https://www.listennotes.com/e/0992201f835744f1aadd0f3d5d5e059b

You're watching TBPN. Today is Monday, June 2, 2025. We are live from the TBPN Ultra Dome, the Temple of Technology, the Fortress of Finance, the capital of capital. How was your weekend? I saw from X, you watched the new movie Mountain Head. Break it down for us. Give us a little Geordie review.

I did. You know I don't watch a lot of movies.

This is your second movie ever after Borat was the other one you watched?

The Borat classic, the cult classic. So I felt inclined to watch this film. It released on Saturday.

And it was kind of pitched. We talked about it on Friday. It was kind of pitched as like very Silicon Valley coded, very, very techy.

So this is by Jesse Armstrong.

Sure.

Creator of Succession.

Oh, okay.

I did watch Succession. I loved it. I thought it's this blend of, you know, dark humor and a good storyline.

And.

Mountain had felt like the exact opposite. It had a really kind of, I'll give you the high level, I'll try not to give anything away.

Sure.

It's four tech entrepreneurs who are going.

To meet up, all entrepreneurs, all founders, no VCs.

It seemed like everybody was a founder, but of course they dabble in investing as well.

Sure.

As one does.

Anyway, so four of them are going for a poker weekend. I found out it was in Park City, Utah because Wander actually helped them get the house.

No way.

Fantastic house. So they all sort of like descend on this house for a poker weekend. I think their deal was no heels, no deals, and no chefs or something like that. Oh, it's like, it was like one of those.

Bros trip.

Yeah, it was boys trip. So they get together, one of them, I couldn't, I'm sure somebody's put it together better than I have, but it was like some, it felt like a combination of Evan Spiegel and Mark Zuckerberg, social media guy running this business called Tram. And he had released a generative AI tool that was so good that it was, it started as they were sort of descending on this Park City home, starting to cause global chaos.

Really?

There was like, you know, you could.

It'S like a doomer movie.

Yeah, the idea, the idea was that the deep fakes were so good that there would be a deep fake of like one tribe in Kenya attacking another tribe and then it would spark, you know, real conflict.

Wow, wait, wait, wait. So that, that screenshot, you, you, you said like, there's, at least there's a new meme format and it's Steve Carell saying he is a D-cell with crazy P-doom and zero risk tolerance.

Like you didn't put that text over that. No, that's actually from the movie. Actually Steve Carell. No way. He is a D-cell with crazy P-doom and zero risk tolerance. Wow. We'll get to that. Okay, anyway, sorry. So the movie they descend on this house.

Yeah.

Tram has launched this product and it goes on. They're basically just hanging out over the weekend as this product gets worse and worse and worse. And I pulled up some notes.

So.

Deep fakes are causing global chaos. They're using a bunch of very awkward buzzwords. You could tell it was written by people that didn't like tech entrepreneurs or bean counters or centi's. There was a funny dynamic where one of the four bros is only worth half a billion.

Oh, okay. That's rough.

At one point they go on a mountain and they write their net worths on their chest. And everybody has a B except the one guy who has just an M. Oh, okay. So it was pretty, you know, he was feeling really bad about it later.

Yeah.

Now this is just a dead giveaway. So I'm not gonna go there. Okay.

But I mean, the PDoom thing seems pretty in-group. Like it's pretty close to like, you.

Know, so, eventually, a couple years old. Eventually, there's a guy, one of the founders has some type of, Filtering technology that that detects deepfakes.

They have to do a deal.

And so the social media guy is like trying to do a deal with him. He says no. Filter guy goes to one of the other guys and says, we got to get this guy like fired basically.

Okay, interesting.

And so that's when Steve Carell goes, no, this guy's a doomer with a crazy P-Doom. And then the three other guys plot to kill the other. So then most of the movie is them trying to kill him. movie.

Yeah. It seems like it's part of that, like, I forget what they called it, like, mansion core. Have you seen this?

Yes.

Where A24 got very, very good at making a movie like Knives Out, where they basically go and rent an incredible house, and then they shoot the whole movie there. And it just takes you around from one room to the next, and you're kind of in this beautiful cinematic environment the whole time. And it seems to be an interesting Takeaway for Hollywood that it's like the higher leverage production because it has the aesthetics of a big cinematic movie and they're shooting it on your own camera. But it's one house.

It's like basically one location.

And so if you rent just one location, all of a sudden you're not like, oh yeah, our second unit is in Tokyo for the scene where Batman jumps off the thing. It's like, that is so much more expensive. Moving around is so much more expensive than just being like, we're gonna dominate this one little house for a month. Yeah.

And then we're just shoot the so that was it. It was basically one location. It was a very nice location. But ultimately it was like this weird combination of like, it was like a critique of the tech billionaires, but at the same time it was like using all this insider sort of teapot language.

Very interesting.

That would have been wildly confusing to somebody that wasn't a tech insider. I came away being like, I don't know who this was for. Like, I understand making a movie for the sort of anti-tech crowd. Or making a movie that's like Succession meets AI. But it came away sort of neither of those things for me. And it was really rough watch. By the end, I was watching purely because I wanted to be able to comment.

Interesting. Interesting. There was a review, I believe, of the movie. It's the End of the World and it's their fault. The tech bros have ascended to movie villain status by Charlie Warzel in the Atlantic. And it has a very attention-grabbing headline because it makes it sound like it's like, you know, tech bros are really bad, but then it's about the movie and I don't know, just like part of the vibe shift, we'll see. I wonder who's giving it rave reviews, I'm sure. I mean, it certainly broke through on the internet.

But anyway, it has 79% on Rotten Tomatoes.

Not too bad. Yeah.

Too bad.

Well, we should talk about Andrew Reed's latest investment. because he's investing in movies now too. Sequoia just invested $100 million into Movie at a $1 billion valuation. He's joining the board and partnering with the founder and the movie team to champion great cinema around the world. You are a movie guy.

Yeah. Are you a Movie user?

I'm not a Movie subscriber, but I'm sure I've watched stuff that they've distributed or published, but probably just on Apple TV or just purchased those films or seen them in theaters. But it sounds like Movie has a much bigger in like vision. And we kind of saw this with what's happening at A24. It seems like A24 is expanding pretty significantly. But Andrew Reid shares a little clip here. What is Mubi? A streaming service, a distributor, a publisher, a curator, a cinema lover, a community, yes. And there's a quote from the Financial Times or maybe from a different article about the company says, but K. Chirakal is the founder, isn't that the founder isn't just interested in topping the box office. He wants to reinvigorate moviegoing culture by creating an ecosystem that extends from streaming to publishing to arthouse theaters, offering movie lovers the chance to see the kind of offbeat visionary work that other studios are afraid to make. In doing so, he's intentionally creating a worldwide community of film devotees that has been neglected for too long. Very interesting. Yeah, we should get Andrew on the show and talk to him. And we have Scott Belsky coming on the show from A24. I'm not sure how much he can talk about A24 strategy right now, but I'm sure we'll get to know him and learn more about how the film industry is changing. It's obviously super relevant to us since we're, you know, trying to bring media to Hollywood and we're building a media company here. And so it's, you know, as these distribution engines change, it's obviously good to keep track of. Anyway, in other news, there's a couple of things we want to cover today. Obviously, we're going to go over the timeline and break down everything, but the big news over the weekend was the Ukraine drone attack on Russia. They shipped shipping containers into deep into Russia, at which point drones flew out of the containers and hit strategic targets. We're going to have two guests on the show today, Soren Munro Anderson from Neros to talk about that, and also Conor Love from Lightspeed to talk about that, and also Defense Tech Investing generally, and a bunch of other a bunch of other things that are going on in his world. And there are a couple other founders and folks coming on the show. Let's go to Cole Rottman and talk about, he made his own version of the minus list, which I think in some ways is potentially, I don't know, it's like in some ways even more, People have been critical.

It's the Rottman list.

The Rottman list, yeah. So he says 16 investors have led two series A rounds that became $5 billion companies since 2020. since 2012, you can see a pattern. Alfred Lin in consumer marketplaces, Andrew Mack in FinTech, Mamoon in B2B work tools, Mickey Malka and FinTech, Mitch Lasky in consumer social. And so he basically, his algorithm, so the Midas list has always been a little bit tricky because funds report the allocation of deals differently. Are you familiar with this?

Yeah, they're gaming it a little bit.

Yeah, it's a game.

It's giving sometimes a little more credit than maybe they should to a certain investor on certain deals in order to.

So if I remember correctly, the MIDAS list has a great data partner through one of the LPs that's basically an LP in every fund. And so they have a really good, they have really great insight into ownership by individual firms into individual companies and valuations of those companies. And so sometimes those marks can be a little bit frothy and it can be debatable because there could be like, you know, you could have made the Midas list for being in FTX and then the next year it's a zero. And so it doesn't feel like it has a staying power. But in general, they have a very good tie between VC firm and company and valuation and entry price too and multiple. But what they don't necessarily know is the individual who is responsible for the particular deal inside the firm because that doesn't necessarily get reported to LPs. And so what funds will do sometimes is they will kind to shift the chips around the table afterwards to give maybe the hero partner that they're trying to boost up more credit for deals that they were only tangentially involved.

If one partner led a certain deal but then had no chance of getting on the list at all, they might say, Hey, look, you're going to take one for the team.

Yeah, yeah. And then there's also the question of like, sometimes partners change from one firm to the next. who gets credit for that deal? If there were two partners at one fund that did a really amazing deal and then one of the partners leaves to go to another fund, then that partner is going to want to take that deal credit with them, but that other fund is going to say, hey, well, it's our position. And so we should now say, it was actually our guy. The guy who left, he left because he was saying no. Yeah, he's on the press release. He wanted to pass.

He wanted to pass.

Exactly. Exactly. So there's a whole bunch of funny dynamics. this is kind of an interesting methodology because it's almost simpler. It's basically just like, who's the Series A lead investor? Who joined the board? Who joined the board, which is probably easier to figure out, although even board seats change around sometimes. Because it's like, oh, your company's doing really well. Well, we're going to make our named partner of the board member instead of the person that found the deal, which is sometimes rough, but sometimes actually some investors are like, no, I don't want to be on the board of a public company, so I'm happy to hand it off to the storied partner at my firm in order so that I can go and be on more boards at the earlier stage. And so the true GOAT list here, Series A lead investors that have had three 5 billion plus outcomes since 2012, you got Doug Leone, Hamant from General Catalyst, Keith Rabois and Marc Andreessen. And then there's maybe 10 or 15 that have done two Series A lead investments in a $5 billion outcome. Now, it is kind of-- I saw.

Ilia Sukhar posted and was like, Just one more deal, then I'll be happy. He's got two from the-- oh, yeah, yeah, yeah. So he just wants to be-- clearly, you know, there's a big gap between the threes and the twos.

No, if Ilia wants to move up the rankings, In the short term, changes name to Aardvark. Aardvark suit car, because within the list of VCs that have done two $5 billion deals, it's alphabetical. And so Alfred Lin's up at the top, so you got to have two A's in front of you to jump Alfred. So Aardvark Suu Car is the ticket. I mean, that's what Bezos did with Amazon, right? He was like, I want to be first in the dictionary. So I'll just I'll use Amazon because it's an A name, it'll show up first.

Legend.

Yeah. Anyway, congrats to everyone that made the Rotman list. Very fun to see this.

This could be the start of something.

Very, yeah, it's an interesting.

It could be the start of a new list, John.

Yeah, yeah, yeah, it's an interesting methodology. Of course, it doesn't really take into account if you had three deals that were exactly 5 billion, you would be ranked higher than one person who got the 100 billion or the trillion dollar company. And so, you know, you kind of miss out that. And also it's like, this is particular to Series A lead. What if you got 20% ownership in the seed? That's probably even better, right? Yeah. Lower entry price. So there are always nuances to these lists, but it's still fun to see all the goats in one place, in one goat herd.

One herd.

And one herd. Give me the rest of the story. There we go. We got a new soundboard today. We got some more soundboard. In other news, you can save time and money with Ramp. Time is money. Save both. Easy use corporate cards, bill payments, accounting, and a whole lot more all in one place. Go to ramp.com to sign up. Andreessen Horowitz and Coastal Ventures are backing a bridge AI for doctors at a $5.3 billion valuation. Let's hear it.

Why are you clapping instead of hitting the size gong?

It's a $5 billion times I like to clap. Animal. There we go. There we go. Yeah, I mean, part of an interesting trend. Obviously application layer, an important narrative, more focus on these verticals, Harvey having a lot of attention. I heard, you know, OpenAI is at something like, you know, 80 plus, 90 plus percent penetration with chat just in terms of users, MDOW or Maodao, like the Maodao ratio. Yeah, just in terms of users, the ChatGPT app has been the runaway success in consumer, But there are still nuances to how medical data is handled.

Yeah, verticalized.

And so the vertical applications are exciting and that's certainly driving this.

Yeah, so this investment will double Bridge's valuation from only a few months ago and underscores the tech industry's interest in healthcare.

It's crazy that this is the first time I've hearing about this company and it was at a 2.75 billion valuation. There used to be when a company makes unicorn, it is big tech news. It breaks through that day. Now it's like, okay. Call us when you're at 10.

Yeah, so Abridge's CEO, Shiv Rao, a cardiologist-turned-founder, said earlier this year that part of his motivation to start the company was that his handwritten notes from patient meetings were often illegible. This inefficiency also made billing and summarizing patient interactions a nightmare. You end up feeling bad about yourself, Rao told Bloomberg in an interview in February. Everybody ends up losing. What's been a game changer is that.

With Abridge, you just walk in the room, have the conversation, and you're present, making eye contact, Very cool.

Found in 2018.

Wow. Even pre-ChatGPT, they've been working for a long time. The company initially struggled. Rao faced a wave of skepticism from his healthcare peers and doubts about the efficacy of AI tools. At one point, there we go. At one point, Rao feared the company's heartbeat was getting more and more faint. Then came ChatGPT and the rapid progress in the capabilities of generative AI tools and a bridge became an overnight success six or seven years in the making. He said, Let's go.

Amazing.

Since then, the startups raised more than $400 million in venture capital funding as investors race to back application layer AI startups that make language models like OpenAI more useful for doctors, lawyers, salespeople and other professionals. Earlier backers include IVP, a lot of Gill, Spark Capital, Bessemer, Union Square Ventures. Good to see all of those people finally having- yeah, it's good to see a lot of you getting in a deal. Yeah, exactly. You're always rooting for them and hoping that they'll come from behind and make something happen. And they did. Let's move on to David Hoyle. He says, At SpaceX, I saw oxygen snow in zero gravity. I saw methane bubbles shimmering like crystal glass and liquid oxygen swimming pools that ripple like silk. I remember sunset plasma vortices flickering through compressed data streams at Mach 25. Pink. You think it glows on its own, but really it is in celebration for those who hurdle monuments through the sky with dreams of earthly beauty everlasting.

Wow.

It's so amazing. He's like, little poetry. He's, yeah, I mean, he's clearly like writing in a way that like, you know, the LLMs cannot. He's like flexing, basically. Yeah. That's the way I see this. But it does seem like he got a pretty incredible tour of SpaceX and probably got to hang out with Elon and talk about robots and all these different world models and how Midjourney can partner with that and whatnot and all the cool things that he's doing. The interesting thing about Midjourney and David Holz is that his methodology is, obviously he's very scale pilled, very AI pilled, but he's not trying to build the average image of what's on the internet. He's just trying to make the images look good. And so he's bringing this artistry to the process and that leads to like a particular mid journey look that can be seen as, well, this all has the mid journey look, but like he's happy as long as like the mid journey look is good, I think. So interesting, interesting way. Anyway, if you're looking for design tools, go to figma.com, think bigger, build faster. Figma helps design and development teams build great products together. You can get started for free at.

Figma.Com, and they now have websites. If you want to make a website really fast, make it in Figma and just publish it. It's pretty great.

Oh, this is hilarious. So I think, wait, was this just a few days ago? Sahil from Gumroad tweeted, if I give you 1 million and you give me the 1 million back, we'll both be at 1 million ARR. And Stephen Tay says, no freaking way someone actually did this IRL. And it's a story about- yeah, really rough.

So this company Builder AI, shut down.

Oh, rough.

They faked business with Indian social media startup Versa Innovation for years to falsely inflated sales. And yeah, so basically they'd just be like, hey, we're gonna sign up for a 50 million ARR contract with you, and then you're gonna sign up with a 50 million ARR contract with us, and we'll both be at 50 million of ARR, even though, you know, it just wasn't real. So anyways, unfortunate. Steven Tay here, very savage tagging both companies.

It's pretty savage. That's rough. Anyway.

Yeah, they say in many cases products and services weren't actually provided to either company for these payments.

Oh, okay. So it was just like completely fake.

There was a contract and then there was a payment.

Yeah.

But it was effectively the same amount. Sounds like they adjusted it slightly.

But because there's always been like this take Kind of going back to like the dot-com boom around how a lot of the dot-com, like internet providers were selling to like fast-growing internet companies and they were doing kind of like equity investments and then the money would flow back to the telecom provider. It was like pretty circular and kind of created a little bit of the bubble narrative. And people were always kind of saying that about like Nvidia and like the big tech companies. But, you know, it's very different when like, yes, And like Microsoft might buy NVIDIA GPUs and then NVIDIA might buy Microsoft Excel licenses to run their business. But both of those are clear value creations and they're independent contracts. And so it's very much more like arms-length.

Yeah, the other thing is if a hyperscaler is investing compute into a startup, the startup isn't, they're getting investment dollars, spending that back with the hyperscaler They're not necessarily like using, they don't use a new investment to drive new investment, right? Yeah. The same way that like if the hyperscaler would spend, had a contract with them and they were sending it back.

That'S quite a different situation.

Yeah, much trickier, much, much trickier.

Anyway, if you want to automate compliance, manage risk, improve trust, continuously go to vanta.com Vanta's trust management platform takes 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 company. Makes me want to hit, the air horn like automated compliance, John.

I love it. I think you threw this one in here, Andrew Huberman, talking about the health indicator that no one talks about, the strong desire to work and build things, whether for the intrinsic love of the work, the rewards, or both. Yes, we need sleep and some need recreation, but drive is at least as important as any other metric.

