Latent Space · 2026-05-18 · 119 min

FPV Drones -The Next War Is Already Here — Yaroslav Azhnyuk, The Fourth Law & Noah Smith, Noahpinion

Brandon AndersonNoah SmithYaroslav Azhnyuk

YouTube ↗

TL;DR

FPV drones have displaced artillery as the battlefield’s main casualty producer, with Azhnyuk attributing 70–80% of frontline casualties to them. Ukraine manufactured about 4 million FPVs last year and publicly targeted 7 million this year; a roughly $500 drone compares with a 155 mm shell priced near $4,000. Artillery remains necessary, but FPVs now occupy the “god of war” slot formerly held by artillery.
Autonomy is the compounding layer that turns cheap drones into software-defined weapons whose capabilities can change overnight. Terminal guidance alone raised one operator’s mission-success rate from 20% to 71% and expanded his kill zone from 3 km to 10 km. Azhnyuk argues full autonomy could multiply accessible operators by 100, mission success by 10, and utility per drone by 10: “four orders of magnitude” more capability.
Electronic warfare does not invalidate drones; it forces a choice among radio, autonomy, and fiber-optic control. Fiber can avoid radio jamming and radio shadows, but a 20 km system can add roughly 3 kg, constrain maneuverability, and become problematic by 30–40 km. AI-data-center demand reportedly drove the same fiber from about $4 to $32 per kilometer within months: “We’re out. We sold it all to the Americans.”
China’s manufacturing base is the episode’s largest strategic tail risk. Against Ukraine’s 4 million FPVs, Azhnyuk posits that China could make 4 billion autonomous, battery-powered systems—including fixed-wing impactors reaching 200–300 km—and distribute millions from ships or containers. He would not declare China the supreme conventional power without combat evidence, but “we cannot count that out.”
The West is behind across four connected layers: autonomy, mass manufacturing, components, and rare-earth processing. Azhnyuk’s answer is to treat Kyiv as a “defense valley,” import its battlefield feedback loops, reform procurement, and build allied industrial scale before a crisis. His stark comparison: Ukraine and Russia moved from drone technology “of 2025” to “2026,” while Europe moved only from winter to spring 2022.
Counter-drone investments must survive the cost-exchange-rate test, not merely demonstrate a successful interception. Azhnyuk contrasts a roughly $3 million laser requiring about three seconds per FPV with 6,000 drones costing the same amount; he doubted it could handle even 600 arriving together. Shotguns, jammers, nets, porcupine-like armor, and cheap autonomous interceptors matter because “the economics matter.”
AI safety and military deployment are inseparable because a faster adversary can make restraint existentially costly. Azhnyuk says the host’s “level 6” AI-command scenario could technically be done now, with deployment across command levels the harder problem; the hosts press him on distribution shift, target selection, and machines allocating human risk. His provocative forecast is that within 5–10 years, using less-precise non-AI weapons “will be immoral” because they may cause more collateral damage.
Drones transform warfare without eliminating combined arms, social endurance, or the need to hold territory. Even with drones causing most casualties, operations still integrate infantry, aviation, tanks, reconnaissance, logistics, psychological operations, and dozens of drone types; Azhnyuk does not see the rifleman disappearing soon. His final prescription is correspondingly broader than a product pitch: “Be prepared for war. Got to invest in defense and security.”

Digest · the substance, structured for research

1. Russia’s invasion converted a pet-tech founder into a defense builder

Azhnyuk and his fiancée landed in Kyiv around 11 p.m. on February 23, 2022, despite a cab driver warning that everyone else was leaving. Eight hours later, bombs were falling; his family packed, improvised diesel containers from emptied windshield-washer bottles, helped friends escape, and spent 17 hours driving west without knowing whether Kyiv would hold.
His route into technology ran through applied mathematics at Kyiv Polytechnic, an academic family, entrepreneurship, and Petcube, which he built while living in San Francisco from 2014 to 2020. His compressed biography became: “I went from making cameras that fling treats to pets to cameras that fling explosives to the occupiers.”
Early work included lobbying American lawmakers for support and Lend-Lease, helping establish the Brave1 defense-innovation cluster, and supporting the D3 fund associated with Eric Schmidt. By 2023, Azhnyuk concluded that the war would persist, drones had created the first software-defined weapons platform, and he should harness the “global techno-capitalist machine” for Ukraine’s defensibility.

2. Immediate survival outranks hypothetical dual-use harms

Azhnyuk describes a period of “soul searching” over building things that enable killing, but Russia’s occupation, abducted children, sexual violence, economic destruction, and stated assault on Ukrainian nationhood resolved it for him. “The only morally right thing to do is to fight back. And it is immoral not to fight back.”
Noah Smith’s pushback — worth keeping: weapons built for a morally defensible national struggle will not necessarily remain there, so their future customers and uses create responsibilities for the maker. Azhnyuk concedes “there’s definitely some responsibility,” but treats that as a governance question shared by knives, fire, and large language models rather than a unique indictment of autonomy.
His decision rule is situational rather than absolutist: a mythical unified government might calculate every downstream consequence before acting, but countries face adversaries developing the same technologies independently. “When you’re in a forest in front of a wolf, you’re first going to deal with the wolf that wants to eat you and then you’re going to go consult Greenpeace.”

3. The Fourth Law is building an integrated autonomy-and-imaging stack

The Fourth Law began with a goal of “massively scalable drone autonomy.” In parallel, Azhnyuk and his Pexip co-founders, partners, and friends started a thermal-camera venture called Ad Systems; the businesses were moving closer together and might merge into what he called a leading Ukrainian team in drone AI and thermal imaging.
The operating stack has three units: cameras, autonomy, and complete drones. Daytime and nighttime cameras plus autonomy modules sell to more than 200 Ukrainian drone manufacturers, while the UAV unit supplies FPV strike drones, bombers, and interceptors to Ukraine’s armed forces and government.
Its interceptor portfolio includes systems against Shahed-type attack drones and ISR reconnaissance drones, but not complete deep-strike drones. The battery-powered Zero interceptor reaches 326 km/h, versus roughly 220 km/h in Shahed cruise and as much as 280 km/h in its terminal phase.
Autonomy spans approximately the last 500 meters of terminal guidance, bombing, target detection, and autonomous navigation across seasons, terrains, day and night, quadcopters, and fixed-wing aircraft. The group also operates simulation and warfighter training and was preparing to construct two plants producing thermal-camera semiconductor sensors.

4. Fiber optics trade jamming resistance for payload, range, and cost

Azhnyuk separates two frequently conflated technologies: “AI replaces a human,” while fiber optics make communications more resilient. A fiber-controlled drone can still use AI, including one system controlling hundreds of aircraft, but the cable itself principally defeats radio jamming.
The overlooked problem is radio horizon. Ukrainian FPVs descend toward ground targets behind forests, hills, and Earth’s curvature, losing video or control in the final tens of meters; a U.S. Drone Dominance competition reached about 10 km, while Ukrainian buyers expect at least 20–25 km and often strike at 30–40 km with ordinary 10-inch FPVs.
Terminal guidance emerged partly as the answer: an operator locks a target from perhaps 200–500 meters away, then the drone continues after the radio or video link disappears. Fiber solves the link directly, but a 20 km installation can add roughly 3 kg; longer systems require larger drones, surrender explosive payload, and become highly problematic around 30–40 km.
Fiber also stopped being unambiguously cheap. Azhnyuk says AI-data-center demand helped push prices from roughly $4 to $32 per kilometer within months; factories told Ukrainian and Russian buyers, “We’re out. We sold it all to the Americans.” Russia scaled fiber first, but he judged Ukraine mostly caught up as of perhaps three months before the conversation.

5. FPVs have dethroned artillery without making older weapons obsolete

The battlefield is saturated: Ukraine manufactured about 4 million FPVs last year, Russia perhaps 20% fewer, and Ukraine announced a 7 million target for this year. Azhnyuk would not disclose fiber-versus-autonomy damage shares, but offered the larger number: FPVs cause 70–80% of frontline casualties.
Artillery once earned the title “god of war” by causing roughly 80% of casualties; FPVs now occupy that ranking. Azhnyuk nevertheless resists a one-weapon thesis: artillery, mortars, armor, aviation, and other systems remain necessary even as their roles change.
His economics are deliberately approximate: a NATO-standard 155 mm shell costs around $4,000 against perhaps $500 for an FPV. Including gun amortization, survivability, and tactical flexibility, he calls the FPV potentially “three orders of magnitude” more versatile—not a one-for-one firepower replacement, but a radically better cost-to-effect instrument.
After Rheinmetall’s CEO reportedly dismissed Ukrainian drones as uninteresting products made by housewives, FPV manufacturer Alexey Babenko supplied the sharper comparison: “The drones we manufacture in one day will be more than enough to destroy all the tanks Rheinmetall manufactures in the year.” Azhnyuk estimates roughly $500 per drone against perhaps $5 million or more per tank, though protective technology might eventually revive armor.

