Jim: Today’s guest is Samo Burja. He’s one of my favorite people to talk to. He’s a real thinker. Samo is the founder of Bismarck Analysis, a consulting and publishing firm that investigates the political and institutional landscape of society. One of their publications is Bismarck Briefs on Substack, of which I am a paying subscriber. The thing ain’t cheap. Ain’t cheap at all. In fact, my most expensive subscription, but I find it to be well worth it. So if you wanna have a deep view into what’s going on, consider subscribing. One of the things I’ve learned as a lens from reading Bismarck Briefs is this idea of live players and dead players. Talk just a minute or two about that concept.
Samo: The difference is one of the most basic observations we can make about the world is that the vast majority of people act on scripts, you know, on essentially roles that have been defined in advance. So this might be something like a professional standard. This might be something like a standard business plan. This might be something like, just, you know, the way things have always been done. Now, no matter how good this script is, if you are in a competitive context with a live player and following just this script, you will end up losing. You could analogize it to a chess game, a game of chess where if you just were following moves from a playbook, I think you would be quite quickly defeated. And because of this, I think a live player is someone that is able to break script and is able to adapt to circumstance and is able to, on the spot, adapt.
Jim: Yeah. This is a hugely important lens. And many of the people I know, they must have all picked it up from Samo, or maybe you stole it from somebody else. I don’t. But, we use it all the time. And a good example to my mind is Kmart. Right? Kmart was the world’s largest retailer back in its heyday, probably reached its apogee. And it just sat there and watched as Walmart came from Arkansas and slowly grew exponentially and just ate its lunch and didn’t do a fucking thing about it. Right? It was a dead player. It had a script. It followed the script. It had no ability to edit the script at all. It did nothing. It just sat there and watched its lunch getting eaten. And while Walmart, of course, then would be the live player, and they came up with a couple of deep insights into the nature of retail, which gave them a sustainable competitive advantage unless somebody else adopted it, which they didn’t. And they just executed relentlessly and adjusted on the fly continuously and defeated Sears and Kmart, the two biggest retailers in the world. Quite interesting.
Samo: Generally speaking, the reason why being a live player is very effective is because circumstances change. No matter how well designed an institution, no matter how high the standard of a profession is. You know, Jim, would you prefer to have a doctor, the best, the most professional doctor from 1950, or would you prefer to have a doctor that reads all the crazy papers today, even if sometimes they go on long rambling tangents about the latest crackpot belief they acquired? Right? So even in professions where we value reliability, such as doctors, lawyers, et cetera, you want people who introduce new things to be done in their field. Otherwise, over the long run, they become totally obsolete. It’s like a computer program that was just unchanged for seventy years. The program might be fine, but the very environment in which it’s embedded has changed completely.
Jim: I give a little tactical example of your live player, dead player examples. They’re not quite dead, but let’s call them zombies. I see this in people that I interact with, particularly older folks. You know, I’m not exactly a spring chicken, even though I don’t act my age, which I am happy to be the world’s oldest 19-year-old. But, there’s lots of folks who are refusing to adopt LLM-based AI in their approach to the world. And man, I yell at them and say, you know, you are not taking advantage of one of the most important things that have happened in all of our lifetimes. And that is to learn the tool. Because if you don’t learn the tool, you can do stupid things with AI. Right? But if you learn the tool and incorporate it into your routine on how you approach the world, it’s unbelievable. Even something as dumb as I want to find the best restaurant that sells fried shrimp within 50 miles of Corpus Christi, Texas, which I actually did using the deep research mode of ChatGPT Pro. And it did this, and it was seamless. And I gave it some geographic preferences, and it came up with the perfect result. It was amazing. Right? But on the other hand, I also get it to help me think through really deep things. Like I’m working on a new way to state emergence to avoid most of the problems with other definitions of emergence. And man, ChatGPT Pro deep research is huge. So if you’re not using the AI tools appropriately in your professional life, if you’re at all involved in symbolic manipulation, you’re at least a dying player, if not yet quite a dead one.
Samo: I think that it is definitely the case that you cannot always rely on the same intellectual tools you’ve always relied on. Say, for example, if one were to, you know, and this was never quite as true as people thought it was, but for example, if one was used to simply trusting the news, trusting the broadcast in, you know, the fifties, the whole Walter Cronkite thing, the seventies and so on, I think today one would be radically misinformed about the modern world.
Likewise, you know, search was this wonderful tool back in 2002, 2005. When I talked to people who are 25, they don’t even believe search used to work. Now, of course, search used to work because the Internet was itself curated. Right? The Internet, and who wrote on the Internet, was self-selected for the more educated, the more thoughtful, the more autistic, right, the more systematizing mindset. And no one had gamed the search engines yet.
Jim: That’s the key right there.
Samo: So if today, you rely on search engines, and even let’s even say that they work just as well as they did in 2015, because of the Internet having been adversarially optimized and also the quality of stuff being lower, I think that you end up having a situation where your cognitive tools work worse and worse.
Jim: Absolutely. That’s a fine example. And then, because around 2005 is when ad support became ubiquitous across the Internet, and that produced a horrific arms race of search spam, horseshit, bullshit. And now that the cost of producing this shit’s gone down to almost zero with LLMs, it’s totally out of control. Straight search engines are utterly useless. I use Perplexity, which is extraordinarily good tool that combines AI on top of its own search and its own proprietary database. Amazing. I use Google like, to find a Mexican restaurant open right now. That’s about it. Right?
Anyway, I thought this was an interesting side digression about live player and dead player, for at least for me, an important lens that I got from Samo. And, you know, people who read Bismarck will pick up cool little tricks like that. So anyway, let me finish the introduction here. In addition to running Bismarck, he’s also a senior research fellow in political science at the Foresight Institute, where he advises on how institutions can shape the future of technology. This is Samo’s sixth appearance on the Jim Rutt Show. Ain’t that cool? Right? I think he’s right up there with Forrest Landry in that he might be an episode or two behind, but he’s one of my most popular guests. Most recently, he was on EP 244 where we talked about lessons from the Russo-Ukrainian War.
Today, we’re going to talk about one of the Bismarck Brief Substack essays called “Anduril’s Plan to Modernize the US Military.” Anduril’s an interesting name. It’s a name of a company that was—we’ll talk about its history—but it comes from The Lord of the Rings. And regular listeners probably know I’m a Lord of the Rings nut, having read the trilogy 35 times and basically have it memorized, not quite, but close. Anduril was Aragorn’s sword, and the analogy works at multiple levels because Anduril was actually a reforging of an older sword called Narsil, which Elendil, the head of the Westernesse people who were loyal to the elves, used Narsil to essentially kill—not quite kill, but at least damage Sauron—and his son cut Sauron’s finger off with the ring in it. And then it was broken and was not rebuilt until the time of The Lord of the Rings. And one could say the silly analogy, maybe not so silly, maybe this was intentional by the founders, was that we used to have a sword good enough to beat Sauron, say World War II. Right? And maybe we are losing that, and maybe we need to reforge a new sword in the lineage of winning World War II, but in a different place. So, anyway.
