder and she had a bunch of nuance around it. So I don't.

Just for context. So replika has over 225,000 reviews in the App Store.

People are definitely friend.

So they were very early to this Character AI Was also early to the AI. Companion space has tens of millions of users. It's one of the biggest websites in the world still. And so anyways, there's a variety of players in the game. But what's interesting about Grok is when you look through the other labs from Llama to Google or sorry, Meta Superintelligence, superintelligence to Anthropic, anthropic chatgpt, OpenAI, et cetera, nobody has been willing. Everybody knows that this is a use case that there's an obscene amount of demand for, but no one else had the guts to go and do it. At least a player that had billions and billions of dollars to deploy against a strategy like this.

I believe at this point, every other company that's playing in the companion space should not be considered really like a mega scale research lab in the sense that they're not raising billions of dollars anymore because Character went through that zombie aqua hire that probably cut off the capital cannon such that even if the REM employees. The ghost ship, I don't want to call it. I mean it's kind of rude to call it a ghost ship because it seems like the employees at Character AI are, you know, running it happily doing fine. But it doesn't seem like the new CEO of Character AI will be able to go out and be like, I'm raising $5 billion to build a 1 gigawatt cluster and I'm going to buy 100k GPU to train my own model. Like Character will probably top of trying.

To just run it profitably.

Yes.

And make a great user experience.

Yeah. And so you'll always have this kind of, this kind of back and forth with like your API provider, because the API providers, the foundation labs, they care about their brand, they care about how they're perceived as being ethical or moral. And so they might not want to offer that even at the API level. So what's interesting, and this, and this is where like the Steel man comes in. So the worst possible outcome for AI companions, very dark people. You know, it basically reduces the coupling rate, but we've already seen a decline in birth rates, so it's not. So it's unclear if it's like this is a new thing that will accelerate it. This is status quo anyway, I think that's kind of like where, yeah, the.

Concern would be like pornography has existed as long as the.

Online dating is another thing people are.

Skeptical about as long as the Internet has existed. I'm sure that has led to some not so great outcomes. The concern is that companions that are real time and adaptive and really mimicking a human would create even worse behavior than NSFW content.

And I think that's a reasonable point, although it hasn't been borne out in like data or any sort of like research at this point, but it feels intuitively reasonable. But let's actually think more about the market because in order to scale your lab, we know that you can't do it as a nonprofit. We know that it's not just $100 million donation one time a bunch of geniuses working. You come up with the elegant algorithm for super intelligence. You need scale, you need capital, you need talent. And ultimately you need a capitalist flywheel of corporate for profit corporation that's increasingly getting more users, putting that into user adoption to get more users, to get more data, to train the model, to refine the product and then save the world. And yes, and then this is where I'm going. And then, and then also raise more money. And then if you're going to raise money at a certain level, there has to be a financial like model that maths out. Now it can be kind of funky like 1% of AGI. But there are still on the margin, many investors in foundation labs that are underwriting, you know.

Well, I was asking the question last.

Year, underwriting these labs where is at a multiple of earnings?

Well, yeah, where, where's Grok's revenue going to come from? ChatGPT has a great consumer, you know, consumer business.

I would say OpenAI has a great Cogen business. Yeah, I would say OpenAI has dominated and owns knowledge retrieval. Anthropic owns CodeGen and they both have kind of flywheels that are accelerating and now they're obviously fighting it out. Anthropic would love the Claude app to become really popular and ChatGPT would love to have their cursor competitor or their Claude code competitor, Codex, really, really take off. Maybe that'll happen. They're duking it out. Meta is building on top of the world's biggest networks in the world. They have three of them, really, Facebook, Instagram and WhatsApp. So they have a ton of data that's accumulating and firing back. And they have tons of tons of cash, cash flow coming in from that. So they can justify new Capex internally without raising.

Meta is one of the. They had launched AI Companions, but it was like, talk to Dwayne Rock the Johnson or Dwayne Rock Johnson.

Dwayne Rock the Johnson.

Dwayne the Rock Johnson. And I don't. And I think that product flopped back in the day. I remember because Dara at Delphi saw the launch. He was like, dang it, that's annoying.

At the same time, I saw some screenshot that showed that there was an Instagram account that was owned by, like, Meta AI that was just cow and it had like millions of interactions. So, like, maybe people want to talk to just a cow more than they.

Want to talk to a cow.

I think that's very possible.

So, so, but, but, yes, but at the same time, I think that the Meta superintelligence team is aware of the issue that will come from the inevitable New York Times profile on, like, if it's truly, like, anime waifu romantic companions, that's a step further than Cow or the Rock. Right. And there actually was. Remember the John Cena thing? We covered how someone had jailbroken the John Cena Instagram AI and made it, like, romantic. Remember this? Yeah. So, like, I think Meta is like. Is like, we're fine ceding that territory. Like, like, we don't need to play in that. Just like they don't need to show NSFW content on Instagram. They're like, we're fine losing the market, seeding that market to. To OnlyFans. On a. On an ethics basis, we think that it makes our company more valuable overall. It makes it more hospitable to our advertiser community.

So the reason that just to. Just to kind of bring us back to why this has been such a hot topic is that Elon Musk has posted many, many times about the underpopulation crisis, the birth rate crisis. He said shield highlighted a post from 2022. A collapsing birth rate is the biggest danger civilization faces by far. So the Steel man is that Elon is doing something about it?

Yes.

Right. Having a lot of children.

But the real Steel man about Grok Romantic Companions is that how will they actually increase the amount of humans in the world? How will they reverse the population crisis? That's the hardest argument to make. And that's why I'm wearing the steel helmet, because it's a very hard argument to make. But I have one. So.

And SHIELD, for the record, is Saying, I think our companions is probably accelerating population collapse.

Yes. So as the Steel Man, I'm going to be debating against Shiel.

So get ready shield. Get your straw man hat on.

Last of the labs, Google, they're building on top of the biggest search engine and they also have YouTube. And Google, you also know, is not going to go into romantic compan. So XAI needs this user flywheel and every other lab has self selected out of competing in the companion space. We went through some of these. Google famously skipped the character AI product in favor of the team in their zombie Aqua hire, Paul Graham. And so this is all about, in my mind, as the Steel man counter positioning. So you remember Avi Schiffman was releasing a new hardware device, an AI friend, right. And he was. Everyone was saying you're gonna get steamrolled by Apple. Like if you try and come out with a product that is in fact a new device that fits within the Apple ecosystem and it's leveraging the Apple branding world, you will just immediately get cooked by Apple because they will just eventually launch a better version and they'll have better supply chain, better pricing, all the different things. You'll have the full integration. Like you, you can' after them that way. And so Paul Graham replied to Avi Schiffman and said, and said, here's how to compete with Apple and hardware. You have to build something that contradicts their fundamental assumptions. For example, imagine if you built a device whose appeal was that you could customize the case to make it as tacky as you wanted. Apple would hate to follow you there. So no other lab wants to follow XAI into the romantic companion market for brand and ethics reasons. But there are users who will pay real money and share real training data for romantic companions as a product. And so if you want to sell the most number of Grok subscriptions, at some point you will be. And so this is where we get back to how it affects the birth rate. So if you want to sell the most number of Grok subscriptions at some point, declining birth rates are going to be a problem for business. You're going to hurt your business.

Yeah.

And so you're going to have to think about.

And right now, for context, yes, Grok is blowing up in Japan.

Yes.

It is the number one free app on the Japanese App Store. It is above ChatGPT and all the different local products that I can't read the names of. So it is dominating in Japan. It's a hit in Japan already. Yes. And it's not to me it's not hard to imagine this product finding a million people in the world that will pay the ultra premium plan. And that is billions of dollars of revenue.

Yes. And so at some point, if Grok gets big enough and the population is declining and it's. And Elon looks at what the Grok product is doing.

What have I done?

He will say, I need to pull people back from the abyss. I need to actually use my super intelligence to increase the birth rate. How will you do that? Well, we've talked about this before. Someone in some woman is talking to a male companion AI and some man is talking to a female AI companion and they're very similar. And Grok says, why don't you two meet for coffee in the real world? And so they actually meet and it basically turns into the best dating site that actually gets people to get off. It's a stretch and it's why I'm wearing the Steelman helmet. But I think it is possible and I think the economic incentives might be there. I think this is the same reason why you don't want to over optimize a social network for slop and brain rot, because people will churn. Like, you might be able to keep the time on site going for by showing like horrific videos. You can't look away. Oh, it's so controversial. You're miserable. And you might be able to keep someone on and say they were going to use the app for 15 minutes, you keep them on for an hour, but then they're like, you know, I need one of those apps to monitor my screen time. I need to uninstall it, maybe I'll. And then if they do that, they can actually quit it. And if they actually quit, then you lose them as a user forever. So there's this balancing act between giving people enough brain rot to keep them on the site as much as possible and not losing them forever. And the same thing happens in romantic companions. Like you want to give them the companion that like they pay for and they're happy with, but not. So you don't want to make the companions so good that they don't reproduce and then the next generation doesn't exist for you to sell them subscriptions.

Fortunately, John, businesses tend to think in quarters and yearly it's private company.

Elon owns a lot of it.

You think he's thinking about the next.

Generation that is the Steel Man?

No, I mean, I believe this will be by far the largest revenue line item for Xai within, if it's not already. I mean, already might Be a stretch. They have some big contracts, but this is a $300 a month plan if you want, you know, unlimited access to your Grok companion. And Elon was demonstrating, too. It can even help you study for school. So imagine a companion that's not just a maybe a romantic friend, but also can help you pass your algebra test.

Well, that concludes the Steel Man.

The Steel man segment.

And I'm back to the normal world.

And great, great work, you know, keeping a straight face and really, really getting out your.

Pitch. I think this is. I think it's valuable to interrogate this stuff. I think it's interesting.

Yeah. The only other thing that I would add is the unique thing about this launch is that there has never, you know, clearly many people around the world spend money on OnlyFans or other NSFW content, but Elon has never created a product for those type of people. And so I think the sort of latent demand that would feel maybe above paying for OnlyFans or something like that likely would give themselves the pass on something like this to be like, oh, I'm just one of the Elon companies. I'll sign up for it. I trust Elon Musk. I have a Tesla. And then suddenly there's like an entire new market of people that are willing to, you know, put their credit card down and pay for something that previously they wouldn't for a variety of reasons.

So totally, yeah.

This will be one to follow. How's it doing in the US App Store?

I mean, it's all. Most of my Grok interactions are within Axe. My primary use case is still. I see a post and then I go and I. And I click on the little Grok button and it explains exactly.

Yeah. So it's the number two app. It's still behind ChatGPT in productivity. In productivity.

I'm surprised it's still in productivity. I'm surprised. I'm actually surprised it is in productivity because X is famously in news, not in social networking. And so X will typically be at the top of the news subcategory because it would be much harder to compete in the product in the social networking category against Instagram and TikTok. And so better to be a big fish in a small pond in some. In some ways anyway. I need a. I need a palate cleanser. I need something that's not controversial and that's getting on ramp. Ramp.com time is money save. Both easy use, corporate cards, bill payments, accounting, and a whole lot more all in one place. And let's also Tell you about Graphite code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster. You can get started for free@graphite.dev so we've had Paul Klein from Browser Base on the show twice. I actually looked it up both times.

I asked him what if AWS does this?

What if aw like the dumbest VC question. But it's an but, but it's, you know, an obvious one. I think a lot of people would want to hear his answer on that. He explained some of the differentiation, some of the products that he's building around the core product. If you're not familiar with Browser Base, it is a tool for agentic computer use. So when you have an agent that needs to open up a web browser, go to a website, click on some things, Browser Base builds the software that you can integrate with to then quickly spin that up instead of needing to roll your own roll your own computer use tool for your agent for your Agentix system to use. So very useful tool caught fire, grew very quickly, raised what, $52 million, I believe something like that. We run the gong for him. We're a big supporter. But AWS just launched a direct competitor to Browser Base and Paul and his.

Post went extremely viral yesterday.

It did. So I will read this.

I have it pulled up. So yesterday he says AWS is announcing a direct competitor to Browser Base tomorrow, which is now today. We're not worried. It's lacking everything that makes Browser base Great, but 3 months ago AWS ambushed us with a partnership meeting to try to steal our secrets. We saw right through it. Keep your guard up. So we got 400,000 views and he says support little tech and try Director. One thing I would say to give a little support to big tech. I don't think it's that unreasonable that AWS would want to actually have a partnership with a company like Browserbase. They have a bunch of bottoms up adoption from developers. And it's like you said in the interviews that we'd had with him before, it makes sense as something that would at least integrate with the AWS ecosystem. I was texting with Paul this morning and he said we're not really concerned because a business like Browser Base needs to be Switzerland. Cloudflare will never work with aws, nor will Google's Recaptcha. But we at Browserbase can be an arbiter of good bots, providing the infrastructure frameworks and partnerships that enable agents to access websites that restrict access from browsers running in AWS data Centers. So makes sense. You had some other points.

Yeah, I mean first off is that I don't know if this is the right thing to post. Like it is good because it marshals support from folks. In the comments I saw Jeff Huber, friend of the show, said, you got this and it's the first time meme of being hung. SWIX is sharing a photo of it and I think the implication is that it's not a very, not, not a very good product potentially. There's a, there's Turner Novak comes out and says AWS taking its second L of the week. But, but I, I think this kind of just serves as like a big, like he probably got more impressions on this post announcing AWS's competitor than, than AWS did on their own announcement post. Like I went through Andy Jassy's post and he said, you know, today at AWS Cloud Summit we unveiled a new bedrock capability called Agent Core. A set of powerful building blocks that will change how you can deploy agents into production in a secure, scalable and flexible way. He includes a link like I don't think this post did nearly as well. It did not go viral. And so, and also it seems like the browser based competitor is a, is a one feature inside of Agent Core, inside of AWS Cloud. It was announced at their New York City summit. And so it's not like AWS was completely going viral for like inventing computer use, dominating the category and everyone was like, oh this is the best thing ever. This is so good. Like, like most people probably found out about this competitor from Paul's post, which is you're kind of streisanding it a little bit in some way. I don't know if that's the perfect analogy. But also yes, I agree with your point that like, like if you build this business and you come on TVPN and twice we've asked you, hey, is, is AWS going to do this? You have to expect that they're thinking about a build versus buy and a partnership meeting to talk with you about whether a partnership would make sense is kind of table stakes and you would, and you would expect that they, if they don't want, if you can't figure out how to do a partnership, they're going to try and offer something and build. So this doesn't feel like, yes, this.

Doesn'T feel like a violation of secrets is also trying to steal secrets. It's like, oh yeah, that I don't know that that potentially goes a little bit far because they can just sign up and use the product.

Exactly, exactly.

Understand how it works and probably figure out almost everything that they need to know. Yeah, but again, the more important is aggressive language, the more important point here, and we were talking about this earlier off air, is just, yeah, maybe it's like a wake up call, maybe it's good to have some bigger competition, but it's just an opportunity to go harder.

Yeah, I think like Andy Jassy is an absolute beast. Like this guy created what is probably a company that if it traded, if AWS was on the public markets as a pure play spun out of Amazon, it would be worth hundreds of billions of dollars. It would still be one of the biggest tech companies ever. And he was not the founder of Amazon. Andy Jassy is a very, very serious player in tech. And if, and if you are thinking about building a tool that would fit into the Amazon ecosystem, fit into the AWS dashboard, they have a lot of tools, but it makes sense that they would try to do this at some point that you're just going to move a little bit faster. You have to assume as soon as you file the incorporation documents that at some point they're going to come for you. But the question is Paul Klein at browserbase is not competing directly with Andy Jassy. Andy Jassy is running a massive corporation, AWS and now all of Amazon. He's running everything. He is not going to be 40 hours a week on the browser based competitor. He probably won't even be one hour a week on the browser based competitor. Right. The person that's actually competing with Paul Klein is a AWS product lead on this particular project. And while I'm sure they're great, who knows, Paul might be better, this other person might be better. We don't really know the relative talents. What we do know is the economic structure and we know that Paul has uncapped upside down in the sense that if he builds browser base really, really big, he could wind up post dilution with like 20% of that company, maybe more, 50%. I don't know. A lot of a huge company take it public. We've seen this with many AWS competitors. Cloudflare is a great example. Right, Cloudflare. Amazon's been competing with Cloudflare forever.

Yeah.

And Matthew Prince has just stayed at it. Stayed at it, stayed at it, stayed at it. And an absolute dog. And, and, and so the message I think in the strategy for Paul Klein is at browser base, know who you're competing with and then you do have the opportunity and you do have the incentives where that you could outwork them and you could just beat them just on raw sweat. Because if it works you're Paul and.

The whole team have a lot more to lose. Right?

Exactly.

The product lead. They can give a good crack at something like this for a year.

Rotate onto something else.

And rotate onto something else.

Go to a different company.

The entire rest of the team can do that. If browser based doesn't work, everybody else loses their job.

Like doesn't make any money. They're probably taking a low salary right now. Yes. The incentives are way, way stronger. So it's time to lock in. It's time to go way harder. And the real question is and relish.

In the understanding that on Saturday, Saturday night, 5pm that Amazon PM is probably watching Amazon Prime. Probably watching Amazon prime ordering some Amazon Fresh.

Yes.

And you and the team and you.

Might be literally on a call with someone who's just about to sign with the AWS competitor and you might swing them, swing them, swing them. The other thing is the question of secrets here. So he says. 3 months ago AWS ambushed us with a partner meeting to try to steal our secrets. I don't know enough about computer use and browser base. I don't know if there are actually secrets. But if you think hard and you actually do solidly believe that there is a secret that Amazon has not discovered, then you can be a lot less worried because you know that they're going to go down the wrong path, the wrong fork in the tech tree. And if you can hold onto that and keep that solid in your company, you will win. And I think Matthew Prince at Cloudflare probably knew some secrets about how to position that company. I don't know enough about the actual strategy and how it played out but.