6. Terminal guidance already rewrites battlefield arithmetic

Azhnyuk’s “full autonomy” does not mean humans never choose violence. A person initially specifies the target area, as with artillery; the drone then takes off, navigates, finds the designated area and target, bombs it, assesses damage, returns, and lands. “A human decision is always made at some point.”
His five-level taxonomy is specific: level one, terminal guidance; level two, bombing; level three, autonomous target detection and engagement decision; level four, autonomous navigation; and level five, autonomous takeoff and landing. It was inspired by self-driving classifications but does not map directly onto them.
The strongest field specimen comes from an operator with the call sign Grom. Adding only level-one terminal guidance reportedly lifted mission success from 20% to 71% and expanded his kill zone from 3 km to 10 km by defeating jamming, radio shadows, navigation errors, and the final aiming burden.
Full autonomy’s claimed multiplier combines 100 times as many potential users because months of pilot training disappear, 10 times mission success, and 10 times utility because a recoverable bomber replaces a one-way kamikaze. Applied to an already cheap platform, Azhnyuk calls that “four orders of magnitude” of additional capability.

7. Battlefield autonomy is an eight-dimensional deployment problem

FPV teams ideally work beneath persistent ISR coverage. DJI Mavics may stay aloft for about half an hour, while larger fixed-wing systems can observe for hours or nearly a day; those aircraft designate targets, confirm strikes, and determine whether another drone is required to finish a damaged vehicle or position.
Azhnyuk maps autonomy across eight dimensions: autonomy level, platform, operating domain, higher-order behavior such as swarms or drone carriers, physical environment, command and control, supporting infrastructure, and distribution scale. A fixed-wing, level-three, ground-to-air interceptor in winter is therefore a different engineering and operating problem from an autonomous quadcopter bomber.
The supporting dimensions are often the real bottleneck. Simulation, data collection, security, software deployment, and coordination across tens of thousands of assets must integrate with battlefield command; deploying 100 systems and deploying 100,000 are “two very different ball games.” A prototype is not a fielded product, and a fielded product is not mass capability.

8. AI safety becomes a race against a faster adversary

A host raises the machine-learning objection directly: fog, seasons, terrain, and novel targets create distribution shifts that could make computer vision fail catastrophically. Azhnyuk answers comparatively—mortars can miss by roughly half a kilometer, whereas autonomous drones may reduce friendly fire and collateral damage rather than introduce precision risk from zero.
His most provocative prediction is conditional but clear: “Five to 10 years from now it will be immoral to use weapons without AI,” just as he expects manual driving eventually to become unacceptable on public roads. The argument is not that AI is infallible; it is that less precise weapons may become predictably more dangerous.
Asked about a host’s “level six” thought experiment involving a unified AI in operational command, Azhnyuk says it could technically be done now, while ubiquitous deployment across platoon, company, battalion, and higher command remains harder. He cites a public claim that, in an operation in Iran, a Patriot system was used for target designation and that AI produced something like a list of 1,027 targets for human confirmation.
The safety dilemma is speed: if an adversary makes decisions “a thousand times faster” with such systems, the side abstaining must catch up or lose. Azhnyuk still names wrong targeting, civilian harm, and decisions that sacrifice friendly personnel as genuine risks; his claim is that safety work cannot be allowed to stop development needed for existential survival.

9. The rifleman survives because autonomy still faces a brutal long tail

A host recalls predicting in 2013 that swarms of autonomous suicide drones would cleanse battlefields of human infantry, despite objections that electronic warfare would ground them or that humans must hold territory. His evidence today is the increasingly sparse frontline, where soldiers hide in dugouts while drones dominate exposed space.
Azhnyuk’s counterpoint is simple: a soldier with a rifle inside a dugout prevents the other side from occupying it. A Ukrainian punk song titled “2030” imagines AI and cyborg warfare continuing while ordinary infantry still freeze in dugouts and hold the line; he thinks that may prove closer to reality than a rapid end of the rifleman.
Humanoid robots remain plausible because the built world is designed around the human body. Without the recent pace of large-language-model improvement, Azhnyuk would have put useful battlefield humanoids at least 10 years away; accounting for exponential progress and cheap local models, he allows that it “might be” closer to five.
The constraint is the long tail familiar from self-driving cars and AI agents. Perception, planning, and control can each use classical algorithms, separate neural networks, or one end-to-end model, but every added behavior creates more failure modes. Even with drones causing most casualties, taking ground still coordinates thousands of people, dozens of drone types, aviation, tanks, logistics, reconnaissance, and psychological operations.

10. China’s latent drone capacity is the strategic tail risk

Azhnyuk begins cautiously: China does not sell directly to Ukraine, does sell directly to Russia, and sits at the beginning of many Ukrainian supply chains; Ukraine does not seek conflict and hopes Beijing remains neutral. Yet his industrial scenario is intentionally severe: “China can produce 4 billion” FPVs against Ukraine’s 4 million.
Those need not remain short-range quadcopters. China could produce autonomous, battery-powered fixed-wing impactors reaching 200–300 km, load millions into shipping containers, freight vessels, barges, or autonomous submarines, and approach coastlines from Taiwan to California with more targets than submarines, aircraft, or anti-ship missiles could preempt.
Countering that mass requires comparable numbers of inexpensive interceptors at sea or in the air. The West presently lacks four layers: autonomy technology, mass-manufacturing capacity, components, and rare-earth refining. “Whoever has bigger manufacturing capacity, that side wins” when the scenario is pushed to its limit.
Asked whether China is therefore already Earth’s supreme conventional military power, Azhnyuk refuses the claim: Chinese drones and forces remain untested, and aircraft carriers, missiles, satellites, electronic warfare, and air forces still matter. But China is not a declining Soviet Union; he sees it ahead in electric vehicles and humanoid robotics and near the West in AI. “We cannot count it out.”

11. The Western answer is allied scale, faster procurement, and credible will

A host pushes back against static factory comparisons: America entered World War II with limited defense production and ultimately outproduced its opponents, while a prolonged emergency could sweep away factory restrictions and mobilize allies such as Japan and Europe. Azhnyuk shares the long-term optimism but wants urgency before an attack, not after it.
His first policy prescription is tighter integration with Kyiv, where “the future of defense has already arrived.” Ukraine’s “defense valley” includes hundreds of firms, combat-tested commanders, government infrastructure, and rapid feedback from soldier to brigade level; U.S. procurement reform and Drone Dominance should absorb that operating system and “scale that up big-time.”
Europe’s pace alarms him. Ukraine and Russia moved from drone technology “of 2025” to “2026,” while Europe advanced only from winter to spring 2022 and the United States arguably moved less; Poland bought roughly 100 tanks and four submarines without training even 1,000 FPV operators. He credits blunt American pressure with waking Europe, but says it remains too slow.
Industrial capacity is only one axis of endurance: populations must accept suffering, political leaders must sustain a defensive war, and alliances must remain credible. Azhnyuk argues the Budapest Memorandum’s failed security assurances taught governments that nuclear weapons may be the only dependable deterrent, predicting more national nuclear programs over the next couple of decades; a host explicitly names Japan, South Korea, and Poland.

12. The drone race has no single leaderboard

Azhnyuk rejects a blanket claim that Ukraine is “ahead.” There are at least 30 categories across air, ground, surface, and undersea domains—reconnaissance, logistics, evacuation, mining, demining, strike, relay, and interception—and each must be scored separately.
Ukraine operates a competitive “zoo,” sometimes with dozens or more than 100 companies in a category; Russia more often centralizes around one to three products and scales them. Ukraine recently moved ahead in frontline strike after catching up on fiber, may be improving in 40–200 km mid-strike, recently gained an edge in deep strike, and has consistently led at sea and, in his judgment, on the ground.
Russia retains advantages in areas including GPS-free navigation and CRP antennas, plus glide bombs reportedly ranging as far as 80 km. Azhnyuk stresses that range is inseparable from launch altitude, aircraft vulnerability, radar coverage, and air-defense geometry—and repeatedly qualifies that he is explaining these matters as a technologist, not a military professional.
Drone interceptions of Russian helicopters are still edge cases rather than a systematic campaign, but he thinks battery-powered interceptors, fixed-wing hunters, or larger drone carriers could make helicopter hunting routine. Frontline attack helicopters may decline; helicopters remain useful, however, including for shooting down Shahed-type deep-strike drones on both sides.