Samo: The analogy that they are making by naming their company that is, you know, self-explanatory. There’s an implicit view that, you know, the sword of the Western world has been, like, broken or shattered, though, of course, there’s been no crushing military defeat. So if we interrogate what would sort of be the thesis that Palmer Luckey and the other founders of the company and the, of course, major investors, like Peter Thiel, are pursuing is that our capability to make these new weapons and these technologically advanced weapons is atrophied. So we can only really think of Anduril if we at the same time think about Boeing and Lockheed Martin and RTX and all of these defense primes, which frankly are still building 20th-century militaries in the 21st century. Right? Still investing in the next generation of the manned jet fighter rather than designing cost-effective drones that can be deployed at scale.
Jim: And as we talked about in our discussion about the Russo-Ukrainian War, there has been a learning curve going on there at the speed of light in terms of thinking about warfare in a very different fashion, and huge bureaucratic organizations like the US Air Force. You know, it’s a truism of military history. The generals always want to fight the last war.
Samo: I mean, not only do they want to fight the last war, but, you know, there’s the saying from the early 20th century that war is a racket. I sort of feel like the last war is the best racket. You can always fundraise for the development of a better version of the weapon that dominated the previous war. It’s much harder to fundraise and convince the generals to use a weapon that will dominate the next war. So it’s like the, you know, tyranny of the previous war. It definitely shackles the imagination. Examples, you know, infamously: the Maginot Line, massive investment preparing for a World War II that just didn’t happen. The World War II we got was very, very different than the First World War. The massive trench fortifications ended up being somewhat irrelevant for that mode of war.
Samo: I mean, not only do they want to fight the last war, but, you know, there’s the saying from the early twentieth century that war is a racket. I sort of feel like the last war is the best racket. You can always fundraise for the development of a better version of the weapon that dominated the previous war. It’s much harder to fundraise and convince the generals to use a weapon that will dominate the next war. So it’s like the, you know, the tyranny of the previous war. It definitely shackles the imagination. Examples, you know, infamously, the Maginot Line, massive investment preparing for a World War II that just didn’t happen. The World War II we got was very, very different than the First World War. The massive trench fortifications ended up being somewhat irrelevant for that mode of war.
Jim: The other one from World War II, just as stark, was the major powers were investing vast sums in building battleships all the way up into the thirties. Oops. Quickly found that torpedo bombers and such from aircraft carriers, or from land, could sink their ass without even coming within sight of another ship, and battleships became essentially irrelevant during the course of World War II.
Samo: Aircraft carriers, which were an adaptation to World War II and have been a symbol of US power for eighty years. So we have a lot of cultural attachment to them. They look great, honestly. They’re magnificent machines. Maybe they’re like battleships. Maybe they are thoroughly obsoleted because when your threat model is, you know, a kamikaze plane crashing into your aircraft carrier, or a plane dropping a torpedo, or a torpedo boat, countermeasures like your jet fighters, or your, you know, basic fighters before jet fighters, or an escort fleet, help. But when the threat model is, you know, a vast barrage of missiles from the Chinese mainland, or perhaps even long-range hypersonic missiles targeted with the assistance of a satellite network that from orbit is tracking the location of your aircraft carrier minute to minute… Well, maybe you’re
Jim: …a sitting duck. I had Tim Clancy on here recently. He’s with another strategy guy, and he sort of on the fly, just because he was programmed to think that way, was, “Well, we can invest this money instead at another aircraft battle group.” Or I go, “Oh, so you want to pave the bottom of the ocean in more iron? That’s a stupid-ass idea. I’ll take 1% of that and I’ll take out your fucking aircraft carrier battle group.” You know, talk about a target for asymmetrical warfare. And there are other ways, you know, smart roving drones, some dumb, but wandering drone submarines that are just cruising around looking for large chunks of magnetic field and then shoot torpedoes at them. You can make those things so cheaply, even if they’re only hitting the target 1% of the time, they’re still going to be a good investment to take out an aircraft carrier battle group. But anyway, let’s get back to the topic here of our… well, this is on the topic, but let’s dig in a little bit more to the specifics. I went to the Anduril homepage and I took their mission statement, and I concocted a slightly Rutt-erized version of it. I’m going to lay it out in three parts, and I want you to react to it from the perspective of your essay.
Samo: They basically described their mission as “pioneering solutions for the software-defined conflicts of tomorrow.” Well, “conflicts of tomorrow” already presupposes you have a hypothesis as to which the conflicts of tomorrow will be. And I think part of the idea is that, at least to some extent, the US should be looking at things that are countermeasures to Chinese capabilities. You know, the future is not going to be perpetually just wars in the Middle East. It’ll involve great power competition with countries like Russia and China. And especially with China, it’ll require technological innovation.
Jim: And software is going to be, no doubt, a big part of that. You know, as Marc Andreessen famously said, “Software is eating the world.” Yes. And I’ll just give a personal aside. I used to do my business thing in Northern Virginia, which is swarming with these parasitical defense contractors, guys who in the DC area, not so much the bending of the iron and making the tanks and the planes, but writing the software for defense systems. And talk about overpriced, underperforming, slow, a generation behind technology. That’s them guys for sure. And we’re not going to win a war that way.
Samo: No, that’s certainly the case. And part of the issue here is, well, you know, the interesting thing about Anduril is that the company has made thoughtful efforts to explain the reasoning behind its mission. There’s an entire 47-page manifesto titled “Rebooting the Arsenal of Democracy” that they released in 2022. And as far as I can tell, they’ve definitely been following that philosophy ever since. It’s the bet that autonomous and software-enabled weapons will be the decisive factors on the battlefield. And I think that this thesis is correct. That still doesn’t mean you necessarily know with what to bet on ahead of time, but it does mean that there is a necessity to deploy software-enabled and empowered weapons on the battlefield. So it’s not just things like the design process. It’s, I don’t know, perhaps in the future, we will be managing thousands of drones, not dozens. And the AI automation aspect of targeting, target selection, fire, perhaps even maintenance, will be crucial.
Jim: Yeah. I had, back in EP 247, we had Sergei Kuprienko on the podcast, and he runs probably the leading drone cluster controller software author in the world. They work exclusively for the Ukrainians. And even then, it was over a year ago, a single operator could control up to 100 drones, which was pretty cool. And that arms race is really important. So let’s go on to point two of the Anduril mission, as coming from their website, which is to build cutting-edge technology at speed. I think I would underline “at speed.” Any of us who played games like Civilization know that you better not get too far behind the technology curve in a technology-intensive kind of game theory situation. And, you know, can the big prime defense contractors and their Beltway bandit subs match the speed of China? That’s an interesting question. I think Anduril wants to say probably not. We need a new approach.
Samo: These companies used to be fast, and they used to be lean and hungry. Right? If you look at the history of the development of something like the B-2 Spirit bomber, these companies would take technological risks. And just because it was necessary for their, you know, economic survival, they would achieve results pretty fast. It’s just been a normal cycle of decades of institutional inertia and decline. Forty or fifty years after a multibillion-dollar company is founded, it is very rare for that to be a fast company still. I struggle to think of a company that’s older than fifty years that is still agile and fast in responding to economic conditions, let alone actual conditions of war. It took, you know, Russia years to retool its industrial base for modern warfare. And, you know, they were paying a bill of tens of thousands of dead soldiers every single year, every single quarter. And Ukraine, on the other hand, perhaps precisely because it had less of a military infrastructure built up, immediately bet on some of the newer companies and newer startups. The U.S. is going to come into any major war it enters with a large ecosystem of basically very stagnant and very set-in-their-way companies building very expensive products. In small quantities also. Very expensive products in small quantities. Yes. U.S. foreign policy is already constrained by what the U.S. can manufacture in the military domain. I don’t know if you’ve heard, but some of the recent weapon supplies to Ukraine have been reduced simply because the U.S. cannot build enough of the stuff, and it’s depleted some of its stockpiles.