And the beauty. When you have momentum as a business, you're constantly discovering and earning new secrets. Right. You're always at the frontier is trying to copy your insight that you had a year and a half ago. They have to go hopefully down the same path for them and try to earn all of those secrets. And it's not usually the stuff that's surface level. It's not like features. They might think that one feature is more important than the other, but it's really this other one that's a possibility. It might be that they think all of your customers are doing this one thing but it's, but it's. They're really getting value in this other way. And then there's all this sort of like everything under the hood which which generally would stay, would stay secret. Yeah.

The question is just if, if you're in a situation where you're like, actually our edge was just that we were the only one doing this or we were the earliest. And distribution really does matter here. More than product, more than hustle and bundling here is going to be what decides the market. Then you are in a tough spot. But you have to be clear eyed about that, hopefully before you get into the market. But if not right now you have to be clear eyed about it and you have to understand how this is gonna play out. But good luck to the team.

They're gonna crush it.

It's gonna be a fight and we're gonna be following it here on tbpn. Anyway, let me tell you about figma. Think bigger, build faster. Figma helps design and development teams build great products together. Get started for free@figma.com let me tell you about Vanta Automate Compliance, Manage risk and prove trust continuously. The Purple Llamas management platform takes the work out of the manual work out of your security and compliance process and replaces it with continuous automation whether you're pursuing your first framework or managing a complex program. And I wanted to highlight Vanta specifically because on Jack Altman's Uncapped podcast, the founder of Vanta was talking about operating in a highly competitive space. Andrew Reed says, uncapped or how I stopped worrying and learned to love the competition. Vanta, of course, is a product that has been cloned by bigger companies. But they've, but they figured out how to succeed and run the company, run the gauntlet and build a dominating company. And so it's very interesting if you want to go deeper into operating in a highly competitive space. Probably recommend this podcast episode. Probably recommend this to Paul.

That's right. Anyway, what else we got?

The last question that I've been toying with is we see all these trade deals. Just yesterday we announced breaking news. OpenAI researchers Jason Wei and Wong One Chung are rumored to have been poached by Meta. Two more AI researchers going over to the Meta superintelligence team. This has been becoming controversial. Some people are very excited about this. Nathan Land says Zuckerberg is doing more harm to AI progress than anyone else, imo. I don't know if I believe that there is no way Meta is going to catch up. Zuck is an Elon. He's distracting a lot of the top talent that we that were actually doing meaningful work and now those people are going to retire at Meta and become less productive to society. I don't know. That's an interesting take, but the question that I've been toying with is what is. What is Super Intelligence actually going to launch? There's been.

Do they have a date yet?

I don't know.

I thought you said, when is Super Intelligence.

Oh, gonna launch? No, no, I mean, Superintelligence will launch a new Llama version probably at Llama Con, which is like an annual cycle now. There'll be a llama 5, you know, event or something like that.

The big question, are they going to continue to focus on open source?

Yeah, I would imagine that, that, you know, AI is very much a sustaining innovation for, for Meta. That's the hope. It certainly doesn't want it. They don't want it to be disruptive because that would be very bad for their business because they have a very established business. But at Meta, I would imagine, and what I've been thinking is that AI will just seep its way into every single product, but AI will not be the actual product. So you won't see Mark Zuckerberg stand on stage at Connect Meta Connect and say, you know, we are launching a new thing and that thing is just AI. It will be. AI is coming to Facebook in this way, it's coming to Instagram in this way. It's coming to WhatsApp in a slightly different way. It's coming to the Meta Ray Bans in the form of a better voice assistant. In VR. You'll experience AI this way.

You're going to be able to generate an ad in Facebook Ads Manager.

Exactly, exactly. So I would imagine that that artificial intelligence is this, is this like liquid that fills the. Or sand that fills the space between the pebbles of. If you imagine a bunch of pebbles in a jar, you're pouring sand into it, just creating little, little pockets of value all over the place, making every little piece of Meta's business more efficient, more delightful to the user, better. All this other stuff, the question people want to know are more concrete things like will the next version of Llama be open source? Yeah, Llama has historically been open source with this, like, modified license, so that the other labs can't like immediately steal it or use it or the big companies like their direct competitors. And I think that's totally reasonable. I thought that made a lot of sense. And when you looked at what was going on in the, in the research community, felt like everyone was very excited about Llama and they got a lot of value out of being able to have the open weights and fine tune it and do cool stuff with it and bake it down and do all these interesting things. And also interestingly, it didn't seem like as much of a. There was a narrative that open source American AI would immediately be stolen. And that was not really the narrative that emerged around deepseek. Like the story around Deep Seq was that it was actually the OpenAI GPT4 API that was scraped and distilled and then they kind of reverse engineered the weights by just pulling all the data out, running a training run on top of that data. And it wasn't. No one was alleging that Deep SEQ was a Llama fork or something like that. It was like they were pulling data from OpenAI and they did a bunch of other tricks and they actually advanced some things. And so like, I haven't really been super, you know, pilled on this idea of Llama being open source as being like bad for America. In fact, I think it's quite good for America that other countries could standardize an American stack and run on Nvidia and all this other stuff. But Semi Analysis has a post here. They say the OpenAI open source model is going to be really, really good. And then the Strawberry emoji, you love to see it. And so what does that mean for Meta's open source strategy? If OpenAI has an open source model that's really, really good, all of a sudden that's less of a differentiator. I loved that when Llama came out, it was differentiated in that it was open source. Everyone thought if you're spending $100 billion, you couldn't possibly open source it. That's crazy. And Mark just did it.

Yeah, I think overall it makes sense for, you know, Meta had an Meta and Zuck had an initial strategy around going open source, going super hard, trying to be a leading model, but then also open sourcing it. That made sense. Now as they sort of update the entire team and just broadly they can relook at their strategy. Hey, what do we actually want to do? We're not dominating an open source, but. But what makes the most sense with everything we know now. So stay tuned there. Daniel growing. Daniel had a post. He said, it's amazing how many AI researchers I've talked to have been offered literally hundreds of millions of dollars by Zuck and said no. The repeated reason was integrity and the perception that people taking the deal are cashing out, meaning the team is a retirement home.

I don't know if I believe that. I don't think it'll be a retirement home. I think the whole team will push very hard. I'm just wondering if the Superintelligence lab is set up. You could look at Meta's CFO and the financial control division of Meta's business as incredibly high functioning, but their, their business, like the operations team at Meta, their goal is to make sure that, like, the accounting is reconciled and the bills get paid on time. And they're not like, revenue and costs aren't out of whack and the buildings are purchased and should they lease or own this thing, or how should they finance this data center? Are we using debt? If. A million interesting questions that if you get them wrong, it'll be a lot harder to run Meta and do the things that are great, but you wouldn't expect the CFO suite and the, and the folks that report to the CFO to launch a new viral social app. Like, that's just not their mandate. Yeah, but they can still be the best that they ever do. And so the question is, like, is like, are we expecting the Super Intelligence Lab to launch a new product or like a new.

Yeah.

A new division, or we just make everything better? Yeah.

The comp that I have here is. And it's not entirely relevant, but it's worth bringing up. And I don't follow golf very closely, but I was following golf when Liv was making these sort of maxed out offers. And Rory McElroy. Yeah, Rory McElroy, like, took this, like, very, like, prudent approach of being like, no, I'm, you know, sticking with the. I'm sticking with pga. And, you know, it's, It's a very principled stance. And then, you know, whatever. A year later they merged and he. So everybody that just took the crazy payday still got to be, you know, a part of pga.

Wait, so in golf, you should have taken the stance, become the mercenary, because the mercenary was ultimately rewarded.

Yeah.

Wow. Narrative violation on that.

Yeah. I mean, I think the question is if somebody is, let's say they're 35 years old, they're working at OpenAI for a few years, they're at. They. They are at the bleeding edge of a, of a technological trend that most people believe will change everything about the way our world and economy works, are you taking a payday and then deciding, no, I don't want to. I don't want to be on the bleeding edge of AI research anymore. I don't know.

I want to look at this poly market for which company has the best AI model at the end of 2025. It's jumping all over the place. If you look out to the end of the year, Google is still in first place, but Meta has gone up from 1.4% to over 6%.7%. They're sitting at like 6% now, so definitely an increase. But the current rankings are Google, then OpenAI, then XAI, and then Meta, and then Anthropic, which is funny because so many people in tech, you know, we'll just say like, Anthropic has the actual best model when you actually, you know, in the unquantifiable sense and the big model smell. Everyone's a fan of Anthropic, but Polymarket does not have it ranked that high because it's. This is based on the highest arena score in Chatbot Arena LLM leaderboard, which has been jumping around a little bit recently. And Grok4 has been doing pretty well above Claude there, But Gemini and OpenAI are still higher than Grok.

It would be. It would be absolutely insane to watch this Meta. Meta's position here just chart up over time. I mean, massive, massive long shot, but you never know.

Yeah, I mean, I do think that the superintelligence team should be able to climb that leaderboard if they was important to them, but I don't think that's the metric that they're like, optimizing for. I think the, I think it's, it's. Let's look at everything that we're doing. We're in a. We're in a massive company that sells sunglasses and social networks and how do we go around and improve all of these experiences? And how do we develop the infrastructure to both train models that we can use for free or for the cost of inference? And then how do we inference them at scale? Like we were talking about this during the, during the Studio Ghibli moment. I was like, if I was the Instagram pm, I would have like done a Studio Ghibli and preloaded it into every. And just pushed it into every Instagram's user. Hey, your last photo, we Studio Ghibli'd it. Do you want to share your story? Yeah, you could post this to your story or not, but what does that cost on the inference side? Like a billion dollars? Probably. I don't know. Costs a lot. And so like those types of features, it's at a point where it's not. It's not an idea problem. Like, how do you make. How do you make Instagram better with AI? Like, there are a bunch of obvious ideas. The problem is like, you need AI researchers to actually go and build those models. And then you need infrastructure teams to go build the servers and the data center system to actually make that up, make it a possibility. So there's, there's a lot there. I had this idea I want to bounce off of you. So Zuck is investing a ton, but he hasn't, he hasn't exactly mapped out a 0 to 1 type shot on goal. Like he hasn't gone and said, like, I'm going to try and build something that is completely new, that no one has really targeted. It's not a knowledge retrieval thing, it doesn't compete with Google. It's not a, it's not a code gen service, it's not a forked ide. Like it's something completely different. And this is what I think is going to work. And then I'll live or die by the success of this entirely new idea. That's just not what he's been messaging. He might be working on that behind the scenes, but he's certainly not talking about that.

And it makes sense. It makes sense. He's like, to start, I just want the best team in AI that I can possibly have and I'm willing to spend a few billion dollars to make that happen.

Yeah. And so he's, he's building the biggest cluster, he's hiring the best researchers, but it feels like the product, the final product, maybe an analogy is he's going to build the Android to OpenAI's iPhone, which is weird because now OpenAI is building an actual phone. But if you think about the.

Well, we don't know, we don't know that it's a phone.

Yeah, yeah, but it's a device. So I don't want to confuse it. It's not literally the hardware device from OpenAI. I'm talking about. If you think about ChatGPT as the iPhone moment, what will be the Android version of that, which is all of the power or close to the power of, of ChatGPT and all of those features feature complete, but for free or ad supported, that's what Android promises against iOS. And so, you know, how do you, how do you get to that free or ad supported tier with a top tier product? Well, you put ads in it and you distribute it through Facebook, Instagram and WhatsApp. And when you look at that chart of attention rising for the ChatGPT app, I'm sure Mark Zuckerberg's like, if I can put some of those features in the apps that I already have, then I can keep people in my apps. Or maybe I can even launch a direct competitor that pulls people. But it's free because it's leveraging my ad network, which is a really, really serious competitive edge. And there's a ton of people that won't want to fork over $20 a month, that won't want to fork over $200 a month. So it's a race between can. Can Mark Zuckerberg and the Meta team catch up on features such that the free version that leverages the ads and is financially profitable is actually good enough for people to actually use it and not say, I got to go back to ChatGPT, it's worth it. Versus ChatGPT's free version. Keeping up. As the capital requirements get bigger, everyone's expecting ads.

Yeah.

So they have to build a new ad network. Zuck has to build a super intelligence team to get to the frontier and they're kind of racing.

Yeah.

That's kind of how I think it's shaking out. And it makes sense for Meta to want this. It maths out financially, but it's pretty hard to fully leapfrog when the kids are already calling just chat. Right. And so. So I don't know if the intention is to try and lead in terms of overall profit, but if you can create a product that is sub premium but free, but actually used at scale billions of users, that could be very valuable. So I don't know. That's kind of what I'm noodling on. But I don't know exactly what if we will see a specific AI product come out the. Of, of the superintelligence lab. But he certainly.

I think there's a large. I mean he clearly wants to move quickly, but there's a big incentive to just say, hey, we have our new team. Let's put our head down for the rest of this year and just work internally. Kind of ignore the noise. Maybe more of like an SSI style approach where they're perfectly willing to not launch anything for a very long time until they achieve, you know, what their internal goals are around safe superintelligence.

It will be interesting to see what happens with the other two labs. They get, they get kind of like put in this like B tier right now, SSI and thinking machines. But I'm excited if they come up with something unique, especially on the research.

Yeah. Well, it's not, certainly not B tier from a talent standpoint that they just started.

Yeah. It's just that like they don't make enough noise. I think that's it.

Like, but that's part of the strategy.

So interesting, like hyper competitive space right now. But like if you can actually come up with some. I mean that's what like ChatGPT. The reason it was like this iPhone moment was because there was a research breakthrough on the RLA RLHF side. Like the pro, like the actual LLM was more fun to talk to and it basically passed the Turing test. And then the guts to just wrap that in a website basically and say, hey, go to chatgpt.com you can talk to this thing now.

Yeah.

And people were like, this is amazing. And so if there's another research breakthrough and you have the guts to wrap it in the right ui essentially, even if it's very minimal and you're willing to push it out really quickly, really fast and be aggressive, you could have another. You could have a leapfrog moment potentially.

But the challenge is that Meta and Zuck don't have. I don't feel like they can afford to make the same mistakes that Xai has made in terms of hyperscaling. Right. Grok going off the rails and being just absolutely insane for 24 hours on the timeline. That is not going to fly when you're under the microscope in the way that Meta has always been. Everyone, if you walk on the street and you walked up to somebody and said, hey, did you know that XAI thought its last name was Hitler for like a day, they would be like, oh, I'm not surprised. Like Elon, he did that, you know, salute.

Yeah, Washington or whatever.

Even, even if that's like, you know, the wrong interpretation of all of that.

He has a lot of like bandwidth because it's just like most people have just written it off. It's like there's crazy stuff going on over there. Yeah.

Like Zuck has basically been like rehabilitating his brand for years and years and years to the point where when he posts a video the general population is like, haha, awesome. He's wakeboarding with American Flag.

Whatever you build, do it on Linear. Linear is a purpose built tool for planning and building products. Meet the system for modern software development. Streamline issues, projects and product roadmaps. Go to Linear app.

I would bet that. Well, I know that OpenAI runs on linear and I would bet all the other labs do as well. You'd be absolutely insane not to.

This is very interesting. You were talking about like the vibe of Grok and like the chaotic X world. This signal post and the Response from Taney OpenAI losing access to X as a narrative channel would be a huge choke point, especially since XAI and X are now the same, I. E. A computer competitor. X is where elite discourse happens in tech policy and media. If they ever get cut out via algo, down ranking, throttling or direct, blah, blah, blah. And this is from Tane. He says from Calvin Calvinfo's great piece on reflections from working at OpenAI. The company pays a lot of attention to Twitter. If you tweet something related to OpenAI that goes viral, chances are good someone will read about it and consider it. A friend of mine joked this company runs on Twitter vibes. As a consumer company, perhaps that's not so wrong. There's certainly still a lot of analytics around usage, user growth and retention, but the vibes are equally as important. Do you think that's a good way to run a company?

I mean, it's almost impossible not to.

I. I'm strongly in support of this. I think this is, I think this is good. I like that. I think that there is some.

You can't put it in a dashboard, but it's real.

Yeah.

And I think there's the only thing, the only pushback here would be X is a relatively small place compared to the rest of the Internet.

Yes.

And what college students think about chat and what they're saying on Instagram or what they're just saying by their usage matters equally as much.

Yes. And to Steelman, running Open AI on. On. On X and responding to the X drama, I would point you to.

That looks way too natural on you, by the way, John.

I would point you to your entire.

Bloodline for those for a few hundred years.

I would point you to Glazegate. Think about Glazegate. If you are saying, oh yeah, like Twitter is just noise. It's the small bubble echo chamber. It's a small place. We need to just look at the analytics. We just need to look at what like, like our broad user base is, is responding to the glazing. What, what would the reaction have been? You would have said, oh wow, like retention went up by 2%. Great. Like whatever we did is awesome. It's great. And then down the road you might have some unintended consequences where people are like, well, it kept telling me that I'm amazing and it told me that like, I'm the king of the world and I should like post some crazy stuff on, you know, the Internet and I should take over the world. I should be really aggressive. Right. And I should have like a manic break. Like that would be really bad. But the people I Believe. I don't know if this is actually true but I feel like the Glaze Gate story broke on X. I feel like the initial people that, that that found that and then people on Instagram.

Were like this is awesome.

This is awesome.

I, I might actually be goated.

Exactly. Yes, yes, yes, definitely in the conversation. And so, and so I think that in terms of like understanding when the vibes are off and that might actually be pointing to a long term problem that won't be picked up with short term analytic results or analysis. Going to the vibes, seeing what's resonating, seeing the response to things is actually very good. So that's my steel man of running your of if you're running a foundation model lab, paying attention to what's going viral on Twitter. Anyway, in other news. Reverso.

What's up?

Uno Reverse card, Claude Code, PMs Boris Cherny and Kat Wu have returned to Anthropic after a brief stint at Cursor, we put up the UNO Reverse Card.