13. Counter-drone economics punish exquisite defenses

Shotguns already down hundreds or thousands of FPVs, but individual survival depends heavily on skill. One Russian soldier reportedly shot down seven attacking drones before eventually being killed; Azhnyuk’s blunt expectation for the “average non-Rambo” caught in the open is that he will probably die.
Electronic countermeasures work perfectly only when they cover the drone’s frequency and the attacker lacks fiber or autonomy. Physical defenses therefore proliferate: roads within roughly 50 km of the front are covered by fishnet corridors, while tanks wear nets and protruding poles like porcupines so warheads detonate half a meter or a meter away from their armor.
Emerging active defenses include handheld, harpoon-like launchers whose onboard AI guides a small interceptor into an incoming FPV, or armored vehicles carrying hundreds of protective drones that launch automatically. Systems such as Rheinmetall’s Skynex can work, but the decisive question remains the price and volume of each effect.
Azhnyuk’s exhibit is a roughly $3 million, 10 kW laser that needs about three seconds on one FPV: the same money buys 6,000 $500 drones, and he doubted it could handle even 600 arriving together. A bigger laser worsens weight, mobility, replacement cost, and coverage. The required mental shift is from exquisite $14 billion platforms toward cheap, distributed mass that iterates like software—hence his closing call: “Be prepared for war.”

Noah Smith

Think about this. Last year, Ukraine produced 4 million FPV drones. Ukraine is not the most industrious nation in the world. China can produce 4 billion of these FPV drones.

Alessio Fanelli

Would you say that right now China is the supreme conventional military power on Earth, given its ability to manufacture and deploy drones in the quantity and quality that you just described?

Yaroslav Azhnyuk

I don't think we have all the information to claim that, but we cannot count that out. That alone should be a big warning sign.

At some point in my life, I went from making cameras that fling treats to pets to cameras that fling explosives to the occupiers. That's the short story. When you think about what your nation and your compatriots are going through, you realize that the only morally right thing to do is to fight back. It is immoral not to fight back. The choice becomes very clear.

Brandon

I'm Brandon. I normally do science podcasts, but today we're going to do something a little bit different. I'm joined by Noah Smith of Noah Opinion on Substack and Twitter, and we have Yaroslav Azhnyuk, founder of Fourth Law and several other drone-related startups.

To get started, it is February 23, 2022. You are running a pet startup, connecting pets with their owners. Let's get a little bit of background. How did you get started in tech, and what were you working on before the war in Ukraine started?

Yaroslav Azhnyuk

It’s good to be here. Thank you. On February 23, late in the evening—around 11:00 p.m. Kyiv time, I think—my wife and I landed in Kyiv. She was my fiancée then. We had come from Lviv, where we were looking at the church where our wedding was supposed to take place.

We got into a cab from the airport to our home, and the driver was like, “You’re crazy. Everyone’s leaving Kyiv. Why do you come?” We were like, “What? Nothing’s going to happen. Dude, chill.” Then, obviously, 8 hours later, the bombs fell on the city.

It was quite surreal. We probably landed on the last flight that landed in Kyiv, or one of the last flights. My background is in technology. I studied applied mathematics at Kyiv Polytechnic. I was born and raised in Kyiv. My parents and grandparents all have PhDs and come from academia, in fields ranging from linguistics to nuclear physics.

I’m an entrepreneur, so I’ve built a bunch of companies. Petcube is the one you were referencing. I lived in San Francisco from 2014 to 2020 while building Petcube, which is one of the leading pet-device companies in the world. We sold a lot of pet cameras.

At some point in my life, I went from making cameras that fling treats to pets to cameras that fling explosives to the occupiers. That’s the short story.

Alessio Fanelli

February 24, I guess a few hours after you went to check out your wedding chapel—what did you do?

Yaroslav Azhnyuk

We had a plan for this situation. My parents and family lived in Kyiv. We were like, “Okay, this has actually started. The worst has come true.” We basically packed our belongings, got in the car, and spent 17 hours driving west.

Most people in our audience have watched at least 1 apocalyptic movie in their life, and this felt exactly like that. Missiles were falling, and there was smoke in Kyiv. My dad and I went to the central part of the city, probably 800 meters from the Presidential Office, to pick up some things at his workplace. He’s the head of an academic institution, so he had to take some things with him.

It was completely surreal. The streets were empty, and the gas stations were out of gas. We found a gas station, but we didn’t have spare canisters with us. We figured that because the car was diesel, we could store it in plastic canisters, so we bought some window wash for the cars, poured it out of the canisters, and poured the diesel into them.

That’s how it was. We were also helping friends get out. A friend of mine had a dog, and my brother was driving in a separate car. We found a place for my friend, who didn’t have a car. It was totally surreal.

Alessio Fanelli

Wow.

Yaroslav Azhnyuk

Of course, we didn’t know then that this would last for so long. We didn’t know whether Kyiv—or whether Ukraine—would be able to defend Kyiv. There was very little information and very little insight into the future.

Alessio Fanelli

What were your thoughts about how to defend Ukraine? You eventually started building drones. What was the process of getting from building devices that connect owners with their pets to building drones? What other things did you do to help the war effort in the process?

Yaroslav Azhnyuk

It’s definitely nontrivial. I didn’t get any military education when I was a student. Normally in Ukraine, you would go to military school even if you were getting higher education in another field. I decided to skip that, which was an unusual way to go. I never thought that I would be involved in the war effort.

One thing you have to understand about many Ukrainians—and I guess it’s also true of most of the people I’ve met here in the United States—is that who you are in terms of your nationality is a big part of your identity. When that comes under attack, it’s something deeper than just the country you live in coming under attack.

On day 1, I figured I was going to fight back with everything I could. I didn’t think on day 1 that I was actually going to make weapons, though. There were a bunch of things we did. We reached out to a number of American congresspeople and senators, advocating for support for Ukraine and for voting for Lend-Lease, which happened in May 2022 but didn’t actually work as expected.

We helped start Brave1, which is now a very important defense-innovation cluster, sort of like the DIU here in the United States. We also helped start a fund called D3. It was started, or co-started, by Eric Schmidt, the former CEO of Google.

By 2023, it was obvious, first, that this thing was going to last a lot longer, and second, that the whole world was shifting. There was going to be a new arms race, warfare was being redefined by drones as platforms, and for the first time in history, you had a platform that was software-defined and could increase your battlefield capabilities in a step change overnight.

It’s as if you could push a software update and give all of your Roman legionaries a new helmet. That had never been possible before. It’s the first time in the history of war that this has been possible.

All of that, and many other things like supply-chain fertilization and the impact that AI is going to have on all of this, became evident to me in 2023. I thought, “I should do what I do best, or what I know how to do best: start a tech company and leverage the global techno-capitalist machine to provide defensibility to Ukraine and the free world.”

That’s literally the mission of the company: to increase the defensibility of Ukraine and the free world.

Then there was some soul-searching and asking myself, “I know nothing about weapons. Am I actually ready to make things that other people use to kill other bad people?” When you think about what your nation and your compatriots are going through, and think about all the terror of places like Bucha, the occupied cities in the east and south, the abducted children, the raped women, all the economic damage being done, and the intention to destroy a whole nation—to commit genocide against the people of Ukraine—you realize that the only morally right thing to do is to fight back.

It is immoral not to fight back. The choice becomes very clear. We’re just passing the ammunition. We’re not doing the actual job. The actual fighters, defenders, and heroes are the people in the armed forces. We’re just support.

Alessio Fanelli

I have so many questions. I know you seem to have a question. Do you want to ask anything?

Noah Smith

No, I’m just listening. Go ahead.

Alessio Fanelli

I do want to talk about some of the moral issues, like you just said.

Yaroslav Azhnyuk

I think there are no issues there.

Alessio Fanelli

Yeah.

Yaroslav Azhnyuk

What would an example of a moral question be in this case?

Alessio Fanelli

Well, no, I mean—

Noah Smith

As you just said, you are creating the tools, but others are using them. I was maybe thinking of having this conversation later, but one of the questions is whether it is actually different when you are building them for your homeland, which is a very morally defensible position, but this technology is not going to stay with you, right?

You will probably be selling these to other people. The future is really where the moral issues may come into play.

Yaroslav Azhnyuk

Yeah, this question becomes easier and more complete if we ask it not about a particular technology or a particular weapon.

If we think that this question actually applies to any kind of technology, right? A knife or fire. You can use a knife to do surgery and save people’s lives, or you can use it as a weapon to take people’s lives.

swyx

Cut tomatoes, too.

Yes, a knife.

swyx

In Japan, they call this sort of knife by the same word.

Yeah. So, you know, it’s the same with any technology—large language models, right? Look at how powerful they are, and yet they’re available to anyone in North Korea or Russia. That’s one side of the argument. The other side is: as a maker, what is your responsibility for how the tools you’re creating will be used? There’s definitely some responsibility, right?

How should the decision process look? Should you try to calculate all the possible scenarios before starting to work on something, or do you create something that’s needed now to save people’s lives and then think about addressing unwanted edge cases later?

In an ideal world—okay, it’s not an ideal world. In a mythical world where there is one governing party that gets to decide everything, and there is no other country that can decide on its own, you could say, “Well, we need to calculate all the consequences and only then maybe build this building by replacing this park, because maybe we need this park in the city,” right? That’s that kind of situation.