Jim: As soon as the Ukraine war started, I said, Biden should appoint a George Marshall to, you know, focus the industrial capacity of the U.S. and the West on scaling up just the raw numbers of the best stuff we have today. Because a good historical analogy was World War II again, where the German stuff was almost always better than the Allied stuff. The Panther tank, the Me 262, the Fw 190, you know, the MG 42. Anyway, their machine gun was great. You know, they had all kinds of good stuff, but we built them, and the Russians built stuff, at staggering scale. I think we built seven times as many planes as the Germans did. Our planes maybe were a little bit worse, but they weren’t that much worse. Our Sherman tanks sucked basically, but we built 14 times as many Sherman tanks as the Germans did. We just crushed them with pure numbers. And, now, that’s not likely to be how the next big one goes, but maybe to everyone’s surprise, the Ukraine-Russia war has shown that wars of attrition might still be a thing, and not having the capacity to build stuff at scale could be a real problem.
Samo: I think that scaling production is very interesting. Did you see any of those videos from the facilities building U.S. artillery shells? Now, artillery shells are about as simple a product as you can imagine. They do it by hand and using some equipment that’s, like, 70 years old. So, actually, you know, the arsenal of democracy we do have looks very similar to the arsenal of democracy 70 years ago. Why? Because the volume of demand for artillery shells was so low because the stockpiles were so large. There was never a reason, a profitable reason, to invest in updating or improving the equipment and automating it. And then you put this side-by-side with China during its current military buildup, and you look at how they make munitions. It’s like a human hand does not touch an artillery shell at any point in its production.
Jim: Yeah. That was my point when I said, George Marshall, stop that, people. And even if it’s a loser, the government will fund the capital necessary to transition to robotic manufacture of, as you say, 155-millimeter shells. That’s a 1910 technology. Right? Those shells, other than their sensor in the nose, aren’t that much different than they were before World War One, and we can’t figure out how to…
Samo: …make them. Exactly. At least not make them at scale and make them fast enough. And I think this also points to a structural problem with the U.S. military. I think there is a lot of investment into developing a new shiny toy and not that much investment in developing a military industrial base. You know? Maybe the problem with a military-industrial complex is that it is no longer a military-industrial complex, but a military-corporate complex where it basically has forgotten the industry part and has replaced it with making the perfect quarterly report. You know? Sort of the victory of MBAs over engineers.
Samo: make them. Exactly. At least not make them at scale and make them fast enough. And I think this also points to a structural problem with the US military. I think there is a lot of investment into developing a new shiny toy and not that much investment in developing a military industrial base, you know? Maybe the problem with a military industrial complex is that it is no longer a military industrial complex, but a military corporate complex where it basically values, it has forgotten the industry part and has replaced it with making the perfect quarterly report, you know? Sort of the victory of MBAs over engineers.
Jim: Yeah. When the suits arrive essentially and everything is about quarterly earnings and big bonus. Right? It loses its soul. It loses its mission. And let me now go on to the third point from their mission statement, which was “created by the brightest minds working on modern defense while partnering with US and allied military forces.” They make quite a point about attracting the best talent. While probably the primes are still getting really good hardware people because there are few places to play with really exotic hardware any better than in, you know, state-of-the-art military. It’s not been the case in software. They’ve tended to get the dregs into the military software. And so, Andrew, by doing a number of cultural things, is attempting to win the recruiting war, at least a sufficient percentage of the recruiting war, to stock itself with the best minds. And probably that’s more salient even in the software world than it is in the hardware world.
Samo: I think that the recruitment of young talent is something that is a persistent sort of advantage of tech culture broadly. People, you know, understand what appeals to sort of like talented millennials, talented Gen Z, and it’s definitely not the sort of like corporate career climbing stability that’s not even real anymore. Right? There is no economic stability there. What people want to do is to like quickly work on stuff that matters, and that’s just something that is much easier to offer at a startup than it is at an established company.
Jim: And you mentioned in your introduction in passing the five primes: Lockheed Martin, Boeing, Northrop Grumman. What does Raytheon call itself now? RTX, something like that? Yeah. And General Dynamics. Why don’t you give the audience a quick pencil sketch of what those guys are like?
Samo: Well, I think, you know, one important example is that RTX, formerly Raytheon, is kind of run under almost a portfolio theory of the firm. It might sound like a parody to say that, you know, someone that is an executive of a major company writes a long doctoral thesis on how the perfect defense company produces a hundred products, each of which is 1% of your revenue, making a perfectly hedged company. But no, that’s just how they think. That’s literally what they’re doing. So in other words, it becomes this strange holding company of completely unrelated businesses, of which many are civilian businesses. And, of course, such a thing, even in principle, cannot have a structural expertise or a thesis for the future of warfare. It’s sort of a toolbox, and this toolbox, at best, assuming the tools are well maintained, would have to be picked up by the Pentagon. And then, you know, okay. The Pentagon’s not actually a technological expert. It’s not actually an expert at producing these weapons. There is in fact no room for an active industrialist. As of 2023, also 43% of the sales of RTX was dependent on international sales. Right? And, you know, these component companies that they have, like Collins Aerospace or Pratt and Whitney, and, yes, also Raytheon, these companies are often very exposed to the entire supply chain of globalization. So they all need components that ultimately are manufactured in China, and there’s almost no way to make the quarterly report look good unless you are building something in China. I think these are massive companies. Right? The revenue of the RTX Corporation is like 68 billion in 2023. It’s the world’s second largest defense contractor behind Lockheed Martin. Lockheed Martin is, you know, has earnings also in this space about like a little bit more, 68 billion as well in 2023. That one is at least focused on defense contracts. Lockheed Martin is at least 96% defense contracts. So it at least understands that it should be a mission-driven company. I do think that the aerospace advantages it has have no natural synergy with AI systems or with software. What they’re good at is aerodynamics, and they’re also quite good at stealth. And then the question is, how useful is stealth actually? They did develop it together with Northrop. This tradition of knowledge that is sort of unique in the United States, it’s not been matched anywhere else even though China tries to reproduce it. But, you know, it’s just, unless there’s a computer company… Okay, let me put it this way. It is crippling in 2025 for not a single defense prime to have originated as a computer company, either as a software company or as a hardware company. In 2025, it is crippling because this has been a new frontier of war fighting. And unless there is a company that joins the defense primes as a major government contractor that receives a hundred billion or fifty billion or even ten billion defense contracts from, you know, the DOD directly or from the Department of Energy, I think the US military will simply be, you know, it’ll begin to lag behind countries like China because China does have that very same cutting-edge hardware and software. Yes. Of course, we can quibble that their GPUs are like two years behind TSMC’s. But let’s be real. Any GPUs the US military is using are more than two years behind.
Jim: There weren’t even GPUs twenty years ago. Right?
Jim:
There weren’t even GPUs twenty years ago. Right?