Basically, grind set number one biomarker to be tracking.

I completely agree. And I think that we asked Brian Johnson about this and we said, Warren Buffett is in the news because he was stepping down, but had a phenomenal run, particularly from age 65 to 90 something. Those are some of his most productive and consequential years of his career when most people spend those in retirement. And also he drinks Coca-Cola and eats See's Candy and eats McDonald's and whatnot and doesn't do biohacking, right? And yet he seems as healthy as you can possibly be. And Brian Johnson was kind of like, oh, well, he's like a single outlier. But I was talking to somebody else and they were like, well, but Charlie Munger had the same diet. This person was like, I went to dinner with Charlie Munger and he had like five glasses of wine and ate the steak and ate all this stuff. And so I think that, yes, a life's work is it's not that it's a complete replacement for a healthy lifestyle. You should do both, for sure.

I think it's that the human spirit can overpower its environment and inputs.

Totally. Yeah, there's some physiological or biological explanation for this. The whole idea of like the- Being built different. Yeah, being built different. But no, I remember you hear about like the mother who lifts a car to save her baby. Yeah. Like that's kind of apocryphal, but it's also like in an intense moment of stress, the body can send a signal to really like refocus and kind of like, you're basically just like shifting around all the energy in your body to like fight this one thing, whether it's like strength or focus or something. And it appears that the brain works pretty similarly. There's some interesting case studies where traumatic brain injury victims, like they'll get hit in the head and they'll become savants. And so they'll be able to remember every single word in a book. And it's just like, why is that? And I think what-- and it's odd.

Because-- Don't tell David Center that he's going to get himself in the head.

No, but the question is, if we have that ability, why didn't we evolve to do that? And it's because we're actually fine-tuned not on just pure recall, although it's extremely impressive when you can see someone who's.

Not as valuable as It's not as.

Valuable as compression and synthesis of information across different domains. So you actually don't need to be the memorizer to really win in our society, but we do have the ability to do an immense amount of rote information. And so there's certain brain treatments and effects in the brain that can kind of shift the focus and the output and the performance of your body to something. And so you could imagine that if your If you're incredibly driven for what you're doing, your body kind of sends these signals that like, hey, we need to be on top of the game. We need to fight back all the bad stuff and deal with the endless stream of Coca-Cola that's coming into our organs. Anyway, very, very, very, very fun post by Huberman. Anyway, Dan Primack has an announcement that Mary Meeker is back. We covered her on the show early on. We did a whole deep dive based on her fantastic profile in the New Yorker. Mary, you're always welcome on the show. She has a new trends report. This time it's focused on artificial intelligence. It's over at Bond Capital. Remember Mary Meeker was on Wall Street sell side analyst for a very long time. The dot-com whisperer then went to Kleiner Perkins, eventually spun out and is running a growth equity fund called Bond Cap. And Ev Randall worked there, former friend of the show, or former guest on the show, current friend of the show. And people are being spicy in the comments. This is the most sell side thing a sell side person has ever done. Bad day for left-aligned text fans, I guess, because the deck is centered or something. But we can share some of this and we can go through some of these.

Yeah, why don't we go through the outline?

So this is for the first time, yes, I'm coordinating with some other guests. So for the first time, I'm going to be hopefully driving this. We'll see. Hopefully this will work. And I don't dox my private keys. I don't have private keys on here. So, Trends in Artificial Intelligence from Bond. Good name for a fund, Bond, James Bond. So, of course, she worked with this team.

And the deck.

There are some crazy cards.

Has crazy aura despite not even being overly designed. Yes. When Mary Meeker makes a deck, people pay attention.

It definitely has classic Wall Street investment bank aesthetics. Wait, so why don't we start with the outline?

I'll read through it. Okay, sure, sure. First section seems like change is happening faster than ever. Yes, it is. Second, AI user users plus usage plus CapEx growth is unprecedented. AI model compute costs are high, rising plus inference costs per token are falling. The performance is converging and developer usage is rising. None of this stuff should be too much of a surprise.

But yes, but they've all been kind of like vibes and whispers and hot takes issued on Dwarkesh's podcast. But now we have kind of like the Wall Street interpretation of the same trends, and it's instantiated in data, which I think is very fascinating. And so we'll go through some of this. Seems like change is happening faster than ever. Yes, it is. Developers in the leading chip makers ecosystems, the number of developers has started absolutely mooning to over 6 million. And so, plus...

This is, why wouldn't she just call this Nvidia?

Well, so this is developers on top of Nvidia.

Yeah.

So you can think about this like CUDA engineers, essentially.

Got it.

People developing on top of Nvidia as opposed to more abstract. And from 2005...

I still was just curious if there was a reason if she didn't want to call out.

Oh, why'd she say, oh, developers... There are so many of those. Just you wait. There are so many of those weird, like, she doesn't want to say this or she's saying this instead of this. A lot of it's just like Wall Street parlance, right?

But so this is okay. Yeah, she's doing it every.

Yeah.

Leading USA based LLM users.

Yeah.

Rocket up to 800.

There's obviously open AI. It's literally the open AI chart. And it's funny because code two, I believe, put out a similar report, which we can maybe go through. But they just say open AI and they have Sam Altman's face right next to it. Yeah. But yeah. So ChatGPT obviously grew very quickly. just this year, they doubled from 400 million to 800 million. It was this huge, huge, huge spike. The question is always the monetizability of those users because as large internet companies eventually reach saturation, the incremental value of the next billion decreases a lot, right? Because you're getting into developing nations and there's less propensity to spend. The advertising dollars go Further and realistically, there's just lots of countries where you can't sell $200 a month subscriptions.

Yeah, or even 20.

Or even 20. And so AI user plus usage plus CapEx growth is unprecedented. Internet versus leading USA-based LLM, total current users outside of North America. And so the internet share of total current users kind of grew slowly. And you can see that LLMs came out and they started at 50% of internet users and rocketed to 90%. And this is a big part of the AI narratives that we already have the internet.

It is the world's greatest distribution engine in history.

Yeah, and like bits move faster than atoms. And even though the internet was a bits movement, it was bound by atoms in the sense that you had to put pipes in the ground and you had to make and sell phones to get the penetration out. And now, when you're leveraging on top of that, you actually, it's like OpenAI might build a device, but they don't have to, to get on 90% of, to get 90% penetration. This is a funny one, you're gonna love this. So AI user plus usage plus CapEx growth is unprecedented. The big six USA technology company, CapEx, big six. That's what people say about big tech. They don't say Magnificent Seven. Yeah. Big six.

Big six.

And they, and she, so it's Apple, Nvidia, Microsoft, alphabet, Amazon, AWS only and meta platforms, just ex Tesla, just writing Tesla out of the mag seven to create the big six. And it's and normally you might say, okay, well, it's like a car company, but like Tesla has an AI division that is like a scale.

Yeah, they changed the name of their self driving. Yeah, from autopilot to Tesla to Tesla. Yeah, exactly.

So a little bit odd and maybe you should put that maybe Tesla isn't investing at the same level as the other big six, I suppose, but it's still just funny to coin kind of like a new to a new term. But obviously, CapEx is increasing very quickly. We track that during earnings to see that-- I mean, it's really just a few of the hyperscalers that are driving this. But there are multiple big tech companies that are investing at the $60 billion a year range now. So totally CapEx.

So, total CapEx for 2025 across the entire business was projected at just $11 billion. So, yeah, that's nothing compared to just.

A Google footprint, which is spelling like Google, Microsoft and AWS are all in like 60, 70, 80 range billion. Yeah. Got to get those numbers up.

CapEx for ants.

I mean, it is when you think about the CapEx that goes into a car plant as well. That's definitely something.

That's what I'm saying. But at the same time, Elon's getting this sort of off balance sheet with XAI.

Yeah. And so there's trends in declining cost of inference. This is interesting, 72 years that you can see the cost curve here. So the cost of it's kind of an odd metric here because we're talking about AI inference cost, which was so high to begin with. It's kind of apples and oranges, but basically electric power went through this, pretty slow driving down the cost over time. Computer memory fell off on more of a asymptotic curve. And AI inference is just basically a line straight down as it got incredibly cheap.

Let's go.

AI monetization threats, rising competition, open source momentum, China's rise, China's obviously growing in LLMs. But interestingly, it seems like went way up from zero percent penetration in February of 2024, by February of 2025, had grown to maybe 10%, but then had fallen, or maybe it was even higher, maybe it was 15%, and then it had fallen. So the pushback against deep-sea is maybe working. Interesting. China versus rest of the world here. Yeah. Yeah. I don't know where else we want to go with this.

Yeah. AI monetization threats equal rising competition plus open source momentum and China's general rise.

But you can go through the whole thing. There's a lot to dig into here. Global GDP, the computing cycles over times, the AI era. It's interesting, it's all focused on just telling a story almost to the public markets.

Yeah, this one's interesting. AI in physical world, Ramps equals fast plus data driven. Doesn't really say much, but they're charting a ride share versus autonomous taxi provider. So this is Lyft versus Waymo in the same San Francisco operating zone market. And so you can just see Waymo really running away with the market.

I was over by LAX yesterday and I saw like five Waymos all in the same block. They're really dominating LA now.

Everywhere. It feels like they're every 10th car.

Yeah, I mean, it makes sense. People prefer them. It's a better product.

Yeah, I was barbecuing with a buddy last night and he was like, yeah, I was, you know, I watched this waymo, Paul was parked in a red zone and then just like ripped out and ran a red going left. And he was like, yeah, like, you know, just me, I sort of don't fully trust them right now. I'm like, you drive around LA, that's like normal moves for people in LA. Like, you know, there's people that go out and they're like, I am gonna break a lot of traffic laws today. And they just do that. So it's like, yeah, It's not great that, yeah, it's not great, but coming along.

I, yeah, I was talking to a friend who over the weekend, who's been saying that Waymo is completely cooked for a long time, and now he's like, no, Waymo's gonna be fine, because apparently there's one guy at the company who's just insanely cracked and is basically solving all the technical problems. And so he's like, yeah, I'm actually kind of bullish on it, but he still maintains that it's, that it's pretty, it's still heavily human in the loop.

You think it's not, that AI is anonymous Indians, as Nepaul would say?

No, kind of. He says that, I think the number he quoted was, I was like, so he was saying, it's not remote controlled in the sense that it's like hands on the wheel, right? It's not like somebody's remote controlled driving, like it's a video game, but they are watching a screen and then they can press a button. to be like, okay, break or go around or throw the warning lights on or, you know, okay, this is fine, or I'll draw on the screen. They can intervene very easily.

Well, I think I said this maybe last year or early this year, if you're Waymo, it's worth $5 an hour to have somebody exclusively focused on the one car.

Yeah, to not have that many-- you- don't have that many disaster.

Yeah.

And so apparently the ratio of support humans to Waymos is less than two, but more than one. And so they've proven that it's not fully just one to one, but it's close.

I think it's like 1.7.

I think it's like 1.7.

They kind of need to make it not one to one. But even I think that pure teleoperation is cool.

Totally.

I would love to be able to pay somebody $5 an hour, $10 an hour to drive my car.

Yeah.

It's also funny because it's like everyone kind of knows that this is happening, that there's some sort of teleoperation involved at least. And most people who understand it at that level Think it's a good thing, but then you'll see like a headline about like, oh, it's actually like anonymous Indians or whatever, and it's framed as like a bad thing, which is very odd. But I've always felt like maybe Google is keeping it as an ace up their sleeves in case there's more pushback on, like, oh, this Waymo's too risky. Then they can say, well, actually we do have a human in the loop. We didn't even tell you, but we're even safer than you thought we were. Yeah. So you should, or maybe they tell it to regulators behind the scenes to.

Say, hey, we're really, really... But that also presents a whole other set of problems, which are okay. Do the teleoperated people need to get, do they have international driver's license?

Yeah, I have no idea. Well, they might not be international. They might be in the San Francisco area.

That's true.

Or they might be, yeah, Wemo HQ or something. Or Vegas.

I mean, we fly the drones remotely from Vegas. I just think we would have had one of those people leave already and be like, Whistleblow?

Yeah. Maybe, I don't know. Maybe the exit package is too delicious. You just can't get out.

You want to give up on that sweet, sweet, sweet.

This was an interesting slide for a few reasons. 260% annual growth over 15 years of data to train AI models led to an absolutely exponential increase in the size of training data sets. which feels less important now that we're in the post pre-training era, I guess. But you know who's up here right at the top? Aramco Meta Brain AI. I love that Saudi Aramco just trained a massive LLM somehow, and it didn't make it to my news until this moment. But there's a whole bunch of these here, and you can kind of see that the curve is changing. But then at a certain point, you run out of of tokens on the internet. There's some interesting stat. All of GitHub is, I think, 10 million tokens or maybe 100 million tokens. It's just not that big of a data set.

And there's pretty massive data sets you can get off Hugging Face that is almost the entire internet. That's like 44 terabytes.

Yeah, yeah.

You put it in a context, which you could get a 50 terabyte hard drive.

Yeah. And so I was talking to some folks about, Is there a world where what we're seeing with VO3 and Google's cornered resource, which is the YouTube data set, could something like that happen with Microsoft that owns GitHub? Because they have the best code data set in theory. And they're just like, well, it's not that big of a data set. It's already been exfiltrated. And so plenty of people have all of GitHub just saved locally, basically. And so it's not as much of a cornered resource as very few companies have have been able to scrape all of YouTube, even if some have maybe tried to do stuff here and there. It's clearly not the same as the resource that YouTube has. What else is interesting in here? I mean, we could go through other stuff, impact of improved algorithms on AI model performance. This is the classic like Chinchilla scaling laws. Just kind of breaking down all the different All the different trends here. But let's go back to the timeline.

And let's go back to the timelines.

I think so. Is there anything else you want to talk in here?

You're really hitting Mary with the boring.

No, there's just a lot in here. And I don't know if we have enough time to dig through each slide. We should have her on to give us the highlights.

Yeah, we should have her on.

This is pretty interesting, the AI milestone timeline. I've covered this before. 1950 Alan Turing creates his Turing test to measure computer intelligence, positing that computers could think like humans. In '56, Stanford computer scientist John McCarthy convenes the Dartmouth Conference on Artificial Intelligence, a term he coined, great coinage, artificial intelligence. Arthur Samuel in 1962, an IBM computer scientist, creates a self-learning program that proves capable of defeating A top US checkers champion, precursor to the chess stuff. 1966, Stanford researchers deploy Shaky, the first general purpose mobile robot that can reason about its own actions. Then there's an AI winter that goes from 1967 to 1996. Unclear these AI winter narratives because that was like the craziest moment.

Space went tech. True, but also- More of a moon landing winter.

Yeah, I guess. But there was an incredible amount of underlying technology built in that period that was directly correlated with artificial intelligence. Even if you just go back to linear regression got faster during that period of time. Computers got faster. This was the era of, this is like the rise of Intel. Nvidia was born in this winter. The AI winter narrative is always a little bit odd because you can always go back and draw a pretty smooth line after the fact, but in the moment it feels very flat. But anyway, the AI winter ends with Deep Blue in 1997. IBM's chess playing computer defeats Gary Kasparov, the world chess champion at the time. In 2002, Roomba, the first mass-produced autonomous robot Vacuum cleaner that can navigate homes is launched. I didn't realize Roomba has been around since '02.

'02. Wow.

It's a vintage robot. And now they got some competition with Matic. The Matic founder came on, he sent.

Us some cleaning robots and they've been working wonders.

It's actually crazy.

Toby Lutke was posting about it. Yeah, Toby was posting about it. It's amazing how my children have completely welcomed the robot into the family.

Oh yeah.

Like they say, hi Matic. Really? They chase it around while it's working. It's totally a member of the family.

That's incredible.

It makes me even more robot-pilled. It already was.

It's great. 2005 Stanford team builds a driverless car named Stanley. It completes a 132-mile course winning the DARPA Grand Challenge. That of course starts the great rivalry.

Between Stanford and Carnegie Mellon. We'll get excited about that. really get excited about Waymo when it can win Le Mans.

Yes, yes.

Like just 24 hours.

Waymo has so much money they really should get an F1 team. Yeah. It would be amazing. Yeah. And just be like, yeah, there's still a driver in there. We just bought, you know, some random team and we're winning because we're putting a lot of money into the car.

Imagine an F1 race, but it's like a roller coaster. So you're not actually like driving, you're just like experiencing the insane. Oh, yeah, I think it would. I think it would be deeply uncomfortable if you weren't properly trained.

For sure. 2010 Apple acquires Siri voice assistant and integrates it into the iPhone 4S model one year later. 2014, Eugene Goustman, a chat bot, passes the Turing test with one of three judges believing that Eugene is human. Wow. I thought we didn't even get close. I mean, that doesn't feel like passing. I thought passing would have to be greater than 50%. I don't know about that exactly, but I thought that the Turing test was basically impossible right up until GPT-3. But I don't know. I guess they were doing some good.

Stuff back in 2014.

Yeah, they never really had an official organization body, like an Olympics of that. Some recognized, we will be the ones to do the Turing test on everything. It's always just been like a philosophical test. 2018, OpenAI releases GPT-1, the first of their large language models. 2020, OpenAI releases GPT-3, an AI tool for automated conversations. Microsoft exclusively licenses the model. Yeah, that's a good summary of the Microsoft relationship. And then in 2022, OpenAI releases ChatGPT to the public. And let's go.

It became Massive.

Fantastic consumer product.

Yeah, I'm interested. I actually, now that we're here, I want to see the entire. So this is the timeline from 2023 to 2025.

So this is 70 years here on this slide, and then this is three years.

So open AI in March 23rd of 20, or sorry, March of 2023. Opening AI releases GPT-4, multimodal model capable of processing both text and images. Same month, Microsoft integrates Copilot into its 360 degree Three, six, five product suite. I remember that day. I remember exactly where I was. It was, it will echo in history.

It will echo in history. You know, the funny thing is that it actually would have if they did Clippy that day. If they had been like, we're bringing back Clippy and Clippy... And he's now God. Yeah, he's God. It's also so funny with the paperclip narrative, you know, everyone's always, oh, do you remember?