But when you’re in a situation where you’re in a forest in front of a wolf, you’re first going to deal with the wolf that wants to eat you, and then you’re going to go consult Greenpeace. That’s the kind of situation Ukraine is in.

swyx

Okay, fair enough. Because this is a tech podcast, I did want to spend some time talking about the tech that you’ve developed and what you’ve been working on. Can you explain, first of all, the problem that you were trying to solve from a technical standpoint? Then maybe go into some of the solutions and the design process that led you from designing a little laser-guided system—guiding lasers with an iPhone—to guiding drones.

Yaroslav Azhnyuk

Sure. My partners and I started one company called The Fourth Law, and its goal was—and is—to make massively scalable drone autonomy. In parallel with that, together with my Pexip co-founders, partners, and friends, we started another company called Ad Systems, which was focused on making thermal cameras. Thermal cameras see thermal radiation and are used to see at night. Those companies are getting closer and closer together, and we’re probably going to merge them.

This group of companies is currently the leading team in drone AI and thermal imaging on a Ukrainian battlefield, and likely one of the leading, if not the leading, in the world. We have 3 business units: cameras, drone autonomy, and drones. The cameras and drone autonomy units sell daytime and nighttime cameras and different types of autonomous drone modules to other drone manufacturers—over 200 drone manufacturers in Ukraine.

The UAV business unit sells the drones themselves to the Armed Forces of Ukraine and the Ukrainian government. There are different types of drones. Those are front-strike drones, as we call them: FPV strike drones and bombers, and then interceptors. There are different kinds of interceptors. We do Shahed interceptors and ISR interceptors. We don’t do deep strikes.

swyx

What’s an ISR interceptor?

ISR stands for intelligence, surveillance, and reconnaissance. Those are basically drones that the Russians use to watch over positions and communicate target reconnaissance. That’s ISR—a classical term for a reconnaissance drone.

swyx

Are all of these battery-powered drones that you just described? I know that the deep-strike drones still have some sort of internal combustion engine. Are all the things you’re talking about battery-powered?

Yeah, everything we’re working on is battery-powered. We don’t do deep strikes.

swyx

You can catch a Shahed with a battery-powered thing. It’s not too fast to catch.

Yeah, absolutely. An interceptor like ours—it’s called Zero—goes up to 326 km/h.

swyx

For reference, how fast is a Shahed?

I think in the terminal phase it could be 280, but in the cruise phase it’s around 220.

swyx

Okay.

Simen Teigre

You can convert that into miles if you’re interested.

swyx

Multiply by 2/3, or 0.66.

Yeah, I see. I was talking about autonomy modules. We make autonomous systems for the frontline, for interceptors, and some for deep strikes as well. There are different levels of autonomy, from terminal guidance—which is the last 500 m, give or take—to autonomous bombing, autonomous target detection, and autonomous navigation.

All of that works across day and night, different terrains, different times of year, and different platforms, like quadcopters and fixed-wing aircraft, and maybe some other platforms. It’s quite a wide variety of products. We also have our own simulation and our own training school for warfighters.

We’re about to start construction of 2 semiconductor plants to make sensors for thermal cameras. That’s super exciting for me as a computer science guy. Doing semiconductors is super cool.

swyx

In terms of the core drone technologies, you basically have one that’s an FPV replacement without fiber optics, and the other is a single-track—

Or without fiber optics. Fiber optics is just a communication module. You can use a classical analog video link and radio link. Those would be 2 separate radios. You can do digital or fiber optic. Fiber optics has its own advantages, but it also adds weight, decreases the distance, and decreases how fast you can turn and move with the drone.

swyx

Do you need AI for fiber-optic drones?

You can use AI for fiber-optic drones. AI replaces a human, right? Fiber optics is about making your communication link more resilient. Those are slightly different goals. You can have AI controlling hundreds of fiber-optic drones instead of having 100 operators for each.

swyx

Okay. I guess I thought that the key reason people moved to fiber-optic drones was for electronic countermeasures, or to counter those.

Yeah, I think that’s a correct assessment from a public-awareness standpoint. In practice, it’s somewhat more difficult, because besides electronic countermeasures, you have issues with the radio horizon for FPV drones. That means that—as I believe—Earth is round.

swyx

Some people disagree.

But basically, if you fly a drone and have a ground station over here and a drone flying over here, if your drone is flying high, you have good direct radio visibility. If your drone goes low—and usually Russian infantry and vehicles are on the ground, and you want to hit them—you need to go low.

The lower you go, the more likely it is that you’ll get behind a hill or a forest, and if you’re far enough, you’ll get behind the curvature of the Earth. You get into what’s called a radio shadow. That’s a real bummer, because for the last 60 or 20 m, you won’t be able to see anything, and it will be very difficult to hit the target.

To counter that, you need to consider the distances at which these FPV drones operate, and they can be quite large. For example, here in the US, there was this Drone Dominance program competition. In Drone Dominance, the furthest distance was about 10 km.

swyx

What was Drone Dominance? What was that competition?

Drone Dominance is a program started by the US government to accelerate the development of drone technology here in the US.

swyx

The longest-range thing they were using was 10 km.

10 km, right. In Ukraine, if your drone doesn’t fly at least 20 or 25 km, no one is interested in it. Many hits happen between 30 and 40 km, and that’s what’s expected from a regular 10-inch FPV drone. At that distance, even at altitudes of 60 to 100 m, you might start losing the link.

One of the earlier AI technologies fielded in FPV drones was terminal guidance technology. That was the first product we ever launched. It helped you as an operator: once you saw the target from 200, 300, or 500 m, you locked onto the target, and then it drove the drone toward the target no matter what, even after you lost the visual connection.

Optical fiber solves that. However, if you want to go 20 km with optical fiber, that adds an extra 3 kg of weight to your drone.

swyx

The cable that you have to unspool as you go weighs—

It is heavy. The spool is about 800 g, so less than a kilo, and 10 km of optical fiber is another kilo or so. That takes away from your useful mass. You need a 15-inch drone, and it can only carry maybe 1 or 2 kg of explosives if you want to go 20 km.

If you want to go to 30 or 40, 30 is probably the max; 40 is very problematic on optic fiber. And then the problem with optic fiber is that it's actually getting super expensive. Do you know why? Because of all the data centers for AI. That's literally the same optic fiber.

When Ukrainians and Russians come to Chinese factories to buy the optic fiber, they're like, “We're out. We sold it all to the Americans.” That's the craziest thing. So, optic fiber went up in price from about $4 per kilometer to about $32 per kilometer in a few months at the beginning of this year.

swyx

Cloud cost is stopping the Russian drone effort here.

And Ukrainian as well.

swyx

Yeah. And Ukrainian. But I read somewhere that the Russians had grown more dependent on fiber-optic drones relative to the Ukrainians, and that's one reason why the Ukrainians have sort of regained the initiative in drones recently. How accurate is that?

The Russians were the first ones to scale that. I think that, as of now, Ukraine has caught up. I think that, as of maybe 3 months ago, Ukraine is mostly caught up on fiber optic.

swyx

What percentage of damage, in terms of FPV drone damage, would you say is now fiber optic versus autonomous?

For obvious reasons, I actually cannot answer that question. I know the answer, but I would not disclose that.

For our audience, I think another interesting fact is that, out of all the casualties on the front line, between 70% and 80% are caused by FPV drones.

swyx

Right. FPV drones are the new universal weapon of warfare.

Yeah. They used to say that artillery is the god of war because artillery used to cause about 80% of casualties. Now, on that ranking, FPV drones rule.

swyx

FPV drones are the god of war.

They've sort of dethroned artillery, but that's not to say that artillery is not useful or not needed. All of these systems are needed. Maybe except cavalry, although the Russians still use it. I don't know. Have you seen the videos of Russians using mules and horses?

swyx

What is the usefulness of a tank on the modern battlefield?

That's where we need Greenpeace to say a word, but they're silent.

swyx

Fair. What's the use of a tank on the modern battlefield?

It's diminishing.

swyx

Diminishing.

Yaroslav Azhnyuk

However, I think there might be technologies that will revive the tank. Look, tanks still provide you armor, and armor is important. You still need armor and firepower, right? You can be an armored personnel carrier that provides you armor.

The challenge that currently exists is that armor is not very well protected against incoming drones. However, there are ways to protect it. We were previously talking about this before the podcast.

Yaroslav Azhnyuk

The CEO of Rheinmetall recently sort of ridiculed the Ukrainian drone industry, saying that there's nothing interesting there, no real innovation, no interest, compared to Rheinmetall or Boeing. And it's all made by housewives. There were obviously a ton of memes about this, with people ridiculing the CEO of Rheinmetall.

One of the best quotes I heard on this topic is from my friend Alexey Babenko, who's the head of and founder of Vyriy Drone, one of the largest manufacturers of FPV drones. They're our partner; they're using our autonomy.

He said that the drones we manufacture in 1 day will be more than enough to destroy all the tanks Rheinmetall manufactures in 1 year. Cost-wise, of course, a drone is about $500, and a Rheinmetall tank is worth probably $5 million or more.