Samo:
Exactly. Is there a custom GPU designed for US military use? Probably not. Is there a GPU designed for custom use by the Chinese military? Probably yes, actually. Because they understand it as a national security advantage. They actually are trying to develop militarized computing hardware because they understand that things like the, you know, chip export ban is intended to restrain them strategically. So they, of course, try to unrestrain themselves strategically.
Jim:
And Anduril, on the other hand, has taken a quite strong point of view on the nature of warfare in the future, particularly around their Lattice software platform.
Samo:
Yes. In short, you know, Lattice is an application that aggregates data from any number of sensors or devices in the field in real time and presents it in a single, easy-to-use interface that’s available on, like, web, desktop, mobile, VR, any specialized equipment you can think of. So the idea that the data from the battlefield should go through this one big platform and channel and be available to commanders and to, of course, to, you know, soldiers and hardware on the battlefield, like, that seems obvious. It seems like something that could have been done in 2005 or 2002 or 2010, but it hasn’t been. Right?
Jim:
Well, it was tried. What was it called? The Warfighter of the Future project or something? Like, $10,000,000,000 flushed down the shitter by some combination of the big Beltway bandits because they’re incompetent. They don’t know how to do software. Right? They rely upon the Beltway bandits who suck. Oh, terrible. Terrible. So, yeah, it has been tried. It was recognized twenty years ago, but it was not executed. So if they can execute, then they could do the play of making the battleships obsolete.
Samo:
Yeah. Lattice is intended to be an upgrade of the current practice of transmitting information and orders through unintegrated communication systems and human chains of command. And note, this is like a small, radical aspect of it, which is the simple truth is every time communication technology has seen a significant upgrade, the military has had to change how it operates. This goes all the way back to the nineteenth century where, during the Crimean War, the telegraph changed the amount of autonomy commanders in the field could expect. They would receive telegraphed orders from London at the highest level. And this really annoyed the generals at the time, but there was a very real way in which they were less autonomous because of better technology. And the US military will have to change how it commands troops in the field once information does not always have to pass through the chain of command. So this is a tricky part. Right? Because you can certainly intentionally cripple information flow to only be the chain of command structure because of military tradition and so on. But you will lose to the first military that understands what a chain of command should look like and what information flow should look like in the modern era enabled by, say, a platform like this. So there is an aspect where the generals are sort of conservative of their own social order. I mean, this even goes to, let’s put it this way: The US Air Force is unreasonably committed to training the next generation of human pilots. Like, they will bend their aim from winning wars very hard to preserve this tradition of, you know, men in fast steel machines zooming from the sky, which honestly, again, I love the romance of it, but that’s kind of like cavalry charge versus a tank. Right?
Jim:
Yep. That it could easily be the case. And it turns out that in dogfighting, G-force is a limiting factor. Humans black out after a couple of minutes of six Gs and about thirty seconds of nine Gs. Well, computers don’t care. You can build them up to take a 100 Gs. Right? And so they could outmaneuver the hell out of a manned plane, or either that the person turns into pudding inside the aircraft. Right?
Samo:
Well, and just to go a little bit back to the software thesis. Right? It might sound boring again in 2025. It’s like, oh, okay. You know, the US military does not know what to do with its data. It doesn’t know how to collect data from sensors. It’s not building a dataset. Why does this matter if, you know, the frontiers of technological development have moved forward so much? The truth of it is something like Lattice is a prerequisite for gathering the data in such a way that you can train AI on it. Right? We train the LLMs on the corpus of all human writing on the Internet. Right? Decades and decades of writing, decades and decades of digitalization. We train the image generation AIs on pictures. We do not have the sensor dataset of what to a machine modern war looks like because no one’s gathered the data. And until Lattice is deployed in the field and until there is an actual battlefield feeding data into a system like Lattice, it will be very, very difficult to apply AI to improving the performance, the battlefield performance, both autonomous equipment, but honestly, also human decision-making. And I think that that could be a true software superweapon. So there is a way in which the ambitious version of Anduril, we, of course, think about manufacturing drones. That’s very charismatic, also very important, and can only be done by a company that is a computing company. But in a more fundamental way, I think there could be a revolution in how military technology is deployed and used.
Jim: Yeah. The machine learning paradigm as a bootstrap. Though, I’ve actually, when we talked to Kurpryinko from Ukraine, he did fess up that even with their pretty high level of instrumentation, the actual amount of data coming from the battlefield is less than you’d think because there’s not that much actual fighting going on in the battlefield. You know, they’ll fight for a few hours and then things are quiet for a while, and that they use a shitload of simulation. And I would suspect Anduril does as well.
Samo: And as we know, self-driving car companies might have driven a million miles in validating their car, but they’ve run 50 billion miles of simulator on their software. So it may be that Anduril can, and anybody else can, move ahead fairly rapidly in doing machine learning with high-fidelity simulations. Well, they will be aware of all of these approaches because some of Anduril’s key technical staff have had a background in working on self-driving cars. The former head of perception was Forrest Iandola, who quit Tesla in 2020 before joining Anduril. Now I think he’s working at Meta in the last few years.
Jim: Got bought off by the big checkbook.
Samo: Right? Of course. I mean, you know, I will say I did, at a time when Zuckerberg was terminally uncool, I did write actually a brief calling him a live player. Because even when I looked at their VR approaches, I realized, wait, they’re actually trying to develop it in a new and interesting way. You know, the optics of it were terrible. You know, we remember that example of, you know, Zuckerberg smiling, walking through an audience of people who had metaphorical and literal blindfolds, who had the VR goggles. For a little bit, that was almost the image of the VR dystopia. So, but it was, you know, technologically, at least, I actually think that they got way more bang for their buck than, say, Apple did with the Vision Pro. So even then, upon looking at it closely, to my surprise, I was like, it’s inevitable that they will adapt to circumstances. And Meta has been considered, like, lagging behind on the whole AI revolution. And now we hear of all of these cases of, like, excellent, excellent technical talent brought in for big bucks.
Jim: Oh, yeah. It’s crazy. I was just talking to a guy the other day who, you know, he’s like maybe a $750,000 market price AI guy in the Valley. He says, you know, Meta’s already throwing $3 million a year offers at him. And he said, “We’ll see. We’ll see how high they go.” Right? It’s nuts. But again, that’s warfighting, really. If you say that this is the strategic pivot of history, even if you spend $10 billion, money well spent perhaps.
Samo: Right. Right. And again, similar to, you know, the other applications of artificial intelligence to the physical world. Right? When we go from the world where AI is writing your email to the world where AI might be serving you a drink over a bar, like, you know, Optimus robot might be, like, you know, might be an excellent bartender, maybe has installed a therapy module or something. I’m not sure. You know? Exactly. You can talk, you can talk to the Optimus robot about, you know, your love life or losing your job or whatever. Like, reproduce, reproduce the barmen, a missing social technology. Anyway, the common point there is that with the autonomous vehicles, you build the hardware because you need to build the sensors. You need to deploy sensors into the world. Now you can sometimes pre-build the sensors in. And I think that I’m convinced that at Anduril, when they are designing drones—both their airborne drones, their underwater drones, and various stationary sensor systems that are designed to monitor and inform countermeasures against other drones—they are packing them with sensors. Right? So rolling out these hardware products and designing them to gather the information that sort of your software will collect, or that you need to train relevant machine learning systems, I think that is definitely part of their strategy, and that’s something that has been discussed in some of the wider material they’ve released. It’s also what I would expect from a modern Silicon Valley company taking a look at what could be done with warfighting in the modern era. And, you know, there’s also this understanding that, hey, this is a market that no matter how good the economies of scale in China are, you know, the U.S. is not going to buy its military drones from China. It would be very stupid to do so.