So this thing has been so long ago. So that same month, Google released Bard. Yeah, it's ChatGPT competitor. Hello, Bard. You don't hear about Bard much anymore because it was rebranded to Gemini. Same month, Anthropic releases Claude, its AI assistant focused on safety and interpretability.

Yeah.

A few months later, in November, 28 countries, including the USA, EU, and China, sign the Bletchley Declaration on AI safety. Sounds dramatic.

Of course. I think the conclusion was that they want it to be safe, not dangerous. Yeah. We all agree.

Which I think we can all agree on. March of 2024, the US Department of Homeland Security unveils its AI roadmap strategy.

I don't actually.

I don't remember that at all. It's hard to say that that was one of the most consequential moments of the last years. April 2024, Meta Platforms releases its open source Llama 3 model with 70 billion parameters.

This is so funny. So there's two stars after the open source. And if you're in tech, you would think that the stars would indicate, oh, well, the difference between open weights and open source code and open dataset. That's the nuance of open sourcing and LLM. And also, they did open source it, but they set this limit where the other hyperscalers couldn't use it. What does this say?

It says open source is AI models and tools made publicly available for use, modification, and redistribution.

Yeah.

So that's who this, I mean, this is again, this is very, yeah, very, very oriented towards the East Coast, specifically the world of finance.

But we love, we love finance and they gotta get up to see on this stuff because they're gonna be able to, they're gonna be able to invest soon in a lot of this stuff. If they're not already, they gotta learn.

So May of 2024, a year ago, OpenAI so fast. Yeah, things are moving quick. Opening AI releases GPT-4.0, which has full multimodality across audio, visual, and text inputs. Same time, Google introduces AI overviews to augmented search.

That's very important for Google because that product has grown super, super fast.

Yeah, and I will say now, I gotta give them some credit that I use AI overviews.

Totally.

Because when they pop up, they're usually pretty good. They're usually pretty good. Very effective.

And they do hallucinate and there's ways to jailbreak them and stuff, but that's accepted. But if you're not trying to do that, you can get a pretty good experience. I did see a hilarious UI mockup that was just Google has the opportunity to do the funniest thing ever, and it was just go all the way back to the original Google search bar, but just have it be a Gemini text box. And so it was the same aesthetics as the original Google, but it would just be a text box that you would put the Gemini, and it would say, Do you want to think, upload PDF? What model do you want to use? And then it would be like, send. And so it would just be actually, you go to google.com and you just get Gemini, which would be aggressive, destroy a lot of the revenue in the short term. But who knows? Maybe it happens at some point. At least it would be interesting if it happened on a per user basis. If they know that this user would retain longer and actually be better monetized by seeing a Gemini prompt, they could serve that to you. Because they have an advanced search product now that you can opt in or out of on google.com, but I think.

If they knew if the PMs really knew how much we love big tech, they would start to experiment with us more because they know we'd give them a chance to really Iterate, improve. We're not going to just try something in churn. I would never churn off Google. Out of respect. Anyways, Apple on July of last year releases Apple Intelligence, an AI system integrated into its devices.

Is that when they released it? I thought they announced it then and then released it over the next league.

Well, I think it was available in beta. Yeah, it wasn't fully released until they released the new iPhone.

Yeah, it was delayed.

On September of last year, Alibaba releases 100 open source Quan 2.5 models with performance in line with Western competitors.

Will Brown, big fan.

Yep.

Loves all the different research models.

Quan guy.

Big Quan guy.

And then December, OpenAI announces 03, its highest performing model.

Ever.

And then, Reasoning Model. We basically are in the present now. We had Deep Seek at the beginning of the year.

Yeah, Deep Seek. And yeah, Deep Seek, I guess that was.

That was really kind of during the fires. Moment for Deep Seek.

Yeah, it was. Deep Seek released R1, R10, open source reasoning models. And yeah, the O3 versus R1 moment was really crazy, 'cause R1 was so free and accessible and O3 was like really paywalled gated. And so, A lot of people's first interaction with reasoning models was R1. And so opening, I had to fire back very, very quickly. Also, Alibaba unveiled Quan 2.5 Max, which surpasses the performance of other leading models. Claude 4.0, Claude 3.5, GPT-4.0 on some reasoning tests. OpenAI releases GPT-4.5, Anthropic releases the Claude 37, Sonnet, and then ChatGPT reaches 800 million weekly users. Let's hear it for ChatGPT.

I mean, that's just, it seriously is unprecedented in so many ways.

Yeah.

10 years from now, some kid will be raising, trying to build a research company, and they'll be like, well, OpenAI started as a research lab and then became a consumer internet hit. It's like, Absolutely wild.

Anyway, there's a ton in here we could go through.

Let's go back to the timeline.

Let's go back to the timeline. I'll stop sharing.

We got a post from Dennis, demand for VO3 has been off the charts. Millions of videos have been generated in the past few days alone, now available on mobile and in more countries, including the UK. So the funny thing here with the timing of Mountainhead is that the entire plot of Mountainhead is based on a VO3 like product being so good that people can't tell the difference. And it causes global chaos. It feels like VO4 will be that good.

Well, depends on the rate limits. That would be an important plot device.

Well, that was part of the plot. It was like it starts saying, you need to contain it more. he's like, oh, the metrics are so good. We got to lift that.

Sure, sure, sure. Yeah. But I don't think the VO3 is causing global chaos with the, at the, Except I did, I did, you can go.

I thought briefly, I mean, I had seen the movie Saturday Night and then Sunday, I saw some of the protests happening in Paris. And I did think for a second this could be, you know, it's shaky video, there's a lot of chaos, smoke, it's at night, If I just glanced at it.

Yeah, I always just wonder how important is fake video? Because there are so many other tools for misinformation that have existed for decades. So just lying in text, like instantiating fake text has always been possible. Photoshop has existed.

I made it up.

Source, I made it up. Also, I mean, one of the classic misinformation things especially during riots and times of misinformation. During a lot of these protests and different moments is they will just take a photo or real video from three years ago or from a different location and be like, look, this is, you.

Know, or another classic.

This is the Los Angeles fires and you're seeing a wildfire from Montana or something.

Yeah, or you just do a freeze frame or it's an image, and then you write text that supports a different narrative than what actually happened.

Exactly. There's so many different tools in the misinformation tool chest. I agree that VO3 and generative video will be one, but it will very quickly turn into, okay, well, I need to treat this video just like I would treat just a block of text or a screenshot or an image that could have been photoshopped.

Yeah, I mean it is it is gonna be a very interesting question. I mean the We were talking with a friend of the show and he was like the worst part of Mountainhead is that it's real this idea that like tech you know, the tech elite are destroying, you know, humanity and they don't care right? Boo, boo, boo But it is, I mean, there's so much about the internet that isn't real or is that embellished in different ways.

Totally.

Anyways, speaking of social media, we have a meme from Eddie Quan. He says, I often think of this, and it is a woman saying, Thank you for ruining my life, and it is the Instagram octopus saying, I'm literally an algorithm designed to maintain your attention by learning from your behavior. and mirroring back that which consciously or not captivates you and the social worlds through which you move. I am literally one of the most fascinating tools for collective and personal shadow work ever created. That is, only if you can learn to recognize that you aren't disturbed by social media, you are disturbed by your own reflection.

So what does it mean that my Instagram Explore page is entirely bodybuilding content?

And Golden Retrievers.

Yeah, I think it's my own reflection. So it's saying, like, I could be a generational bodybuilder. Yeah, that's what I potentially am.

And I've told you that many times.

Yes, I've told you. Elite level genetics, fantastic insertions.

That's the only thing that if you said, I'm going to quit the show.

I'm going to be a bodybuilder, I would be like, yeah, that is your purpose. Yeah, 100%.

Exactly.

It's that and supercars and watches. It's really so ridiculous. Anyway, let's do an ad for 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. Let's go. I love that new soundboard. They got Linear for agents. Start building.

Do it.

Check out Linear.

Do it. We are pushing linear to the limits internally.

Yeah, we're excited. Antibiology says, gave a talk at Edge Esmeralda on the scientific history leading up to Ray Peat today. Gonna post soon. Stay tuned. Very excited.

I think I included this more as a mental note, but I want to do a deep dive on Ray Peat just because he's influenced so much of a lot of the think, the current thinking around health and we should have anabology on.

Yeah, we should have a few different people in that space come on while we do some deep dives. Yeah, exactly.

But I met anabology at this event last year.

Cool.

Friend of Justin Mares.

So yeah. Oh, back to Mary Meeker, Bouko Capital Bloke says, I asked Gemini and ChatGPT to construct a portfolio based solely on the new 300 plus slide Mary Meeker AI trend report. Here were the results. who do you think wins? And so we have Gemini says 15% Nvidia, 10% Microsoft, 10% Alphabet, 9% Amazon, 6% Meta. Key enablers and differentiated AI leaders. They got TSMC, Apple, Tesla, growth in AI apps and infrastructure. Salesforce, Oracle, pretty good company. I don't know. This seems interesting to me. What do you think? Waiting.

Waiting Meta at 6%.

That does seem low. But again, it's like all of the, I mean, there's some stat where like the Mag 7 grew revenue at something like 30% annually over the last like two quarters. And the rest of the Fortune 500, if you average all of them up, they grew at like 5% or something, 8%. And so like, yeah, yeah, you can say like, yeah, Meta feels underweight there, but you got to take it from somewhere because you, you know, like, Every point you're taking from your adding to meta has to be out of Nvidia and that's a great company or out of Microsoft and that's a great company. But yes, I agree with you. Very interesting. Anyway, let's go through a real quick overview of what we're about to talk to our first guest about. The Trojan truck, how Ukraine just made nowhere safe across 400 kilometers on the show today.

We can do it tomorrow or something.

Across 400 kilometers of Russian territory this morning, delivery trucks completed what looked like a routine stop near five Russian air bases. These Trojan trucks with cargo containers disguised as garden sheds opened upward to release clusters of first-person view quad copters into the bright morning sky. Minutes later, over 40 aircraft were burning, including irreplaceable strategic bombers that form a.

Core component of the Russian military.

So they released this image.

This had to have just come out.

Of this is like where they prepped, I think.

Yeah, I think it's like, but it's an interesting for Ukraine to be like marketing totally.

Totally. To show, hey, we did this and we're giving you enough to actually tell the story completely because we want to.

Send this whole plot extremely crafty.

It is for sure. Well, we have a great guest to discuss it with us today. We have Soren from Neros who builds drones and has been to the Ukraine and so we'll bring him into the studio and Ask him how he's doing. How you doing?

There he is.

Welcome.

Great. How are you guys?

We're good. We have new soundboards, so expect some wild cards.

Wild stuff.

But yeah, we missed you on Teal Fellow Day. I hope you're doing well. Maybe you can kick it off with just a brief overview of what you're tracking in the news. I mean, we just covered this incredible Ukraine operation. What information, how are you processing it?

Let's have him do a quick one minute intro for anybody that didn't hear him the last time and isn't familiar with your background.

That's great.

Yeah, absolutely. So I am the co-founder and CEO at Neuros. What we're doing here is building massively scalable defense systems starting with drones. My personal background is I was a professional drone racer for a long time. I've been building and flying FPV drones, first-person view drones, for about 10 years, competed with Team USA, won the world championship, started a company in drone racing, and then got really, really pulled into defense a couple of years ago. And now, Nirox has been around for about two years. I've been over to Ukraine many times. We have a lot of products deployed there, and we are now the highest rate drone production line in America. So we've been really trying to ramp production. looking at what the Ukrainians are doing and taking a lot of inspiration.

And how much of the drones right now are FPV versus fully autonomous? Is that an important distinction for what you're building?

Yeah, it is an important distinction, although the line is starting to get more blurry. The vast majority are FPV. And in Ukraine, too, the vast majority of drones in general are FPVs, and the vast majority of those are completely manually piloted.

Yeah.

I had no idea that you started a drone racing company before this. Did you ever get that to scale or was it still just like benchtop? I remember the first time I saw your facility, you were kind of hand assembling. It sounds like now the supply chain is much more robust, but what was that early experience like?

Yeah, so the company, it's actually still operating primarily. We were focused on, at the beginning we were focused on building the materials for race courses. So there weren't really There wasn't a really good place to go for serious racing pilots to buy the gates or what they called what you actually fly through. So this is a lot of actually working with the Chinese industrial base figuring out where we can go to get these better materials. And then a lot of the arbitrage was actually in the shipping, figuring out how to do shipping all over the United States of these very heavy items. without driving the cost super high. And then we moved into other types of components. We like to collaborate with kind of the top racing pilots in the world. So we make like frames and motors and other things that are really tailored to the needs of the best pilots, but it's still operating. I've passed it off now, but it was a really good experience also to see how easy it is to go on Alibaba and, you know, get something done in a matter of weeks. Basically like zero to full product in a matter of weeks versus you know, working with US suppliers that would maybe, maybe get you a sample.

Yeah, those gates are somewhat tech enabled, right? Because I've seen like the LEDs on them and do they actually have a sensor that can tell if the drone went through the circle?

So they do. That's usually on the kind of hobbyist setups. That's more on the like just start and finish gate and it's a separate system. Ours were purely just a fabric But that is a key part of it as well.

Yeah, that makes a lot of sense.

Before we dive into the news, specifically the news over the weekend, could you give us a high level overview of the history of drone warfare in Ukraine? Because I understand it's been progressing super rapidly on both sides and it'd be helpful to understand kind of the different stages.

Did they ever have like predator drones, like the Global War on Terror type drone or do they jump straight to quadcopter and kind of like leapfrog the technology?

So, you know, you've had this this Russian aggression war in Ukraine since 2014. Obviously, the full scale invasion was 2022. But even during that that period before the full scale invasion, there was some usage of drones for surveillance and dropping explosives. These are primarily still like small drones like what you're seeing now. But this was not a proliferated technology. Then when the full-scale invasion happened, within a few months, Ukrainians started thinking about all these ways that they could use inexpensive drone technology to get an asymmetric advantage. And that is where FPV drones started becoming a really, really big deal. So they pioneered really this idea of putting an explosive on a racing drone and using that as a precision strike weapon. there were instances of this happening in other places, but they really scaled it and they've really refined it. And then Russia was was much slower to take it seriously, although now they tend to in some ways out produce Ukraine and they have a much, you know, more direct line to China where most of these components are coming from. But since 2022 and FPV is just starting to get used, now it's reached an unbelievable scale. It's estimated Ukraine is going to produce four and a half million FPV drones this year. And those are ranging from, you know, ones that are this big to 15 inch propellers, fiber optic controlled drones, many different types and sizes of warheads, different configurations. And I can talk more about the drones that were used in in Operation Spiderweb as well, because those were really interesting. But what we've seen is just this vast technology landscape where new clever ideas like fiber optic are gonna be the hot thing for a few months, and then they sort of just become another tool in the tool belt, and it's just this constant arms race.

Yeah.

Talk about this attack was unique in a bunch of different ways, but is this something that had been to your knowledge, or just more generally known to be something that had been attempted multiple times, or maybe like, I'm curious to know, yeah, kind of the backstory on this type of attack, because it seems, you know, it's a massive difference to be using this technology way behind enemy lines versus using it, you know, at the front line.

Yeah.

So primarily FPV drones are used on the front line, say the kind of 30 kilometer band across the zero line. What was so unique here is that it was FPV drones, short range drones being used 4000 kilometers inside of Russia. It was this unbelievable application where you've seen Ukraine using long range one way attack drones that are going 1500 kilometers to strike targets deep inside of Russia. But here, these were small drones actually driven in on trucks, basically in the tops of shipping containers. and I don't know of any operations that were similar to this beforehand. I think it was not something they wanted to give away. And the drones were actually operating on cellular. They were not operating on local, like the, the normal low latency local radios you use for fpvs, typically. And so I think, you know, this is going to be something that a lot of people are going to look at and, and see. if you have drones that are operating on cellular, you can't really tell them apart from cell phones. That's really hard to defend against, really hard to detect. But now it's going to. Going to be part of air-based defense is thinking about drones that are operating on cellular being piloted from basically anywhere in the world.

Yeah.

Talk about. Talk about the Russian response, the immediate response to this incident from the footage that I saw, and I think most people saw that tracked it, it seemed incredibly challenging to respond to it quickly, right? By the time you could sort of organize a response, a lot of the core damage had been done. What do you think the question I think that every country is asking themselves now is how do you defend against this type of attack, whether you're at war like Ukraine and Russia are, or you're just, you know, thinking, you know, long term, Yeah, this clearly poses a.

Massive threat to critical infrastructure. I mean, being blatant, the US does not have any defenses in place that would stop this from happening. We already know there's already news stories about drones that are flying over our Air Force bases and we can't do anything about it. And I think the only approach here has to be a multi-layered system where you're looking at all of the different types of electronic warfare, and also considering things like satellite communications and cellular communications where you're basically able to turn those off on the flip of the switch, which is a huge a huge inconvenience and a huge thing to build into the infrastructure. But clearly that's going to be required.

And you mean that just just to drill down there, you mean asking Verizon or AT&T or every cell provider in a certain area to turn off, you know, cell coverage because a drone took off. You know, is, is. Or a fleet of drones. Or a fleet of drones. But either way, it's hard to tell if they're a threat or just, you know, some recreational use.

I mean, I'm sure many people on the internet will tell me why this is a very stupid idea. But when I think about this, it's clearly the Russians were not equipped to jam drones operating on cellular. That is totally possible. And we could have better cellular jammers. But if they had been able to recognize this threat immediately and turn off all the cellular networks in that Location, then it basically would have stopped this this operation completely. That's where my mind immediately goes with something like this.

Do you think they had electronic warfare or or radio blocking equipment set up at that air base? And it just you're saying it wasn't functional or it wasn't for that particular band.

It wasn't it wasn't blocking cellular. It might have been blocking low latency. I forget what was the what's the actual frequency for the radio that you were using with the heads up display.

That we were that we were playing.

It depends.