Shawn Wang

Don't mess with those housewives.

Lorenz Meier

Yeah, drone wives.

Shawn Wang

Drone wives.

Lorenz Meier

And that's it.

Shawn Wang

There's a classic saying that everyone always fights the last war. How did we get to the point where tanks became irrelevant, at least for now, in a matter of just a few years?

Yaroslav Azhnyuk

Well, look, I think it's the same way: How did we get to the point that calculators became irrelevant? Now we have iPhones. Why would you need a calculator? Technology progresses, and its obsolescence grows nonlinearly. It's all exponential.

I can tell you that full autonomy, when you put it on a drone—if you think about a tank and a drone, it's not a direct comparison—but even a drone and an artillery shell, in terms of cost per kill, are different.

A 155 mm artillery shell, which is a standard NATO caliber, currently has a market price of about $4,000 per piece. Compare that to, say, $500 per drone; that's 10 times more expensive.

If you account for the amortization of the artillery gun, how vulnerable it is, and the tactical capabilities it gives you compared to a drone, you'll figure out that an FPV drone is maybe 3 orders of magnitude more versatile, more useful, and more capable than artillery in many of the classic artillery cases.

There are different types of artillery, not just 155 mm. You have mortars, you have all that. But, give or take, roughly 3 orders of magnitude, maybe. Again, it doesn't have that firepower; it's not a 1-to-1 comparison, but still.

Now take that FPV drone. When you put full autonomy on that FPV drone, which can be not very expensive—systems that we were producing were in the hundreds of dollars in pure bomb cost—

Shawn Wang

Just to interrupt, you said “full autonomy,” but just a second ago you were saying that the autonomy here is guidance, right? It's not decision-making.

Lorenz Meier

No, I was saying that that's the first and sort of easiest piece of autonomy that was fielded by us. But if you add full autonomy to a drone—

Shawn Wang

I think he's asking: What is it? Can you, for the listeners, explain what the term “full autonomy” means?

Lorenz Meier

Yeah. Basically, I think a good way to think about an FPV drone is like an iPhone of warfare. It's very inexpensive, very mass-producible, and very versatile. You don't need a bunch of other things when you have an iPhone in your pocket. You don't need an MP3 player. You don't need a calculator. You don't need other things, all right?

So, after we draw this analogy to an iPhone—or, okay, Apple, please don't sue me, it's a smartphone—when you add autonomy to it, it sort of becomes like Uber or ride-sharing.

What it means is that instead of actually being a trained pilot who has this complex remote-control device, which requires a couple of months of training to actually pilot a drone, and then having to pilot it for 30 minutes, flying toward the target, etc., now you basically have your smartphone.

You have a drone; you pick up your smartphone and say, “We are here. The bad guys are here. Go and get them.” The drone goes up, flies in the given direction, localizes itself on the map, finds the designated area where the bad guys are supposed to be, sees the bad guys, bombs them, returns, and does damage assessment. It returns back, lands, and then you can pick it up and watch the video if you didn't have the radio link, right? That's a bomber drone. That's full autonomy for a bomber drone, right?

Shawn Wang

So you're saying that no human decisions are made in this entire process?

Lorenz Meier

That's not the same as no human decision being made. A decision was made at the beginning of the process, in the same way as you would fire artillery. When you fire artillery, you don't stop it 500 m away from a target and ask it whether you want to strike or not. So that's exactly it: A human decision is always made at some point, right?

Shawn Wang

Okay.

Lorenz Meier

So when you do that, that's full autonomy. Such full autonomy is happening as we speak. Such full autonomy increases the capabilities of an FPV drone, which is already 3 orders of magnitude more powerful than an artillery shell.

Full autonomy increases its capabilities by 4 orders of magnitude because now you can have 100 times as many people who can use it, because you don't need to train those people. And this is important. You can have a 10 times higher mission success rate.

And you can have 10 times the utility per drone because now, instead of being a 1-way kamikaze, it can be a bomber.

Shawn Wang

Now wait. You said “10 times mission success rate,” which means that fully autonomous bomber drones succeed in their missions 10 times more often than human-piloted bomber drones do. That's an important thing to know.

Lorenz Meier

Okay, but maybe to push back on—

Shawn Wang

They're superhuman. They're 10× superhuman.

Lorenz Meier

They're not vulnerable to electronic warfare. They don't care about the radio horizon. They don't lose track during navigation. They are not susceptible to human error when an artillery shell or another drone blows up beside you and you're like, “Hell no, I'm getting out of here.” Right? That doesn't happen to an autonomous drone. All of those things.

We have one of the brigades that's using our drones with just first-level autonomy.

Shawn Wang

So what's first-level autonomy?

Lorenz Meier

First-level autonomy is just terminal guidance.

Shawn Wang

Okay.

Lorenz Meier

By the way, we have a video of that. We can watch that.

Shawn Wang

Terminal guidance means a human gets it nearby and then the AI takes over.

Lorenz Meier

The human flies it all the way, like 30 km, toward the target. Obviously, the target was probably given to that human by someone who's flying some ISR drone, some reconnaissance drone, right? So the human flies all the way to the target, and once they see the target from a distance of 500 m, they do the target lock. From there, the drone flies autonomously.

So, just that feature alone has increased the guy’s—his call sign is Grom—mission success rate from 20% to 71%. It also increased his kill zone from 3 km to 10 km. That means there is a certain area around the front line designated as a kill zone. Whenever the enemy goes into that area, it’s almost guaranteed to be destroyed by a drone. Obviously, the drones are not launched from the zero line; they’re usually launched from minus 10 km.

Shawn Wang

What is the zero line?

Lorenz Meier

The zero line is a certain imaginary line of control between 2 conflicting forces.

Shawn Wang

Okay. It’s important to explain these things to a lot of the listeners who aren’t familiar with what they’re—

Lorenz Meier

Absolutely. Absolutely. Yeah, yeah, yeah.

Shawn Wang

Right. So, you said that level 1 autonomy—in other words, just terminal guidance, where a human gets it to the finish line and then it goes over the finish line—increases mission success from 20-something percent to 71%, or something.

Lorenz Meier

And it increases the kill zone from 3 km to 10 km.

Shawn Wang

Actually, real quick, can we define mission success and, maybe, in a way, what the failure modes of missions are? I have a guess at what mission success is.

Lorenz Meier

I mean, well, yeah, but I could—

Shawn Wang

Get them.

Lorenz Meier

Yeah. Yeah, I know, but that’s a very good question, in fact, because even if you fly into the target, first, the target can be damaged or destroyed. Those are 2 different modes. Then there can be different targets. A sole infantryman is 1 kind of target. A dugout where there are supposed to be some enemies is another kind of target. Some mechanical equipment is another type of target. Radio-emitting equipment—which is often the kind of target that the military wants to get more than anything else—could be an enemy radio tower or some small radio dish. That really makes life difficult in that area, in that combat area.

Those are different targets, right? The target can be destroyed or damaged. Sometimes the drone hits but doesn’t explode. That happens.

There are other failure modes. You didn’t even reach the target because, A, you were jammed by electronic warfare; B, you lost control over the drone because of the radio horizon; or C, you were jammed by a different type of electronic warfare that happens way before you hit the target area. It’s impacting your video receiver, so jamming on video and jamming on control are 2 different types of jamming.

Then something malfunctioned on the drone—a mechanical malfunction. Maybe a motor broke, or whatever. You may have gotten lost while navigating to your target. That happens, too.

Shawn Wang

Okay. So, level 1 autonomy basically means you manage to point the drone in a direction, you go there, and then the last mile is the drone taking over.

Lorenz Meier

I define that, but it sort of got picked up by the industry. We define 5 levels of autonomy. Level 1 is terminal guidance. Level 2 is bombing. Level 3 is autonomous target detection and engagement decision. Level 4 is autonomous navigation. Level 5 is autonomous takeoff and landing.

Shawn Wang

Those are good things to know.

Lorenz Meier

Those are 5 levels of autonomy. Now, if you want—

Shawn Wang

What level is Tesla running on right now?

Lorenz Meier

Tesla?

Shawn Wang

No, sorry.

Lorenz Meier

Well, that’s a very good point. It’s exactly what it was inspired by: the levels of self-driving autonomy.

swyx

Waymo’s level 5, right? You just tell it where you want to go, it picks you up, and then you go.

Yaroslav Azhnyuk

If you look at the classic definitions of self-driving cars, Waymo is still level 4 because it still requires, even remotely, human control. If Waymo gets in trouble, there is no operator who takes over and resolves this. That would still be level 4. It doesn’t map directly, but it’s also 5 levels.

swyx

Can I interject a question? In terms of an FPV drone that’s like a suicide drone that’ll just blow itself up, killing something, how do you know what it hit? Does it transmit back, or do you sort of lose track of it and hope it hit? What happens to that?

That’s a great question.

swyx

You mean another drone?