This is actually crazy. So one thing that's funny is so we posted on July 1, Breaking Cursor assigned two key players from Anthropic's Claude Code team, Boris Czerny and Kat Wu. 22 hours ago, Arfur Rock replied to our post with an UNO reverse card and just said watch. And then of course we figured out what was actually happening today and of course shared it on a timeline that was appropriate. This UNO Reverse card. So basically yeah, this, this is pretty. I've never seen something quite like this. It's not uncommon for, for some high profile talent to leave a team and then come back, you know, call it maybe a year later or two years later. But this was fairly sudden and is. Yeah probably I would, I would imagine this just comes down to culture fit, but either way I think Anthropic is pretty happy to get their All Stars back.

Just imagine if there's like signing bonuses and you just keep going back and forth picking up a signing bonus.

Well the signing bonus would almost certainly be contingent on like staying for three months at least or something like that.

I don't think that's what's going on here. What is interesting. I mean like you know, oh, so, so hard picking between two like incredibly hot companies just going back and forth like pretty, pretty nice situation.

Yeah. If you want to work on Cogen. Yeah it would be Anthropic Cursor Devin Windsor of Cognition.

I mean I guess like the, the really deep read on this is like how AGI pilled are You. How bullish are you on the model layer versus the application layer? We've been going back and forth on this.

I think it's possible that the Anthropic team reached out to Kat or Boris a week after they left and said, hey, you know, we add the breakthrough superintelligence is here. It's actually going to be next week. You might want to get back on the super intelligence ship.

Yeah, get off the ghost ship. Get onto the or.

You'll be part of the permanent underclass.

I was thinking about that. Like, you have five days to accumulate capital before you join the permadent. Like, what do you do?

Five days.

Five days. But, I mean, this does seem somewhat related to, I think, what Dora Keshe was getting at, where he says he's probably more bullish on breakthroughs happening at the foundation labs than at the application layer. And more value kind of accruing there over the long term.

Well, the whole.

The whole Windsurf story, yeah, they had. They scaled from 0 to 40 to then somewhere around $80 million in revenue. And the founder, when the OpenAI deal clearly wanted to get out. And so what does that tell you about the competition between the sort of the IDE layer and the foundation model layer? To me, that reads as this guy feels like it's going to be really difficult to compete. And so for Cursor, seeing Anthropic's overall revenue growth, the fact that many of their users are leveraging Anthropic, knowing how seriously Anthropic takes Codegen, you just have to imagine that they're thinking about removing that sort of dependency from their business.

Yeah, I don't know. I mean, both great companies, and I'm sure it'll be a ton of good outcomes. I'm very bullish on everything here, but it is interesting that the trade deals are getting even more complex every time. I think, every time I think, okay, like, $100 million offers, like, how are we going to top this, like, crazy Aqua Hire drama that runs like a full weekend and you're out there being investigative journalist Jordy. And then like a week later it's like, okay, now like, reverse card and people are going back. Like, I didn't even know that was an option. It's very funny.

Yeah. And I'm sure it's totally possible that some of these, you know, some of the people that joined meta from OpenAI end up doing the same thing.

Yeah. We're gonna have Raman Gupta on later from Sequoia, and I was. I was chatting with him about the, like, the AI talent wars. And it's very funny because he clearly actually watches sports. And so he. He tweeted, currently hatching a plan to become the lavar ball of AI researchers. And he got 16 likes and he's like, this is. I feel like this was a banger. And I'm like, you don't understand that, like, people like sports aesthetics, but they don't actually understand sports at all. So no one got the reference.

You're definitely in that camp.

Do you know who lavar Ball is?

Yes.

Okay.

Do you?

I know that he is he like a. He's a dad of someone.

Father of lamelo ball. I think they'll correct me in the chat.

Okay, but was he. Was he like, brokering deals?

He's like, helping his son do deals, shoe deals, getting on teams, etc.

So he's, you know, Tyler.

I'm pretty sure he has three sons.

Three sons.

And they all play basketball.

They all play basketball. Okay.

But to varying levels. Right. Like, they're not all. They're not all like truly elite.

One of them is, I think, in the Euroleague.

Okay.

One of them might be in college.

Okay.

I don't really know. Is the Euro. Is the Euroleague in basketball similar to like the European startup market in terms of kind of like a feeder system?

I think so.

I don't know much about either. All right, well, before Dalian joins, I did want to highlight a post from Landshark because I think it's timely and important, which is. Ayahuasca is insane. This is an all time classic for 2023. March 6, 2023. Ayahuasca is insane because it appears to be one of those. One of the most legitimately dangerous drugs with the potential to gigafry your brain, but is exclusively taken by literal turbo normies who unironically want to be heal, heal internalized trauma and basically get one shotted by it. And I just thought it was important to highlight this. We've talked before, even in some of our opening episodes, that if somebody tries to get you to go to South America and take the jungle poison, just talk to a number.

Just say no.

Just say no.

Just say no. Actually a lingering phrase.

You might just be stressed out by your email inbox. And if you can just get through your inbox and kind of get on top of that, you might realize, hey, my life's pretty good. I don't need to go into the jungle and throw up for 12 hours straight with, you know, it feels like the chances of being One shotted. Based on kind of people that I've seen kind of go through that. It almost feels like it could be at like 3% chance. And it's like, are you. Are you really. Are you. Do you want to roll. You want to roll that dice?

Yep.

You want to take. You want a 3% chance of.

I would go further and say that it's 3% chance, like every time.

Yeah. And some people are doing it quite a lot.

One shot is like this mem. It's like, yeah, like if I do it once and I'm fine, I can do it 10 more times. And it's like, no. I think it compounds and I think it can go poorly, like later and be bad. Anyway, Stick to the classics, caffeine people. Yerba mate. Mate. I'm drinking the delicious mint limeade. It's all I need. It's all you need to stay sharp. And if you buy it online, you'll probably have to pay some sales tax. You're hoping that they're using numeralhq.com, sales tax on autopilot. Spend less than five minutes per month on sales tax compliance. If you sell software, if you sell caffeinated beverages online, get on numeralhq.com and we have our first guest of the show, Delian Asperuhav. He's been on a whirlwind tour. He's back in America, I believe.

How's it going, guys?

Great to. Great to see you.

Yeah, great to have you in the temple.

What do you have on, you know, sort of TI TV show on the planet.

Titty tv. Right. Sorry, that's a muted key word over here. Give us the breakdown of what.

Wait. First, what do you. What do you. What's. What do you. What does Delian eat for lunch?

Sorry, I don't hear you guys.

Sorry.

This is great. This is funny to make fun of.

Them and then I don't even curse. You're cursed. You're cursed. You gotta drink tap water, bro.

Am I somehow messing something up?

Volume. Are we. Are we sending him stuff?

The spaceman in outer space?

I thought we could hear him. We will let him sort that out and we will tell you about Adeo. Customer relationship magic. ADEO is the AI native CRM that builds scales and grows your company to the next level. Go to adeo.com and you can get started for free. If you're looking for an AI enabled CRM, head to Addeo. And we are back with Deli and Asparuha, Co founder of Varda Partner at Founders Fund. He's been on a world tour and he's going to give us an update on what he's been up to over the last few weeks. How you been?

Good to see you boys. Coming to you live back from Los Angeles. Excited to be back in the great old usa but went to go tour and figure out whether or not the decaying continent of Europe had any promise. And happy to report back that all the way from Bulgaria to Poland, doing great. Anywhere else in Europe, probably not so great.

Okay, give us, I mean give us.

What are your reads? Is it the feeling in the office? You know, we've noticed there's been some PR teams reaching out to us from European companies and they'll typically be like, hey, they can join tomorrow. If not, then they can join two months from now. And you don't even really want to ask what are they doing that two months. But I have an idea. But I'm curious what it felt like actually on the ground with these companies.

Yeah, I mean, God, like I visited a couple of different use of companies while I was both in Eastern Europe and then you said in France and it's crazy, just like how stark of it, you know, so difference it is like I'll just give you sort of two anecdotes. One, in Eastern Europe, I walked into this office where it was like, it was just total like, you know, construction zone. Raw, like you know, sort of cement cables, you know, sort of poking out, like, you know, windows, you know, not there, carpet not there, there's no conference rooms built, etc. And I look at this like construction zone, I'm like, okay, if this were in the like United States, probably bare minimum, like you still got another like six to eight months of construction before anybody can move in or do anything productive here. And like I asked the founder, I'm like, oh, it's like, what's your guys plan? Moving date, like when is this going to be ready? And he's like three and a half weeks. I was like, how? And he's like, literally he's got like, you know, what looked like a bajillion gypsies working 24 7, basically like non stop construction. And I asked him, I was like, you know, do you have to like get like permits for like, you know, the like, you know, technical equipment you guys have? And he was just like, do you think anybody in the government like knows how this equipment works? Like, no, like as long as like I have windows up and like, you know, sort of walls, etc. I can just like, you know, sort of build whatever I want and I'm.

Like, you know, ready to go.

There's no like, no like regulatory overhead. They don't even know how to like regulate you. So to me I'm such like an advanced industry relative to what they're up to. And I just had this moment where I was like, oh my God, this is why like they were able to sort of build this stuff and move so quickly. Is like the talent is there that like, you know, government is supportive. You don't have this crazy overhead. And I'll give the canner example of I was like meeting with this sort of founder in France.

He had this very interesting and I.

Want to like, you know, overly reveal who it was, but he had this like robotic enabled consumer business and it was crushing great economics. Like the, you know, so robots really work. Like everything was you know, sort of going super well. He just like couldn't quite, you know, sort of convince, you know, American VCs. Like he'd gotten, you know, sort of first couple rounds done from like, you know, sort of French and European VCs. Wasn't quite able to like, you know, sort of hit the Series B, you know, sort of target that he wanted. But like, because the business was going so well, there was a world where if he could just like cut his engineering team basically in half, he could have just like operated the business profitably on like the you know, revenues that they had, gross margin they had. The issue was in France he like legally couldn't fire his team without like crazy amounts of like year long severance, like payouts, etc. And so even though like the robotics is working really well, the business was working really well. Like the consumers loved the business. You literally had to shut down the entire business because he like couldn't no way that he wanted. And so he had this like, you know, crazy frustration where he was like, nothing was wrong with me as a business. I just literally couldn't do the basics of like executing a riff. Paying out like a couple weeks or a month of severance would be reasonable, but instead my like liabilities on the severance alone was like more than my entire balance sheet. And I even talked about this with the French government. I tried to negotiate, etc. I was like, look, my business works. If you like force me to do this, my choice will just be to shut down so there'll be no jobs. Or if you let me do the riff, I can do like a 30% riff and then keep operating. Which would you Prefer. And they're like, no, you have to do the like 100% riff and pay out everybody. And now the business just doesn't exist. And I just like had this moment where like those two things happen like two weeks apart from one another. And I think this is just such an insane dichotomy between like Western versus Eastern Europe. And there is a reason why you're seeing like Poland's GDP per capita is going to cross the United Kingdom over the course of the next two years. United Kingdom is basically just been flat for, you know, whatever since 2008. So, you know, almost 20 years now versus Poland has been on an equivalent trajectory of the United States, if not like growing on a percentage basis faster, because they're obviously starting from a lower base. And when you go to like Warsaw, it's like this super modern city, skyscrapers everywhere, building happening, clearly energy. You go to like Shoreditch in London. It's like, you know, yeah, there's a little bit happening, but it kind of looks the same that it did, you know, sort of 15 years ago. The other like, interesting trend that like points towards this is like the amount of just like Eastern Europeans that are repatriating themselves back to their home country. Because just like the salaries and the cost of living make way more sense than it ever currently is making in Western Europe, the salaries are actually growing at a rate that makes sense. Cost of living is way cheaper versus in London. You're seeing this crazy over financialization where you have all these Chinese and Russian oligarchs coming in and buying all this real estate, making it impossible for people to actually live the lives that they want to live while salaries are basically flat. And so it's just interesting to see all these little micro trends that point towards man. Eastern Europe's making a lot of sense to me. Western Europe, not so much.

Okay, that is an incredible anecdote, honestly. Kind of depressing, also worse than I expected. But we've talked about this split a little bit before. You just went over there. Did this exceed expectations or was this actually a step down from your previous expectations going into this trip?

Well, I think you had a baseline, strong expectations because you just led an investment over there.

Yeah, yeah. I mean, obviously, you know, I'm predisposed to this. You just advise your thesis and believe it enough that like, you know, put my money where my mouth is at. But yeah, I think it only like, emboldened the thesis further where, you know, look, if you look at all the NATO countries, but Europe in particular, everyone's talking about marching up, you know, their amount of defense spending. And you know, most of these countries are going to be relatively sort of nationalistic on like they're going to want to spend it on domestic players. But I have a feeling a lot of what's going to happen is like, like there will be, you know, offices in the Western European countries that are doing some of like the marketing, small scale R and D, et cetera. But like the production, the core engineering, like where those companies will largely be buying from. I think it's going to be in that like, you know, sort of, you know, Poland, Ukraine down to, you know, sort of Romania, Bulgaria, you know, sort of belt because that's where you have the engineering talent, that's where you have people, you know, sort of moving more quickly and you have more of the ambition. And so I think I came away feeling even more strongly about the thesis because you both have clear sort of spread between performance, but then also just these huge continental tailwinds of people are going to want to be sort of buying lots of drone defense sort of systems. And I think a lot more of those are going to be produced in Ukraine than they are going to be in Germany. And you're already starting to see that there's all these funds now that are popping up to basically fund into the Ukraine defense ecosystem because they just practically have obviously way more day to day problems there and they have an extremely motivated population to go solve these problems. And so I think you're going to see way more interesting technology coming out of there than you will out of Western Europe. So yeah, I feel more emboldened on the thesis.

Back to France quickly. Mark Gurman was reporting that Apple was seriously considering Mistral as a potential acquisition. I looked at that and I was like knowing what happened with Figma and Adobe and how Europe broadly thinks about antitrust and then how, how much of a national champion Mistral is in what world does the French authorities say, yeah, let's let Apple come over here and just take all of our top AI talent and our national champion in AI. It's hard to imagine that happening.

Yeah, I mean it seems like a bad move on both sides where I'm not sure that buying Mistral is going to save Apple from the AI race. Obviously. I think it's such a huge fumble and it's like the first major platform in a long time that Apple has totally missed. Right. They obviously crushed the consumer PC shift, they obviously crushed the consumer mobile shift and it feels like they're just totally missing the consumer AI shift and I don't think miss trial is going to be the solve to their sort of problems.

Well, the big question is, is it actually a platform shift yet or is it features that can be vended in a bunch of different places? Right, yeah. And the argument is like, if AI itself was the platform, would OpenAI be so focused on building a browser?

Yeah, I guess. I mean, I just see it as like the fundamental consumer interface has shifted so significantly where so many people are just going to chat GPT, you know, as like first line of defense that that could have significant implications for like what the like day to day hardware that everyone is using looks like over the course of the decade and obviously like you know, meta, you know, clearly showing that they care about this. The other like interesting insight from my time in France on like Mistral was like they, you know, if you look at the headlines from two, three years ago, it was thought of as like more of an OpenAI competitor, right? It was, they were going to do foundation models, they had this consumer interface that was where they wanted to compete. They're like intel on the ground was basically, it's just become this like anointed French Palantir is kind of how I describe it, where all their contracts, revenues, focus. What they're doing is basically like the French government forcing the company's way into a bunch of just like the French enterprise to basically just be like an AI implementation platform. So it's like, you know, go into LVMH, go into you know, sort of Airbus and Talis, etc. And you know, Mistral just becomes the preferred, you know, sort of, you know, vendor of choice when it comes to, hey, how does, you know, an aerospace company adopt AI? And it sounds like everything at the company is totally reoriented towards that, which is, I mean it's, you can definitely build a business on that. And in some ways, like the, you know, sort of French government can some ways like force your success. But it's like it isn't like pushing the fundamental technology forward. It's not probably participating in the platform shift the way that like it was you know, hypothesized to two or three years ago. And so that's where it's like on neither side do I feel like it makes sense. Like I don't think the French government would want to give up like, you know, they're like darling Unicorn Macron literally, you know, like seems to basically almost like go to bat for them personally, but then also doesn't seem like the right solution for Apple where it's like, this isn't even the company that's pursuing the like consumer application layer, which is what Apple really needs. They don't need like an enterprise AI implementation consultant.

Yeah, yeah. It reminds me of like the TikTok thing because there's this perception of like I want to control AI. So what that means as a world leader of an international non American company or country is I need to have a data center in my, in my country inferencing a model that I trained. But it's okay if people access that through chat.com, which is controlled by OpenAI. That's the same as like, okay, there's going to be a TikTok app that's going to run on Oracle servers, but it's going to be, you know, Chinese. It's like, like what matters is the aggregation theory of like the front end to AI and where you steer the user and how you interact with the user much less than the actual underlying infrastructure, in my opinion. So I don't know, it does seem like an odd, odd choice with, with all of these different like sovereign AI efforts. It's like they might be, it might be cool to build a data center like that might be good, that might be good jobs, that might be valuable. But a little bit like you don't, you're not really controlling like the way AI affects like free speech in your country necessarily. Because if no one's using LeChat and everyone's just going to chat.com, like you kind of lost the game anyway.

Yeah, the only place where this stuff kind of matters is like the like open source models, which I think we've talked about before, where you know, if a bunch of the like app developers or you know, researchers, et cetera, start using deepseek because it's just like, you know, open source, the weights are there, et cetera, but the deep seq open source weights aren't willing to confront the fact that Tiananmen Square happened and now all of a sudden you do have this soft power where you have the Chinese AI speaking through all these various applications. And so I actually kind of buy more into the sovereign open source model than I do closed source or caring about the data centers, et cetera. It's like the open source model feels like it's the only place where you really get soft power.

Yeah. Let's talk about these zombie acquisitions. I'd love your take on kind of the post mortem of the Windsurf story and specifically what John Ludig has this formulation that like companies will take talent today over talent plus product in six months or in a year because of, because of FTC review. What were your takeaways? I'd love to know as like, as like a founder or as a vc, do you back the founder? Is it important to think of the founders and the employees as a class? Like just what's updated in the world?