So sure. That's the that is the point. So basically you can actually see in the videos the drones take off and they have GPS. And once the attack starts, they lose GPS, which means GPS is getting jammed, which makes total sense. That's that's a common way of defending against these one way strike drones. So air bases are going to already be set up with GPS jamming. And, you know, so is Moscow, so is Kiev. Like that's common. But they're not set up to just obliterate cell phone usage on airbases. And they probably do have jamming for like control links in the 915 megahertz range, video links in the 5.8 gigahertz range. There's other common ones that are probably, they're more well equipped to jam. But I think what caught them so by surprise was the use of cellular.

I mean, we've talked about some of the different, you said there was kind of like an ensemble approach to stopping these types of attacks. Walk through some of those, because I remember seeing like, oh, we're gonna train eagles to catch the drones. We're gonna just have guys with shotguns that shoot them, or anvil drones, or nets, or electronic takeover. There's so many different approaches, and it feels like, at least in the US defense tech space, There's a whole bunch of startups going after different counter drone, counter UAS strategies. And it feels like there might be an ensemble one, but what do you expect to be the mix or the roadmap or the tech tree that we go down to prep for defense against this type of thing?

First off, our strategic bombers should be inside of hangars. I think that would be a great start. But as you mentioned, there are these really basic things like nets that do help against FPVs. you know, that's not going to last very long.

Right. Sure.

I think you're going to see a, a mixed layer, especially with electronic Warfare combined with interceptors, combined with the sort of kinetic defeat, like the bullfrog system, where you have an automatic machine gun turret that's able to, to shoot drones out of the sky. The last line of defense is really guys with shotguns, and that, that can actually work against an FPV drone, but. you don't want to be relying on that. So I think it's one of the things that worries me the most is that I've never seen a jammer in the United States made by a US company that can reliably take out FPV drones. And I think we're very, very far behind in the practical application of electronic warfare, especially with these local radios. So I believe this is a huge area where we need to start investing dramatically and putting way more attention.

Yeah, okay, so I want to walk through a little bit more of this. You mentioned GPS gets jammed, then they're going over the cell network. There's probably a way to provision almost like licenses to operate on the cell network that might make it harder. I imagine that they had to effectively buy Russian SIM cards and set them up in the drones. But talk to me about the actual flight experience of flying remotely over cell. because I imagine that it's high latency, low bandwidth. So is this, like, when you fly, it's remarkable. It's super precise. I imagine this is a little bit more jerky and a little bit slower. And so you could probably take advantage of that. But can you characterize the type of flying that we might have seen if.

We were on the ground that day? Absolutely. So I spent a lot of time looking at the flight videos and the pictures of the FPV drones. And these are not normal drones. These were very special. special builds to accomplish this. So one of the key characteristics of these systems that are operating off of cellular where you do have high latency control is they are fully stabilized. So they're doing altitude and position hold, which is not actually normal for an FPV drone. Normally you're running on a firmware called Betaflight, which was developed for racing applications. And it's really precise. It flies really well to a human pilot. but it doesn't really have built in stabilization. They were using Ardupilot, which is another open source firmware, and it's used widely across drones, but not usually for racing drones or for FPVs. And so you can see that it actually shows in the onscreen display from the videos that they're doing that position hold. And you can tell it's not like a normal FPV drone where they're putting it at kind of any angle. And what was also really interesting is they set these up to not fly into the target forward like you would with a normal FPV strike drone. They descended flat onto the target and they had a camera on a gimbal that can point down. So what you would usually do is have a warhead that's pointing basically in the same direction, forward direction, as your FPV camera. And on these drones, they had two warheads that were basically pipes between the motors. That way they could descend with that camera looking straight down and descend onto the area with the fuel tanks. of those bombers. And the other advantage of that is that those drones inherently were flat and stable when they're sitting in the container. Instead of having one big round warhead on the bottom, they had these two smaller ones on the sides, which made it much easier for them to pack them flat inside the top of that shipping container.

Interesting.

And so having a relatively larger target than a FPV drone would typically had enabled that.

Right.

They just had to land kind of within a, I don't know, I don't know what the surface area was, but 20 something feet to really have the effect.

They were landing on a pretty precise location, but it was, it was, they were able to do that, especially because they knew the targets were stationary and it's pretty easy to descend, you know, onto a flat wing versus if you're going after a moving vehicle or a soldier or anything like that, descending flat onto them is going to be really, really hard, especially in that full stabilized mode.

Yeah. What do you think the pipeline is for identifying targets? Are they using satellite imagery to see that the bombers are not in hangers and then they can clock that in? Yeah.

Wasn't there, there was a treaty at some point too that required bombers to be parked outside.

Oh, really? I don't know about that. No.

I think it was, I'll look it up.

Well, that might be rolling back soon.

No, I think it did get abandoned already, but there was some historical precedent for nuclear assets needing to be visible.

Yeah, yeah, I want to talk about target identification and then I want to talk about just essentially complete offline drone flying and targeting with computer vision and kind of closed loop, basically doing geo guesser on the fly and just popping up and realizing like, look around, okay, I'm in, I'm like, I can kind of guess that I'm a thousand miles outside of Moscow. There might be A target somewhere nearby, I'm gonna fly over there. Okay, I see, I can identify a.

Hangar, I can identify a bomber. Before we dive into that, let me give some context so I don't leave people hanging. So there was a treaty, the Strategic Arms Reduction Treaty, START I, signed in 1991 by the United States and the Soviet Union, included provisions for transparency and verification, such as placing strategic nuclear delivery vehicles like bombers at declared satellite observable locations to ensure compliance. with the treaty's limits on nuclear arsenals. So it was broadly, it was suspended by Russia in 2023, but there was still this sort of global infrastructure for storing these types of assets in a way that you wouldn't store them today if you were sort of building systems from first principles.

Yeah, so to answer your question, John. I have to imagine that satellite imagery and also just, this was a very well planned out plan for a year and a half operation. And it was clear that Ukrainian SBU was operating inside of Russia on the ground. And so it was pretty easy for them to gather that intelligence of where the bombers were and when would be the right time to strike. So I don't think that's the, that's not the main challenge here. But there was a lot of talk on X about using AI and training models to identify the planes and people saying that this was autonomous drones. To me, I watch these videos very closely. There is nothing to indicate that these were autonomous drones. It actually, to me, looks very obvious that they're fully manually piloted. And that also makes sense where you have a one-shot mission and you spent so much work to set up the cellular network or these drones that work on the cellular network, it would not make sense to trust that to some kind of unproven terminal guidance on a completely new set of targets.

Totally.

Where we will get to is drones that are able to accomplish missions like this completely autonomously. And what you were kind of alluding to is like this GPS denied navigation world that's getting a lot of attention right now. where basically you have your known map and then you have what the camera sees and you're able to match those together and say, look, this is where I am. Typically you need to do that in a pretty confined space because you have to preload all of that map data.

Sure.

But that's really, really useful, especially on these one-way strike drones, the larger drones, when you're trying to do deep strike and hit a precise target and you've accepted the fact that GPS is just going to be completely useless because it's.

Going cellular as well, right? Right. So total signal jamming, but you maybe don't have to load all of Google Street View from all over the world, but you could load in just this thousand-mile region in Russia, and then you kind of know that we're going to start here, so you need to be grounding, and then you can compress that ideally with some AI maybe, or just some general compression to load as much as possible on the drone. But I'm wondering about if we're, If we can't do that without an Nvidia GPU on board, that's going to change the weight and cost and all the economic equations around this. But it sounds like we're maybe close to that already.

I can speak from the nearest perspective, which is we are specing all of our autonomy to work on a computer that doesn't kind of ruin the inherent nature of an FPV drone. It needs to be small enough, it needs to be low enough cost. that doesn't tend to be an NVIDIA GPU for the systems we build. But you can still do a lot with that. And there's also these much more traditional forms of navigation. You know, inertial navigation has been around for a very long time and it works quite well. And so what you want to do is actually combine these different things together. And it's going to depend on the user and the doctrine if they will be okay with drones just fully autonomously going after targets. But it's we are getting very close to the point where you could have a low cost FPV style drone that's completing a mission basically on its own, flying to a certain area, scanning for targets, identifying those targets, and then basically just saying, like, go or no go, and then you just have to click a button. That's that's what I think we're actually approaching quite quickly. Yeah.

Are there international treaties around that? I mean, it feels like the the decision to maybe not destroy a military asset that doesn't have a human on board, but certainly to take a human life that feels like a pretty distinct Rubicon that would be discussed in the global order at the UN before it happens. And yet it does seem inevitable as technology progresses. But where are we in that type of discussion and Geneva Convention stuff?

I think the line gets blurry with AI because we've had weapons for a long time where you press the button and once it fires and it can be fine for hours, you're still never calling it back, right?

Like a cruise missile, right? Or even just like a mortar shell, right?

Exactly. And so my opinion here is that the systems that NeurUs builds and these systems that are enabled by AI and drones in general are are much more precise and cause much less casualties of civilians.

True.

And that idea, I think, is starting to proliferate. But it's heavily debated. I don't think anyone could give you a perfectly clear answer of this is exactly what everyone's agreed upon, because when you get in a situation like Ukraine, where you are just defending your territory and it is an all-out war, those things don't seem to matter as much, right? You're coming up with the most clever solution to get it done with your limited resources. And so I think it's really the only thing that matters is what is actually going to happen in a real conflict. And we can look to Ukraine for that and we can think about what potential adversaries the United States has in the future.

Yeah, it does seem like if the Ukrainians had a choice to send a couple more mortar shells and kill 50 Russian people, they would have much rather just destroyed the capital assets of these exquisite bombers that are very hard to build and are not, strictly speaking, more valuable than human life, but certainly more strategic to the war effort. So yeah, it makes sense that a precise instrument like a drone would actually favor targeting the bomber, the asset, as opposed to the human. which is potentially how good.

How are you seeing, you mentioned a little bit about how the warhead this time around was different and that it was flat so it could be packed into the truck. What's the general evolution that you're seeing on the warhead side?

So the classic image of a Ukrainian FPV drone is a 7-8-10 inch FPV drone with an RPG-7 warhead. yeah, duct tape or zip tied to the bottom and some type of homegrown initiation board to actually trigger that. And then there's a blasting cap that just goes in the back of the RPG seven. That's like the most common setup. And there's a lot of images of that for a long time now. Ukrainians and Russians have been building purpose, purpose made FPV warheads, and these can be, you know, basically from scratch. we on our systems have warheads that are very purpose-built for the intended target. And so that's really just depending on the mission, you're able to quickly swap on, swap off various different warheads. And I think that's the ideal scenario where you can support a wide array of different effects and swap them out quickly.

Can you talk about the economics between the drone and the munition? It seems like the like the disposable drone was quickly adopted. And is that a function of the fact that the warhead costs more than the drone, or they're roughly 50-50? What's the evolution been there? Because you could imagine if you wind up in a world where munition cost is driven to five bucks and the drone is still hovering around a couple hundred, well then it might make sense to release the munition and try and return the drone, even if it's somewhat low probability. Yeah.

The costs are roughly 50/50 on an FPV system. And it really depends. But I would put it roughly 50/50. Why you don't want to release the munition is because the point of an FPV drone is that it's the cheapest and most precise guidance system you can have. So as soon as you decouple that, even if it's very close to the target, you're losing precision. and it's not a great cost to actually blow up the drone.

Right.

And you we do have bigger bomber drones, which makes sense because those are much more costly and they're able to carry very, very heavy warheads. But to me, the reason why an FPV is so good is because it is actually just kind of a guidance system around a warhead.

Yeah. When you say we you mean the US Armed Forces, not near us. I mean, come and say Ukraine, the collective. Yeah, yeah, yeah. So yeah, what is in the Neuros portfolio right now? How is scaling production been? And I'm interested in specifically knowing, like, what are the downstream, what are the frustration points that you've experienced building that manufacturing line that you maybe even expect someone to build a startup around to make it easier I've been talking to a lot of defense tech friends who are employees and I've been telling them like, if you just go into the most hyper scale, the most aggressive startup, like if you want to be a founder but you crush it there, like you might discover something that needs its own business instead of like looking at a market map and trying to decide that way, like go experience the pain. So where has the pain been? Where has been the opportunity? And how has that process been scaling up the manufacturing line?

So speaking to our current products, we produce Archer, which is our fpv drone built on an allied supply chain. It's certified by the dod for. For usage and to be cyber secure and supply chain secure. We have Archer strike, which is the version of that where we actually integrate the warhead system. And then we have our various different ground stations for different use cases. Crossbow is more tactical and portable. Longbow is our max range, max anti-jam ground station. We have other things in the works, but the main focus has been scaling the production of Archer in the ground station. It was quite painful earlier this year. We went through sort of the first version of a production ramp, which I think for any company ends up being a really, really painful time. For us, it was all of our, not all, but we have all of these custom electronics that we bring in from a board house in Arizona. and we were finding these really high failure rates in some of the designs. And sometimes it wasn't even because of something we were doing. And so we had to spend a lot of time to get those components to a really high first pass yield. So for a while it was lots and lots of testing, lots of drones failing at the end of line test. Now we're in a really stable spot and that is going quite well. We're shipping very consistently about 1500 drones a month. ramping that actually very fast. On the ground station side, there's also been a lot of engineering challenges. But I think what's more interesting, what you were alluding to is the supply chain. And one of the big things that I think you'll probably hear a lot of drone and defense tech people talking about right now is the supply chain for brushless motors. We have worked with a partner outside of China to scale up their capacity, but even right now, that's still a game where, you know, we're pushing them to go faster, pushing them to make new specs, and they are not prepared for the volume that we're doing and that we're planning to be doing quite soon. So this is one that I've seen a lot of folks getting interested in. I think there's some really good efforts that are starting to appear within the US and and allied countries for making brushless motors. But there's probably going to be need to be a lot of people serving this because there's many different sizes, many different specs. One of the challenges here is also the neodymium that goes into the motors and the other raw materials are also typically controlled by China. So there's a many, many layer supply chain problem that one single drone company isn't going to fix on its own.

Yeah, I talked to a friend who is doing business with an international founder who had experience in the brushless motor industry in China and was thinking about setting up an operation in America. and was asking this other friend, go, yeah, we want to get set up, but of course we want to be where the action is, because we have a supply chain. We'd love to set up in the brushless motor district in America. Wherever the district is where all the brushless motor companies are, we'll go set up there so that we're just walking distance. So if we need a specific material, and they had to explain, no, no, no, America doesn't even have a district for that. That's not even a concept. we don't have any companies, but we also don't have the rest of the supply chain. And China's really, really done a great job of creating not just the power law outcomes like the DJIs of the world, but also all the minor supply chain companies. They're all right next to each other. So if you need some piece of equipment, you can just go across the street. It's kind of like what's happening at El Segundo right now. You guys are building it up where you can go over to Cameron at Rangeview and say, hey, can you help me with this part or something like that. But we're a lot earlier on that curve. So hopefully it's solved. I know people have flagged the brushless motor industry quite early and so people are definitely working on it. Yeah. Anyway, thank you so much for coming on.

I'm sure you're gonna have a very busy week. Yeah. I'm sure we weren't the first people.

To give you a call.

Shoot you a text. We appreciate you coming on. Ask, get your thoughts.

So thanks for coming on and breaking it down. And thanks for doing what you do. Yeah. We'll talk to you soon. Cheers. Thanks, guys. Thanks. Yeah. Quickly, let me tell you about numeral sales tax on autopilot. Spend less than five minutes per month on sales tax compliance.

What is this?

It's just a cool sound effect.

Go to numeralhq.com, put your sales tax on autopilot. Next up, we have Conor from Lightspeed coming in. He's been on the show before. We're going to get an update from him on all things in the defense tech world, what he's thinking about, how he's seeing things in the government.

Welcome to the stream, Conor. How are you doing?

I'm good.

I'm doing all right.

Good to be back, guys. Yeah.

Oh, you got a suit this time.

Oh, looking great.

I won't say I dress up just for you, but, you know, I would have taken the suit off far before.

This, you know, if I wasn't coming on. Fantastic. Good, good. Thanks so much for jumping on. Have you been tracking the Ukraine story closely? Any insights there? Anything in the portfolio that's at all relevant in the defense tech world? do you expect a response from the US government or guidance or change to any strategies? Really, any takes on that?

I mean, first shit, what a time to be alive. I mean, you know, I'm sure your Twitter feeds and your group chats were blowing up, you know, pun intended, over the weekend. I mean, it's pretty crazy. I mean, let's be honest, like, first, I'm not shocked that the Ukrainians did this. I mean, the execution seemed to be flawless from what we can pull from open source intel. I do think though, I mean, again, it's not a surprise that the Ukrainians have been mastering drone warfare for the last handful of years. And you want to call it that, they called it spiderweb. This was their Trojan horse. This was their Israeli beeper. And the outcome is pretty impressive, to be honest. I mean, from the outside looking in, the Russians woke up over the weekend and they thought they were getting their $4 Temu orders. And what did they get? They got a thousand FPV drones blowing them to smithereens. So it's pretty impressive. I mean, my takeaways from this are really twofold. Please. The first is like, there's never been a clear signal of where warfare is going. And to be clear, what, you know, what I, when I view this from, you know, both the entrepreneurs in my portfolio, but also from my perspective, I mean, the world is about, you know, cheap, attritable, a lot of times, autonomous systems. And that's, you know, playing out in warfare that's playing out in other areas of life. And then the second thing is, candidly, it's really hard to defend yourself at the pace at which things are changing. And again, I know we do some things here in the United States and are trying to be on the front end of a lot of this innovation. But when this happens, I think this almost just resets everyone again and says, All right, how do we respond to it? And I think it's, to your point, it's not a direct US response. It's more of, Hey, what do we need to buy? What do we need to develop? for our own fight in some way, shape or form?

Yeah, what do you think? Obviously, you're a venture capitalist, not a geopolitical strategist. But what's the right Russian response to this? Is it, hey, we suddenly need to be wary of having cell coverage anywhere near strategic military assets? I mean, it seems like Ukraine in Ukraine's perfect world, they could run this style of attack a bunch and copy and paste and hit other targets. but it feels like something that was dependent on cellular technology that that's something that the Russians can revoke fairly quickly. Sure, it'll be inconvenient, but I'm curious if you have a take.