The current battlefield in Ukraine is saturated with different types of drones. Obviously, you have all the FPV drones. Last year alone, Ukraine manufactured about 4 million of these, and Russia’s maybe 20% less than that. For this year, the publicly voiced target was 7 million on the Ukrainian side. We’re getting into serious numbers here.

Besides those, there are different reconnaissance drones, or ISR drones, as we call them. There is tactical-level ISR, where both Ukrainians and Russians usually use a Mavic drone by DJI. Then there are a bunch of locally produced fixed-wing drones that can stay in the air much longer than a Mavic. A Mavic can stay in the air for maybe half an hour, and then there are drones that can stay in the air for many hours or even up to a day. Those drones are more expensive and have more expensive cameras.

We hunt the drones that the Russians launch. The Russians hunt our drones, and so on. Ideally, when you’re a group of soldiers operating an FPV, you’ll have someone in your company or platoon who has an ISR asset that will do target designation for you. They’ll say, “There’s a Russian vehicle over there. Go and get him.” You go there, you get it, and they’re like, “Okay, confirm.”

They have their own drones in the sky. They have a carousel of drones because 1 drone maybe cannot stay in the air for more than 30 minutes, so they constantly have 1 surveilling the battlefield—almost every spot on the battlefield.

But it’s not always the case. Sometimes you will not have a surveillance asset, so then you would launch another drone just to confirm that there was a hit. If you see there was a hit and you’re not sure whether it completely destroyed the target, you may hit it again for good measure. That’s how it works.

But I was about to give you another piece of taxonomy. You have 5 levels of autonomy, right? Then you have 8 dimensions of the autonomous battlefield. It’s crucial to understand how autonomy evolves in a modern battlefield environment.

Dimension 1 is the level of autonomy: what capabilities your asset has. Dimension 2 is the platform you’re operating on. It can be a quadcopter or a fixed-wing drone, different types of long-range or short-range drones. But it can also be a missile. You can have autonomy even on an artillery shell, a ground vehicle, or a sea vehicle. All of those are different platforms.

Dimension 3 would be the domain. Is it ground-to-ground, ground-to-air as an interception, ground-to-sea, or sea-to-air? There are all the nuances with different domains.

Then, dimension 4 would be high levels of autonomy, such as swarming, drone carriers, and drone nests. And now, when you’re—

swyx

When you’re saying “level,” you’re talking about dimensions, not about—

Sorry. Yeah, I mean the dimension. I was supposed to say “dimension.” I say dimension because each of them works with another, right? You might have 3rd-level autonomy, a fixed-wing drone operating in land-to-air, and stuff like that, operating in a swarm or from a nest.

Dimension 5 is the environment. Is it day or night? Is it summer or winter? Is it humid, cold, or dry? What kind of target is it? Is your target hiding in a forest, behind a hill, or within buildings? All of that is the environment.

Dimension 6 is command and control. How are you dealing with tens of thousands of those assets around the battlefield? How are you coordinating that at the higher levels of command? Are you collecting data? All that.

Dimension 7 would be infrastructure: things like simulation, data collection tools, security, deployment mechanisms, and so on. All those systems have to be developed separately and then integrated with all the others.

Finally, dimension 8 is distribution. Have you deployed 100 of these systems or 100,000 of these systems? Those are 2 very different ball games. That gives you a broader overview of how autonomy propagates across the battlespace.

swyx

As someone who has done machine learning and had things go out of distribution and go horribly wrong, several of these axes for thinking about drone warfare seem like they could be very susceptible to some sort of distribution shift if you start making things autonomous.

Likewise.

swyx

Well, I mean, first of all, I’m very interested in the sorts of scenarios that you’re thinking about. The most obvious one is, if I assume these are computer-vision-guided systems, at least for the last mile, how do you ensure that—well, you now have some fog rolling in or something, and the drones just attack the wrong thing? Maybe—I mean, it probably will not turn around and fly back and attack you.

Yaroslav Azhnyuk

The question is: how do you assure that your mortar fire hits the right thing? Mortar fire, give or take half a kilometer, could be plus or minus. So maybe you fire 1 and then you fire another. Drones are actually much better at being precise in those scenarios, and I think, to your point, 5 to 10 years from now it will be immoral to use weapons without AI. Weapons without AI will be more likely to cause collateral damage or unwanted damage, in the same way that it will be immoral to drive your own car manually on a public road because it's more likely to cause unwanted damage.

Alessio Fanelli

Well, I never considered that might be immoral.

Guest

Really?

Alessio Fanelli

Yeah. That's definitely coming. Anyway, it's an obvious thought. I agree with you. No, I mean, they're obviously not going to let you drive once most of the cars on the road are autonomous.

Guest

That one I believe.

Alessio Fanelli

No, you were talking about drones, right? Or weapons, right?

Guest

Friendly fire and collateral damage and stuff like that is all minimized with AI.

Alessio Fanelli

Okay, so here's my question. Let's go to Level 6 autonomy. Let's take all of the target selection and all of the battlefield data, integrate it into one big AI, and have that big AI basically be in command of the battlefield and agentically do target selection. That's a general AI. You've cut humans out of the loop, except maybe as dexterous robots repairing drones, fastening things to drones, or something like that, because you don't have those robots yet. How soon are we there?

Guest

Mm-hmm.

Alessio Fanelli

AI general.

Guest

Well, the most important thing to ask ourselves is: who will be faster to that, us or our adversaries?

Alessio Fanelli

I assume us, but how fast will we be to that? I hope us.

Guest

Mm-hmm. I hope so, too.

Alessio Fanelli

But how fast can we—when are we looking at that in terms of horizons, in years?

Guest

Technically, it could be done now. The question is, of course, that there's some engineering work to be done. The bigger challenge is deployment.

Technically, in an operation in Iran, it was publicly claimed that, I think, the Patriot system was used for target designation. It is not exactly as you say—the AI makes all the decisions—but basically, AI goes through all the data you have, gives you 1,027 different targets, and says, “Please confirm. Press okay.” You look at the targets and you're like, “Yeah, sounds right,” and press okay. I think that's where we are now already, or where we were a couple of weeks ago, as we're recording this on April 10.

Another question is how massively deployable it is. Is every decision being made like that, or are just some of the decisions made like that? Then there are different levels of command and control. You have the platoon, the company level, the battalion, and so on.

But the tricky thing, when we get into that territory, is this: if your enemy is gaining the advantage of being 1,000 times faster than you by deploying such systems, what do you do?

Alessio Fanelli

Yeah, if the enemy is 1,000 times faster than you at deploying the systems?

Guest

Yeah. If the enemy starts deploying Level 6 autonomy as you call it, and you have not started doing that—

Alessio Fanelli

You're in trouble.

Guest

Yes, exactly. So you have to catch up. My point is that it is very important to think about the safety of these systems, but that thinking should not slow you down in developing them, because they are critical for your existential survival. The person who doesn't think doesn't get to think about the ethics of war. That person surely doesn't get to think about it.

Alessio Fanelli

What would be the safety risk of such a system?

Guest

Well, of course—

Alessio Fanelli

Just wrong decisions, right? Maybe AI decided to use a drone army to kill us.

Guest

These decisions will not only be made about drones. They are likely to be made about what the human should do on your side as well. Obviously, some environments are more like the Russia-Ukraine war, where you have—

Alessio Fanelli

You have to choose to risk lives. You will have to choose to sacrifice human lives on your side.

Guest

Of course. Some environments are just dead zones, and there are no civilians. There are virtually no civilians close to the front line because it's super dangerous; everyone has evacuated from there.

But there are other environments that are more like, “Okay, this is a counterterrorist operation,” or there's a group of terrorists with a group of civilians, like the recent operations in Iran. I imagine that the U.S. and Israeli forces do not want to harm civilians. They will only target the military targets there. So in those situations, it's a different level of responsibility for that decision-making as well.

There is such a big variety of military missions, and I'm not even educated in military science enough to tell you about all of those scenarios. You would need to put a general beside me. Maybe a Ukrainian general and an American general would have told you very different stories about these things.

Alessio Fanelli

Got it. Can I ask a few more questions?

Guest

All right.

swyx

So, in 2013, one of my first paid articles ever was about how the era of drones will change human society. I was just sitting around bored, thinking about things.

You were way ahead of your time.

swyx

And I said the following would happen.

It's real. I read it.

swyx

I said small, autonomous suicide drones will cleanse the battlefield of human infantry. Human infantry will not be able to stand against swarms of AI-powered suicide drones. I didn't even know about AlexNet at the time, I think.

Yeah. And then—

swyx

Yeah.

Guest

But he's just an avid sci-fi reader.

swyx

I mean, I'm an avid sci-fi reader, but it's not like there will be a way to do that. It's a nonlinear, multidimensional search problem. If you get enough compute, you'll find some search algorithm that will get you there.

I think that one sentence describes the bitter lesson right there. It's just a multidimensional search space. You search it somehow. I don't know. Figure out some search algorithm. It's not that hard.