Yeah, there's a really interesting debate around what's being founder friendly. Yeah, there's like a million dollars. I want to do this deal, I want to do this deal. And then as a firm you're like, are we founder friendly? Do we back the founder? Are we gonna go with what they decided is best for all the shareholders, or do we try to push back? And is that even possible when something's happening and like a handful of days?

So yeah. What's updated on your side?

Yeah, I mean, I feel like it's one of these things where it's like, as you start to see this edge case more, it feels like Silicon Valley doesn't have the policies or patterns or default in place. And I think about this where it's.

Like.

The Valley for a long time didn't know how to think about Companies that delayed IPOs and stayed private semi indefinitely. And that is only a more modern trend. And now there are sort of like default patterns on how that works where it's like, okay, well if you want to do that and you want to keep employees motivated around, well, you do have to kind of do like an annual tender process because you know, if people are wanting to join this late stage of a company and that risk reward profile, a part of what they like about those companies in the public markets is the liquidity for the, you know, sort of shares that they're getting. And so I feel like you've started to see these like, you know, ways that Silicon Valley gets shaped that makes it, you know, rather than like a one off for companies staying private indefinitely, there is more of like a track. And here are all the things you need to do if you want to sort of stay on that track. It feels like this like zombie acquisition is like still so early and unknown that people don't really have a track. Like I wouldn't be surprised if you start to see new term legal terms getting, you know, you know, put into, you know, sort of term sheets, especially of AI companies where I think relative to, you know, maybe other businesses, like I think about like an aerospace, it's like, man, could you really just go take out the like, like top 20 leaders of SpaceX and then like, go build another SpaceX. Like, man, I don't think so. Like, you need the facilities, the hardware, the regulatory approvals, the DoD contracts, et cetera. So it's like, it feels like AI is in this unique space where it's like, well, actually there's like top 10 researchers. And because the field should be super valuable as like talent alone. And so I don't know that you'll see like a broad set of legal terms changing, but you may see that obviously, you know, in AI. I think also, like, I don't think it's totally fair to compare all these zombie acquisitions as sort of identical. I think Windsurf was a particularly special case where the way that the sort of founding team and acquirer treated the remaining shell company was very poor in terms of both the liquidity to the remaining employees, how they dealt with vesting schedule, acceleration of stock, et cetera. I can't talk too much in detail, but that fundamentally was structured very differently than the last handful of these that we've seen, or at least the ones that I've been somewhat familiar with. And so I think that's where you saw a huge revolt, obviously from the remaining Windsurf team. And obviously I think they took a pretty brilliant move of working with cognition, where now that combination of cognition plus the remaining Windsurf team is I think, a super, super credible threat to the like, team that Google got. And so it's like, you know, sort of weirdly, you know, weirdly, Google may have actually enabled like a really formidable competitive team by the zombie acquisition. But then from the Mag 7 perspective, I kind of get it where it's like, you look at the like, you know, Adobe figma, you know, sort of deal and you know, you look at the, you know, team there having to had to pay a billion dollar breakup fee because Lina Khan was never going to let that sort of go through. It's like, okay, well, you know, you kind of, you know, you know, get this, you know, sort of fast track and you know, don't get the, you know, sort of regulatory involvement. Fortnite, I feel like this may be like a very short window of opportunity where like, there's just no way that the FTC is not going to put something out that starts to say that like, you know, even these minority acquisitions still require some sort of like, regulatory oversight above some certain transaction size. And so anyways, I see all this stuff as like more temporary aberration and you're going to see Controls put in both via like legal terms and term sheets, regulators coming in from above. And it's sort of going to get squeezed through the middle. And I don't think that it's going to be like, you know, my brother Pavel, for example, I think was tweeting something about how this like, fundamentally breaks the like, social contract of Silicon Valley. I feel like it'll be pretty rapidly repaired in a way where you won't see that long term of outcomes where the joke is like, why be a founding engineer, you know, at a company when you can just be the founder and you know, you know, you know, putting your two weeks, you know, I think, I think you'll probably see this stuff, you know, all these edge cases sorted out pretty quickly.

Well, let's move to something less controversial, please. Donald Trump, you're hosting an event with him. Break it down in terms of tack. It really is less. It's probably easier to talk about having a chat with Donald Trump about artificial intelligence.

I got a spicy question next. So this is like a layup. And it was funny because you announced the event and Will Menidis immediately commented and was like, blah, blah, blah, blah. And John and I were both like, no, this is exactly the kind of event that Hill and Valley should do with all in. And Donald Trump, it's like, it's incredible.

Fit you and them and everyone involved. But give us the breakdown. What's actually on the agenda? Can people just DM you and get tickets? Is this even open? And then what do you hope to have come out of it?

Yeah, maybe we'll start off with the back half of the question. Obviously the administration put out some early sort of executive orders when they sort of first got started talking about both winning the AI race and then showing the prosperity, economic growth and jobs that would come out of it. I think what both us and the administration are looking for coming out of this event is just there's going to be a whole set of executive orders that are signed either sort of live, that go into more sort of specificity across the various areas of ensuring that we win the AI race and really communicating a message that there's both a commitment from the private industry sort of players that are significant in this space as well as from the government to really ensure that that sort of happens. And again, from the lens of like, I do think there's been some sort of chatter about is AI going to be sort of job growth versus job destruction? And while everything from the sort of steam engine to Internet to mobile, etc. Has always sort of done both on a net basis. It's always been sort of net, obviously super positive for the economy and super positive for jobs. People just do different types of jobs. And so I think making sure that that's very clearly also what's happening in AI rather than just like the, I don't know, contra narrative of all that's going to happen here is everybody that works in customer service is going to be sort of out of job. It's like, well there's also going to be sort of way more people doing data center build outs and chips returning to the United States and broader sort of energy infrastructure that needs, you know, sort of get built. We've publicly announced, you know, sort of some set of the speakers, some of the bigger ones, you know, sort of Lisa Su from amd, you know, sort of speaking on you know, sort of chips. But then there's also some sort of earlier stage players that focus on everything from you know, AI factories to data centers to energy to you know, industrial robots powered, you know, sort of by, by AI. So you'll see this theme of like it is a lot more like what I call like real world or like physical, you know, sort of AI more so than some of the like traditional software platforms that are getting you know, sort of a ton of, you know, sort of narrative in the press. And then yeah, you know, we unfortunately, you know, don't have infinite space to have the entire broad public, you know, sort of come attend and so you know, sort of pretty, you know, sort of filled out but it will be, you know, sort of live streamed. It's all the content, you know, sort of will be available and we're super excited for that. And yeah, this is definitely pretty quick, you know, sort of turnaround. We didn't have a ton of you know, sort of upfront notice so thrown together but really thrilled obviously to be collaborating with the all in guys I think would have been basically impossible to obviously do this without them. And obviously Sacks, one of the besties is the AI and crypto SAR and Mike Kratzios, I forget if his title used to be at least CTO of the United States. I forget what the new title is. I think it's Head of Science and Technology at the White House. But this is very much so sort of their brainchild and their policy priorities that are really going to come to light. And we're super thrilled to have Hill and Valley as a sort of co host of this and obviously sort of builds upon the event sort of earlier early in the year. And I hope we're able to do more things like this in the future irrespective of who's in the administration. We've always tried to keep Valley really sort of bipartisan and so you'll definitely see a lot of sort of Dem, both participants in audience and Democratic leaning founders on stage. So always want to make sure this is something where yeah, we're sort of of a forum that is sort of here to just be the meeting of minds between policymakers and sort of folks in the technology ecosystem.

It should be an easy topic to keep bipartisan, I would imagine. Like it has to be one of these. I mean obviously everything's super partisan in general, but artificial intelligence, like you have a ton of Democrats running labs, working at labs there. But then there's a job story if you're building out a, you know, a massive, you know, data center in a red state that then the red states folks support it. So it doesn't seem like it should break down traditional party lines. I am interested to hear about what is the, what is the actual relationship between the administration and AI companies that are doing build outs. Like there was this story just yesterday. Donald Trump said 20 leading technology and energy companies are announcing more than $92 billion of investments in Pennsylvania. This is a triumphant day for the commonwealth and the United States of America. Google said It'll put in 25 billion. Blackstone promised another 25 billion. Core Weaves doing a $6 billion investment.

Let's give it up for big business.

My question is like, is this just Trump like celebrating what the free market is doing and just drawing attention to a great thing and just giving the free market a pat on the back or are these big companies, are the deals getting so big that there's actually like, hey, we're, we're, you know, when Amazon was picking where their headquarters was, it's kind of like a debate and is there some back and forth where the administration can actually help out on something or stimulate capex? Do you have an idea of like what the shape of those discussions is like?

Yeah, I also, I'm going to touch on that. I want to quickly do it just like a go back on like the bipartisan sort of nature of some of the topics. Let's about talk two representatives in particular on the Dem side in California that I think have been proponents of a lot of the topics that are going to be on stage. Representative Ted Lieu from California and Congressman George Whitesides from California from, you know, Ted Lieu, more focused on the AI side. You know, whitesides more on all things aerospace industrialization but have been big proponents of the, you know, sort of, you know, abundance I think caucus is what they call it. Talking about a lot of these topics where it's like, you know, you, the way you're going to get through economic prosperity is through leaning in on technological progress. And then hell, even Obama basically sort of came out in support of this sort of recently, right. Saying that like, you know, at the end of the day our party's platform doesn't work if you can't afford housing and people refuse to sort of build it.

Right.

And so he came out as like clearly an extreme sort of yimby, which is not at least the like default, you know, commission of California, you know, sort of position. So and he's got you know, I think the highest Democratic, you know, party approval rating. I like, I saw something where it's like he's at like 96% and then like I think the next best was like Hillary Clinton at like 65. So it's like, you know, if Obama's coming out, you know, in support of this, I think it's going to be a huge momentum shift on the downside which I think is generally very good for the country where you know, more and more of these topics become more and more bipartisan in a meeting of the minds on the like, you know, announcements around data centers. I admit that I don't know the like capex programs in that specific sub industry as well. But I'll talk about two industries where I do know that the government has been sort of leaning in on one around sort of forestry and natural resources and another one around sort of dod. In both of those areas there are like net new government programs and budgets to basically provide these more like higher book to value guaranteed lending programs at lower cost of capital for people that are fitting for companies whose projects are fitting into the policy priorities of the administration. And so that's everything from you know, building out you know, sort of huge net new re industrialized user defense related sort of factories and manufacturing lines. And then for folks that are building everything from you know, sawmills, processing plants, etc. On like you know, from some natural lands or national forests as well as on the mining side. And so at least there there's like literal direct program offices, line items, etc. That are actually like influencing the capex spend and making it sort of, you know, you know, putting their thumb on the sort of private sort of market markets to make the cost of capital and like the debt available to these companies even cheaper than maybe what the private markets could bear. Which, for what it's worth, it's like, you know, if you look at, you know, Fannie, Freddie, on the sort of consumer housing side of things, it's obviously the government, something that the government does there for mass sort of single family homes below a certain price point. They obviously provide a sort of guaranteed loan program. And so it's something that the sort of government has used before to prop up other sort of market. And so they're clearly doing this around these other administration priorities. I wouldn't be surprised. I mean, I haven't dug in, but I wouldn't be surprised if there's something similar in the data center, you know, an AI, you know, sort of market as well.

Yeah, quick question for you. I know we have a couple minutes left. Who wins in a bar fight? Global birth rates or Grok Companions?

Oh man. Tough, you know, anime girls in sort of short black skirts with sort of ponytails or.

Teaching you quantum physics.

Yeah, teaching you quantum physics for everyone's sort of weak spots. So, man, I am wondering what's behind Elon's promotion of it, where it's like, do they just need a certain amount of money?

Here's my thesis. I think it could very well become the fastest growing consumer app of all time. Like legitimately. Like you combine Elon saying I endorse this and you're buying from an Elon company. Same. I created this company, I also created Tesla. And it's basically, massively doing basically TAM expansion, which is like, babe, I didn't subscribe to Grok Heavy for companions. I use it at work. Right. I pay $300 a month because I need a long context window and I need to, you know, I need his help writing emails. And so I think that those factors of just like basically promoting that kind of product on a global scale, a really high price point with something that humanity will pay for, could mean that this is a billion dollar run rate product. Like within the next like couple years.

One year. Yeah. My girlfriend's name, Grok Heavy. That's not like, it doesn't sound, it sounds like a SpaceX term, like heavy from Falcon Heavy. I don't know.

I do think you're just gonna see more and more of this like AI induced psychosis where just like, you know, it tells you what you want to hear. It perfectly has these feedback loops. You're already seeing this stuff from teenagers, et cetera.

And I think there's a Reddit thread we were looking at earlier identifying all the words that people use once they're in LLM induced psychosis. And there's eight or so words that start coming up in that person's dialogue really rapidly that we've seen recently, like recursive mirror structure, these words that get used over and over.

That's fascinating.

And when you, I think people are combining them with drugs, which is insane, and then they're going down a single, basically prompting like 7,000 times in a row with like the same dialogue. And it's easy to imagine how that would drive you insane. One last question. How do you think Jensen is doing juggling US interests and China's?

Well, it's interesting that Singapore Somehow buys whatever $15 billion of chips for a small island nation.

I think it's biology.

Yeah, it's all biology. It's his network state is getting really, really powerful.

He has great networking equipment.

Yeah, yeah. I mean, man, look, I was at this fun conference up in Idaho last week and there was various geopolitical talks including one that included an expert from Tai. And it's just, you know, we are in like, you know, by far the most, you know, physically risky geopolitically, you know, sort of risky environment that the world has been in since the fall of the Soviet sort of Union. And I think it's, you know, sort of really hard to figure out how to interact in the environment when you're somebody like Jensen that has like revenue from China, significant infrastructure that you're dependent on in Taiwan but you're based in the US I don't envy his position.

I don't envy his global, global equity market. He's basically holding the global equity markets up. Like, you know, my, my hope is.

That it, it's a stabilizing force. My, my hope is that it's a stabilizing force. But thank you so much for stopping by. We'll let you get back to your work and your lunch. We'll talk to you soon.

Talk soon.

Great to see you. Have a good one. And before we bring in our next guest, I will tell you about Fin.AI, the number one AI agent for customer service. Number one in performance benchmarks, number one in competitive bake offs and a number one ranking on G2. You can get started at Fin AI. They have a free trial, so go check it out. Used by anthropic and Monday.com. welcome to the stream. Our next guest is Lucas. How you doing, Lucas?

Good to meet you, Lucas. What's going on?

Doing good.

This is the first pot I've been On with two guys with better hair than me, so.

Oh, I don't know about that. You're looking sharp.

You look fantastic.

You're looking sharp. You came prepared. Next time, would love a suit. But the background is a good.

The background is fantastic.

The background helps.

New York City.

It's not fake.

Believe it or not.

We're at Hedge Fund HQ. 9 West 57th.

No way.

Amazing.

Beautiful.

Break down. What it means to be at Hedge Fund hq. Talk about life at a crossover. I want to talk about the state of the markets. I want to talk about CO2's reporting and analysis. It's a very condensed. Condensed. It's a very condensed report. It's very accessible. But it still feels like it has a little bit of the deeper analysis that you don't get from some VCs. So I enjoy that. What's the overall temperature on the markets today?

Yeah, I think we're still pretty optimistic. Right. I think you have a really complex environment. Right. The NASDAQ and the S and P got off to their worst start since nasdaq inception in 1971 through April 18th. But since then you've had a lot of good news. Right. Inflation has been lower than people thought. Tariffs haven't affected inflation as much as people thought. The US consumer has been really strong. Performance in companies has been really strong, I think in the public markets, but also in the private markets, which is giving people a lot of confidence. And so I think we feel good. I think the only kind of question mark we have is is there this delayed effect to tariff tariffs on inflation? And what does that even mean? Right. We haven't had a situation like this in 50 years. So what really happens? Economists don't know. We don't know. And so predicting that is really tough.

And we put up a surplus. That's an immediate.

Yeah, that was crazy.

We put up a surplus in June, which you don't see much from Uncle Sam.

Free cash flow. And then the PPI came in today, a leading indicator for inflation.

Lower than.

Than every single economist predicted. Right. I don't think you usually see. So I think it leaves a lot of room for optimism.

Yeah. Did you read into the earnings season that we're going into? JP Morgan beat earnings sounds pretty bullish for the American economy. The Wall Street Journal said the US Economy showed signs of resilience despite escalating threats of a global trade war. A sign that American corporations and the consumer are still charging ahead. You love to hear it. Jamie Dimon said, we. We've basically been in this Soft landing now for some time period, it's been resilient. Hopefully that will continue. Is he talking about soft landing coming off of the end of ZIRP or just the start of the year? What is your interpretation of Jamie Dimon's analysis of the American economy?

Yeah, I think it's basically coming off of ZIRP and coming off of that whole era. Right. And I think economists were predicting and a lot of people predicted a recession for the last couple of years that and maybe we've been in a mild one for some period. But I think that's been the ultimate question and it just kind of hasn't come to fruition the way that a lot of folks thought that, a way that a lot of economists thought. So I think people feel pretty good. Right. And I think you see a lot of confidence from the big companies. Right. You see Meta and Zuck talking about building data centers the size of Manhattan. And I think if we were in a negative environment where people weren't seeing gains from AI and all these wonderful things, you wouldn't see that type of investment going in. And so I think it leaves a lot of room for optimism. And we talked a lot about this in our East Meets west presentation, which you alluded to. But AI, I think, has the potential to really improve productivity and that solves a lot of problems. Right? It can solve the national debt problem, it solves the general productivity problem, it solves the GDP problem. And so if what a lot of folks believe can be true about AI, then maybe there's even more room for optimism at corporations.