Yeah, I mean, to be honest, when I think about how do you defend against this, I think there is, I wouldn't call this the easy answer of just turning off the cellular network. I actually think the only way to do it practically is in layers or in a multitude of different ways because, you know, yeah, the, you know, the reality is if you looked at how the Ukrainians carried out this attack, they did so on the local, you know, Russian cell network, which again, I don't think any Russian kind of defense unit, any of these basis was ever thinking that they would have to turn off their own cell network. And then there's just the practicality of how you do it. I mean, I think there was, what, four or five different attacks that hit all at the same time. What do you do? You turn off the network for tens of thousands, hundreds, thousands of people. And oh, by the way, this is like a dirty little secret that nobody talks about. You know, yes, you have your military systems that are protected and all that, but a lot of coordination is happening through WhatsApp, a lot of coordination on. And so all of a sudden, you turn off the cell networks, you're actually inhibiting your own defense, your own response, the, the first responder, you know, the, you know, getting your own people out of there. So I think it's a bit more complex than that. And then the last thing I'd say is just like, Even if you do this in layers, you need to be resilient in a way, but you're not going to stop everything. I mean, this was just brilliant master class of, again, if maybe there was a plan, we didn't know this, but maybe there's a plan for 100 bases and we only hit five of them. And so if you think about just the broad geographic coverage you have to have, I think to be 100% certain on anything, it's just you, It's impossible. You can't do it.

It's interesting. There was like that Huawei narrative for a while about 5G towers and potential backdoors. It was always about it. The narrative is always about spying. And I think now you have to consider sabotage, not just espionage. And the idea that if you could even just provision a tower that has some sort of just just if this SIM card comes through, just let it go through. you know, don't, don't worry about it. This, this, this SIM card can always communicate. That's, that's very scary.

So I wonder, this is a funny aside, but it's real. I mean, back when I was in Iraq, not that long ago, we would go buy, like, burner phones to be able to, like, check the news and, you know, catch up with your family on, on obviously non pertinent things. And every now and then, I was, you know, I was in Northern Iraq, and every now and then I would go just pull up my Google search on this burner phone. And my location happened to be Tehran, Iran. Almost every time that I would go to search there and I'd be like, oh, That's just weird. That's exactly where I am. So again, I think, I think it's really hard to do in practicality. I think the manner is like you just have to have resiliency. You just have to have a bunch of different options, both on the defense and the offensive side. I mean, look at what the Ukrainians are actually using for the majority of their drones in Ukraine right now. Fiber optic cable, they're not even using networks at this point too.

So can you actually, what is that in? Can you go a level deeper on what that fiber optic infrastructure?

Yeah, Soren mentioned that, but I don't actually understand, is it like a cable that flies through the air?

100%.

I mean, think about it. It's arguably unlimited amount of fiber optic, really thin filament cables that all it's doing is transmitting data to the drone. And you're inhibited arguably by how much fiber optic cable, how much kind of filament can you lay out?

So it's actually like a spiderweb that they're just drawing across the sky.

100%.

So you could have, and it's, I'm so curious to actually understand, you know, we don't have to talk about it today, but the actual mechanics, is it like effectively on a spool that's just running out?

That's literally what it looks like. I mean, I haven't been on the ground in Ukraine, but when I talk to my portfolio, you know, founders and companies that either are or have kind of partners there, I mean, you literally drive around the front lines of Ukraine right now, and all you see is just miles and miles of clumps of all this severed fiber optic cable. So again, I think that just to zoom back, I actually think the takeaway here is that the pace at innovation that's happening both in Ukraine, but let's be honest here, what we are watching in Ukraine and in Russia is a precursor to what life looks like in the Pacific. And I think that's maybe sounds a little bit doomsday in a way or not.

Other.

But if you don't think that the Chinese are watching what happened to the Russians and, and either changing their own kind of defensive plans, you know, building new kind of technology and Equipment, you're just, you're wrong.

Like, they are.

So know that for a fact.

I, I remember because it was somebody at anderal was responding to one of these posts, but a Chinese official was posting this video, a render. It was a render. They were getting roasted for using a render. But it was like this mothership style aircraft drone that was launching hundreds of drones off the side. And so that's like the scary thing is, could you have stealth aircraft at some point that can actually do the same type of deployment that these trucks did?

Yeah, like a new B-52 should be dropping FPV drones instead of just untargeted bombs. They'll just be a lot more targeted.

I mean, I think there are many-- I mean, again, I think the the framing that's happening both on, in my mind as an investor, but kindly, like also in the dod and the, in the government side, is before we used to think about deterrence just by like, hey, let's buy the big new Exquisite system. Let's buy the thing that if we stack, you know, head to head, like, again, I think I, I shared this with you guys before, but back when I used to work in the Pacific, we created this really kind of popular and widely distributed unclassified you know, slide that was literally just all of the US ships stacked up against the Chinese ships. And just like you look at the math, it's this crazy, crazy metric where, you know, grossly behind by, you know, hundreds of ships. And that's just one example. I think the frame, the mindset we had then was just like, oh, well, then if we get more systems and we get more of these things that could have these big effects, then we're going to win. We're going to deter conflict. We're going to win. I just think the paradigm has changed. Now it's about, okay, What are the means that we actually have those effects? And it's not just about, I hate to say it, buying another F-35 or buying another aircraft carrier, squids system. It's let's think about this as a parts of a whole thesis and say, okay, if I can build a missile that costs a tenth of what we used to have, if I can use a drone that costs a thousandth of what we have, but have a thousand of them, have 10,000 of them, you actually achieve, if not a better end outcome in just a different means, and by changing the modicum of how you deliver it, the enemy has to change how they defend against it. So again, I just I think that's the future of where we're going. It's not going to look just like what happened, you know, over the weekend in Russia, but it's going to rhyme.

It's going to look very similar to that. Do you think this speeds up DOD procurement in any way? I mean, it's certainly such a such a visceral and like tractable problem. We've seen Anduril working in counter UAS for years, but there's also a ton of other startups taking different approaches. We had the Secretary of the Army, Dan Durschlag on, and he was mentioning that they are ready to see demos from even early stage startups. They want to see things. And you think about not even in great power competition with China, but just if there's an army base abroad, they are going to want to fight against this. And they're going to want a really robust ensemble of counter-UAS technologies.

Well, yeah, why don't you give us, I'm curious how you think of the market map of counter-UAS technology. Totally. because we've had Steve Simoneon from Allen Control Systems. That can be a great solution if you're on a battlefield, right? But if you can't protect against an NFL stadium for counterterrorism by shooting rounds into the air in the middle of a city, right? Unless you're in Texas.

I think the way to, back to kind of my defense in multiple layers comment. I think the way you look at this market is almost like a spectrum. On one end of the spectrum, you have what I would define as like the non-kinetic stuff that in a weird way can be kinetic. But this is your EW jamming, this is your laser-based systems. And then on the other end of the spectrum, you have something as simple as shooting a bullet or shooting a shotgun at something. And I think again, there's a middle ground here which you start to take both EW resilient things, but then also like, you know, you mentioned the Anvil drone from, from, you know, Anderol of a drone flying and hitting another drone. I also just think there's, there's this layer that no one else really talks about. Um, and when you look at this, you know, what happened over the weekend, there's a full, like, how would you have solved this in the first place? You can make, oh yeah, we can shut down the cell network. We can get a bunch of shotguns. We can get a bunch of, you know, different systems that shoot down systems. Or you could just do better counterintelligence and understand in some way, shape, or form. And that's hard to do, to be really, really honest, because one truck gets in, you still have an effect in some way, shape, or form. But in the end, I think the US military is not just going to buy one of these systems. They're not going to be like, ah, let's go with Andruil's Anvil. They're going to buy a Litany of these things. And if you think about just the sheer complexity of the locations at which we are at and where our critical infrastructure sits, a single solution doesn't solve the problem. I mean, you know, go talk to, you know, I know you guys were talking about the cell network earlier, but you guys should get a famous John Doyle, who's now building a business called Cape, but used to be an early Palantir guy. You know, he's building a private cell network. And, you know, part of what he's doing is, is enabling for when, you know, we call it a pace plane in the military, when you shut everything off, you still have a way to kind to communicate and talk. But yeah, in the end, I think you're going to see this litany of many buys as opposed to like one solution that solves them all.

Yeah, I wonder if there will be almost like adversarial camouflage for computer vision based models. Because I mean, I've seen situations where people were worried about facial detection. So they would wear specific makeup with triangles and it would look very odd to a human, but to a computer, it would read as just, I can't process this at all. And so imagine like-- It's crazy because.

In a way, this isn't new. Like, go back and look at World War II. Like, what did we do? We built literally wooden hounds across the channel because when people were looking or doing like, you know, it looked like it was real. Now it's just, you know, the technology is fast forward so far. But, you know, in the end, I think, again, it's one of these things where it's like, you know, there's going to be no single solution that fits them all. I do think that, and this is worth commenting, I think a lot of times when we look at we as investors or entrepreneurs building, or even just the general public, thinks that the DOD has truly autonomous systems. And the reality is, is like you have humans in the loop on 99% of what's happening. And that's not necessarily a bad thing. I just think the scary thing is like, call it our doctrine, call it our Geneva Convention law, call it whatever you want to call it. But the reality is when shit hits the fan with China, do you think they're going to have a human in the loop and making decisions of how to kill or not kill something? They're just going to kill things. And so to me, it's like we, you know, we need to be thoughtful about how we do it. We are a beautiful democracy who cares about life and kind of human nature. But at the same time, like we're not really doing autonomous systems yet. And that just makes me a little scared, to be honest.

Yeah, put this in the context of DJI and TikTok. The TikTok ban was always a little abstract. It was like, well, it's kind of brain rot for kids and that's not great, but people watch Instagram Reels and that seems maybe equally as bad or maybe fine. But maybe they're steering the algorithm to influence our... It was all very like four steps away from something bad, and so it kind of didn't catch ground and didn't really get off the ground. But DJI has been in the news as a potential banned target. This seems like it concretizes the feeling of danger so much more intimately than TikTok, where it was like, oh, well, like the, you know, the CCP could find out that you are into luxury sports cars or something, and they could do something with that. They could blackmail you. But that's very different than drones explode and so maybe we don't want drones. So do you have any take on that? I think even more important, it's like.

Look who's using those drones. I mean again, this is what we did. I mean, I was in the military not too long ago, but I remember on some of my first training exercises, this new team, and I don't even know who they were off the top of my head, but they came to us and they're like, hey, we're gonna try this new thing that's drone warfare. I was like, oh shit, we developed a new smaller predator or something like that. they would have just pulled the DJI out of a box and they said, Hey, go screw around with this in the woods. And like we did, it was like super cool. We innovated on some stuff. But I think you're right. It gets super scary when we're taking the feeds and the data and the information from that drone, you know, not necessarily touching it, putting it into our classified system, because there are controls for that. But again, even some of the, you know, again, you should talk to, you know, Soren, you should talk to the Orca guys, you should talk to, you know, Andy from Vector about like, There's a lot of open source tools that are actually used as I think everything from navigating the fight computers of these drones. But even in those situations, we just have to be thoughtful. Again, I am on the side of, again, when I was in the military, everyone gets so scared about classifications and no one wanted to get in trouble for this reason or the other. The reality is if you're so scared about using some type of tech, then you're going to lose. It's going to be the different way. It's not going to be the information leakage way. So there's some middle ground of like, we need to just set a policy and hopefully the policy is broad enough. And then, oh, by the way, hopefully US builders are building at the same time so we can build a capacity that actually meets it. But the story is not clear here in the future other than we need a US version of that to be able to operate in similar ways.

Yeah, the capacity issue is so interesting because it's like even if DJI is Just a friendly consumer drone company by virtue of buying $100 million worth like, like demand pull this massive demand signal to scale the supply chain Results in an industrial capacity is dual use, even if the products that reach America are never dual use.

I mean, this is, you know, again, I said I'd put on a suit for you, but I was talking with that, with that, with the member this morning, a member of Congress this morning. And this is actually what we spent 90% of our time talking about, which is just okay, say we all agree and say the DOD in Congress fixes our acquisition pathways, say these are the right things to buy, you can't just snap your fingers and build these things. I mean, this takes time. I mean, go look at what Anduril is doing in Ohio. I mean, I think they're doing it at a pace better than anybody else. But like, you know, same thing with Saronic on the boat building side too, but it takes time, you know what I mean? You can't just have thousands of systems tomorrow. You need to make decisions today that are doing everything from the supply chain to, again, Congress gets a little scared sometimes, or the DoD gets a little scared sometimes, like, well, if you're not a prime, we can't trust you to give you a bunch of money up front to develop it. And finally, that's changing, that culture is changing, and saying, actually giving $100 million to Castilian or Andoril or whoever in advance of them building it is probably a 10 times better answer than giving it to a prime. But we're not there yet. We're still pretty far off.

Yeah.

Soren said they're the highest producing US manufacturer right now in their class. And they're putting up 1500 a month or something.

Yeah. But the Ukrainians built Millions. Yeah.

I mean, the Ukrainians are building thousands a day. Like, in just a. And again, it's, it's a, it's a different ecosystem. People come to me a lot and we're just like, well, isn't the us dod gonna do this? And, you know, part of me says, like, no, because I know. slow we move. The other part of me just thinks that like there is some value in, you know, what we call interoperability. And again, I think that's kind of somewhat an old school way of thinking, but like you need to know that if you give a thousands of one type of drones to, you know, one unit, that the next unit is going to fight in a similar way where, you know, the tactics and techniques kind of all align. Because in the end, like what we are talking about with these drones, What we aren't talking about with these new missiles or systems is like there's an entire downstream integration and there's a training that happens. It's not just like, okay, once we have them in the warehouse, we press a button and we win the war with China. Like, no, we need to train, integrate. And again, that's just another problem. Again, like there are some companies trying to trying to go attack this. Like Andy Ad Vector is doing this. He calls it like warfare as a service model, which I think is like really interesting. It looks more like training with with a little bit of tech. But yeah, I just think there are big gaps. And, you know, I'm optimistic about the future, but, you know, I have a, you know, a one-year-old daughter. It's like, this is what I think about. Like, this is what keeps me up at night, man.

Talk to me about Nominal, any news there? And really, like, how does that type of product and interface with, like, just the general news and tenor in DC right now?

Yeah, I mean, I, I know you guys have had cam on your show a couple times, the, the founder of nominal. I think, I think what we're, what's, what's super interesting is, like, we are seeing this wave of new Hardware development in the US, and historically, it's been in a manner and in a way that's very much like, you know, hey, here, Lockheed, here's your hundreds, millions of dollars. Go build, go iterate slowly. And now we're getting, like, you know, what looks more like the SpaceX, like build the thing in flight. This is what Castilian is doing. This is what many other companies are doing. And so there needs to be this, like software, you know, both, you know, telemetry and observation layer. But it's also just like the tools that if we want to get to this end state of having a million drones in the hands of our war fighters, having hundreds of thousands of, you know, autonomous systems, having thousands of missile systems, you kind of need to build all the infrastructure and enable it below it. That's both on the Supply Chain side, like I was talking about, but also on the Software and Development side. So I just think what you're seeing, you know, nominal is one example, but there are many other examples of this is like you're seeing the picks and shovels built out for this ecosystem. I guess there will be big winners in like the hardware categories that's owning the product. But I think there's going to be just as many winners on the downstream both, you know, infrastructure side and call it supply chain, call it software, call it whatever. And again, like I just, I'm optimistic about the market. I'm not necessarily optimistic about the outcome of our security and safety in the world, but that's my job as a venture capitalist, I guess.

It's rough.

Always paranoid. What was your reaction to the Meta andor news last week around VR? I'm sure it wasn't a surprise, but I'm also curious if you think we're going to see more, if that gives, you know, other big tech giants kind of full permission to lean into that kind of partnership.

Yeah, VR is a weird one for me personally. And again, like back in my time in service, there was this, I'd call it a research project through Army Research Laboratory, where they were trying to develop, and again, really old school, you know, bad, you know, VR at the time, but use it for like planning purposes. And the idea is like, in the military, you get out these two-dimensional maps and you look at something and say, like, hey, this is how we're going to plan. This is where forces are going to be. And it's really slow. It's obviously not optimal when you talk about terrain and kind of other things. So I was part of a unit in the Pacific, actually, where we were testing this back then. And I remember taking it to, like, the brigade commander, the guy who, I don't know, was in charge of 5,000 or so soldiers. And, you know, I have this. I still have this photo on my phone of him with these glasses on, like, like, standing there being like, what do I do with this thing? You know, and again, like, it's part of this is a cultural thing that we need to shift. But, you know, I think when I see the, the, you know, Palmer, you know, going back with Mark, and I put us, it puts a smile on my face because to me, it's like, this is the best of what we need. We need the best technology companies and the best defense technology companies working together in some way, shape or form. And the reality is where, you know, 10 years ago or more than that, when I was in and we tried this and Technology wasn't there. I know for a fact that, you know, the solution and the hardware and the software around it is 10 times better now. So I look forward to see what Andro comes out with. I mean, shit, if anybody is going to do it, Palmer is probably the right guy to do it. But, we'll see, you know, I haven't.

Seen the outcome yet. So I'm excited to give it a try. Yeah, I really just feel like if.

It'S anyone's game to win, it's Palmer. It's like the perfect merger of everything he's done in his career. No, in my view, he was going to win that category. It was just whether or not he would have to reinvent the full production capacity, reinvent the technology that he had.

Created, patented there, and give them to them. What is your take on Golden Dome? What companies or industries are you tracking related to that? Obviously, we have a missile defense system, but we're planning on ramping it up. What do you think is interesting if that project gets off the ground?