Anyway, the point is that human infantry on the battlefield will be gone. At the end, I wrote that in 2013. Many people on social media laughed at me for that, called me hysterical, and said things like, “Electronic warfare will knock all the drones out of the sky,” or, “You need humans to hold ground.” That's something you still hear from a lot of people on social media today.

I feel that this article I've written has never been directionally wrong. It has gotten more and more right steadily over time, and we're very close, reading the battlefield reports from Ukraine, where human infantry are basically just a few guys hiding in dugouts for months. I'm not sure what—

On the Russian side, that's just like a Zerg rush.

Alessio Fanelli

The Zerg rush, and then they just die. But they have some guys in dugouts too, right? Hiding in dugouts for months?

Guest

Yeah, they do.

Alessio Fanelli

Okay, but what are those guys doing in the dugouts? Are they providing frontline reconnaissance? What are they doing?

Guest

If there is a guy in a dugout with some bullets and an automatic weapon, the other guy cannot come and take the dugout. So they're establishing control over territory.

Alessio Fanelli

I see. So there still is a use for human infantry on the battlefield as of today. How long will that last?

Guest

I think it will last for a while. There's this funny whole layer of modern Ukrainian culture built around war-related stuff. There is a punk rock band called S Z C, which stands for something like a deserter or something like that.

Anyhow, this band has a song titled “2030.” It's basically about the year 2030, and the war still goes on—like, the whatever Third World War or whatever. They sing about AI and cyborgs and everything, but simple infantry is still needed, and we're still getting cold in those dugouts and still doing our job. That's sort of the theme of the song, and it seems like that's actually what's going to happen.

Alessio Fanelli

Ground robots will not replace humans in the dugouts soon.

Guest

You know, I'm very much interested in following the whole humanoid robot theme.

Alessio Fanelli

What about, like, a dog? A dog robot.

Guest

Or just mobile control or something.

Alessio Fanelli

Yeah.

Guest

Into a crab.

Alessio Fanelli

It'll be a crab robot.

Guest

A humanoid—

Alessio Fanelli

Personalization of warfare.

Guest

There is a lot of utility in humanoid robots because the world is designed around humanoids. So I would not 100% disqualify the possibility that, 10 years in the future, humanoid robots will actually be fighting. That's an actual Terminator kind of scenario.

swyx

And in the first Terminator movie, you look at what they've got on the battlefield: flying bomber drones and humanoid robots.

Mhm. Yeah. The cost of running large language models is getting so low. You can have an inexpensive computer running what was a state-of-the-art model a year and a half ago locally on a device with an open-source model, which also means that the Chinese can have it, the Russians can have it, the North Koreans can have it, and so on.

That is already possible. If not for the acceleration of large language models, I would have said that I don't think humanoid robots will be able to be useful on the battlefield earlier than in 10 years. But if you account for the exponential, it might be 5 years or so.

The problem with all of the autonomous systems—and it starts with self-driving cars, and even with all the modern-day AI agents—is that to make them really useful, you have to solve such a long tail of edge cases that it's really difficult to make them useful. We were promised self-driving cars in 2007 by Sebastian Thrun and Google, and even before that. Elon, of course, told us it was going to be 1 year from 2014, and now we still don't have self-driving Teslas everywhere.

We have Waymos in San Francisco and some other places, but they're still not perfect. So I expect something similar from self-flying drones and fully autonomous drones. We saw that firsthand: with each level of autonomy that we're adding, there is a very wide distance between a prototype, something that is ready to be scaled to millions of units, and something that has been scaled to millions of units.

But the race with AI coding tools is just insane, so things might accelerate very fast—faster than we can imagine.

swyx

So, I think your point is that, due to this long-tail behavior, level-one autonomy as you've defined it is actually very natural. You're basically just solving an image-recognition and tracking system.

Yeah. It's actually interesting that you say it that way. I thought about this the very same way. We have this joke that there are like 200 companies in Ukraine trying to solve last-mile targeting or terminal guidance. It seems like we're the only company that actually solved that, because even that's trivial, but it's at least something that you imagine us and Eric Schmidt's companies solving.

Eric Schmidt's companies are pretty good. I actually have lots of respect for what they're doing. They've been practically influential and helpful on the battlefield, and they have good engineering.

swyx

I wasn't saying it's trivial. I'm just saying this is something naturally adaptive, based on things that we know work well. But some of the other domains where you do have to make decisions and you have a long tail become much harder, and you worry about edge cases more.

The more complex the behavior you're trying to simulate, the more edge cases there are, right? The more ways to do it wrong there are. And then there are different approaches.

If you read academic papers about robotics, the robot is represented as something that has sensor input, and then you have 3 levels of logic or decision-making, which are perception, planning, and control. Then you have actuators as output.

Before neural nets, you would do perception, planning, and control all with classical logic, right? Then, with AlexNet and computer vision, you could do perception with neural nets and the rest with logic. You can currently do each of those separately with neural nets, each of those separately with logic, or you can just have 1 huge neural net that takes lots of sensory data—not just pixels; it could be sound, an accelerometer, everything—as input and just outputs the controls. Some of the self-driving car companies are doing that, or experimenting between different ways of doing that.

You can also think about that. The way you implement those features also influences how many degrees of freedom the system would have, right? For control, you can do classical algorithmic control with Kalman filters and PID controllers, et cetera. Or you can do a neural net that was trained in the gym with reinforcement learning, et cetera. Those would be 2 different behaviors of the system.

swyx

Yeah. Maybe my point was just that, if you go high-level, you can try to have whatever Fei-Fei Li and folks who are doing physical world models—Physical Intelligence—they're trying to make these big models and sort of understand the world. Then, supposedly, you have such a model and you can tell a drone, “Okay, go over that hill, find the bad guys, and then get them. Or make me a video, make me a photo of the guy smiling, and get back to me.”

That's 1 way. Another way is to have these subsystems: one is navigation, another is finding the person, and another is getting to them to take a photo. Those are again very different behaviors.

It's not that 1 is necessarily better than the other, and we might have more technological ability to do 1 or another. But all of those systems will exist, and you should always keep in mind that it's not only the good guys who are developing these systems; the bad guys are developing these systems as well.

swyx

Yeah. I guess where I'm going with this, back to our original thought, is the end of the soldier—or at least the end of the rifleman.

Yeah, I'm not seeing that very close. As much as I'm a lover of sci-fi and all of that, and a technologist, the more I try to have a certain humility about these things, the more I realize that the military domain has so much human history, blood, and tears dedicated to understanding war, perfecting it, and so on. There's so much knowledge in there that I don't feel like I've even started to comprehend a lot of it.

One thing that I really understood is that even though drones are now making 80% of the casualties, you go to the actual officers—you talk to the actual brigade commanders, corps commanders—and they explain to you how all of it fits together. When you're thinking about an operation that involves a couple thousand people to get this piece of land out of the enemy's hands, to deoccupy it, it is so complex.

It involves dozens of different types of drones, and then land operations, reconnaissance operations, psychological operations, aviation, tanks, logistics, and all kinds of these different assets. Modern warfare is really very complex, and the fact that drones are the latest, coolest thing and AI is essentially the latest, coolest thing doesn't mean that now it's all that and only that.

Whoever's looking into that should realize that it's not just what the press talks about. The reality is much more difficult and much more complex.

Shawn Wang

Let's talk about China and China's manufacturing capabilities. Suppose that the United States went to war with China.

Yaroslav Azhnyuk

I hope not.

Shawn Wang

I hope not as well. But suppose that drones were very essential to that war, of all the types of drones that we're talking about here. Suppose that China said, “All right, well, you need X, Y, and Z to make those drones to fight us, and we control the production of X, Y, and Z. So we're just going to cut you right off, and now you have no drones.”

I know that a number of countries, including Ukraine and Taiwan, have been making moves to China-proof their drone production so that China couldn't do that. Examples of things they might be able to cut off might include rare earths, fiber-optic cable that you were talking about before, and various other things that—even if they don't control 100% of the production—they control enough of the production that it would be extremely expensive to produce them without relying on Chinese sources, or the market is fragmented enough, and so on.

What do you see as China's key bottlenecks, and how easy are those to overcome in terms of China-proofing drone production in case of a war against China?

Yaroslav Azhnyuk

Mhm. Although China does not sell directly to Ukraine and it does sell directly to Russia, a lot of Ukraine's supply chains start in China. We're not in a conflict with China, and we would not want to be in a conflict with China. We would hope that China stays a neutral power between Ukraine, Russia, and the US as well.

That said, the scenario that you're describing is much, much, much worse. Think about this: last year, Ukraine produced 4 million FPV drones. Ukraine is not the most industrious nation in the world. China can produce 4 billion of these FPV drones.

China can make them not as drones with propellers, but as fixed-wing drones, which go not 40 km far, but maybe 200 or 300 km inland. They’re slightly more expensive.