Jim: Yeah. To the point about sensors, Anduril’s first products were sensors. Right? They originally had a product that was used on the southern border, for instance, and they still offer a number of both land- and sea-based, standalone sensors that can integrate into their Lattice system.
Samo: I’m honestly surprised that given all of the ICE talk and even the talk of expanding ICE’s budget and Border Patrol budgets in the latest bill, in this administration, we’ve not heard as many moves from Anduril to pursue that aspect of sensor development. Because if one imagines a, you know, 24/7 surveilled border, I think that’s possible technologically in 2025. Just no one’s really had a reason to do it so far.
Jim: Yeah. To the point about sensors, Anduril’s first products were sensors. Right? They originally had a product that was used on the southern border, for instance, and they still offer a number of both land- and sea-based, standalone sensors that can integrate into their Lattice system.
Samo: I’m honestly surprised that given all of the ICE talk and even the talk of expanding ICE’s budget and Border Patrol budgets in the latest bill, in this administration, we’ve not heard as many moves from Anduril to pursue that aspect of sensor development. Because if one imagines a, you know, 24/7 surveilled border, I think that’s possible technologically in 2025. Just, no one’s really had a reason to do it so far.
Jim: Well, I got a simple explanation to understand how that fits in as the right solution. You need an IQ of 130. The number of people with IQs of 130 in the Trump administration is approximately zero. Now, one thing that was interesting and quite surprising to me, having been mostly a software and networks guy in my career, that Anduril is heavy in hardware. They make many different kinds of hardware. They have airborne stuff. They got underwater stuff. They’re working on something to be the wingman for fighter planes, etc. Tell me a little bit about your perspective on them actually being in the hardware business. You know, is that necessary for them to bootstrap their company because you’d have the problem of software without hardware? Or do you think that hardware is a long-term integral part of their play?
Samo: I honestly think you have to sell the right hardware platforms to make any money off software. Right? If you think about it, the smartphone is such a brilliant thing and was such a brilliant thing for selling completely new kinds of software. And I think that the companies on the other side, right, the legacy defense primes, they would not have made it easy to make a pure software play. So I actually think Anduril, in this way, I’m more optimistic about Anduril than Palantir. Palantir was kind of almost a pure software play, and it was supposed to just integrate with the computers the defense world is already using. Here, though, I think that there is also, you know, I think the market is a bit different. This is not consumer-focused. This is a monopsony, which means, you know, the opposite of a monopoly. In a monopsony, you have only one customer. That customer is the U.S. government. If you added up all the other NATO militaries, you would still not get a customer that’s as good as the U.S. military in terms of acquiring new equipment and new technology. So, in the entire world, assuming you only sell to allies and U.S.-aligned interests, which, of course, you have to as an American company and you should, you end up having this one customer you have to make happy. And, you know, I think the Pentagon loves shiny toys.
Jim: Yep. That’s gonna be tough. And, obviously, the Grumman’s of the world love their big, gray, heavy toys as well. So it does make sense. At least at this introductory period, they kinda have to do both sides.
Samo: And, also, they do acquire a lot of these companies that have formed to try to modernize the U.S. military before. Like, in 2023, Anduril acquired Blue Force Technologies, which is notable as the designer and manufacturer of the Fury drone. Now, the Fury drone is supposed to be the sort of wingman concept. I’ll be real with you. I do not understand what is the use of having a human pilot, a, you know, F-22 or an F-35, and to be accompanied by four Fury drone jets. Why not just send the Fury drone jets on their own?
Jim: I suppose the theory is the operational awareness at the actual point. The OODA loop is faster with a man-in-the-loop. That would be the argument. Whether it’s true or not, I don’t know.
Samo: I mean, you know, presumably, the pilot is busy flying their own aircraft. I mean, I would assume the F-35 is probably already maxed out the human ability to process information in the moment given the speeds and the complexity of the sensors they’re dealing with there. I will say, though, this seems to be just, you know, if you develop the Fury drone to be flown by a pilot in the air, and it turns out that’s not practical, but the Fury drone is built, well, it’s easy to remove the pilot from the air. So I think it’s that kind of ploy where you’re…
Jim: Oh, actually, another little detail on the Fury. I did look it up to see. The Fury is subsonic. There’s a mismatch there. Right? You have these supersonic, high-performance fighter jets. How does a subsonic wingman actually work in really total kick-ass dogfighting where their guys are gonna be going supersonic all the time? So it may well be this is a Trojan horse just to find an excuse to build the subsonic, which is good enough for many missions.
Jim: Oh, actually, another little detail on the Fury. I did look it up to see. The Fury is subsonic. There’s a mismatch there. Right? You have these supersonic, high-performance fighter jets. How does a subsonic wingman actually work in really total kick-ass dogfighting where the guys are going to be going supersonic all the time? So it may well be this is a Trojan horse just to find an excuse to build the subsonic, which is good enough for many missions.
Samo: We shouldn’t underestimate. While I stand by the evaluation of the defense primes as relatively dysfunctional, they are often homes to traditions of knowledge that were decades in the making. Right?
And, you know, it’s smart for a company like Anduril to try to acquire expertise from a company like Blue Force Technologies. Because, you know, again, aerodynamics and supersonic aerodynamics especially, not trivial. That is, in fact, a different deep branch of the tech tree, or the knowledge tree, and there is no natural transfer from software.
Now, of course, you could say, “Oh, but can’t you use software to simulate them?” And I’m like, “Okay. Well, then we’re talking about setting up a supercomputer and getting the turbulence simulations right. That’s its own company.” You could have a totally great startup on just building a computer system and an AI system that solves unsolved physics problems with fluid dynamics. It really is like an art and a technology at the very edge of our scientific understanding of what’s happening with the movement of air at supersonic, let alone hypersonic, speeds.
Jim: Yeah. Lockheed Martin and Boeing both have great expertise in that area, best in the world, probably.
Samo: Exactly. Exactly. It’s just a problem. That expertise is locked in a dysfunctional management structure. So, you know, I honestly think that one of the best reasons the US government should fund these new startup companies and honestly defund the old defense primes. (I’m going to go there because, you know, I’m not working in defense right now, so I can just say these things.) It will force the engineers to seek new employment. The new companies will be able to buy the engineering teams and the expertise and even some of the hardware from these old firms. Again, these old firms that are not really exposed to market discipline. Like I said, it’s a monopsony.
So, you know, it’s preferable to keep an old dysfunctional company alive if the alternative is that your tradition of knowledge will die. Like, if the US lost the ability to make stealth aircraft, that’s catastrophic. Right? If the US had lost the ability to go into low Earth orbit, which it actually did for a while before SpaceX picked up the slack, to bring people into low Earth orbit, that is catastrophic. But the US is fortunate enough to have other options. And I really do think that prioritizing making the SpaceX of aerospace or the SpaceX of naval warfare or the SpaceX of US drone manufacturing. That’s what the US government should be looking at. That’s the only way to do it is to create these new companies.