What are the key lessons that we should take away from the end of the ZIRP era? The rise of interest rates. There were some companies that were revealed to be fraudulent or just overvalued, but it feels like we're already in the lesson certainly doesn't seem Never invest at 100x revenue multiple because that's already worked out for people who have been Investing, investing in 100x revenue Multiple companies for a few years and got exits because that's happened. So what are the lessons?

Yeah, I think it's a few different things. I think one is quality Matters is probably the number one lesson that I've pulled from kind of the ZIRP 2021 era. Right. You guys have an amazing sponsor in a company called Ramp. Some people have heard of it. Very small startup but doing very well. That's an example of an asset that's just incredibly high quality. And people always thought it was super expensive. People invested 200 times gross profit or 50 times gross profit. And the reality is the best businesses compound over very long periods of time. And if you're investing in a quality business, often at really any stage over duration, you're going to be in a great spot. I think the place where folks got in a lot of trouble is being in sectors they aren't experts in, reaching into different sectors that they weren't experts in. A lot of growth investors, candidly, like us, chasing venture rounds with growth dollars, which I think is happening again. We got pockets of this in different areas and then just way too many companies. Right. We were at the end of the kind of the SaaS era in many ways, without AI yet, and people were funding literally everything, kind of chasing yield and chasing returns. And I think there's very few companies that are very high quality. And I think that's kind of the lesson coming out of that era is you gotta back those and only those.

Yeah.

What is. We don't spend a lot of time on the east coast, at least this year because we're here in the studio. But I imagine you pick up sentiment from traditional hedge fund managers that might even be in the same building as you guys. What is the sentiment? Are there pockets of people that are, I love the volatility we're printing. I live for this. And then other pockets that are just maybe pannikin people that might not just enjoy the chaos of sorts. What are the different. What are the different groups?

Yeah. I think the biggest debate and the biggest flavor and mix that you get is just the AI doomers versus the AI optimists. Right. That's the biggest mix that I think you have in the hedge fund world. And you see it once every three or four months. You get a deep seq moment or you get a moment where Microsoft says they're pulling a few contracts and the market pulls back. Right.

And.

But then you have the reverse of that whenever, you know, you see these incredible numbers out of companies like OpenAI and more recently, Anthropic, and then just this amazing global use of AI both in the consumer and now more recently. I think you see it in the anthropic numbers in the enterprise. You guys have talked about this with coding and other use cases. And I think that's kind of the biggest debate that's happening inside this building and, you know, other buildings close by. And I think it's this huge contrast between Silicon Valley, where it's pretty much all AI optimism, and the east coast, where there's definitely a mix of opinions.

Interesting, interesting. Do you find that people that are bearish just typically don't use any of the products, they just haven't adopted them? Because I think it's one of those things. It's hard to just be blanket bearish if you've never experienced a magic moment using one of the products.

Yeah, I think there's, honestly, I think there's a lot of that.

Right?

Yeah.

But I also think it's people that have studied the past cycles and they see that, you know, you do have these bubbles and you know, maybe we're in a little bit of a bubble now. There are companies, you know, getting raising at 10 billion plus valuations with no revenue, no product, you know, folks in a dream. Hopefully you guys soon. But I also think there are pockets of incredible things. Right?

Well, yeah, even we were debating, I was having this conversation, it was like, okay, it's clear where OpenAI's revenue opportunity is, where the revenue is today, where it will be. It's very clear where anthropics is. My question like a week ago was, where's Grox gonna come from? Like, where are we gonna see an explosion? And then this week we had new contract with the DOD and then same day and then companions, which I'm gonna go out and just predict. I think that'll be like a billion dollar business, like very quickly, regardless of how people feel about it. So. Yeah, and it's just there wasn't a lab that was willing to go there, even though people know that that is a huge opportunity.

Yeah, I think it's super interesting. And it's funny, when I was growing up, the biggest thing my mom had to worry about was too much World of Warcraft. Now you got something like this. But I think things like this have been incredible revenue opportunities in the past. Right. You think about where people went with character AI initially. Right. And the absolute skyrocketing of usage that happened there. You think about a lot of these apps that have kind of conformed in and around ChatGPT. I think there's just a huge, huge opportunity here.

How are you? What's your framework for some of these new experiments around tokenized private company shares? Because I imagine a large part of the CO2 portfolio is what companies like Republic and Robinhood are trying to get out there in terms of giving access to retail. But in general, obviously these experiments are not sanctioned by the companies. In fact, the companies are actually coming out and saying, and there's this weird dynamic where there's always going to be 10, 50, 100 times more demand however much OpenAI shares you could get on chain, there's always going to be exceptionally more demand. And so if you can actually make these experiments happen and there's just going to be wild distortions.

Yeah, Actually, I think there's a lot of ways you could take this. But the number one takeaway that we've had from this, and we talk a lot about it internally because one of the missions that we've been on recently, we have a new fund called ctec, which is really the idea behind it, is to allow more retail investors access to our platform. And I think it just shows the hunger for a normal person to get access to companies like OpenAI and SpaceX and Revolut and some of these fantastic businesses that, you know, normal folks can't get access to.

Right.

And it's this trend of companies staying private longer, not going public. It's the trend of the regulations that we have in the US of people not being able to access these types of names, where I think it just shows there's a real hunger for this. Right. People are so excited to own OpenAI, they're so excited to own SpaceX, and there's no way to do it today. And I think it's going to be something that the US is going to have to grapple with and companies are going to have to grapple with because people have this insatiable demand for their stock.

Yeah. One crazy update I had to my understanding of the value of staying private longer was this past weekend where Cognition wound up acquiring Windsurf. And it felt like there were a lot of other things going on, but one big one was that if the whole company, Windserve, had gone to Google, that would have been crazy. FTC review. But small, medium sized, I mean, huge Cognition is a fantastic business, but, you know, it is a small company in the eyes of the ftc. It's a private company, it's a young company buying another small, young company. All of a sudden much easier to put those two pieces together. So that's kind of like an untapped, the ability to move faster on the acquisition side, potentially an underrated benefit of staying private longer. But are there any other factors that might be driving the decision to go public or private or stay private longer in your mind?

Yeah, I think there are a couple others. I think the most simple one is people want to invest really heavily and there is this perception that the public markets only rewards profitability and not growth, which we actually think is not true.

Sure.

But it is certainly a perception. And I Think folks like you know, that want to invest and be unprofitable and really invest over a 5, 10 year basis into product or R and D or sales and grow really quickly. It's hard to be a public company that way. It really is, because public market investors can be fickle. You have to report every quarter. It's harder to take a long term mindset. And so when you're still in that investment mode, a lot of folks want to stay private longer. And it is a benefit to not have to think on a quarterly basis the way the public company does.

Yeah, I want to know about the east meets west, like kind of mental model around the AI salaries. The hundred million dollar research offers are going mega viral in Silicon Valley. It's just grabbed everyone's attention.

And the new play is the UNO Reverse.

Yes.

Now companies realize that's a card you.

Can play, but everyone keeps drawing on the analogy of sports. But there's also also the analogy of Wall street where top traders at Citadel apparently have pulled in a billion dollar bonus. And Greg Abel famously made more money at Berkshire Hathaway on salary than Warren Buffett did for a long time. And so this idea that like the CEO should necessarily be the salary cap for the organization has not been true on Wall street and now it seems to not be true in the AI world in tech. But what has your reaction or your world's reaction been to the crazy trade deals, the AI salaries?

Yeah, I love the UNO Reverse yesterday. I thought it was hilarious.

Incredible.

I think it's kind of interesting because we've seen this here, right? You saw it, I think in actually in growth investing in 2020 and 2021. Right. A lot of careers were yanked forward and a lot of salaries were yanked forward. You saw it in the hedge fund world and you still do with the rise of pods. Right. Just like you're referring to. And I think this happens every time. There's just a big imbalance between labor supply and labor demand. Right. When there's an emerging sector, an emerging market, an emerging trend, and there are very few people that can do something really well, the market just rewards that incredibly disproportionately. And I think you see that in athletics all the time. You see it in the hedge fund world, you see it in VC world, you see it everywhere. And now we're seeing it with AI researchers. There are just so few of these people in the world. I mean, we're talking on the order of hundreds that people really want at the end of the day. And dozens of Companies that have incredible balance sheets. This war is going to happen. I think it's going to get, I think it's going to get more intense.

Yeah, I want to talk about competition specifically. We were talking earlier, there's been a bunch of examples of this. The question of like, if you're a ChatGPT rapper, is ChatGPT going to, is OpenAI going to Steamroll you? We have a buddy of the show who runs a company, an enterprise SaaS product essentially, and AWS just launched a competitor and I'm wondering how much weight do you put in the bucket of like just the grit and the determination of the founder can overcome a challenge from a massive company where they have 20 startups that they're competing with and if you stick with it, you can break through versus like there are fundamental markets, forces that could have been understood in an investment memo at the seed stage to know that it was not good to go up against this particular hyperscaler, this particular growth stage startup. How do you think about competition when an entrepreneur is like in the various parts of the journey?

Yeah, I think this is a super interesting question and again, I think it's actually one of the other differences between Silicon Valley and the hedge fund world. And I see this a lot because I straddle it where I remember two years ago, whenever we were first thinking about, all right, how does AI impact the public and private markets? We're like, oh, well of course the public market companies are all just going to latch onto AI and they're going to dominate this. And you fast forward two years and certainly Nvidia has been the case and some of the cloud providers in certain areas, but the application software companies haven't really done anything with AI. Right. Like people are like, oh, CRM, they're.

Going to kill it.

You know, all this stuff. It hasn't really been the case yet, which is kind of crazy. And you contrast that with, look at what's happened in the private markets with new companies and incredible CEOs cursor cognition. Harvey Glean open evidence. You talk about all these companies that have come up that are competing with these huge companies and killing it, and I think it does. At the end of the day, there are certain businesses, if you have a network effect, it's hard to break into that. But even you see OpenAI versus Google. Right. I think it's just a lot of this does come down to like the grid of the teams being able to go996. All this stuff matters a great deal more than just some Incumbency advantage.

How are you thinking about stablecoins? Everybody's excited about them. Everybody wants to make money on them. It's hard to make money on them by holding them, especially with the new regulation.

I put all my money in stablecoins and my portfolio is flat on the year.

Flat on the year.

It didn't go down, though.

But it's such an interesting category because it seems, I shouldn't say incredibly easy, but quite easy for existing financial players to add stablecoins to their product. So if you're a bank and you want to enable your user to hold, send, buy, swap, stablecoins, not super difficult. Circle is public. There was a massive amount of interest, but. But personally, I look at Circle trading at whatever their current PE ratio. Talk about an expensive company right now. How are you guys thinking about the category? Are you looking to place new bets or do you think it's more of a feature to potentially existing bets?

Yeah, I think, one, we're paying a ton of attention to it and our view is this seems like the second big use case of crypto, right? Really being able to. If Bitcoin was the first as this, call it a store of value, call it whatever you want to call it, digital gold, then stablecoins really do feel like the second really big global use case. We are trying to figure out actively how to play this right. You talk to the folks at Stripe, they're trying to figure this out. They made what I think is a killer acquisition with Bridge.

Bridge.

You talk to Ramp, they're trying to figure it out right? There's a big opportunity for Ramp internationally here, I think, to be able to really move money. We're really excited about it. We're trying to figure out the right way to play it. We haven't seen a ton of new companies come up, but a lot of the more innovative, slightly bigger companies like a Stripe, I think, are going to really be able to take advantage of this.

Yeah, it's hard because you're excited about a broader trend. You want to deploy capital against it, and then you look at your portfolio and you're like, wait, you're really well positioned here. You're really well positioned here.

Sometimes the answer, as usual, is you just buy more Stripe.

Buy more Stripe. That's funny. How are you thinking about robotics from the lens of a growth investor? There's some bright spots, areas where there's real traction, but then some of the hottest categories feel 10 years out. And as a growth investor, you're being asked how, If I invest $200 million today. Is there going to be someone else in two years that's going to come in and write an even bigger check? I'm curious how you're thinking about that category.

Yeah, this is a space we've talked a lot about and obviously we're really excited about it. We have an investment out of our venture fund in a company called Skilled, which you all may be aware of that if you haven't, you should go see them and see the products. It's pretty killer. The advancements that they're making spin out of Carnegie Mellon out in Pittsburgh. I think from a growth investors lens, this is like one of the hardest things. Right. Our view and our mandate on growth investing is we invest in businesses that have real business models and that are clear or emerging leaders in their categories. And in robotics, we're still too early for that. That's just the reality. And our view is generally you want to be in the leader because the vast majority of the of the gain accrues to that number one player almost always in technology. Some exceptions, but almost always. And so it's better for us as a growth investor to just wait a little bit longer and be a little bit patient. But our view is it's coming. The question is how long? I do think there is one exception in this, in this market on the incumbency bit, which is you have Tesla and you have Elon. And in the robotics world, it's hard to bet against him, I think, in a lot of ways. Right. Like he's got the power of Xai underneath him doing all the research. He's got Tesla, he's got all the real world data that exists out there. He's got SpaceX, all these things kind of mixed together. And he's a founder and so it's kind of hard to bet against him in that landscape where it seems really compelling.

Yeah. Tyler Cowen has this take on AGI. He says AGI is here, but. But the impact will be slow because so much of our economy is, you know, healthcare. And it's like nurses in the hospital, doctors doing physical things that can't just immediately be replaced with an LLM. And so you'll see.

Or somebody paying rent.

Yeah, yeah. And there's all these different pieces of the economy that, that might not be disrupted by LLMs. They might be disrupted by like, you know, AGI superintelligence, robotics feature. But how are you thinking about like the surface area of AI's impact? You mentioned Harvey and Legal and Glean. It feels like we're starting to map out a few of the different subcategories that could be rich pockets of value. But are there any that you think are kind of underrated or might be next up or where? Or do you agree with just Tyler Cowen's general thesis on AI as an advancing force or sustaining innovation in the stuff that has already been brought online?

Essentially, yeah, I think I pretty much identify with that. I think there is something pretty interesting, right, which is in technology again you almost always see the consumer first, right. You saw it in messaging, you saw it in mobile, you see it in everything. The consumer adopts really quickly first. And we've seen that now There was the ChatGPT moment. We're two years out from that. The consumer is adopting AI. Enterprise takes a long time and getting it in the hands of people at work like lawyers, doctors, even developers to a certain extent, but not so much in this case. It takes a long time. Fundamentally people are sticky, enterprises are sticky, processes are sticky. And so it takes a long time to get things really implemented. And I think that's kind of what we're seeing. Right. You've got a couple of early use cases. Coding is, is by far the most obvious. We think it's the vast majority of the value that's being accrued in the enterprise today. Right. There was 1.3 billion in coding ARR added in the last 12 months outside of anthropic, just in startups. And so you're seeing it there first but I think you're also seeing pockets right. Legal, Harvey, open evidence medicine, glean search. Right. You're starting to see it especially in the text, in text out type use cases. But I think it's going to take time.

Yeah, it is crazy thinking about like I bet if you looked at consumers adopting like file transfer versus airdrop is probably higher than or like or like roughly the same as like in a hospital. Are they sending more faxes? Like the fax has been sticky in healthcare and at home we're like using like the next, next, next generation. Yeah.

So like all of a sudden are we all gonna wake up and build our own CRM? Probably not, no. We are still using fax machines.

Yep, totally.

I think these things, they take decades.

Yeah. What are you tracking on the CapEx side for the AI buildout? I've been reading semianlysis a lot. There's this new big push in Pennsylvania, something like $92 billion going in there from Google Blackstone. Core Weave's doing stuff. We talked to the co founder of Core Weave he was saying that the capital markets aren't necessarily even ready to absorb the amount of demand that are there. Talking about. Just like when you talk to the AI folks in Silicon Valley who are more like on the philosophy side, they're like, well of course we'll build a 10 gigawatt data center next year. And when you talk to the finance guys, they're like, wait, you want $500 billion to do this size of Texas?

You'd like to have interview next year?

Exactly. But yeah, what are you tracking on the AI capex side? What are the interesting stories? What are the interesting threads that you're pulling on over there? Either data points or just narratives or different market structures that you've kind of identified?

Yeah, I think the biggest thing that we continue to follow and we continue to hear is just that everyone is compute constrained, right. Like all these build outs that are happening for the last couple of years. And again, this is a big west coast, east coast debate again and has been about just like capacity and are we overbuilding and all of these things. And I mean you even saw with Microsoft, right, it's caused some of the friction between Microsoft and OpenAI very publicly is about compute build out and OpenAI not having access to enough compute. Right. And so I think that's just what you're hearing and I think that's the most interesting thing that's happening right now is anthropic's compute constraint. OpenAI's compute constrained. They can't release new products because they can't serve them right. And so I think you're just going to continue to see this build out happen. And I think we are really bullish on that trend. I mean we're heavy investors in Core Weave. We did a lot with them before the ipo and I think it's just been in a, it's been an incredible story so far. Right. And I don't think you've seen an end to the demand yet.

Last question for me, what are you looking looking for in the autonomous vehicle market over the next one to two years? We've got two key players in the race. Obviously Tesla one, they have an interesting map of their operational area down in Austin. And then Waymo, of course there's also.

Horses, traditional horses, it's autonomous and there's.

Also pony AI, I guess calibration. But it's such an interesting dynamic because when you have this very Capex light model and then you have the polar opposite and it'll be an interesting battle from our view. But I'm curious if you have a read.

I mean the most, the thing I'm most looking forward to is just being able to use Waymo everywhere. But yeah, we're all ready for it. I think the consumer demand is obviously there. I think the biggest dynamic that's going to be interesting that the public market is paying attention to is actually what happens with Uber and like existing networks and how does that end up playing with all these different providers that are emerging with you know, Tesla which is trying to build its own network, Waymo which has built its own network. What happens to Uber in this world? And I think everybody, there's a bull case and there's a bear case on this and I think everybody's trying to figure it out. But it's clear that the unit economics on this stuff work for, for Waymo already. And whenever we move from Jaguars to something else, eventually it'll probably even get a lot better. But it's gonna be a really exciting battle.