I won't bury the lead. I don't think anyone knows what Golden Dome is or isn't, which is okay. I mean, the thing that I give President Trump a lot of credit for is he has these big audacious things and he just goes and wills them into happen. And again, I'm not an expert in from inside the DOD, but like that's how you have to get a lot of stuff done. I mean, you have to almost make it not too big to fail, but put the cart before the horse in order to align both Congress and DOD support behind it. I think from a from a principled kind of like what can it do perspective, it's an absolute no brainer. I mean, again, like to look at the technology that has happened in Israel and to understand that like one Israelis developed that. But, you know, we had, you know, a say and an input in that as well, that we don't have that in the United States just seems that again, other things that make me feel uncomfortable. I think the reality is there's there's going to be a lot of different components of this that still need to be built. I mean, there's There's an entire space component. There's an entire, you know, kind of missile and effector component. I mean, in a perfect world, when we can see and know everything coming, we actually need to be able to do something to it. And if you think we have enough missiles to respond to, you know, everything we see, potentially, you're wrong. So again, I think it's, it goes back to my statement before. These things take time. And I'm glad that President Trump has these audacious goals and getting it done before his third term. I just think it's arguably, in my opinion, if done correctly, it will be a Manhattan Project-like undertaking where it's going to take multiple different departments. It's not just going to space missile defense and saying like, hey, solve this problem for us. I mean, this is going to create a new line of budget. It's going to create new kind of work streams, going to need new congressional support. The beauty is we got a bunch of awesome space companies and a bunch of defense companies that can respond to it. I just don't think anyone knows what it looks like yet. So excited and optimistic, I would say.

It's interesting. As soon as we were talking about Golden Dome and the space-based ICBM weapons, the big guys, now we need a golden spider web defense system to stop.

The local drones and do- I mean, it's layered, like I said. You know, to some extent, a missile defense system. I mean, probably not with, you know, I don't have the full specs, but probably not with the height of those drones we're kind of fine at. But you fly a one-way drone 1,000 kilometers or 3,000 kilometers, you throw it up high enough in the air, like 100% something like Golden Dome can help with that.

It feels like a missile attack, honestly.

And again, this is the last thing I'd leave you with, which is I do fundamentally believe back to my both spectrum on defense. There's also a spectrum on offensive weapons too. Everyone thinks that the answer is a $10 million missile and that's the most explicit best thing that's going to save us every time. If we can get better both, you know, on the low range, excuse me, on the low cost high performance missiles and in drone based systems, I actually don't even think we need those, you know, $100 million, $10 million systems in the end. But again, something optimistic to look forward to. It's great.

Well, thank you so much for coming on. This is fantastic conversation.

Always a pleasure.

Yeah, guys. All right, we'll chat soon. Have a great trip. Thanks, you sonner. Bye. Let's tell you about public.com investing for those who take it seriously. They got multi-asset investing, industry leading yield, They're trusted by millions, folks. And we have Melissa in the studio. She has some massive news, and we're gonna need to hit that size gong. Can you introduce yourself?

Wirts.

Welcome to the stream. How are you doing?

What's gong?

We're gonna have to hit the real size gong. So break it down. What are you announcing today, and why should we hit this gong?

Yes, we are coming out of stealth with Netik today. and announcing our $20 million fundraise from...

Woo!

Let's go!

20 million! And Mike Wolpoe amongst amazing other investors and visionary founders and growing our community to serve.

Fantastic.

Hit the wide, hit the wide.

Hit the wide. Let's go for it.

That's one of the class.

Here we go, back to background.

Let's go.

Congratulations. Thank you, thank you. Now that we've gotten the important stuff out of the way, what does the company do?

Yes, we work with the backbone of America. We're building for the essential services industries like home services. Podcasting? Yeah, exactly. You guys are essential. That's why I'm here, right? Yeah, I could be doing a lot of different stuff today. But, you know, we give them an AI revenue engine so that when they are, these are industries, even though they're super important, they're affected by labor shortages, external, circumstances and demand changes, physical infrastructure. So when they have high demand, we help them handle it so they get all the dollars and help their end users. And when they're low on demand, we help them generate demand so that you can always maximize the revenue.

What is a good example customer? We hear about these SMBs in this abstract context, but is it actually like one guy who owns an LLC who has a few employees? Is the median customer more of a small business or medium sized business with dozens or even hundreds of employees? And then I want to walk through how they actually use the product. Is it replacing or augmenting an existing kind of sales force or someone that they have internally? Or is it just kind of helping individuals do more with less?

So we work with actually pretty large customers that would be owned by private equity from hundreds of of millions of dollars to a few billion dollars, as well as smaller companies that are around like 10-20 million dollar revenue size. So across the board, we started with the larger ones to really demonstrate the impact of our platform. At the end, this is also a business. But now, actually, this is a life passion for me because I do want to help with smaller companies as well. That's how I came across this problem as a customer myself or H back in my own home. So we just announced a very large partnership with NextStar Network to be able to help a thousand companies that are more on the mid-market side as well. And for the smaller, you know what? We want to support everybody. I think we would start with some materials for them to get ready for AI before maybe they roll us out. But, you know, we want to really help across the board. So one example would be how are they using it? They're actually completely integrated with our platform. so we would use first party signals about their customers as well as third party signals like we find from weather to, you know, really property information or different types of things to help any of the customers that are coming in so they get help immediately. And these are complex jobs, right? Like there's actually quite a bit of information you need to collect to be able to deploy the right labor and then Based on that data, we predict the user's next need, you know? So I'm almost building for myself. I'm like, I would never think about these things. For me, I wish somebody told me, Melissa, you're gonna need this next month. Because, you know, I don't know, a storm is coming or San Francisco weather is again terrible, so you've got to fix this. So, and then we turn that into predicting the next need for their end users so that the companies can really cultivate the relationship with their customers, but based on need versus random promos, right?

So everyone wants their HVAC system tuned up before the big heat wave hits.

Exactly.

You can do essentially outbound sales to existing clients and potentially new clients before that hits with the correct information, the correct pitch.

Exactly.

What is the actual medium? Are you sending text messages, emails, phone calls, all of the above?

Yeah, for inbound we're really integrated across the board on all channels. So voice, text, online widgets, web chat, third party integrations because revenue doesn't come from one channel, right? It comes from all channels and you know, cut the same customer. If I'm in a meeting right now, I might be texting an essential service provider right now even though I'm busy, but a few hours later I might call them, right? On the outbound campaigns we really do see a lot of success with especially text since it's much more respectful. So we started from there and we'll be rolling out more channels.

So I'm intimately familiar with all the tooling in email, you have MailChimp and all the different AWS SES and all the different systems that have built up to make that easy. And I can imagine how you would automate that. Talk to me about automating a phone call. It feels like with Whisper, transcription and then text to speech, we're now kind of past the Turing test on that. But are you building stuff yourself? Are you training your own models? Are you partnered with other AI companies to provide that piece of the stack, or is that so integral that you're handling it yourself?

Yeah, this is actually the integral part of this is that certain technologies have passed the Turing test, but actually for really, you know, mission critical workflows like this where you are absolutely like utility and you can't drop, all the work and, you know, the engineering really focus goes to the orchestration, right? So what does it mean is that like when depending on whatever workflows that you're handling, which models are the best? When they're not the best, how can you actually fine tune only for that task. And how can you make sure not only it looks good in a demo, we actually, a culture, we don't ever do demo. We show real deployments and let you test. But it's very different out there when you're talking to somebody with an accent or that person doesn't even know what they need, right? It's actually very complex. So all of the engineering work not only goes improving and integrating machine learning advancements, but also doing our own orchestration fine tuning as well as evaluations so that you can be like utility and really reliable when you're replacing these systems and augmenting the teams that rely on this.

What's the process like selling AI to SMBs? We've had a bunch of people on the show that sell to larger corporates, Fortune 500. There's an excitement from the Fortune 500 to just buy AI even if they're not un, unclear of what the value is. I could imagine you see two scenarios, one where owners or operators are excited about potential efficiency or more revenue, more leads. But, and then another side, which is like, well, I don't know about this AI stuff.

Why are you doing an accent, Jordi? Why are you doing an accent?

Actually, you know, you'll be surprised. This is why I'm building for these industries. These are, probably the best founders and entrepreneurs I have ever met in my life. And I'm in the heart of Silicon Valley and I worked very hard to get here. And now all these founders truly, they are extremely ROI oriented, they're extremely customer oriented and incentives are very aligned, right? Like they only do better if they serve their end customer better and I only do better if I help them do better. so it's extremely aligned. So we actually don't see, you know, they might ask questions, but everybody is extremely open about it. They know that that future is here. They can't stay behind. It's not like the enterprises you're a little bit alluding to where let me just put this on my board deck and have like a few million spend here and there to show how AI oriented I am and then they won't actually use it. No, these people actually are absolutely incredible and partner with us so closely to see in numbers, right? We let them track it. This is how much you generated from only AI handled jobs or interactions. So we've been very happy about it. And it's the same from a $20 million revenue business to a $1,000,000,000 revenue business. Obviously, there might be a few steps here and there, but I am pretty happy about how incredible of entrepreneurs I'm working with in this industry. I think I choose to do that.

Yeah, I always think it's funny when people in Silicon Valley look at these maybe trade businesses and they think, oh, I can just get a better software. And it's like, you realize the person that was operating this business for 20 years is like a fixture of their community.

Yeah, they are like expected. And also extremely hard working.

They're working 14 hours a day, seven days a week.

Success comes over 20 years.

Yeah.

Not just like various, I don't know, and tonight in my bedroom I built.

This and now I'm- yeah, yeah, yeah. I went viral and now I'm the hot kid on the block and all the VC's are calling me.

And many times it's their families, right? Everybody really investing in this and even when you work with a larger company or owned by it, everybody's truly invested in this. To give you a sense, I have brainstorming calls with my customers on Sundays. They'll call me with really cool ideas and I'm like, the next day I'll follow up with a try this one, try this demo. Let's see if you like it and then we'll actually launch it this week for you, right?

So I absolutely love that. Do you guys leverage your own tools at all or do you have your own internal tooling to be more efficient? From what I can see, you don't have very many employees. Yeah.

What do you use for outreach?

I know. Hopefully you'll help with that. We're hiring across the board, especially for our engineering, but also go-to-market teams now and products. So we actually do utilize various tools internally just to increase engineering output, but also still, especially when it comes to various coding tools, I will say because we ship for really mission critical workflows, you need to make sure that it might like accelerate you and augment you, but you still have to make sure everything is ready for production. Because we're handling a lot of large volume customers that are really deploying these services to help somebody maybe after a storm, like a few weeks ago in St. Louis, it happened after the tornadoes. So we have to make sure everything is really getting that check from us. And like the best engineers we have hired across MIT, Stanford, Scale, Palantir, HRT, But yeah, I mean, if you're a fool, if you're not using augmented tools for your team. And you are right, we are a small team, and that is by design, because I think all of us are also here to create the best things out of nothing, right? And really keep our culture and grow in that way.

You mentioned you worked really hard to get to Silicon Valley. Can you give us a little bit of the journey and the background to get here?

Yeah, I will say building a company is almost as hard as getting to hear from a small town in Turkey. I grew up in the Mediterranean. This is like a tiny town actually in Turkey on the western coast. And yeah, I think it is an incredible community, but there is absolutely like no opportunity, especially for someone like me who's focused on math and computer science. So I tried to first leave actually, I went to boarding schools that I got in starting age 13. So I haven't lived at home since 13. And then spent some time in India when I got into United World Colleges actually for two years. And I thought that's how I would learn English because my English really sucked. And I did, you know, I'm doing pretty fine. And yeah, from there it's been my dream to come to Stanford. And I didn't know anything about it actually. I only saw it through a summer camp I came to on a I have a scholarship. And even my flights were funded by thousands of emails I sent to businessmen in Turkey. And one of them responded and sent me here to California. And when I saw it, I was like, I have to be here. And since then, there's no, you know, easy way of explaining this. It's you, I think probably I fought tooth and nail to be able to really get here. And I'm glad because I love this place. And mainly for me, it's about the people. Everybody's interested in your ideas. Everybody is interested in making them possible for the world. And if anything, I wanted to change one part of it when it came to my own company, is that I want to build for the real world, not just for Silicon Valley or other startups. I want to build for America. And then. And then the world, like this country that gave me a lot of opportunity, and then hopefully globally, the real world economies that run everybody's lives.

Yeah, I know you worked at Scale. Can you talk a little bit about human in the loop or business process outsourcing or any sort of like the centaur model of AI and how that might play a role in your current or future roadmap?

Yeah, so actually labeling does not play a lot of role for our roadmap. Like right now, obviously it was big, especially when you're improving these models, there was a lot of labeling and then it turned to expert labeling. Now you really focus these models in terms of how is it going to be better in physics or coding, et cetera. But for us, Human in the Loop actually comes from collaborating very closely with the teams of our customers because they will have actually many times what we see in these industries that they won't have like BPOs, what we call, or outsource call centers. Sometimes they might have, but they'll have also internal teams which are very valuable resources. So what they can do with us is that they spend their time only for mission critical interactions when they need to take over and actually see the context from the conversation about the person, about the home, so that they don't have to repeat anything. They can build rapport immediately. And for the remainder of time, since AI is really handling everything else, that they can focus on more important and high leverage tasks. whether you're helping with accounting, whether you're helping with actually making intelligent business decisions about where to grow. And believe it or not, a lot of people start from there to build their careers in these industries. For example, our engagement management lead, she actually joined us from our earliest customers and she started as a CSR. Don't worry, they're amazing still, Gaspar and you know, she was making a location change. We did everything right. I love, you know, like immediately coming into that.

Did you do anything?

No, no, no.

I mean, it's great. It's like you could just say, we're gonna take your. Your star employee, but we'll give you the greatest software that you've ever experienced.

And we're gonna go. Now they also get to interact with her, but she's supported by us.

Off balance sheet r d. It's off balance sheet r d now.

Exactly. I'd like to think that they got a lot of tools from that. But she started actually as a call center employee 20 years ago and she's so smart and incredible that really built the domain knowledge and went to consulting from there and a VP of customer experience and really built her career. That's what we want for the people, right? Like actually focus on the things that you can shine in and build your career in, not just the menial tasks. asks you don't want to do and only spend time on customer intake.

Sure. Talk about business model, pricing model, Salesforce, and Mark Benioff has been talking about cost per resolution in the customer service perspective. There's other companies that are doing consumption-based pricing, seat-based pricing. What's working, what's not, what are you seeing resonate with your customers best? Yeah.

I think we don't do cost per resolution. It's actually an interesting pricing model, but I think works better for customer support companies for tech when it's all about tickets. And even then, I think it's a little bit iffy because, I mean, what's the resolution? You're able to close the ticket. How many of us have been in a situation where I'm, like, getting this automated email that my ticket is closed, and I'm like, well, I am not close. I need that help. So for us, we really work in Enterprise contracts with our so they get on a, there's a platform kind of package that they choose based on which products they want to use and how expanded of channels that they want. And then on top of that, they have a volume package that they add so that they can spend it across any channel that they liked. And this also gives them that ability that we're not just signing and goodbye because in a lot of these industries or quite frankly in any company, AI is not just like I got on it and now it works, it doesn't. You really have to make it work, ensure that it is working for their operational workflows. So it gives us the ability to do that and closely partner. And then as we go to mid-market, that was something that's why I was super excited about this NextStar partnership because they have decades long experience in these industries and they have accumulated so much knowledge that today we're able to deploy a Netiq tenant with that knowledge for their members so that, you know, a lot of the, like, maybe they won't have any lift to do. They can really roll it out directly from the business insights they would be getting from an amazing membership like them anyways.

So, yeah, so is there was like a meme for a long time about like the search fund, gonna buy a business, gonna do a roll up in these SMB markets. Is that, have these markets been already rolled up? Are we past that?

Yeah, there's quite a bit of rollouts. I'm sure you'll be interested in it next.

Yeah, no, we're gonna roll up all the podcasts.

I'm sure you're looking into it already. Yes, there's quite a bit of rollouts, to be honest. It really became popular recently because there's quite a bit of capital out there, as you also know, even from venture capitals. So much capital that they're like, yeah.

VCs are crossing over and being like, yeah, we'll play in the card. What could go wrong?

Yeah, only either deploy to startups that don't need it or do rollups.

Yeah.

So, yeah, there are quite a bit. But I think, again, the winning strategy becomes, even if you have a roll-up, right? It's a lot about how are you creating value? What tools are you using? What tools are you not using? What, where was the company or or this combination of companies you have gotten when you got them, and how are you really implementing changes to serve your customer better? So we work with various companies that are owned by incredible private equity partners.

Let's hear for private equity. Thank you. Yes, finally getting some recognition. We love private equity on the show.

Yeah, they don't get any recognition. No, nothing.

It's a thankless job.

Private equity is really the backbone of America.

I agree.

Because it finances.

It totally is. It totally is. I'll tell you what happened. I think 20 years ago, maybe, you know, it was all around in all of these industries. It doesn't have to be one. A lot of companies to be picked up, right?

Yeah, totally.

So today, that's not the case.

Yeah, pretty much.

Like, you're not really looking for that perfect business. Everything is stellar and like waiting for you to be picked up and somewhere in the heartland of America. so I think private equity actually have to be very innovative. And I think they're obviously already numbers focused. They are looking for true partners. And that's how many of our engagements started with these companies. Like actually wasn't even about our company was about talking about AI honestly, straightforwardly, what's going to work for you, what's not going to work for you. And I think they are really seeing that for value, they have to change. It's not just about using one playbook and it works for you for 40 years. And they realize now that a lot of them have to make it work for these companies with the right partners and take good decisions. So I think I love that we are coinciding with that change, right? So it's kind of harder to compete there as well because everybody has capital and not much to really roll up around, right?

Well, let's hear for the capital. At least there's a lot of capital. Congratulations on your round. Thanks so much for coming on. Tremendous momentum. Conversation.

Thank you for having me. And we're so excited to be here, especially for the two people who also work very hard. And especially you, John, have built companies that optimizes performance left and right with Soylent and Lucy.

Left and right.

Hopefully we take that from that culture.

Yes, yes, yes. Thank you. Well, we'll talk more about that. Thanks so much for coming on. Thanks so much for coming on.

Thank you.

We'll talk to you soon.

Next up, we have Jordan Schneider from China Talk coming on.

Jordan.

We're gonna ask him.

An absolute dog.

Is China important to talk about? Bring it down for us.

Who's talking about China?

Yeah, who's talking about China? Jordan, welcome to the show. How are you doing? Can we get a sound effect? Oh, we got the horse.

Oh, the year of the horse.

What year is it? In China?

I don't know. We're playing around with sound effects. You're playing around with backgrounds, too.

I think it's the rat. You're the rat. That seems. Last year was.