Jim: That’s the perfect analogy. Yeah. I think the analogy of SpaceX is much better than with Palantir. You know, Palantir really doesn’t sell much to the defense; their intelligence and law enforcement is where most of their market is. And a lot of that is indeed information collation. In fact, I heard from somebody I know very well that is in this industry. He’s not working for Palantir. And he said, “Oh, yeah. Palantir: Salesforce for killing people.” I thought that was perfect because it’s basically a data management tool.
Samo: Yeah. I mean, to be fair again, the US government is bad at data management. And I think a lot of it is just this problem where people defend existing ways of doing things in existing hierarchies.
You know, in Estonia, they first fired all the bureaucrats because the Soviet Union ended, and then they used computers to replace some of them because it was a small country. You couldn’t hire that many people. And in Belgium, they didn’t fire the bureaucrats, but they adopted the same exact software and hardware tools that Estonia first developed. The end result is that Belgium has something like twice the bureaucrats per capita than Estonia does. You know, firing the bureaucrats and replacing them with computers or building computers that could replace them, these are two separate steps undertaken for very different reasons. Indeed.
Jim: Let me pick on the point you make, the monopsony factor that the big elephant in the room is the US Defense Department at soon to be a trillion dollars a year of spending. And it is, you know, hidebound, dominant, control-freaky, etcetera. But what about this new move that NATO is going to double its defense spending over the next few years? Is it possible that Anduril could have more traction with these European defense projects that are going to be growing at a much more rapid rate? And it’s easier to capture new programs than it is to retrofit into older programs. And also frankly, they’re so far behind making a quantum leap to command and control hierarchies, mesh-oriented networking, smart weapons that are much cheaper. If I was them, I would seriously consider aligning with somebody like Anduril. Is that feasible in your mind?
Jim: Let me pick on the point you make, the monopsony factor. The big elephant in the room is the U.S. Defense Department, soon to be a trillion dollars a year of spending. And it is, you know, hidebound, dominant, control-freaky, etcetera. But what about this new move that NATO is going to double its defense spending over the next few years? Is it possible that Anduril could have more traction with these European defense projects that are going to be growing at a much more rapid rate? And it’s easier to capture new programs than it is to retrofit into older programs. And also, frankly, they’re so far behind, making a quantum leap to command and control hierarchies, you know, mesh-oriented networking, smart weapons that are much cheaper. If I was them, I would seriously consider aligning with somebody like Anduril. Is that feasible in your mind?
Samo: Well, I think that it is most probable to work in the Eastern European countries that have newly increased their military budgets because they will not have… you know, if you create a new budget and a new company comes in and is partnering, if, say, there were an Anduril campus in Poland or an Anduril campus in Estonia, this could be a point of national pride, especially if you allow someone that used to be prominent in an Estonian or Polish context, either through an acquisition from the business world or through someone that used to be in government, to be the face of this, you know, subsidiary company there, then I think it’ll be fairly easy to justify the new budget being spent on Anduril. Meanwhile, if you try the same thing in Germany or France, the new money already has old customers, old providers waiting for that government money. There are already German firms that need a bailout. There are already French firms that are in line to receive that funding. So it really depends on how mature and old the defense ecosystem is. And I think in Eastern Europe, they’re more keenly aware of it. They have much less money, but not as little money as you might expect. Poland’s military spending is serious. Right? Like, Poland has actually become one of the top NATO spenders in military terms. Their new defense budget, I think, for 2025 is $45 billion. Like, that’s a lot of money for a small country like that.
Jim: For a country with 30 million people. Right? That’s moving up towards U.S. level of spending per capita. Not quite there yet.
Samo: They are spending 4.2% of GDP on it. Right? And they are the only European economy that’s growing. So, I don’t know. Let me say this: I would not be shocked if in the year 2040, Poland is spending more on defense than Germany is, both because Poland has actually become by then richer than Germany and because Germany, it just… you know, I think we all wanted a peaceful Germany, but maybe it worked too well. And now they’re kind of insane. They never take the reasonable middle ground. They can only be complete pacifists, complete hippies, or, you know, the opposite.
Jim: Or they could blow up their nuclear power plants while claiming to be green. What the fuck? Right? They just seem to have something in their temperament. I say that Rutt is a German name, so I can hack on the Krauts. In fact, my rather mongrelized American heritage got all kinds of shit in it, but probably the two biggest pieces are Irish and German. So I have coined the only known Irish-German joke, which is, “What is an Irish-German’s idea of a good time?” What is it? Get drunk and take over the world.
Samo: But to go back to Anduril’s hardware offerings, right, they also acquired Dive Technologies, which was a manufacturer of unmanned underwater vehicles. Now the Dive-LD and the Dive-XL are small autonomous submarines, and they can pair with larger numbers of Anduril’s Copperhead underwater drones, which themselves can carry both explosives or reconnaissance and surveillance equipment. And, again, this sort of blanketing of the battlefield with sensors and surveillance, I think this is very important to solve this problem of the extreme scarcity of data coming in from the modern battlefield. The vision of these small autonomous submarines is to just lay sensors all through underwater and then launch autonomous torpedoes as necessary. This seems honestly a compelling update to what naval warfare should look like in 2025. I, for example, find it completely plausible that surface-based naval warfare is just obsolete, that it’s just over because surface ships are too vulnerable to drones and missiles. Like, if you look at what happens to the Russian Black Sea Fleet, right, the Russian Black Sea Fleet, a third of its 80 warships were destroyed by Ukraine, not even with super expensive missiles, but just with drones. And you know what? Let’s say that I believe American ships are about four or five times better at intercepting drones than Russian ships are, I do not believe they are 100 times better. So if you can sink a ship that costs many, many, many millions with a $30,000 drone, I mean, the economics of a surface ship just don’t make sense anymore.
Jim: Yeah. In fact, as we talked about in our last podcast, this may be moving us into this classic what’s happened again and again through history as the balance of warfare moves from favoring the offense to the defense. You know, World War I was the machine gun and the high-speed, small battlefield cannon. They moved things to the defense. World War II, the radio, the tank, and the coordinated air-land Blitzkrieg favored the offense. Maybe it’s looking like smart weapons are favoring the defense. You think about the things you described that dive and similar things can do, and think about Taiwan. Right? You give the Taiwanese a few more years, and it’ll be impossible for the Chinese to do at least an overwater invasion of Taiwan. Let loose 20,000 cheap underwater drones that will just sink any ship they find. And there’s no way you’re going to get across 80 miles of open water. I mean, what…
Samo: What I’m curious about in general is to what extent will the underwater warfare—to what extent will the underwater warfare—just be the sort of, like, you know, kind of a deployed-ish task force, or whether you’re just going to have lots and lots of basically mines, autonomous mines waiting in the sea, just blanketing particular areas of water like the South China Sea. Might it be viable to just blanket the South China Sea in sleeping drones that just wait under the water until they activate? What a blockade might look like could be very different. Can you imagine an island completely isolated from shipping because it is surrounded, like having piranha-infested waters, with, you know, taking a swim as a human? Trying to sail through them might just be sort of physically impossible. It’s not that there are humans on ships chasing you. It’s that, you know, every mile, there are like 20 drones waiting and trying to find your ship to just crash into it and sink it. It might make naval blockades radically more effective if that is the future of warfare.