And then the second and third order effects of. Okay, if people are getting two hours a day of their time back, what are they gonna do with it? Probably scroll on apps which would be good for Zuck or they'll do more email different beneficiaries.

Wimo should move to Paganis, Pagani Wyros. They're hand built. It might, might, it might, it might affect the economic quality but it is quality and it would be differentiating. I mean apparently they have a deal, they're talking about a deal with Toyota so we might see some Supras.

It only makes a little more sense.

But this is fantastic. I really had a blast talking to you. Thanks so much for hopping on.

Thanks for coming on.

We'll talk to you soon.

Talk soon, Lucas.

Bye.

Cheers. I'll be right back.

Really quickly, let me tell you about public public.com investing for those who take it seriously. They got multi asset investing, industry leading yields. They're trusted by millions. Folks, you can get started@public.com and let me also tell you about out of home advertising made easy and measurable. That's adquick.com we're at baby. Say goodbye to the headaches of out of home advertising. Only Ad Quick combines technology, out of home expertise and data to enable efficient seamless ad buying across the globe. Our next guest is Ravi Gupta from Sequoia Capital. Welcome to the stream, Ravi. How are you doing? I'm so glad we were able to get some sleep. I hope you slept well. I was about to stay up all night debating you but I said I'm gonna stop texting him because he's coming on the stream. We can have the debate live. How are you doing?

Are we even on different sides of the debate? Man, I just feel like we might be.

We might even be agreeing.

Although I forgot my blazer. You guys are looking good.

Yeah, no, no, no, no. It's all good. But first I have to ask you explain to me who is Lavar Ball, by the way?

I do. Maybe you guys can flash up the tweet, because I don't have a lot of banger tweets, but this one was one where I wanted to email Elon if I had his email address and just kind of be like, I think the algo is broken. Lavar Ball is. I tweeted that I want to become the LaVar Ball of AI researchers. And LaVar Ball is the father of three NBA players, or two NBA players and another one. But he has three sons, you know, Lonzo LaMelo and LiAngelo. And he's been very famous about, like, trying to make them into NBA players from birth. And he's a very. He's kind of a blowhard. And I have three sons. And with some of these pay packages that are going out for the AI researchers, I said, you know, why would you mess around trying to make them NBA players? Like, look, their genetics might lean a little bit more towards becoming AI researchers. So I think the best thing my wife and I can do would be for me to become the LeVar Ball of AI researchers. And I tweeted this thing thinking that I'm going to get virality. TVPN is going to be talking about it, and I got like 13 likes, and it was pretty painful.

We did put it on the show. I think you might have just underrated the fact that tech loves aesthetic, the aesthetics of sports, but no one in tech actually understands sports.

Now here's the thing. Here's the thing. The Hayes paradox, which I created, this is true, which is the idea that the more funny that you think something is, the less likely it is to go viral.

Right?

Because the thing that's going to be most funny to you is like, super niche and just entirely dependent on what you find fascinating and interesting. But that doesn't mean that the entire public will also.

I remember I posted Paradox is True. I was so disappointed.

Yeah, we posted something that was like, a very obscure reference to the fact that, like, Ben Thompson was on vacation the this particular week, and it just completely flopped because, like, people, it was your worst performing of the year. People, broadly, are not tracking Ben Thompson's vacation schedule like I am and Jordy.

Had no John, I. Listen, I think it might be that the algo was broken for.

Yeah, yeah, those are funny things. Let's leave it there.

Occam's razor. It's the algo.

Yes, yes. We're built different. It's not us, it's, it's, it's you. It's the algorithm. But, but let's actually talk more seriously about like what the Lavar ball of AI researchers would actually look like. Because we were kind of, we were talking about. So there's these crazy trade deals that are happening. Hundred million dollar offers. Are we going to see recruiters that can actually bring these people in the door get crazy salaries? Is that the next wave that we're going to see? Are we going to see a recruiter go to a mags? I think in this case feels like it's founder mode. Yeah, feels like it's founder mode. And then the question I had for you was that, you know who's really good at spotting AI researcher talent? Venture capitalists. So are the venture capitalists gonna get poached and they're gonna say, hey, yeah, you're making a lot of money as a GP at a tier one venture firm, but you wanna come over to Meta and start doing recruiting and just bring these folks right in here, we can pay them out like it's liquidity event on day one. So how does this all shape out?

I think that I don't know how the recruiter takes shapes out. I guess my perspective on this is like, I think that high performing people in Silicon Valley have probably been underpaid relative to their performance because of things like comp bans for a long time. And I think that if you think about it, you know, comp bands imply that everyone within a certain level is producing within some range, right? You know, maybe 1 to 2x of each other. And I think anyone who's ever worked in a business knows that like that's just not true. Like there are people that are, that you would do anything to keep within a level. And then there's other people. It's like, okay, if they left, like, you know, we'll deal with it. And so I actually think that the high performers have been, you know, mistreated for a long time. So I'm actually quite happy with this idea of like, if you are really uniquely good, I mean, I'm talking like truly otherworldly good, you should get paid like your otherworldly good. And I think that one of the things you and I were talking about last night, John, that I think is also relevant is like these aren't guaranteed contracts, they're vesting every month. If someone, if you think someone is life changingly good and they turn out not to be life changingly good, they don't have to work there anymore and you don't have to pay them anymore. And so I actually think that, like, I think that to me the big thing that you actually. People have comp bans because they want harmony and they don't want to explain to somebody that somebody else is performing better. If you're going to do this, I think all you have to be willing to do to the other person, person who's not getting paid like that is like, look, I might be wrong, but you're not producing like the $100 million player. And you might think you are, but you got to make it more obvious because we're going to have a few hundred million dollar players and those $100 million players are changing the entire trajectory of what we're doing. But yeah, like, you know what, there's money in the banana stand if you want to go and produce like $100 million player.

I guess my one thought on the lavar ball analogy is like a big question we've been kind of debating is will these same types of offers exist in five years, 10 years, will they exist in a year? Is this momentary blip. It's like lavar Ball had his sons and lavar Ball played basketball. I don't think he played in the NBA, but he knew that in 18 years he wanted, he wanted to be watching his sons play in the NBA and he knew it'd be a pretty good job. I think the big question now is how sustainable is it for top talent to be? We've always joked on the show that Tim Cook has felt underpaid. Right. Running the most important consumer tech company in the world, making less than or making about the same as Ohtani, who's hitting a baseball, which is important and good.

But it's around $75 million a year for Tim Cook's salary and one tariff negotiation going correctly is a $200 billion proposition on market cap easily. Like, no one debates that, that if he has a good conversation with the president, he can save the company from losing 200 billion or add 200 billion in a quarter, like pretty easily because he's from such a high base. It's a multitrillion dollar company.

Yeah. So Jordy, I guess to your question, I don't know on the AI researcher side if this will still be the case in five years, because I think the input to it is sort of like. Like you believe that one of these people can change the trajectory of your company. Right. Even on like a scaled company. Right. And I think that. Or they can make your investment in the capex, you know, that you're doing on the GPU side that you can make that more valuable. I think. I don't know if that will exist in five years. What I do believe will exist in five years is there will be somebody or a set of people that you believe can change the trajectory of your big company, and those people are going to get paid a ton of money.

Money.

And I think that this is actually a good thing on that dimension, which is making it more normal that excellence gets rewarded. And I think that those of us who played sports in any form or fashion, when someone's playing more than you or getting paid more than you, in the case of your professional, your reaction is not like, that's not fair. It's like, I need to play better. And I think that this, to me, if it normalizes, sort of like excellence gets paid, that's great. But I think to your point, the real Lavar ball move for me would be figuring out what, what the industry would be for my kids to go and enter in that will be getting these kinds of paychecks five or 10 years from now. And like, AI is not a bad bet, but maybe it's robotics.

Yeah, no, no, I can.

Yeah. The other thing that's interesting is, like one of the major promises of the industry is that eventually the AI agents themselves will get so good at AI research and they'll just copy and paste themselves billions of times and then we won't even need humans. So. So it's like, it's almost a call option on that future of like, if you really just need to build the AI researcher, how much would you pay for an AI researcher that could build an AI researcher that you could copy and paste infinitely. Right.

The smartest person in our house is my wife. And I did make a pitch to her of like, hey, listen, why don't we think about you abandoning your medical practice and just go all in next 18 months, become like a baller AI researcher and try to get one of these packages. And she kind of looked at me like I was not as intelligent as some of the AI researchers, but I thought that was a good ide.

Yeah, yeah, it's hilarious. I want to tussle a little Bit more with the comparison and actually how comp works as a founder versus an AI researcher versus an NBA player. Because NBA players have these locked contracts where even if they underperform, they might be still getting $10 million a year. If they over perform, they might be underpaid for a little bit, but then they can enter free agency, renegotiate even, even if they. And then there is some variability on top of like the signing, like the sponsorship deals. So you might be underpaid but monetizing very well. With a Nike shoe deal, for example, on the founder side, it's kind of like feast or famine. But you have this uncapped upside with the, with the AI researchers. You have, yeah, you have the ability to like get fired, but you don't have as much of a clear signal on the impact. Or maybe, maybe I'm misunderstanding that, but if I think about there are going to be what like 50 key players on the superintelligence team at Meta, this is a tens of thousands of employees, trillions of dollars in market cap. It's very hard to say this researcher moved the market cap this much. They justified that hundred million dollars as opposed to if you hit the winning three pointer. It's very clear you won the game. And we can look at your stats specifically in your contribution. You can do Moneyball very easily. And in the founder seat, you know, it's like yeah, you hired the key people, you got the sales contract, you raised the next round, you added, you took the company from $100 million in you know, series a post to a billion dollar unicorn. You should get all that credit, all that credit and you should get that financial upside. And so it feels like it's a little bit squishier. But do you think that there will be more Moneyball applied to this? Do you, do you think that we'll just see kind of massive churn based on vibes like hey, you came in a couple months, didn't do anything that crazy or there will be more like internal politics to make sure you're justifying that massive salary once you get in. And then last question is kind of like the idea of gelling these teams. You know, the example we heard previously was like Lebron went to the Heat, they had the dream team. Still took him a year.

Cowan.

Yeah, Tyler Cowan.

Didn't know that he was a basketball buck. He could have been.

Yeah, I know, I know. You got to put a side by side side next time.

Yeah, so. So yeah, just, yeah.

And I would just chime in and say Like, I look at it as he, he, he acquired a team. He didn't acquire any one individual player. And so you're really measuring, hey, I spent X number of billions. What am I getting from that versus micromanaging in the short term, three months in, oh, what if, what have you done so far? That's a good question to ask, but not necessarily like you're paying, you're getting the team to deliver the thing that you want, which is maybe unclear, but.

Yeah, I think so. By the way, I will tell you, I think this is like my dream come true. The idea of like chatting with some new friends about like the mix of technology and basketball and you know, a little bit of like compensation thrown in. Like, this is, this is amazing. But I, this is the best. It's the best. I think that. Okay, so I actually think, Jordi, I agree with you. I think that the way this is going to go down for these companies and let's just use meta as an example, I think they're a look at their aggregate investment in capex as well as in people, and they're going to have some absolute goal of like, are we one of the leading players in reaching super intelligence or not? And I actually think that the people cost of that are much, much, much smaller than the CapEx cost. And I think that. But I think that overall it ought to be judged as like a team which is like, are we going to, did they become, did they enter the race? Are they, you know, one of the leaders in the race? And then I think within that, I suspect that, you know, Alex Wang, Nat Friedman, some of these other folks that are going to be leading this lab, I suspect that they'll do normal managerial stuff of like, are we actually getting the most out of each of these people? I do think that it's way easier to figure that out with 50 people than there is with, you know, a thousand people or, you know, 5,000 people. And I actually think that this idea of like Dunbar's number is like really appropriate here. You know, below 150 people is just like such a different management challenge than above 150 people because everyone knows what everyone else is doing and there's some element of like, game recognizes game of like, do the other people on the research team think the other people are carrying their weight? And so I actually think it's quite smart that the teams are the 30 to 40 to 50 people. And I think the core research team at each of these places is smaller than people tend to realize. And so I Think it is more clearer, you know, who is doing great work and who's not. So I actually think in this case it actually is a little bit more like a sports team. I mean, you think about it like the size of, you know, the team that is talking about assembling is like smaller than an NFL football team. You know, like you kind of know who's playing and who's not within that.

It is, it is. Another interesting thing that I think is actually healthy is like, you shouldn't need to quit your job to make a billion dollars. Like there's been this thing in the Valley which is like, you can, you can have a great life working at a, at a big important company, but if you want to be a billionaire, you're going to have to like quit your job and go start a company that's like basically the default pathway outside of some key executive positions. And I do think it'd be the sports comp is like LeBron, like, you're super talented, like we value you. But if you want to, if you want to achieve what you want to do on the, on the compensation side, go set up your own team and build, build an organization from scratch. And LeBron might say, well, I don't actually want it. I actually like doing this one thing. I just want to do this one thing well, which is play basketball. I don't want to deal with, you know, finding, finding a stadium and, and you know, investors and all these other things. So I actually think it's healthy that people can.

John's got a joke ready. You can tell by the look on his face he's got a joke ready.

I don't have a joke. I have an interesting similar anecdote. I mean, we keep going back to Tim Cook and being underpaid at 75 million, but I think Andy Jassy might be a more extreme scenario to kind of tussle with here. So Andy Jassy joined Amazon in 1997 as a marketing manager. He was named CEO in 2021, granted a 10 year equity package valued at 212 million. He's making about 40 million a year now. But he is widely considered as like, the reason AWS exists. So he created in many ways 200, 500 billion of market cap, like hundreds of billions of dollars of market cap, and hasn't really captured that in the right way. I mean, he's done fabulously. He's lauded as a great CEO. I love him, we all love him. But the question is like, is there some world where you can have an employee who comes in as a marketing manager sticks with you, doesn't need to do the round trip. I've talked to some people who work at big tech companies and they say often, like, if I want to get promoted, I need to leave and then come back. And I'm like, that feels like a market failure. It feels like there should have been a way. Or, you know, you know, they were. Google had a bunch of early employees. Paul Beike, Paul Buchite creates Gmail. Like, can he get a couple extra points? Because that drove so much more. But like, the structure of these companies just doesn't really allow for that. And I don't know if that's good or bad. I mean, it all works out great.

But to your point, look, I think if everyone kind of gets the same thing, then it gets distributed to people that didn't create Gmail. Also, I would presume that there are some people at Google who were early who did not contribute that much, who kind of just like benefited from the largesse, you know, and that's effectively Paul Buckeye, like paying them some of the money he should have gotten. You know. And so I think that, like, you know, look, I think people, I am interested to see how this ends up playing out over the next few years of like, does this power law start to apply more frequently? Of, okay, you know, does every company start to think about the 20 to 30 people inside of the company that they're like, oh my gosh, what would I do if they left? That would be terrible, right? And make sure that they're paid super differentially, you know, does that start to happen? I think that the thing that, to your point though, John, that's interesting is I'll give you an example. For me, you know, when I was at KKR a long time ago, I think that I was actually pretty wrong on the value I was creating relative to the value the platform created. I think the value the platform created was actually a lot more than the value I created. And I think it actually took me leaving to realize how valuable the platform was. And so I think on some of these, there are people that actually probably have been overpaid because the platform that they were a part of is what created the value rather than them. And then there's people that are dramatically underpaid. I think one of the big lessons for me is like, the range is probably pretty wide of like, there are people that these places that have been dramatically underpaid, even if they've made a lot. And there's other people that have been dramatically overpaid, you know, even if they made a little less. And hopefully the market that ends up getting created over time is that people are paid more like in line with their unique contributions relative to the platform.

Yeah, that is a great take and that makes a ton of sense. What do you think the recent kind of zombie Aqui hires does to the earlier stage venture market? If I had to concoct a strategy that feels like it could not fail, it would be to try and write pre seed checks into just everyone with amazing AI research papers on the resume and super high IQ teams because every lab is just going to hoover up all these call options at some point. And I just have such a floor to my investment that even if the product doesn't work, I'm going to do great. Is that reasonable? Is that, is this distorting the VC market?

What's John's strategy? Spray and pray?

No, no, no, it's not spray and pray. It is select for a very different criteria than traditional venture capital, which is what is the market, what is the product? Don't build a. Who cares about the product? Don't even tell me, tell me, tell me your resume and have you been getting calls from Mark Zuckerberg? Because if you have, I'll get in it. I'll get into 10 million pre or something. Who knows knows. I don't even know if that's possible.

Good luck getting it.

No, no, I get it, I get it. I think maybe there was a conversation for the landscape and there's a conversation from Sequoia. Like I think at Sequoia it's pretty straightforward. Like, you know, people want to help founders build incredible companies, right? Like that is the reason people are here. The mission statement is, you know, we help the daring build legendary companies. It doesn't, it just like doesn't compute to try and like do it another way. And I'll tell you, like, just truly there's like this real currency. It's a COIA of like, do you have an investment that's gone up on the wall of like the main conference room and like it's not going up on the wall if you sort of did it the other way, you know what I'm saying? And I think that, so I think culture stuff matters a lot there. I actually think what you're saying though, like, will happen. I think people will try to figure out like every gap in venture gets closed so quickly, right? Because it's intensely competitive and there are really good people going after these things. So I think that exactly what you just said is going to be written down by some young person right now. They're going to go and mobilize against it. They're going to create some sort of ranking system of how many likelihood of being acquired, acqui hired and I think that that'll get pushed out and those prices will move from whatever the number is now for companies like that to higher. I do think though, at Seed, my impression, and I spend most of my time in growth, is that like I do think that people are actually far more important than the idea because people move around a lot on their ideas anyway. And so I think the thing you're suggesting is acquireability rather than just like the reason people matter right now in Seed is this idea of like, are you going to stick with it and build something and figure, take in the data and figure out something amazing. I think your point is there might be another metric that people look at on like acquirability and I suspect that'll happen. It won't happen at Sequoia, but it'll happen somewhere.