Last year was Dragon, so it's like it's worse after.

Okay. Okay. I want to start with a victory lap. I said. I said, tariff war gonna not be a big thing. We're not gonna be talking about it in four. Four years. Push back.

We're not gonna be talking about it in four years?

So when you came on last time, we were talking about how it seemed like complete doom and gloom. Like the Trump tariffs were absolute chaos, it was complete disaster. And I was saying that this is within Trump's control, and so he could potentially roll them back and the market would go back up and everyone would be breathing a sigh of relief. Like, okay, that was a crazy time. but we're not in this insanely high tariff regime, so things aren't that bad. And we would look back on it like we looked at the previous trade war, which was a crazy time, but did not like permanently change the structure of America. And so it feels like we walked to the edge, and then we walked back from the edge, and it's less of a story now, but am I right or wrong?

Let's be clear, who's walking him back from the edge? The justice system. Supreme Court is going to take the decision out of his hands.

Sure.

And then it'll just kind of be slower and more awkward and his lawyers will actually have to do work to write these section 232 investigations, which like, I guess they'll just have chat GPT, right? Because like there are no lawyers right now and USTR. But come on, man. I mean, look, yet two days ago, like some reporter asked him about the taco trade and then he said, fuck it, we're going back up to 50. So look, I mean, it's the story, the saga is not done. We have four years of this. Old men don't change their habits all that often. This is clearly the thing he enjoys about being old.

No, I've taught old dogs new tricks all the time.

I got a ten year old Newfoundland at home. He just learns to balance the ball on his nose. He's doing great. No, I hear you.

Yeah, yeah, yeah, yeah, he should.

Well then, what else is driving the news? Obviously, the Ukrainian drone attack is very interesting in the Ukraine-Russian theater. Is there a reaction from your community in China world yet, or are we waiting to hear how that plays out? Already we're seeing defense tech companies talk about counter-UAS more seriously. DOD procurement modernization. You could imagine that this brings DJI into focus, but how should we be framing it within the China lens? I got.

A line from a group chat.

Please.

Don't ask a woman their age, and don't ask a Teal-funded defense tech startup where they get their batteries from.

Okay.

Spicy.

I don't know.

It's a brave new world we're walking into, and I'm worried that we're not ready. I think there are definitely a lot of advantages that the US has, but one of the big ones is manufacturing.

And.

Speed and agility of procurement. And if you look at what's happened in Ukraine over the past two years, the amount of iterations that all of these drones have gone from in the different electronic warfare, like responses and counter responses and this sort of the size and scope of these different drones and being able to scale up manufacturing of them. And then now we have fiber optic drones and now we have things to cut the cables of the fiber optic drones. I mean, just the speed at which you have to innovate when you are fighting and dying is so much faster than what the US It does. And because the wars in Iraq and Afghanistan were not against high, you know, we're not against like great powers, thank God. The new things that the enemy was able to bring to the battlefield was much less challenging and you could kind of be like fat and lazy and still mostly be okay and not have to worry about.

Well, I mean, even IEDs massively transformed the battlefield. at the time. And that was another kind of asymmetric trade in some way where, you know, a small homemade device could take out, you know, cause huge loss of life. But your point is totally taken. It wasn't like four million drones a year being produced and flying through the sky.

There was still so many IEDs that it did kind of breed private sector defense innovation in the sense that like, One of the first use cases for Palantir was map all the IEDs, see that if the ones in this area have nails and the one in that area have TNT and dynamites over here and C4s being used over here, you can kind of cluster those and see that the bomb maker must be in this city and then you can go and find them. Yeah. And so, I mean, yeah, it is this like cat and mouse game, right?

But I think if you look at like the story of the procurement side of all that with the MRAP, and the other sort of like physical, like hardware changes that the US government needed to supply its troops, to supply the troops, like it was so fucking slow.

Sure.

And, you know, not the best showing, I think, if like even like Al-Qaeda can kind of get you on your toes when it comes to technological innovation. So I don't know, it's a bummer, but hopefully folks are waking up and doing their best on this.

I mean, I, Yeah, you mentioned the drone battery thing, but it feels like people woke up to that after Skydio, the Skydio battery ban. And so I would imagine that there's probably dozens of companies working on reshoring drone batteries right now, right?

I hope, I hope. I mean, I'm sure they are, but it's some, yeah, I mean, it's, look, it's an obvious thing, but what's interesting, I think, is to what extent on both the Ukraine and the Russia side of this, there are just like Chinese parts that are driving both of.

The armies.

Okay, let's switch to a kind of a less contentious issue, immigration. What's going on? What's going on in terms of- the students. Students, research labs. What's the high level update?

Well, I want to challenge you guys a little bit. Because you have a lot of CEOs and investors in the Silicon Valley broader ecosystem, and they, as well as anyone else in this country, understand just how important international talent is to the future of American science and technology flourishing. So I would be really curious.

To.

Get these folks on the record to just even talk about their experiences kind of like coming to America and getting their first H-1B visa and the sort of the challenges and the opportunity that you have to, and the sacrifices you have to make.

I mean, the woman we had on right before you was from Turkey, made it to Stanford, had some random Turkey businessman pay for her flight so she could just get here and is now building, you know, incredible company. so we certainly recognize the importance of Inter International Talent. And I'm broadly in favor of brain draining the world, as I think you are as well. I don't, I don't follow immigration, you know, law or Trends nearly to the level that I imagine you do specifically around China. And it's an interesting time. I mean, we had some Stanford students on that had written that article that was... Pressure on the CCP? Yeah, well, it was widely read and it was also controversial because many of the sources were anonymous, but it also seems to be something that people just take as fact that there is a lot of, you know, sort of low-level espionage information gathering happening on campus. But I was more less trying to get your take, your take more trying to get like the late, the core 101 development.

So I think there are two, well, it's, there are a number of things going on. First, the Trump administration is threatening, like revoking Harvard's ability to take international students. We'll see how legal that is. But that is a remarkable development. and something that I think is sort of like spread if it like ends up being a tool that's used across a lot of different universities is really dramatic. Like if you look at a lot of the sort of top 10 lists of publications, there's like Harvard and then there's MIT and then there's eight Chinese universities. And the fact that, you know, they're going after like what is the, like one of the, four most important centers of research in America is really concerning. And I think part, you know, if you look at these, if you look at the STEM programs in America, oftentimes, I think CS is like over 70% foreign. And the programs aren't stupid. Like they want more American students because they understand that there are a lot of challenges by having you end up having this sort of exposure, but the fact is just the talent isn't there. And you want to have the best and brightest in your programs go on to do amazing research and start amazing companies. And so this is sort of the world we live in where American primary and secondary education is not going to fill up these slots fast enough, particularly for CS is kind of an exception to the rule. A lot of the hard sciences are grueling. They don't pay particularly well, and the sort of the options to just like, you know, be a CS major in college and like go get a degree for the past 15 years have been much more lucrative than like studying mechanical engineering or electrical engineering. So kind of if we're trying to sort of like re-industrialize and build the future, like you need PhDs and you want the best PhDs and scaring them off by having the head of the, the nominee for the head of the immigration process say that he's in favor of revoking OPT, which is like the ability for students once they graduate for any major, I think you get a year, and if it's a STEM major, you get three years where you just have a blanket work authorization. So you don't need your employer to, you know, get you in the H-1B lottery and sponsor you is a big part of the value proposition. for going to school in the US alongside being able to go to the best research universities on the planet. So you have on the one hand, from like the demand side, I guess, students being a lot, you know, just having to price in an uncertainty factor of whether or not they'll be able to stay in the US, much less state to finish their program. And then on the supply side, I guess, The Trump administration having the NSF spend half as much money as it's authorized to spend and kind of blanket cutting off universities, which are on the shit list for whatever reason, to the tunes of billions of dollars, just like, sorry, Hopkins, sorry Harvard, sorry Penn, which is leading to layoffs and worse research. And I'm annoyed because the future is going to come slower. And America is going to be worse off because we're taking these incredibly unforced.

Have we seen any pressure on the 01 program? We talked to Sigal Wen Teal Fellow who's been trying to speed up the 01 process for those extraordinary candidates. And it seemed like he was pretty optimistic about that program continuing and flourishing, but maybe that's just not enough in your mind.

When.

A year, right?

Yeah.

And other things and like, I mean, we have 500,000 foreign students in America. I mean, aside from like a, like this is the thing that is funding the universities. Like there are, there are like maybe 10 schools that have billion dollar endowments. I mean, you're going to start seeing a lot of universities go under, which is like just sad in general. But this is like education is not a zero sum thing. And I think that's kind of like one of the more sad talking points. It's like, why aren't there more slots for Americans? It's like, no, like the American slots in all these universities are being subsidized by all the undergraduates and master's students from abroad who are paying full ride to this where, you know, we have like, you know, in-state tuition and whatnot. I mean, it's a different ball game.

How would you think about changing higher ed? I mean, there's been this idea that Harvard, for example, was founded hundreds of years ago and was servicing a population of maybe 30 million people, and now there's probably a billion people that have heard of Harvard or maybe in the candidate pool. But the class size hasn't scaled. Would you scale up? these elite universities and try and get the elite pedagogy into more hands or into more minds? Or are there other things that you think we could do? Because this feels driven by some sort of dissatisfaction with the results of higher ed. I don't know if you agree with them. Maybe you think it's perfect, but certainly there's the first question of like, there might be a flaw. There's a lot of student debt. There's a lot of people going to schools and taking on debt to pursue degrees that don't necessarily pay and they don't make economic sense. How would you change higher ed without disrupting the brain drain equation? Sure.

I mean, I think at a macro level, it's important to recognize that anti-intellectualism beating up on the universities has a long and storied tradition in American history. Going back to McCarthy and even before, So like the fact that politicians are making hey shitting on academia is not like something that is particularly novel. Should, you know, if I was like the secretary of education and I wanted to use a stick, I would do the exact thing you said, John, and say like, you need to spend down a percentage of your endowment every year if you want to stay tax eligible. And does that mean, you know, growing your class size by 25% every year? Sure, absolutely. Like there is a glut of professors and there is a demand glut. And like who is gaining by Harvard staying small? Like the few people who get in who like get to feel more special about themselves.

Yeah. What is your model for higher ed? I mean, like Tyler Cowen kind of mentioned, he posited that higher ed is a bundle of goods. It's both a daycare for parents to get their kids out of the house. It's a dating service. It's also, you know, a series of textbooks that you are forced to read.

It's also like a music festival.

It's a bunch of different products kind of bundled, and it's prestige and signaling and essentially a one-word summary of your SAT score. And you could potentially unbundle those. I don't know if that's necessarily-- SAT score. Right, right. Like if you say Harvard, people know, okay, potentially you've been filtered for IQ at some level. And so, you know, is that the correct model? Do you disagree with that? Or do you think that's a good thing? Because I'm hearing higher ed is perfect, and I don't really buy that. But I'd like to know what vectors you would improve.

Well, I mean, I think.

There is.

A real, Golden Goose aspect to what we have, particularly when it comes to science research.

Right? Okay.

So, you know, look, I was a history major and I think a lot of the sort of like soft, they don't teach us any, anything like real stuff, like critique baked into what you said is, is a lot more applicable to humanities. I mean, thinking back, like, could I be where I am today? Would I have the mind I have if I just like read all the books they assigned me in college? instead of taking the courses? I think so. But could I become a biophysicist without having access to a lab and professors to train and tutor me? I don't really think that's the case. So particularly when it comes to sort of science and engineering disciplines, there is a real aspect of mentorship and handholding that needs to exist.

I agree with that. I agree with the bio side. CRISPR came out of academia. The Transformer did not. And in fact, the one academic lab that is listed on the Transformer paper attention is all you need, Canadian. It's a Canadian lab.

Elizabeth Holmes had never been in a.

Lab before she started her company.

To be clear.

If you actually go back to the Transformer paper and look at the authors, most of them, by the way, have PhDs. And most of their PhDs were in labs that were funded by the National Science Foundation. So, I don't necessarily think you could write off AI as not being something that had government funded basic research behind it. I think that was absolutely a crucial thing. And we've had a lot of AI winters over the past few decades where people have sort of, industry has basically given up on the technology and the only folks that were funding it and doing the research were in government.

Given.

Government money and working in universities.

Okay, that's fair.

It's the place where not, where the stuff that you can't get venture back funding happens.

And I think that- Everyone can get venture back funding now. There's unlimited venture. There will never be enough venture capital. Let's just venture fund everything, even the basic research that won't return ROI.

Yeah, I'm sick of VCs preferring a deck. Don't even prefer it.

Yeah.

Idea.

Yeah.

Check.

This is the long-term solution.

I want to go through a couple. You got executive order for DJI ban in the next two months. Yes or no?

Sure.

Sure.

I'm worried about my drone racing league.

Yeah, I hope.

I hope they.

Do you actually race drones?

Chinese. Last one. Chinese AR VR. Should we be paying attention to oh, interesting.

Yeah, man, I wrote a whole feature about the Chinese AR ecosystem. It's a really interesting development because basically what happened in the US was Meta consolidated it. And you had a handful of startups that all kind of gave up because there was a trillion dollar gorilla in the room. But China is a much more dynamic ecosystem. You have like six or seven players who are all exploring all these different hardware trees of like where to put the battery and do you need the screen on the glasses? Do you not need the screen on the glasses? so, you know, ChinaTalk.media, it's one of our more recent articles kind of looking at, at, at Rokid and a number of other, a number of other Chinese startups, which, like, all have products that are 300 to 600 dollars and are really cool. Can I pitch, can I show a book, please, before we go to the next course? Okay, so it's coming out tomorrow, I believe. The party's interests come first. Joseph Teridian, it's a 500 page biography of Xi Zhongxun, who is Xi Jinping's dad. And it is this incredibly detailed, like wild ride through this guy's life. He started as a, he joined the Communist Party when he was 15 and 1926. And the first half of the book is like all these crazy war stories where he is like fighting and executing nationalists and Japanese. And then, you know, he becomes, he's like the, He's like the highest flyer in the 1950s. Like he's promoted faster than everyone in his 30s. And then the Cultural Revolution hits and his life gets completely ruined. Xi Jinping is 15. His dad is this like black stain on the party. And the amount of sort of like family trauma that the dad and then the son by proxy like ingests over the course of their life is just a is kind of an unfathomable thing that you can only really experience by just like living day through day through this guy's life. And the fact that he was able to write this book, you know, I see books with this level of detail about folks like Stalin because the archives have opened at this point. But Joseph did an incredible job like reading all these memoirs and talking to people and digging for stuff. And if you want a sort of window into what Chinese elite politics looks like, it is a really a really special piece of scholarship and something that comes around really rarely in the in the China studies space. So, Heather, you guys get them on the pod.

I would love to talk about these things, Dad.

And I highly recommend everyone check it out.

Thank you for the plug. That sounds fantastic.

Thanks for coming on, Jordan.

We'll talk to you soon.

Always a pleasure. Cheers.

Next up we have Max from Arena Magazine.

We are surrounded by journalists.

We have Arena Magazines right here.

Always. You always have to keep working.

Welcome to the studio, Max. We always keep your arena on you. How you doing? Oh, he's got it behind him.

With the tie. He was ready.

Bolo. Looking good. What's new?

Well, we We just moved into a new Arena Magazine world headquarters.

World headquarters.

Hence the bolo.

And, you know, we-- It's not Arena Magazine headquarters though. It's the Intergalactic Media Corporation of America, correct?

Wait, don't you have a new website too?

Oh, yeah.

That's the other big thing. We launched our new website during the first hour of the show.

Congratulations.

Actually, it may have been the second hour. It wasn't ready during the first hour. We wrapped it up in time for.

The third hour of the show.

That's great.

Let's go. So we took a look at all of our favorite websites for consuming long-form text. And Silicon Valley won again. We couldn't find any superior form than the software documentation website. And so initially we actually thought about hosting the website on GitHub and just using Markdown files for all of the essays. Wow. We didn't literally do that, but we built a site that's based on engineering docs. So it's in dark mode. It's a science-y vibe, and we think that people are really going to love it.

Wow, yeah, this is great. Oh, very cool.

I like how I can decide if I want to focus or not. There's a little button.

That's right. Oh, focus just puts it right in the center. Okay, cool.

You can get rid of all of the prompts for you to send us money. by subscribing or you can leave them in there. We got a lot of fun stuff. It's sort of almost a relaunch of the magazine. After we did the first four print issues, we're now going to really make a big push to get a lot bigger. And that happens mostly on the internet.

So what is the flow for specifically windowing? I know Taylor Swift is very good at this, where She goes on a tour, and then there's documentary that comes out later. So you have to see her in person, and then you go watch the movie in theaters, and then you can stream it. And a lot of the streamers made the mistake of allowing you to binge all 10 episodes right up front, and it doesn't create these shelling points, these moments. Are you thinking about gating articles to the physical magazine first, then putting them on the internet later? How are you thinking about that?

It's great that because our readers who pay us get the print magazines first, we really don't have to be in a rush to put things up on the website that are in the print magazine. Until now, there's basically been 100% overlap between what's on the website and what's in print. That's going to change a lot because we're going to do a lot more stuff online. We had something very funny happen in the first few months of the magazine, which was subscribers emailing us upset that we had posted an article online before they had gotten a chance to read it on paper. which is sort of strange but also really made sense. And so yeah, we put things in print first. And we're not really sure how the cadence is going to go. We'll figure it out.

Are you going to increase the frequency at all as you go bigger or just trying to go bigger with the stories?

For us, quarterly is about the right pace. We might go more than that. But the truth is, in order to publish every week or even with us every month, you'd have to reduce the paper quality to get the printing time down. It takes us multiple weeks to go from submitting the files to the printer to the magazines ending up in mailboxes. Whereas something that's arriving in your mailbox every single week on what I call glossy toilet paper, which is the sort of very, very light paper where it's falling apart, it's been stapled together, you can do that at a very fast cadence. It's no good. So what we're going to do, increase the online volume quite a bit, do a lot more stuff there, but keep the print magazine spare in four of them per year and super high quality.

You recently wrote a profile about Brian Schimpf, CEO of Anderall. What stuck out to you about him? What was the most interesting takeaway from spending time with him?