Jim: Yep. And, of course, it cuts both ways. When people talk about Taiwan, I always like to point out China’s got a horrid, horrific Achilles’ heel when it comes to people playing the blockade game: the Straits of Malacca. Seventy percent of China’s oil comes through the Straits, and it’s like one sub could block the Straits. You know, two troops of Boy Scouts with Neptune missiles or Harpoons could stop—could block—the Straits. So, if the Chinese start playing the blockade game, we have a really easy counter to that. So I suspect that the idea that China can easily blockade Taiwan and win a non-kinetic victory is probably wrong. But we shall see. We shall see. We’ll have to bet another bottle of Slivovitz on that sometime.
Samo: I think one point to raise here, though, is that, you know, the US military has 1,300,000 active personnel, and it has 225,000 land vehicles and 20,000 aircraft, and it has under 300 ships. This means its ratio is 100 humans to less than 19 weapon systems. And then, reportedly, the whole US military only has 11,000 drones. I think people do not understand: when we talk of mechanized warfare as if it is something that has been achieved eighty years ago, we are actually still at a very low rate of mechanization of warfare—100 humans for 19 weapon systems. You know? When did farming reach that level of automation? Probably around, like, 1890, maybe 1900, you know, 100 farmhands for 19 pieces of equipment. Right? We are actually very early in what should be thought of as the automation of warfare.
Jim: Yeah. Indeed. And let’s go back to the 10,000 drones, and this also impacts Anduril. I looked up the, you know, approximate pricing for some of the Anduril hardware, and it ain’t cheap. Right? It’s less expensive than the prime stuff, but Ukraine and Russia are each producing on the order of a million drones a year. A million. And, you know, some of these drones are a couple hundred dollars that the Ukrainians are building. And even the ones that are relatively long-range—a few hundred miles—cost $10,000, $15,000. I mean, and in a world where the equivalent Anduril products cost a million… And after the war, both Ukraine and Russia will have the capability to flood the world with unbelievably vast quantities of lower-end, you know, drones and self-piloting undersea torpedoes, and things of that sort. What do you think that does to the Anduril business model?
Samo: I think that it actually stimulates the Anduril business model because, you know, while Russia will sell to countries like Iran, Ethiopia—basically, it will sell to many of the same Third World countries and conflicts—the US side will be very reluctant to fund, I don’t know. If there’s a civil war in Libya again, or a continuation of the previous one, and the US is funding one side (which, of course, inevitably, it will do; maybe it’ll even fund both sides), they’ll be reluctant to give money that ultimately ends up in Russian hands. So, if the only drones on the market were the Russian drones, in a way, when Russia deploys or sells drones to one side in a war or conflict, this will create, as a condition of US support, demand for a US drone that responds to it. So I think war is kind of interesting, right? It’s not like you sell cars and eventually you saturate the demand for cars. But, you know, if these are autonomous tanks destroying each other, then if I sold a million to my allies, then my enemies get to sell a million to the enemies of my allies as well. So, in a way, it’s almost a collaborative business model. It produces an arms race, and it induces demand.
Jim: Yeah. But let’s imagine it’s
Jim: Yeah. But let’s imagine it’s
Samo: not just Russia. It’s Ukraine also selling to our side, of course. So perhaps Ukraine will beat… will beat Anduril. But, you know, the fun question is, like, the U.S. is not always kind to its allies when it comes to defense procurement. Right? The French, they got a little bit burned when they were going to sell these submarines to Australia. And then Australia and the U.S., after, you know, France had already invested all of this money into building up the tooling, the equipment, the development, they snatched the contract away from them to the long-term detriment of the French naval industry. Once Ukraine—if Ukraine ends—if the Russia-Ukraine war ends, it’s not clear that the U.S. will want to subsidize the Ukrainian industry rather than just say, “Oh, thank you very much. We’ll build our own now.”
Jim: Let’s see. I suspect that they have so much learning curve now that they can hit the market at a very low price. You wrote a fair bit towards the end of the piece about what impact Anduril is likely to actually have on America’s ability to fight the true twenty-first-century war. Why don’t you talk about that a little bit in general, and then let’s pivot specifically to some hypothetical conflict with China.
Samo: Well, I… I think, you know, Anduril is likely to succeed at bringing cutting-edge, you know, modern military technology to the U.S. military. So, in other words, I expect they will be able to catch up to both allies and enemies in the domains of drone warfare or deployment of sensors to the battlefield. This technology was not innovated by U.S. defense contractors. Actually, it was not even the startups that did it. It was sort of like, you know, everyone from the Ukrainians to the Iranians. Of all things, Iranian military R&D units, like, managed to get the cost of their drones very low. But, ultimately, it was also just a lot of China’s hardware startups where they sold dual-use stuff, technologies used by both civilian and military applications. So just commercial drones kind of ran ahead of military drones, and I think that’s something that’s good to keep in mind. It was sort of, oh, it turned out that weaponizing basically civilian drone components would get you further and be more cost-effective than anything descended from, say, the early 2000s, like Predator drone programs and so on and so on.
You know, I think Anduril is also likely to succeed at bringing cutting-edge Silicon Valley software to the U.S. military. And, you know, I think this will combine to shore up U.S. power in a wide array of contexts, against many potential adversaries and opponents. I do think one thing that I’m not sure that they will succeed at is building an industrial base that could compete with China. Because that’s sort of almost an unreasonable mission to give to a single company. How can a single company outproduce the world’s factory? Right? But to be credible in a serious, longer-term war with China, the U.S. would have to have a comparable industrial capacity. So I think we need to start talking not about state capacity, but industrial capacity when it comes to manufacturing these key weapon systems.
And here, I don’t know, I could see them eventually developing something similar. And, and, you know, they’ve talked about it, right? This kind of big factory-inspired facility, highly automated, large-scale, that is just producing military equipment. But I think the ultimate vision of it will be perhaps something closer to U.S. reclaiming dominance in drone manufacturing, starting with the military application. I think that would take, like, a live player. I could see, you know, if this had been Elon Musk’s sole goal in 2015, I sort of see he could do it. And an example of this is Tesla. Tesla is still, if barely, competitive in the Chinese EV market, electric vehicle market. So it is possible for a U.S. company with a strong, singular focus to compete on mass production with China. But I do think it’s challenging. I think it really is to sign those sort of, like, top talent, like, barely, barely doable, barely doable competition, just because the economics are so brutal. Right? If you build nothing, building something is much harder than if you’re building everything. For every extra marginal thing you build, it’s just easier and easier and easier and easier to get the economics working.
Jim: And, of course, Tim Cook said about building the iPhone in the United States, and he said, “Well, one of the problems is China has two football stadiums full of industrial engineers. All the industrial engineers in America would fit in a medium-sized theater. Basically, we have a few thousand, they have hundreds of thousands.” And so we have a whole bunch of just learning curve to go on to reclaim the ability to mass produce at scale. That’s going to be… unfortunately, we don’t have the political vision or the political will to make relatively modest investments, say a hundred billion dollars in, you know, refiring up the ability to mass produce, you know, good, high-quality military equipment in this country.