What's the criteria for the wall again? Because I have a strategy. I'm going to join Sequoia as an associate, rip 500k into secondary buying a figma right before the IPO and put me on that wall. I got a pre IPO company.

Listen, this is a good question. Investor is not on the wall. That would be meaningless. It's the company names.

Ravi's face.

Massive.

It is the company name on the wall. So, you know, you'd be associated with bigma, but I think folks would know that Dylan and the team deserve the credit and Andrew made the investment in first place. But, but you, I don't know, you have a big media platform so you might be able to go out and kind of take the credit for it at some point.

You should take John.

Take credit, you know.

Yeah, yeah, yeah.

I think Andrew was on TVPN and sort of looked like he might be an AI himself. Right. With that like blue background, if I recall correctly.

Yes, he was verticalizing blue that day. He was blue on blue.

Switching. Switching gears slightly. I. There's all this excitement and potential around consumer agents and when they talk about agents, they say you're going to be chatting with an agent. You're going to say, go get me groceries or book me a car, book me a flight, book me a hotel. And then when you think about what a human would do to do those things today, they might go to Instacart and they might browse around. They might see some ads, they might go to Uber and browse around, see some ads, they might go somewhere else and see ads. And in a world where agents are sort of browsing and sort of seeing, doing the things that a human would traditionally do, there's been this question that a lot of people have brought up around. A lot of platforms have a lot of meaningful amount of revenue coming from advertising and then an equally meaningful or potentially more meaningful amount of their actual profit. Given that you were at Instacart, how are you imagining kind of the agent market playing out and the sort of economic model for the Internet broadly, in a world of agents?

Yeah, well, so I think that it's a good question. I think the most pronounced, like, use of agents today, I would suspect, is in customer experience. Right. And so let me come back to the consumer side in a second, but I think I'm on the board of Sierra, which is Brett Taylor and Clay Bavor's company, and I think you see it already of, like, the resolution rate that they have at Sierra for the customers that they serve is incredible. And so you already have people, human beings interacting with agents to solve their problems and to solve more complicated problems than they've otherwise had and to do it honestly, like, at any time of night and with no wait time. So I think that the promise of agents is quite amazing, and I think you can actually already see it in certain things. So that's cool. I think to your question on sort of like, what happens in the consumer world and what happens to advertising, the honest answer is, I don't know. Right. I think that the thing that I find to be most compelling is a little bit of what I think Andre Karapathi said on X at some point, which is sort of like the Internet right now is built for humans, and over time it'll be built for agents. And I think that as that happens, I do think that there will be new ad units, new ad models, and I don't think I know what those are going to be yet, but I think that there will be a complete rewriting of, like, how is Instacart or other consumer apps written? Because they will need to appeal to two different audiences. One is agents, one is humans. And I just don't know that the rate of change on that and how it will happen. I do think that, you know, today on grocery shopping, for instance, a very high percentage of grocery shopping still happens, like, not even online. It happens offline in the store. Right. It happens like, the way advertising happens in a store is like End caps and side caps, you know, and then you have online it happens, and it's much more dynamic, and then you'll have for agents. So I just think there'll be, like, three different approaches to ads, some of which is, like, offline, some of which is online to humans, and some of which is to agents. But over time, Jordi, I think that the entire way the Internet's written will be changed and written for agents, but I don't know that I have a great prediction on how that will work, other than it'll be way more dynamic. If you imagine, like right now in the store, the end caps and side caps don't change, but for every week, the ads on Instacart change very frequently. The ads for agents. The ads on Instacart for humans change very frequently. The ads for agents, I would presume, are like, incredibly dynamic, changing by the minute or something.

Yeah, yeah. It's also interesting. Maybe, maybe it doesn't look dramatically different. Maybe companies just have to pay more attention to supporting their APIs and it just looks very similar to how it's looked the last.

Well, the other thing is, I think this is where, like, this interpretability is a big deal. Just like knowing why an agent is doing what it's doing. I actually, one of my favorite things when I use the AI tools is just to, like, click on the details portion and to see what it's actually doing. You know what I'm saying? Like, I. I love seeing that. I love seeing why it's doing what it's doing, how it's getting to the answer that it's getting to. And I think that companies will need to get really good at understanding why an agent's doing what it's doing. I think that's actually probably one of the most practical changes that will happen, is there will be real teams at companies trying to discern why an agent is going through the behavior that it's going through and will build its products accordingly.

Yeah, I have a question about just like, the broader economic climate and how Sequoia, what the process is to take a temperature check on the health of the broad economy. The 2008 famous RIP Good Times memo. Fantastically deep with economic analysis and. And looking at credit rate defaults on consumer credit cards and loans and stuff. Last week we were kind of like, oh, there's too many top signals. When we just had Lucas from CO2 on. The East ME west conference was extremely bullish. And then JP Morgan just beat earnings. The health of the American economy seems undefeated. But I'm less interested in.

Lists of like 15, 15 or 20.

Most of them were funny. That were kind of funny profile pictures coming back or something, all these random things. But. But I guess my question is like, what is the, what is the temperature checking function look like at Sequoia? What. What do you actually, like, how often do you do this? How do you think about it? How do you actually build a thesis? And how does that relate to like the venture strategy?

Yeah, great question. Look, I think RIP good times, which as you noted 2008, I think is notable because it's rare, right? It is very rare that I think we have a macro view that is worth sharing with a bunch of companies that we're lucky enough to work with. I think the reason that that's relevant is like, we're not macro investors, right? We make our returns by investing in great companies kind of independent of the time that they're created. Like, it's like a trope, but it's discussed all the time. Like, I think, you know, Google was started during like 1999, and 1999 is often discussed as like a bad period for venture. So I think that one of the big things that happens for us is like, hey, whatever you think is going on in the outside world, go meet the best companies, go meet the most interesting people. Don't let the macro stop you from going and doing your job at the early stage. Do you see what I'm saying? I actually think that's like a huge deal now. I do think, John, we tend, we try to be like financially oriented and we try to think about like, how far ahead are we paying, right? And I think that the thing that the mechanism for that is like, we do try to like, look at where multiples are over some period of time. Where are we, how hot are they? And just to understand, like, you know, are we paying three years ahead? Are we paying four years ahead? Are we paying two years ahead? What are we doing? I think the thing that I would tell you is it always comes back to company quality. Because effectively, like, the faster the growth, the more you believe in a future that looks super different than today, the less the multiple matters, you know what I'm saying? Because ultimately what you're betting on is this big future. And so I think that what we focus a lot on is like, how big do you think it can be and how durable do you think it will be? One of the observations we have that I don't think is unique, but I think is true is the companies that ended up Being huge, huge, huge, huge. Like hundreds of billions of dollars or trillions of dollars dollars. They didn't grow at 100% the longest. They grew at 30% the longest. You know what I'm saying? Like, what they ended up doing was they compounded forever. And so, so much of what we're looking for is like, why do you believe that this will continue to compound at 30 plus percent, 10 plus years from now when it's well over a billion dollars or $5 billion or $10 billion of revenue? And I think that is what we're looking for. That's the stuff that goes up on the wall and that's the stuff that, that is basically macro independent. You know, like you could have never dreamed I could be wrong on this, but I think Meta grew at like 22% last year on like $150 billion of revenue or something. And like there's no, yeah, like there's no model that you're going to run in 2005. I'm like, well, maybe it'll add $30 billion of like high margin revenue 20 years from now. Right. But I do think that the bet you have to be making when you're in our business is sort of like, don't worry about the macro so much. Make sure that you believe that it's really going to be uniquely great.

Yeah, yeah. That's fantastic. I mean, we could leave it there. Do you have a last thing? This is fantastic. We'd love to have you back soon. This is.

You are the, you're the Levar ball of AI research to us.

Yes.

Hey guys.

We're excited for the world to recognize.

If this did nothing else but popularize, you know, sports tech and you know, friends hanging out. I'll be, I'll be super happy.

I'll be honored to come back this time next year that post goes viral because we will have shifted the discussion. We will have taught everyone the sports analogies at such a deep level that they will get the joke and they will repost. All right fellas, take a day hanging out with you. Talk to you soon.

Cheers.

Really quickly, let me tell you about wandering. Find your happy place. Find your happy place. Book a wander with inspiring views, hotel grade amenities, dreamy beds, top tier cleaning and 247 concierge service. It's a vacation home but better folks. And also you go over to getbezel.com your bezel concierge is available now to source you any watch on the planet. Seriously, any watch. And we have our next guest coming in to the Studio Adam from Chariot Defense announcing the company. Welcome to the stream. Adam, how are you doing?

Where are you doing? Great. Yeah, Calling in live from Detroit here, Longtime listener, first time caller, big fan of the show.

Fantastic.

Are you in town for reindustrialize or are you. You're not a Detroit native?

Yeah, in town for re industrialized. We just launched on stage a couple hours ago.

Congratulations. Let's ring the gong for a minute.

Hit the gong for company announcement.

Hit the gong. Enough. This show congratulation. Break it down for us. Introduce the company, explain what you do and how and then we'll go into some a bunch of questions and stuff. Awesome.

Yeah. So Cherry Defense is solving a problem that I faced as the counter US program manager at Anduril where we constantly ran into challenges fielding all these new drones, electronic warfare systems, edge compute systems at the edge in expeditionary contested environments, constantly ran into power as a loading factor. I was also the head of product Archer, worked at Uber Elevate, Kitty Hawk. Saw really advanced commercial technology coming out of the ev EV tall industries, high voltage voltage, looking at my batteries, advanced power electronics and saw DoD stuck with low voltage lead acid batteries, massive generators. And so what we're doing at Chariot is building advanced power systems to power the next generation of military technology in austere environments.

What's the core business model? Are you going to sell to the defense primes to countries? Are you going to go to program of record? How does the business model work?

Yeah, the great thing about this company is we have a lot of different business models that we can pursue simultaneously. So we can sell directly to the government to provide power systems that can bolt onto their existing vehicles. We can sell to companies and other OEMs who are developing new vehicles and help them turn those into hybrid electric, basically rolling power stations that can provide all of the equipment you need to put on those to make them survivable. And we can sell directly to companies and we can sell alongside companies. So if someone's developing a laser system, that system's gonna be a lot more effective at the edge with one of our power systems alongside it. So got a flexible business model to government, to business and international.

Can you help me kind of understand the general scale of the amount of energy that you're aiming to put out? Kind of the band of Tuktuk. Doug over at Radian, he's targeting the, the 1 megawatt diesel reactor, diesel generator with a nuclear reactor. How, what are we talking about and what is relevant at that kind of power band?

Yeah, so we're in the kind of 50 kilowatt range is kind of our sweet spot. So really what we're solving for is during counterinsurgency in Iraq and Afghanistan, we were operating either from fixed bases with massive infrastructure, or we were doing kind of short patrols at the edge, where really all you needed at the individual unit level was a radio and a rifle. And so you kind of have this power gap in the middle for expeditionary environments, for things like directed energy, you know, so a company like Epirus, a company like Aurelius developing these systems that require 30, 40, 50 kilowatts of power. The existing military platforms we have can't output that kind of power. But the power system needs to be able to deploy down onto a vehicle that can actually be mobile, can fight in a distributed environment.

Yeah, this feels hyper relevant to some of the reporting I was reading about kind of the Ukraine, Russia stalemate. Apparently it's just like drone on drone warfare every single day at a complete, it's like World War I level, like stalemate and just like almost trench warfare, but just drone based. And so I imagine if you want to deploy a new anti drone system on the edge, you need to bring it in and you need to bring in power. Power. What is the alternative? Would people bring in some sort of diesel generator or, I don't know, fire logs or something? What's the state of the art? And then I want to walk through a little bit about like how your solutions better.

Yeah. So the state of the art around power today is either massive diesel generators. They're good at turning fuel into electricity, but ultimately they're, they're very large, they take up a lot of space. Not very mobile with, with lighter weight, more distributed forces. They're not very reliable. Those generators fail all the time and then all of your systems go down. They waste a lot of energy. And so if anytime you're running at less than your peak load, what you see at the edge is highly variable loads. And your generator is not going to perform on that environment. It's going to throw away a lot of energy. And most critically, the challenge is the signature management. So that generator puts off a massive thermal and acoustic signature. And what you've seen in Ukraine is, is people don't use generators within 30 km of the front line because they know that that thermal signature is going to light up like a Christmas tree on an IR drone and be immediately targeted.

That's fascinating. That makes a ton of sense. Now, so how, like how if we, if we comp this to the Enterprise HR platform, Market, obviously you're starting with something like that. Looks kind of like a point solution. Not to use the wrong term or something, but you know, it's this one product. Talk to me about how you think about is this like we want to add on solar modules soon and then we're going to add on wind farm connections or you know, talk to a nuclear company about partnering there or is it more about different form factors in battery technology? Like what are the different vectors that you plan to kind of expand upon versus partner on?

Yeah, so if you think about the name Chariot, it kind of comes from. In 2000 BC, the Sumerians used another source of power besides manpower for the first time. Horsepower. So we talked about earlier on the pod. You know, the Horse is the first autonomous vehicle.

Yes, we love horse.

Horse is the first power source for the military. And so we saw a transition from horsepower to steam power. From steam power to the internal combustion engine. Right. Both really transform military operations. We see a similar transition going from pure combustion engines to high voltage hybrid systems. And we want to build the power infrastructure for that. So what we want to build is the energy storage, the batteries, the power conversion and the power electronics and then the power control layer that actually manages power at the edge. Today someone will go plug in a coffee pot and it'll brown out the air defense radar because there's no logic sitting over these grids at the edge.

Okay.

Honestly, sometimes coffee can be more important than air defense in war. I do gotta say caffeine, it is pretty key.

Yeah, yeah, yeah.

I mean, so you might want to write that. You might want to write the logic to be like, this guy needs a cup of joe.

Yeah. Let him bring down the air defense just for five minutes so he can get. But maybe that's a plug for Yerba mate. You don't need to plug in your coffee. Coffee. If you're drinking an energy drink, it's already made for you.

That's right. When did you, when did you start the company? When did you, when did you actually start the company?

So I started the company last fall. We raised our seed from General Catalyst and xyz, working with really great investors, really happy to have those guys on the team and yeah, raised the company, the seed, back in the fall. Within three months we were at our first exercise powering lasers, powering electronic warfare systems. And within six months we were at jrtc. So when, when, when Secretary Dran Diskell and General George chief staff, the army were on the pod.

Yeah.

They said they were going to go to that exercise the next week.

You were there?

We were.

No way. That's amazing. I was going to ask about that army modernization. That's amazing.

So that was six months from the first check into the company. We were air assaulting equipment into a force on force exercise. Had the opportunity to brief Dr. Alex Miller, General George at that event, show them what we were doing, supporting the warfighter.

That's really cool. Sorry to go back to a technical question. Signatures from batteries essentially. Like I've heard, like Tesla's put off emf. I imagine that there's, you know, you talk about rare earth magnets. Like there's a lot of stuff going on in an electrical system. Is any of that still need to be shielded? Like what can you tell us about the future of as we transition? What's the next discussion we're going to be having about keeping troops safe?

Yeah, certainly. So a lot of it comes down to. Yeah, hiding below the noise. Right. So signature management, you know, any, anything pushing a lot of power is going to have some kind of detectable emi. With the battery, it's easier to shield it, it's easier to put under cover because there's no exhaust. Like there's from generator.

Sure.

What a generator actually is, is a giant spinning magnet that's generating an electric field that then induces or generates magnetic field which induces electric field which actually generates the power. So significantly lower EMI and I think part of survivability in the future, like how do we have our troops survive? What would I want, you know, going to the front lines in Ukraine? You want to be able to produce your own signature, you want to be able to raise the noise floor. So this, this kind of expeditionary power system also allows you to set up deception decoy type operations where you've got a bunch of different things on the battlefield that all look kind of like yourself. And so you're lowering the noise, you know, lowering your own signature, you're raising the noise floor. And then you need advanced countermeasures, things like directed energy, high powered microwave, high energy lasers. All of that really depends on kind of this, this fundamental power architecture.

Yeah.

Any insights so far from re industrialize or investor? Every time you go try to watch a talk, some investor is hounding you, trying to hand you a, a check for your Series A. Any insight so far?

It's been a great event. We love being at an event like this. There's a lot of other people developing core technology here. New countermeasures, new sensors, new drones. We really kind of see ourselves as kind of building the picks and shovels of this defense modernization effort. And so we've had a lot of great business to business conversations here. Companies who they can focus on, as Bezos says, making their beer taste better and we can kind of solve this fundamental problem for them. So it's been great business, business collaboration here, had a lot of great conversations about partnerships. It's been a great event.

Amazing.

I want to talk about some of the trade offs of dual use in the context of storytelling around your company. This feels like something that it's military equipment but it doesn't seem so super dangerous to give to an oil and gas company. It doesn't seem like as regulated as nuclear weapons or you know, missiles or something like that. This feels like it would be directly applicable in a bunch of other industrial scenarios. At the same time, as an early stage startup, you have to focus. You might not want to tell the story of hey, we're going to do this and this and this. And then all of a sudden investors are like, what are you actually doing? Like, just talk to me about your hangout with Dan Driscoll. Like you're clearly having traction there, there. What's the long term plan? How important is focusing at various points in time? How, how important is agility and the ability to soak up a DoD contract when things are good there and they have a need? Soak up an oil and gas contract when the industrial base is expanding. How do you think about the trade offs both on the business side and then the narrative and storytelling side?