He's definitely a genius and a stand-up guy as well. I mean, I just thought it was sort of funny to write a piece about a man that most people are unaware is the CEO of this company, but they all know. And even some of the people that I was discussing the piece with in advance were like, who's the CEO? And all of the Anduril co-founders are emphatic that Brian is the CEO. Brian has the best decision making. Brian is absolutely in charge, as everyone expects the chief executive officer to be. I mean, Trey said something like, I trust his judgment more than my own. And there's a lot of fun stuff in that profile. Even stuff about Palmer. Palmer has zero direct reports at Under Armour. And it was Brian who told me that. And I thought that that was amazing. Zero.

That's fascinating. I mean, it's probably the perfect situation. He can just go around the company and invent and do what he does best and evangelize and tinker and also just drop into certain projects, be an individual contributor if he needs to be, be a manager if he needs to be, but doesn't need a standing staff. Right.

One of the other things that this is not about Brian in particular. In fact, Brian thought it was sort of funny when I pointed it out to him. He must not have noticed before. But everything on the Ondorel campus has been set in the same typeface with the exclusion of the government mandated parking signs. And so it's like everything, the signs on the gates, the room labels, the stationary, it's all been set in Helvetica now, which is sort of a 2000s recut of Helvetica. And you don't see design discipline like that anywhere. But now that I'm a magazine man and I'm thinking about letterforms and typography all the time. It's like you notice it very viscerally walking around Ondorel that they've done everything to like exacting specifications. And they also use the same typefaces.

That's real brand. Brand is not a logo or a website. It's like showing up with that level of consistency, which makes it consistency.

And you know, it's the idea that like, don't you dare try to design something, let the design team do it. Because there's like they're protecting the identity. And even the drones, the missiles, the tanks, the submarines, they have the exact same typography as the stationery and the meeting room names. And I just, I had never seen it before. You go on like an airline or whatnot, the typography is all over the place. And that was my sort of one of the standout things from visiting the Andor campus.

How are you thinking about growing the magazine in terms of balance between full-time writers and contributors. It's an interesting place to send a thought or an essay or something. And so I imagine that it's more tractable to be a contributor at Arena than have a column somewhere else. It's probably a little bit more manageable for someone who's maybe not writing all the time, but at the same time, you probably want a steady heartbeat of writers. how are you thinking about balancing those things?

Yeah, well, so first of all, we want anyone and everyone that's got something to say about, you know, technology, capitalism, civilization to send it to Arena. We're very, very good editors, and it's, you know, it's very useful to have editors, and we provide that service to anyone to help them, to help them get their word out. Some of it will end up in print, some of it will just be online. You know, we have some full-time people, we'll have some more full-time people, we have some contributors, we'll have a lot more full-time full-time contributors. I think that it's on the extremely polarized ends of this spectrum that you start to get into weird stuff, which is way over budgeting for having full-time writers. You can create this Frankenstein, where it's some full-time people and some contributors that will create the most interesting balance. It's not going to be the most interesting balance if it's all random contributors or if it's all full-time staff. And so we mix it up.

How do you think about the different types of pieces that you want to, like, what is the shape of the different type of coverage? Obviously you have profiles, you probably also have op-eds, you're not really doing breaking news, or investigative journalism. What is it called?

Not news.

You're saying not news.

Not news.

Anything but news.

But what about, what are the other areas that you're interested in exploring? exploring like Forbes famously has the Midas list and the Forbes 400. Are they going to be like list.

Products or anything like that? One of the things that we talked about at the very beginning was like how to do lists in a non-awful way. Haven't figured it out yet.

So we're not doing lists yet. Well, maybe there's a future contributor listening right now who can come up with something.

Yeah, I don't know how to figure it out. If someone can figure out how to make a better list, and I'm all ears.

Switching gears. What's AI adoption like in Iowa?

You know, it's well, Google was just yesterday put $7 billion into a campus there, but it's just for server racks or whatnot. I'm not sure that the people are really following suit.

That's similar in Abilene, in Texas. There's Stargate's going to be staffed with like not tens of thousands of people, like hundreds maybe.

Yeah, yeah.

It's a very, very small organization. Yeah.

You know, my mom just retired as a teacher and all of her teacher colleagues are complaining about the kids using ChatGPT, so that's good.

There you go.

The kids are up to it. But I mean, I sort of doubt whether any of my neighbors are paying attention to it. I sent one of them a poem that I had written with Claude. He was like, How's this so funny?

That's like the biggest alpha right now. It's just using models to generate super thoughtful, creative work for people that aren't online.

Yeah, I mean, I assume that there are some people in Iowa that are on Facebook that are looking at the sort of AI images and being, Wow, that's beautiful.

Yeah.

Did you watch Mountain Head yet? Did you have a reaction?

I haven't, I'm sorry to say. I heard you talking about it earlier.

I guess I- well, yeah, it's just, it's, it's, the main kind of narrative of the story, besides being a critique of the tech elite, is that, you know, it's, it's this global catastrophe because AI has gotten so good that nobody can tell what's real and what's fake.

And- no, I so disagree with the framing. People already have a lot of trouble figuring out what's real and what's fake. AI is going to be an improvement over the status quo in certain ways.

Okay. How are you using AI@Arena? I imagine that you're not using LLMs to actually write whole articles, but what about proofreading or even just... Well, they.

Trained a model to remove mdashes after.

They generate Yeah, yeah, yeah, yeah. And then problem solved. And also the word del is just to find and replace. But I imagine that the problem of typesetting and transforming text from just a big block of text into something that fits nicely in columns, that feels more tractable and more tactical than the artistry that goes into writing an actual article. Is that useful? What about AI images? for collages? Has that been useful? Anything like that that's kind of popped up? Yeah.

So on the writing front, you know, I'm prepared for the day, which it is like better, but right now it's not.

Yeah.

At least not compared to what we can do. And based on the way that the like, that the LLMs operate, they tend to be pretty, they'll use the same sentence structure over and over again.

Sure, sure.

For me, it's a tell.

Yeah.

I'm not saying I could like judge it 100%. A, no one wants to pay to read something that's taken zero marginal minutes to do. It can be very useful for brainstorming. It can catch some of the copy edits. My mother tends to be better at copy editing, though, than any of the GPTs now. But yeah, we have subscriptions to all of them out the wazoo.

To.

Use them. But yeah, I mean, people are coming to Arena for a bit of an analog thing in the first place, which is a print magazine. And so it's like, to the extent that the AI can help us do more with less, it's great. But the core sort of writing work is something that is, you know, we're also trying to keep that art alive.

We were just talking to Jordan Schneider at China Talk about higher ed and some of the problems there. How would you kind of diagnose the problems, if any, in higher ed right now?

I mean, I think that people tend to focus on the elite institutions because of their sort of cultural power. But I mean, clearly the biggest catastrophe in higher ed is that we agreed to indefinitely fund higher and higher loans at the federal level for students to pursue degrees from universities. And it's not, it's really not the Harvards that are the problem in that equation. It's the, it's the universities that can't offer much, but that we're allowed to sort of way over inflate their budgets with the, with the, with the federal loans. And so this is, this is a, you know, related to the China point. I read a funny story that the University of Illinois took out an insurance policy in 2018, hedging against a decline in Chinese enrollment.

Oh, wow.

But the lawyers messed up the contract, and so it was invalid in December 2019. And then it took them like five months to renegotiate it, during which time the COVID pandemic happened.

It happened, yeah.

Literally nuts. So I always have the lawyers read the fine print. I think it's possible that Claude would have done a better job than those lawyers in that maybe her, maybe Harvey.

And so, somebody, somebody, you know, that wants to go super risk on in China should bet on a rise in American students because right now there are, I guess, roughly 800 U.S. students in China, which is just very low, unfathomably low.

Yeah.

I have no idea how to diagnose what goes on inside China or why people would want to go over there. I don't know how to, I don't know how to price that one.

You think we're getting a DJI ban in the next couple months? You feeling excited?

I think that people will be upset if they ban the best product available. I say this with a fervent desire that we have one that's amazing, but I'm not sure that for random civilians it's going to fly.

Yeah, it's hard if it's not popular. Anyway, anything else, Jordy? I think we're good.

This is great. Congratulations.

Congratulations on your new website. Good luck.

We go to ArenaMag. We actually always keep, we've kept an ArenaMag on our desk at all times in the entire history of TBPN.

So we got it.

We love physical media. It's a very good desk object. And I'll just say, we're going to.

Have even better desk objects in the future. Oh, can't wait. Yeah, I'm excited. I mean, I know you can take it up a notch to where it's more of a Book.

Totally.

Those are sort of like chapters in technology and industry. Well, I have no, I legitimately have no inside knowledge. I'm just speculating. Excited to see what you have.

Anything.

Anyway, thanks so much for coming on the show.

Wait, before you leave, before you leave, we got some new sound effects I want you to hear.

We're surrounded by journalists. Hold your position.

We're working on these. Anyways, have a great afternoon, Max. Thanks for going on.

We'll talk to you guys. Bye. I see multiple journalists on the horizon. I still don't know where these came from.

These are from Cod.

These are like, yeah, but like, how do they change the voice? Is this like AI generated? Is it effectively like the Captain Price?

Okay, but you can just do a Captain Price generator.

But it's Ben making.

Is that Ben's voice?

That's Ben's voice. Wow. I see multiple journalists on the horizon. Market clearing order inbound. I like the kill streak. Those are great. I see a large IPO on the horizon.

That's so good.

I love it. Good, really good impression. I love it. Anyway, thank you so much for watching today. We did not have more timeline. I'm gonna be doing.

Oh, yeah, yeah, yeah.

Let's do some timeline. Let's do some timeline. That's great.

Yeah.

Congrats to Jacob Kimmel. We generated the most visually striking data of my career at New Limit this week. We have a real opportunity to create that had healthy years for everyone. Jacob's been on the show before, of course, started New Limit with the help of Brian Armstrong in partnership with Brian Armstrong, founder of Coinbase, and really pumped us up. But he barely teased it, but it seemed like something happened that was very good there. We also need to tell you about AdQuick, out-of-home advertising made easy and measurable. Go to adquick.com, say goodbye to the headaches of out-of-home advertising. Only AdQuick combines technology and out-of-home advertising. I'm about to buy every billboard in SF. Market clearing order in effect. I love it. We covered the attack. James Cadwaladr, I don't know, Cadwaladr, says GG Tri-Ramp, which startup in the US is known for shipping new features the fastest? This is ChatGPT.

Ramp gets the top spot.

These are hotly debated. This is the generative engine optimization, the G-E-O that and Dreeson's been writing about the AI SEO. Like, you gotta be.

Well, this is what. This is what James coming. You remember we had James on.

This is profound. Oh, this is.

Okay, that makes sense.

Yeah.

And they're absolutely cooking.

Very. This isn't.

This kind of thing isn't by chance, you know? Sure, sure, sure.

Yeah, that makes sense.

We have a post from Kareem.

Car.

Car says, this is wild to me. How much of a nothing burger AI has so far.

It's been 2.5 years and the most tangible effects of AI are students cheating more and slightly higher, more realistic Facebook slot. I don't want vibes or speculation or demos. I want one concrete real world achievement from the current generation of AI that's not potentially a big deal, but actually a big deal right now because I got nothing. And yeah, it's just very funny because obviously like LLMs have been vended into like every enterprise everywhere and are like.

Constantly- yeah, basically just constantly- you-'re not a business business person.

Yeah. But also just like day-to-day use of, for a lot of people, like OpenAI has just replaced Google. Yeah.

Kareem clearly doesn't respect how insane it is that people are using something else for search besides Google.

That is crazy. That is the entire Microsoft, which is currently, I think, the biggest company in the world. has tried for two decades to unseat Google.

Yeah, and they now own a large part of a company that actually kind of did it.

Yeah, exactly. And yeah, pretty remarkable. Let's also tell you about Wander. Go to wander.com, find your happy place, book a wander with inspiring views, hotel grade amenities, dreamy beds, top tier cleaning and 24/7 concierge service. vacation home, but better folks. $50 million.

It's great that we can talk about it now. It basically leaked weeks ago, but John Andrew was on and broke it all down and he's going for 300,000 homes.

Let's go.

In a decade. Love to see it. We had Patrick asking, what is the lightest, thinnest, most comfortable and simple watch? And the best recommendation that I saw here was from Will Minidus, what do you say? Recommending a Richard Mille. And I agree with him. I think it's a great option. You gotta have a Royal Oak.

Yeah.

Extra thin.

Why not a Graff Diamonds Hallucination?

Yeah, he's getting up there.

You know, it's not the thinnest, lightest, or most comfortable watch, but you know, it makes a statement.

Will recommended the RM 66 manual winding flying tourbillon is a good entry level piece.

Yeah.

Skeletonized, it is.

I think Patrick should go with an Urwerk. That would be interesting.

That'd be interesting.

But I mean, seriously, if he's looking for something in that category, it's probably Patek Philippe Calatrava, Vacheron Constantin, Patrimony, or JLC Ultra-Thin, probably something along those lines is going to be probably what he's looking for. He didn't really specify dress watch versus sports watch in the thin. But I think you'd look good with a dress watch. So hopefully you can pick one up on bezel.com. your bezel concierge is available now to source you any watch on the planet. Seriously, any watch. I think that tweet is just permission for us to introduce Patrick to the CEO of bezel over text message as soon as the show wraps.

I think we should just-- you- guys should really talk.

You guys should talk. It's not a dumb opt-in intro.

It's just happening.

Zero opt-in. Yeah, you posted about it. You got 140,000 views on this post.

It's happening. It's happening. It's happening. We got a post here from Gabe. He says, quote, LMAO has survived and even thrived over years, but its cousin ROFL, Raffle, has faded into indignity. The cruelty of 120,000 likes.

Wow. Oh my God, I didn't notice that. that is wild.

A 120 banger is really good. Really good.

But it's so true.

It's so true.

There were all these different acronyms. Let's bring it back. Rafflecopter, LMAO, really stuck around though. Yeah, there were a bunch of them. Let's bring it back. Hey, yeah. I mean, TB Nation.

I underrated is that LOL is still around, you know?

LOL made it through. See, I actually, I adopted LMAO.

I underrated LMAO. I underrated LMAO. I underrated LMAO.

Yeah.

Like, within the last two years, I just. I never had a huge amount of respect.

Yeah.

But now I've.

No.

Lmao.

Yeah. But are you. Do you. Do you draw upon lmao more frequently than LOL?

Less frequently.

Yeah.

I'm sprinkling it in.

You know, I think I like. Like, the. I. I also like the, like, it's honestly variation. The hahaha is also good because you can go, ha. That's, like, not good. Haha is like barely trying. Once you get into three ha's, it's like, okay, I'm actually giving you some positive feedback here.

LMAO is usually, I usually use it when something's actually ridiculous. Exactly. Like somebody is doing something that is just silly. Yeah, yeah.

It doesn't really have any of the eye roll that the LOL does. The LOL can just be like, oh, it can be kind of laughing at it. LMAO is like a little bit higher. But yeah, maybe we got to bring back ROFL. Raffle.

Yeah.

Rolling on the floor.

Well, we got to tell you about our newest and today, our greatest sponsor on the greatest sponsor of today's episode in many ways. Adio.

Yes.

Pull it up. Adio is the AI native CRM that builds scales and grows your company to the next level. We do a lot of, I mean, we have an interesting show and that we're like, We're basically booking five sales calls a day in the form of guests. And then we're also running an advertising business. I've used Adio over the years. They've been around for a while, and we were excited about the opportunity to partner with them. They are the backbone to the revenue side of our business, and we are super pumped to partner with them. You can go start with a 14-day free trial of their Pro plan. and you're going to love it. It's fast, it's flexible. It reminds me of if the Linear team was going to build a CRM product that they feel similar in many ways. And yeah, also Red Point backed, you know, so you know it's a banger.

Yeah, you got Logan Bartlett in the deal. Just attributing, he's like the main guy now in Red Point. Of course, attributing every deal at Red Point.

Classy.

No, Jeff Brody.

Polytechnic alum.

Let's go.

What's that?

The founder of Red Point went to my high school.

Oh, no way.

Yeah, Jeff Brody. And so, what a dog. Yeah, I've never met him. But just being like, shout out Red Point.

Yeah. Shout it out.

I would die for Red Point. I would die for big tech. And also I want to give a shout out to Captions App, the founder's been on the show. Big fan of captions. They introduced Mirage Studio, powered by our proprietary Omni Modal Foundation Model. They can generate expressive videos at scale with actors that actually look and feel alive. Our actors flinch, laugh, flinch, sing, rap, all of course per your direction. Just upload an audio, describe the scene, and drop in a reference image and create energetic content in minutes built for marketers, creative teams, and anyone serious about crafting great narrative videos. I'm excited to get that started and check that out. It's actually a new URL, mirage.app. So they're kind of like forking the two products.

A lot of companies have been doing that with, I feel like that's a more recent occurrence where they'll spin up kind of a new brand and app to kind of test something. Yeah, yeah. More news. Apparently Elon is doing some type of share sale, $300 million share sale that values the company at $180 billion. 13 billion for XAI.

Oh, XAI. And.

The official Neuralink round actually got announced today.

Oh, that was just leaked earlier. It's actually out now. That's great.

Yeah.

Well, congrats to everyone at Neuralink. I believe it's 9 billion now. Is that right?

Yeah, crazy.

Speaking of things that start with nines, my sleep score 91. Back in the game.

Back in the game.

Yeah, you did. get an eight sleep, five-year warranty, 30 night risk-free trial, free returns, free shipping.

I had to fuel your best days. I didn't issue the power went out.

In the middle of the night Saturday in my neighborhood. And so I don't care about Saturday. I want Sunday. I got a 90.

90.

Oh, let's hear it for me.

Let's hear for the big dog.

Once a week, I'll give it to you.

Once a week.

I got back to back days coming up. I'm.

I'm. I got you.

Let's see what you tonight. Let's see what numbers I put up.

Anyway, nobody outsleeps me. Anyways, thanks so much for watching. We will see you tomorrow. Go leave us a five star review if you like the show. And we appreciate all of you. We're excited for tomorrow.

See you later.