And to that, actually, when you were speaking, I had a thought. One possible way that Anduril could punch above its weight is if it developed an ecosystem similar to that of the car industry. A company like General Motors used to employ 500,000 people, now it employs 50,000 and still sells the same number of cars. And how is that? Is that it has basically developed a whole ecosystem of parts manufacturers. I don’t know how Anduril does this, but if it is able to gain the ability to scale, there may be a whole economic ecosystem that could evolve around being a components and parts manufacturer for Anduril. And Anduril is like the equivalent of General Motors and does the final assembly, or Apple. Apple makes chips basically. Right? And they don’t even assemble it. They hire somebody else to do the assembly. And that’s a way you can scale faster with less need of capital.
Samo: I think there are things in this space. And I think also if there was a good way to do a version of friend-shoring, that is, you know, partnerships with various aspects of this extended economy the U.S. has, then the competitiveness maybe becomes a factor. Because, you know, if you think about it, NATO-plus, kind of like NATO-plus, let’s say, Japan and Australia and New Zealand and South Korea and Taiwan, you know, that is about a billion people. And that is then also an economy that by GDP is twice the size of China currently. So even if we’re very pessimistic about what GDP numbers mean, the combined industrial capacity of the U.S. plus its allies could match that of China. And you could have, you know, even some of this version of friend-shoring allow you to have all of these parts be made in places that have kept various aspects of this industrial capacity around. Perhaps less and less Germany, but still more and more South Korea. Right? South Korea can make American-designed nuclear reactors cheaper than anyone else in the world. Actually, they make those reactors cheaper. Their nuclear reactors are cheaper than the Chinese do their own, which is an impressive industrial achievement for South Korea. It is true South Korea would be not too far from the front lines and would probably see some bombing if there was, you know, a World War III type situation. Still though, I think that there’s a lot that could be done in this space. And, you know, if we think of even some of these products of the defense primes like the F-35, it’s branded as American. It is American-funded, but extensive partnerships with the French, Spanish, and Japanese. And, of course, the Japanese buy the F-35 too. It’s not just that they make components for it. It makes the economics of developing such an expensive piece of equipment better. So I do think that there is some hope for matching the economies of scale of China, but it requires this, like, very clever reform of government spending and actually some live players to pursue these economies of scale. And truly, like, the future of warfare is that quantity will be a quality of its own.
Jim: Indeed. As Philip Anderson famously said, “More is different.” Right? And it was a fundamental statement that started the complexity science movement, essentially. And particularly when—oh, this is an insight I had this morning while taking a shower. Always a scary time to be thinking—which is, one of the interesting things about the smart battlefield is that training is hard, right, of humans in the military, particularly for the Europeans where they’re still conscript-based to a substantial degree. So they have a lot of turnover in the lower ranks. Even the U.S. has a substantial turnover in the lower ranks, not nearly as high as the conscript companies. But one of the beauties of anything digital is you can make a copy for free. Right? And if you can get one copy to work right, all the equipment can have the new copy in a day or two, or a month if you want a staged rollout. And that’s going to be another kind of game theory advantage of moving as much smarts as possible out to the devices and not relying on the humans.
Samo: Yes. That sounds right. One of the things that is important to understand is that, you know, when we’re talking about these numbers like a million drones, these numbers are only going to go up in any sort of war between countries with industrial or technological capacity. I would not be shocked, for example, if, say, Pakistan and India were to wage a war in, like, ten years or so, if those numbers might not be 10 million drones or 20 million drones.
Jim: Yep. Those countries are big. Or if the U.S. and China… I still think that the China, Russia, U.S. war thing is probably overstated. It just doesn’t make any sense for either side, but should it happen, man, the number of drones that China could produce, God knows how many. You know, they could make 100 million a year probably. Right? The U.S. could be at a big disadvantage there. We talk about Anduril here. Do you see anybody else on the horizon or appetite for other investors to build other 21st-century
Samo: …economically significant defense contractors? There’s certainly appetite from investors to pursue them, and there are many smaller companies and firms that have, like, specialized theses. I think that Anduril is just the first in what will be a new industry. And perhaps, you know, perhaps they’ll end up buying their competitors. The U.S. government does prefer a larger contractor. Right? Like, the mergers, like, Northrop Grumman was a merger of two companies. Right? That’s why it has two names. Right? And Lockheed Martin. Exactly. Right? There was a consolidation in the 1990s. And I think for now, the U.S. government might continue pursuing it in that way. I do think, however, that almost the hardest barrier to entry is to simultaneously have the right contacts to make the sales work—so the right contacts in the Pentagon and in D.C.—and have the ability to discern and recruit good talent from Silicon Valley. Doing both at the same time is difficult, and especially the first one takes many years to develop. So I think there’s an incumbency advantage here where I think both Palantir and Anduril, they both have an advantage in their respective field. There will be a second or a third company that will eventually come to compete with them. Someone will make it. It will be, you know, some company somewhere. However, there’ll be a lot of startups that try to make their way, eventually run out of money, and then maybe they get acquired by Anduril. There’s just nothing wrong with their technology; they just could not make the sales work.
Jim: Yeah. And the other one, as you point out, for good reason, something like the US Military that needs to guarantee continuity of supply over decades. Do you really want to bet on a company with a $15,000,000 capital base? Probably not. Right? A diversified company with a much bigger capital base is a far more reasonable bet for something that you may need to depend upon for ten or twenty or thirty years. So one could actually see that dynamic. So, last question for you. Let’s pop up all the way to the top. When we think about things like—this was a term the Russians used in the eighties, one of the reasons they decided to fold—is a revolution in military affairs is occurring. What do you think that this game of new smart weapons, totally new paradigms, is going to do to the global geopolitical competition? Let’s frame it from the top down. What do you think that all these changes imply? Anduril and its competitors, and the Chinese, and the Ukrainians, and the Russians. What’s it mean for geopolitics, say, over the next ten years?
Samo: I think that the US will have a persistent advantage in terms of, essentially, I think the US will have a persistent advantage in terms of AI scientists. I think the US will actually have a significant advantage there, but a lot of the advantages in the software engineer space and so on will kind of vanish. And I think we will see a greater and greater dependency of Russia on Chinese-provided services and equipment. I think they will not be able to develop sort of their own stack there.
Jim: Okay. But what about the combination of all the various factors? For instance, the Chinese industrial muscle, you know, does that give them a potentially coercive ability if they wanted to use it? Which I’m not sure they do, by the way. I do think there’s an awful lot of China bashing that goes on. But anyway, assuming that things do get ugly, what other advantages in this space does China have that may offset the American benefit in AI expertise?
Samo: I think that the industrial advantage is something we talked about for a while. I think it’s very much still there. It’s true. And in the entire world, I do think that it will be China that will be closest to the US in AI capabilities. Right? So if, you know, any sort of super weapon the US can build, if the US builds a super weapon, but China does it cheaper and faster, if a little bit later, I think that’s, you know, it can match US capabilities one to one. And I think that’ll be a very different world than the world where the US had military hegemony, which is the world we lived in until quite recently.
Jim: Alrighty. Want to thank Samo Burja for an extraordinarily interesting dive into Anduril and some context on what it may mean for the big picture going forward here. Thank you very much for coming back on The Jim Rutt Show.
Samo: Yeah. Thank you for having me.