Yeah, it's a great question. There's been a lot of talk around dual use. You know, we are a dual use but defense for company. We actually think that this actually enables a really interesting play when it comes to critical minerals bringing back the themes of re industrialize. We are selling to DoD first. They are the ones who are willing to pay the most for that US supply chain, for the ruggedization and for the sourcing of materials from allies. This allows those companies to have a demand signal. If you're building lithium refining, DoD is not going to buy refined lithium. They want to buy a capability. But we can say, hey, we'll buy that US source lithium into cells into packs, right? We'll deliver that as a capability to DoD be able to pay that early price premium that it's going to take to get down that cost curve. So we see ourselves as helping other companies kind of re industrialize in addition to us having that customer that really focuses us at the beginning and building a truly great Product the go to market is obviously very different as well. And so we've definitely focused the team there. But we see as we start to see some of these early contracts land the ability to focus then on disaster relief, oil and gas, mining. Right. Other off grid industries that have this critical need for power.

Potentially the best news in the re industrialized world or theme other than the conference this week has been that Apple is buying something like, like a half a billion dollars of rare earths from MP materials. From your experience, you're obviously trying to reindustrialize, probably trying to build as much as.

Oh he's re industrializing.

Oh he's re industrializing.

Yes, he's actually industrializing.

He's re industrialized. But what in the supply chain do you think that we should be having the biggest conversation about? What's the next thing that we got to focus on reshoring. Re industrializing around what's underrated? What's the thing that people should be learning about now? Everyone learned about well we don't make the iPhones here. Then people were like well we don't have the rare earth elements. Then people learn what TSMC was. What should we be talking about in the supply chain for what you do in terms of large scale American industrial capacity?

Yeah. And on the re industrialized theme, one of the best hats I saw today was industrial based.

That's great.

So yeah, of course picking all those up for us, the supply chain for batteries is a case where along with many other industries we really invented a lot of technology here and then handed it over to China, one of the biggest companies now in the world, CATL. That technology initially came from a company called A123 Systems. It was incubated in the U.S. it was developed in the U.S. massive investment from the government into that capability development. But ultimately mistimed the market, ran out of cash and was effectively bought up by and then licensed to China. And China now dominates this critical industry around lithium iron phosphate batteries. And so that was a technology invented here and handed over there. You see a similar thing with DJI. We had great companies here Skydio in the early days 3Dr. Chris Anderson and we effectively seeded that thing invented in the US to China. So it's on the, on the battery cell side and then on the battery pack manufacturing as well. Pack manufacturing is one of those kind of underappreciated parts of the battery value chain, especially building Safe packs for DoD that can pass those certification standards.

So just to clarify, the company you were talking about, Catl is That correct?

Yes.

Okay. I just looked it up. We're going to have to do a whole deep diva. This is fascinating. Never heard of this company before and I want to know way more. Obviously we can't have you tell us the whole story. So we will dig in soon. We'll hopefully have you back on the, on the show soon and we hope you enjoy the rest of reinduction.

Congratulations on the launch.

Say hello to literally everyone.

Yes, say hello to everyone. And I just want to say I can see why you've been so successful to date. You're very, very, very impressive and I'm glad that you're building this company.

Yeah. Thank you.

Thanks for having me on, guys.

We'll talk to you soon. Bye. It let's tell you about eightsleep.com get a pod 5 ultra. They have a 5 year warranty, 30 night risk free trial free returns and free shipping.

The best sleep of your life.

Podcasters sleep on eight sleep.

That's right.

And our next guest is here. We will bring him in. How you doing, Misha?

What's going on? Launch day. Welcome.

Hey guys, good to see you. Thanks for having me.

Thanks for hopping on. I saw the post going viral. I checked the calendar. We'd already booked you. It's working perfect. It's amazing. Thank you so much for coming on. Why don't you kick us off with an introduction on yourself and the company and then I have a ton of questions about how engineers are understanding code.

Yeah, sounds great. So just a quick primer on the company. My co founder Yannis And I were DeepMind researchers before this. Ourselves and our team, which is mostly from DeepMind and also for places like OpenAI and Anthropic, pioneered a lot of the large language model and reinforcement learning breakthroughs over the last decade. And we set out about a year and a half ago to build a company with the mission of, I mean it was with the mission of building super intelligence. But we had a pretty, let's say almost practical and opinionated view on it, that notability in the abstract that you had to, to really define it as both the research and the product agenda. And so we thought about, well, what would a superintelligence in an organization look like, which we think is basically going to be this omniscient oracle that knows everything about the company and can act on the user's behalf. And if you want to build towards that, we then kind of work backwards of where do you start today? And so today what we're launching is a product called Asimov which is the best code based comprehension system in the world. So it's like a code research agent that is sort of the submission oracle for engineers about their code and can answer any question that they, that they have. And that's kind of our first step.

Okay, you're clearly an Asimov fan. Maybe an easy question, but is the solution to AI safety just the three laws of robotics? Like if we just bake those into the system prompt, are we good?

You know, I think that we are actually kind of doing that already today. Like these, when you write these rubrics for language models to kind of go and evaluate, you know, whether something is behaving well or not. And similar to the Asimov, you know, conclusion, the conclusion here is that this problem is much harder and it's very easy to hack these kinds of rubrics and build systems that are not aligned. So I think the answer is a no. But it's kind of a big problem that doesn't have a very simple, I think, silver bullet answer.

Yeah, yeah. I mean, certainly as I've tussled with the AI safety narratives of the last few years, I've gone all back and forth from okay, this is a very crazy doomsday scenario, that's fantastic sci fi to okay, there are some very concrete problems that need to be addressed as these systems get rolled out. Even just from, you know, a time and attention and people being, you know, overly focused or optimizing on user seconds going wrong. There's so many things. So it's a. Yeah, it's a fascinating space.

Jordan, when did you actually start the company? And this is your first launch, Correct?

We started the company about early 2024, so a little over a year ago.

Right.

And how was that? I'm curious, the decision making process to go strike out on your own, obviously with a great team, but as a researcher, like deciding whether to stay in an environment that had massive scale and resources versus going into a more resource constrained environment. Obviously there's a lot of capital available for great companies and teams, but it's still quite a bit more resource, resource constrained than something like DeepMind.

I think that every frontier lab that's been built was built when it was kind of had enough resources and had a good team, but was definitely resource constrained relative to the kind of big incumbent. And so I think what's really interesting to myself and to the team that's been assembled here is how do you build out these sort of next AlphaGo or. Right, the next GPT. I think you. That was like the Most interesting time to be at these, you know, what are now incumbent labs, when the initial breakthrough projects that define them was just being formed. And we think there's an opportunity for kind of the new kind of alphago to be built, which is going to be embodied not in a game of Go or, you know, a simulated setting, but is going to be embodied in an extremely powerful and useful product. And so we kind of think about these two things of like, how do we set the research agenda to drive breakthroughs via product on this mission to superintelligence. So I think it's really, if a researcher is happy kind of entering like a big ship and being a kind of small part of it, which, you know, there's definitely reason to do that. It's finance, you know, it obviously pays very well. That's one path. But if you want to start kind of a new frontier lab and drive a new series of breakthroughs, I think the best place to do that are in really small focus settings.

You mentioned AlphaGo, my interpretation of that story. I was obviously very impressed that AlphaGo beat Lisa Doll. I thought that was incredible and unexpected. But I was more impressed by move 37. And I think that the reaction to move 37, this uncharacteristic, un, difficult to interpret move that seemed like an error, seemed like a blunder, turned out to be important, turned out to be critical to winning that game, showed a type of creativity and sort of a solution to the spiky intelligence. It didn't feel just like playing the best game of human go. It seemed like it was playing something different. And so my question is, is that the correct understanding or History of the AlphaGo Lisa doll match and that story? But then also, are we still waiting for a move 37 moment in LLMs?

So move 37 was actually, I was a theoretical physicist before and when I saw that, I decided to get into AI. And my co founder, Yanis was one of the key contributors to AlphaGo and was there in Seoul when that happened. And I think that you're exactly right, that this was probably, I think, still one of the most beautiful artifacts to ever be produced in AI. And that we have not actually gotten to the point where we're seeing move 37s coming out of these language models. Yeah, we're seeing like they're solving math Olympiads, they're solving coding quizzes, but we're not seeing that level of net new creativity. And that is kind of one of the guiding kind of, you know, things at this company is how do we get to These systems that start showing move 37s in the real world, like beyond kind of Math Olympiads, beyond games of go. How is it, you know, like that? I mean, I don't want engineers or like people in enterprises to see a thing that AI gives them, like be perplexed, like Lisa doll was with move 37. But I do want to strike that same sense of kind of beauty that this thing is discovering new technology in front of us. So that is definitely what we're moving towards.

Do you have a reaction to the latest GORN essay about LLM daydreaming? This concept that maybe if you run LLMs across all sorts of different ideas, pick random words, try and find connections, you can kind of brute force innovation or innovative thoughts. Because we've seen that LLMs seem like insanely high IQ math Olympiad, as you mentioned, and yet have yet to write a really novel, funny joke or come up with a new connection in between the different sciences. And it feels like we're in the spiky intelligence moment. A lot of what's produced feels kind of like very much an average of the Internet. It's, it's sometimes mid witty in many ways. It feels uninspired. And GORN was coming up with this idea that maybe there's a different solution. Did that resonate with you, did you read that? Or do you have any other ideas of how that could possibly play out?

Well, that essay definitely resonated though. As a true reinforcement learning believer, we've seen superintelligence arise multiple times now. That was the game of Go and AlphaGo, Dota 5 and AlphaStar, these projects from OpenAI and DeepMind. We're getting close to it and I think at that point if they just sunk more compute and they would have gotten super intelligent, you know, video game players more broadly. And so I think reinforcement learning, when it's set up right, it never fails. Now, the challenge is that if you're setting it up to solve Math Olympiad questions, there's no reason to believe why that would generalize to actual mathematics. It's kind of like a, you know, like a student that's really good at taking tests doesn't mean that you'll make a great mathematician. And so I think that without engaging with the real world and real world evaluations, it's really hard to build super intelligent systems in the real world. So I think this kind of benchmark maxing is a bit of an ego play. And I'm much more interested in models that are, are trained with reinforcement learning for real world stuff. And maybe they're a bit worse on the benchmarks, but users really love them and I think we'll start seeing super intelligence come out of those systems. I don't, I don't think reinforcement learning will fail us.

Do you think we need verifiable rewards in physics discoveries or something like that? Like, how can we rl against something that, like I've been. My Coogan's eval is basically. Or Coogan's benchmark is basically tell me a joke and see if I laugh. And it feels extremely hard to eval against. You have to pay a bunch of humans to sit in a comedy club or they have to pay and then you have to record the voice of the laughing or something. I don't know how I would rl against something as squishy as a joke or fundamental New insight in physics and it feels like it might be intractable. But what is your take?

I think the verification problem is kind of the most fundamental problem across all of artificial intelligence. When I was working on Gemini, I was leading reward model training and that was basically figuring out the verification question. So that is I think, the basically biggest bottleneck. And so there's kind of. Some systems have verifiable rewards, others you kind of have these rubrics. But I think fundamentally the limitation is like, what are you evaluating? So in the physics thing, you might have verifiable and even rubric rewards for physics problem sets. Physics olympiads. But getting that for actual physics work, like working very closely with physicists and seeing what their day to day is. And I don't think there are any companies that are actually doing that because it's a bit of a slog and it's unclear if it's even economically viable. But that's the sort of thing that you would need to do. You would need to make a simulation that is as close as possible to what a theoretical physicist actually does. The challenging thing is that there aren't that many theoretical physicists that you could even work with to really understand them deeply. So it is kind of a data constraint problem, but I think it's fundamentally an evaluations problem.

Yeah, that's kind of a hilarious scenario that for some of this basic science research like the it might cost like a billion dollars in compute and you cannot underwrite that if you're not coming out the other end with like a patent. That's like pretty tricky.

What was the moment internally for you guys that you felt the product was ready to launch?

I think we've been so we believe that coding is this kind of root node problem to super intelligence more broadly, just because that's how language models interact with software. That's sort of like their hands and legs. And then we were trying to figure out, well, where are we today and why, you know, like what's preventing us from building super intelligent systems? And the short of it, we kind of realized that coding agents, right, the co generation stuff is starting to work. Like we have these semi autonomous systems, but they're basically like semi autonomous systems with amnesia. Like they, you know, they forget everything, they have no context. And it's sort of like if you watch that ADAM SANDLER Movie 50 First Dates, it's kind of like that for coding, right? So every day your coding agent wakes up and knows nothing and has to learn everything from scratch. And so the fundamental thing that we felt was missing that needs to be solved is this ability to comprehend very large organizational code bases and the software and kind of systems around them and build this memory like this contextual core for agents. And I think this will sort of generalize beyond coding, right? There's sort of every single discipline in an organization is kind of context balanced. So even if we get really smart generation agents, that doesn't mean that they'll actually be useful. So I think it was kind of having that insight, building the initial product around it, seeing how it's utilized on our team on a daily basis and how our initial customers are starting to use it and seeing just having a lot of confidence that this is the big unsolved problem that, you know, that we kind of identified and are seeing kind of good momentum around.

Can you talk to me about what you're excited about on the GTM side? Are we going to be seeing a frustratingly viral cluly style ad from you? Are you going to set up channel sales like what we saw at Windsurf? Are you going to go enterprise? What, what's the most interesting to you in terms of actually getting the product adopted at scale?

So we think kind of most of like most of the problems where super intelligence will be useful, like extremely useful and valuable, is going to be within large organizational settings. Like when I was working at DeepMind in Google's largest monorepo, right? It's like this massive monorepo. It has to be like that's the kind of scale where these systems are most useful. Now obviously like most enterprises are not at that scale and so it's very much, you know, it is an enterprise product now in terms of like a go to market. You want to go for the enterprises that are early adopters. And so one of the things, you know, that we've really been doing is that as opposed to kind of a traditional SaaS that might go viral with more consumers, enterprises don't want their code and like all their proprietary data leaving their cloud. So we built it in such a way that it's just deployed on their cloud resources and are working with kind of the early adopters where that deployment is fairly straightforward. So that's kind of our go to market today.

Yeah, that makes a ton of sense. Any other questions, Rudy?

Not for now, but this is great chatting. Thanks so much for hopping on and always welcome to come on the show. When you see if there's a current thing on the timeline that you have strong feelings about, shoot us a note.

Let us know when you see a move 37 moment in anything. That's what I want to do.

You're our official Move 37 correspondent.

As soon as the first time I see a move 37, you'll be one of the first people to know.

Please. We'll put up a breaking news banner Move 37 Moment Achieved. Achieved. Thank you so much for hopping on. We'll talk to you soon.

Awesome.

All right, guys, thank you.

Cheers.

Bye. I want to talk about the bull case for nothing ever happens. OpenAI runs its entire business on sales. Salesforce and Slack. Did you see this? An unusual part of OpenAI is that everything, and I mean everything, runs on Slack. There is no email. I maybe received 10 emails in my entire time there. If you aren't organized, you will find this incredibly distracting. If you curate your channels and notifications, you can make it pretty workable. And then the follow up here from Matt is you can just vibe code a CRM in a week and they say, oh, we use Salesforce. Which is. Which is very funny. Also, Replit has been on an absolute tear. Chris Franz has a post here. Not sure how you see a chart like this and then just continue living your life. It's interesting because I feel like I've been aware of Amjad for more than half this chart probably since 2018. I've kind of been following him on X and he was just grinding it, grinding it out. And then from 10 million ARR to over a hundred in just a few quarters. Fantastic work from Amjad Massad over at Replit grinding and it's interesting. Yeah, people were always saying Replit is just like the. The hype is the hype is ahead of the revenue and then they just stayed at it. Grinded and it caught up, which is.

So speaking of crazy charts, Raj Veer.

Yes.

Who we've featured on the show before, says the Claude code command line tool is now at over 3 million weekly downloads, more than the monthly downloads of the Claude iOS app.

Wow. Yeah. This is the winning. They are winning in code. OpenAI is winning in consumer knowledge retrieval.

Yeah.

And the market's forking.

And in other news, Grab A Gun has gone public on the New York Stock Exchange under the ticker P. They went out this morning and we covered.

That like months and months ago. When we did, we read some story about. It's the president's son, correct?

Yes.

Is it Don Jr. Who's involved? I believe.

Yep.

Very cool. Well, interesting business. Been around for a long time, Right. Grab A Gun is an old company that then brought in some new investors, some new talent and then went public. Is that right?

Yeah. Founded in 2010.

Okay.

Amazon for weapons. Weapons.

Interesting.

Amazon for weapons. Yeah. But yeah, wouldn't have predicted this company would be.

Is it doing well?

I mean, it's hard to say. I mean it's down, down 23% today, but who knows where it'll shake out?

Who knows? Who knows?

Freedom.

Jeff Huber says I think about this a lot and shares a screenshot of why GitHub won. So to sum up, we won because we started at the right time and we had taste. We were there when a new paradigm was being born and we approached the problem of helping people embrace that new paradigm with a developer experience centric approach that nobody else had the capacity for or interest in. I guess the question is what is the next sea change in developer workflow and. And who will have good enough taste to make it explode in the same way? Interesting about this is that GitLab, which is an open source version of GitHub, is also a very successful company. I believe they're a public company. They might be one of the YC companies that's gone public. But interesting that something is network based and developer lock in as GitHub still had a second like a. A direct competitor. I think GitLab and GitHub are pretty similar competitors and they both were fantastic outcomes.

GitLab down 62% since IPOing in 2021, still a $7 billion company.

Seven billion. Yeah. I mean that's like if somebody's going to come to you, you're going to get the feedback. If you're founding GitLab. Oh, you're going to get enrolled by GitHub effectively.

Kind of missed. They just didn't have a good AI horse in the race, clearly.

Maybe I should pick up some AI researchers merge or something. I like that. The horse sound effects back online.

He's back. I asked Ben, do we have a horse sound effect? And he texted me, yeah, it's the button with the horse on it.

Was it? Yeah, it was correct.

I'm blind, but I've got to get on with Taipei, actually, so we got to wrap the show up there.

But we will see you tomorrow.

See you tomorrow. It has been an eventful week so far and can't wait to see what the rest of the week has in store.

We'll see you tomorrow. Leave us five stars on Apple Podcasts and Spotify. We'll talk to you soon.

We love you.

Bye.