Shawn Wang

With internal combustion?

Yaroslav Azhnyuk

No, no, no. With battery.

Shawn Wang

Battery-powered fixed-wing drones?

Yaroslav Azhnyuk

Yeah. Battery.

Shawn Wang

What’s the propulsion system on those? Propellers?

Yaroslav Azhnyuk

Yeah.

Shawn Wang

I don’t—I just don’t know how that works.

Yaroslav Azhnyuk

Yeah. You have that. They can also make them all fully autonomous. They have DJI, the world’s most advanced drone company. They can make them fully autonomous without GPS, without anything.

And then they can put those drones on maybe tens of thousands of fully autonomous underwater submarines. Or maybe not even that, just on shipping containers and barges that ship goods, freight ships. Then they show up with millions of drones packed onto those sea vessels.

They show up to any coastline in the world, be it Taiwan or California, and they have millions of long-range impactors targeted at a piece of land. What do you do with that? There are not enough hunter submarines. There are not enough anti-ship missiles, anti-ship planes.

They can produce these assets in tens of thousands of factories because they’re so simple to produce. Even if the FBI director picks up a phone and calls the president of the United States and says, “Hey, the scenario Yaroslav was warning us about is beginning to unfold. We need to do a preemptive strike,” you wouldn’t have enough assets to do preemptive strikes because there can be tens of thousands of places where these things are being manufactured.

To counteract a scenario like that, we would need to have a similar amount of mass.

Shawn Wang

You mean a similar number of drones.

Yaroslav Azhnyuk

Yes, to intercept that, whether at sea or in the air, at a similar cost, right? So economics should work out. I’ll tell you that currently, we in the West—and we in the United States—we don’t have the technology to do that. We don’t have—

Shawn Wang

What key technologies do we lack?

Yaroslav Azhnyuk

Autonomy, mass manufacturing, stuff like that.

Shawn Wang

Autonomy technology?

Yaroslav Azhnyuk

Well, I think so—

Shawn Wang

Because our computer vision algorithms are not as good?

Yaroslav Azhnyuk

It’s not only about the computer vision algorithms. If the group of companies Eric Schmidt founded 2 or 3 years ago and my small startup—it was maybe not that small, but also founded 3 years ago—are 2 of the leading companies in the world, and maybe a couple of others are capable of something like that, but not really on small drones, I do think we’re behind China in technology.

So we lack technology, we lack mass manufacturing capacity, we lack the components, and we lack the rare-earth materials. So there are 4 layers in which we’re behind in this challenge.

That’s why it’s my point that we in the West, and especially in the States, should have far more smart people working on defense. There should be more funding if we want to keep the semblance of our good past life.

Shawn Wang

That’s really important. Would you say that right now, as things stand in conventional terms—not abstracting from strategic nuclear weapons—but in conventional terms, China is now the supreme conventional military power on Earth, given its ability to manufacture and deploy drones in the quantity and quality that you just described?

Yaroslav Azhnyuk

I don’t think we have all the information to claim that, but we cannot count it out, and that alone should be a big warning sign.

We have not seen Chinese drones in action. We’ve seen some Iranian drones in action, and then Russian drones in action—not Chinese, really. We have not seen Chinese forces in action.

Obviously, hopefully this never happens, but in a conflict on the scale of U.S.-China, there are many classical assets that we should not discount, as we just discussed. We should not discount artillery in a land war. We should not discount aircraft carrier groups, the air force, long-range missiles, electronic warfare, satellites, and so on.

But there are also things that we, at least as a general public, don’t really know about China. I’m sure there’s a lot of information that U.S. intelligence has about Chinese capabilities.

If you get back to the scenario that I just described and take that to the maximum, you basically see that whoever has the bigger manufacturing capacity, that side wins.

Shawn Wang

That’s just a typical law of conventional warfare, and has been forever.

Yaroslav Azhnyuk

Sort of.

Shawn Wang

You read Noah’s blog?

Yaroslav Azhnyuk

Not as often as I would like, but I read Noah’s X.

Shawn Wang

It’s a thing where you—

Yaroslav Azhnyuk

Don’t read my X.

Shawn Wang

Yeah, it’s just for fun.

Yaroslav Azhnyuk

There’s no opinion about anything.

Shawn Wang

Okay, so here are, I guess, 2 questions. The first question is: could the United States and other countries allied with the United States even develop supply chains that are independent of China to make any of these drones? The second question is: could they do it in sufficient mass?

I think the answer to the question of whether they can do it in sufficient mass is, today, no. But in an extended, prolonged war situation, things change a lot. All the development restrictions that we put on new factories go out the window in a sense of urgency. Ukraine obviously wasn’t making all these drones before the war, and so if America had the same kind of urgency that Ukraine has now, things would happen. Things would move.

Of course, America has allies too—or had allies until recently—and may have them again in the future. America has, or had, allies that would also scale up very quickly, like Japan and European countries, if we ever ally with them again. A lot of things could then change in terms of the actual mass.

So, in terms of looking at China and saying they have all these factories today, and looking at the history of conventional warfare, America had very little defense production capability on the eve of World War II and ended up easily outproducing everyone else, even the Soviet Union.

Yaroslav Azhnyuk

Maybe not easily.

Shawn Wang

Not easily, but I mean by a long, long way.

Yaroslav Azhnyuk

The benefit of not being attacked.

Shawn Wang

That’s right.

Yaroslav Azhnyuk

That helps.

Shawn Wang

Yeah. Who knows how secure they are now, or what we know we’re cyber-vulnerable to?

Yaroslav Azhnyuk

I totally agree with your sentiment. I’m even less doomerish than you are—or at least, it seems to me you’re a little bit doomerish—but in the long term, you’re bullish.

Shawn Wang

Doomerish. I’m thinking about what we need to do. I’m not thinking, “Oh, we’re doomed.” That’s not my point. It’s never useful saying that. If you’re doomed, then just don’t go on podcasts. Go pet a rabbit and play video games or something.

Anyway, we’re not doomed, but I’m saying: step 1, what are the key choke points that we need to address tomorrow, besides rare earth, which we already know about? What are the other key choke points that the West needs to free itself from in terms of Chinese supply chains in order to manufacture even 1 drone free of Chinese supply chains?

Yaroslav Azhnyuk

There are companies here that are doing that. We have good friends at a company called Neos. I know they’re down in El Segundo, somewhere in Southern California.

Shawn Wang

What are the most pressing choke points besides rare earth, which everyone talks about?

Yaroslav Azhnyuk

That’s one of the pieces. We do thermal cameras. That’s actually a big one.

Shawn Wang

Thermal cameras?

Yaroslav Azhnyuk

Yeah. Then there are the motors. You need the special—

Shawn Wang

After you have the magnets, then you turn them into a really good motor?

Yaroslav Azhnyuk

Yeah. You need these special magnets, and then that’s sort of your rare-earth component. Rare earth isn’t about metals that, for some reason, God put only under Chinese territory and nowhere else. They’re distributed. There are plenty of them around Earth. It’s about the refining capabilities and investing in that, and so on.

Lorenz Meier

Frankly, at some point, we don’t have that many humans. That’s where humanoid robots help. China is a populous country. The population of the united West is comparable to that, but the population of the U.S. is much lower than that.

I definitely think that the whole West should get its act together because, as the saying goes, ubi concordia, ibi victoria: there is always victory where there is union. As free nations of the world, we should get our act together because freedom is what unites us.

I’m also pretty mad at what’s happening in the European Union. I think the current U.S. administration is probably the best thing that has ever happened to Europe since World War II—or since the post–World War II period, because World War II wasn’t the best thing.

Shawn Wang

Trump withdrawing the image of omnipotent American support forced the Europeans to get their butts in gear, unite, and—

Lorenz Meier

Yeah, and also doing that not in the nice way.

Right? When Vance came to the Munich forum 1 year ago, he wasn't super nice. He wasn't saying, “Oh, please, our European friends, could you please increase your defense spending?” He was somewhat pushy, I'll put it that way. I think that was a necessary measure.

I've been thinking about that. Could he have been nicer? I was like, “No,” because the voters of European countries would not have understood this. They would not have gotten the message. I think the message got across, but Europe is still slow to wake up, I would put it that way. Things are getting better, but I'm not happy about the speed at which they're getting better.

When I would go to some of the European capitals, I would get back pretty depressed from talking to their military officials and their entrepreneurs, et cetera. I've been in the US for the last month or so, and I'm not depressed. I'm actually excited. I still think you should 10× the effort in making sure that you remain the strongest power in the world and can defend your values, et cetera.

But I'm very optimistic. Definitely, once we are in danger, I think there are lots of very smart people in the West who can figure these things out. But people in China are also extremely smart. It's very different even from the Cold War situation: the Soviet Union was an economically declining power. China is not like that.

When you look at the electric car race, I think China is ahead of the US and ahead of the whole world—definitely ahead of Europe. Europe used to be a car superpower. When you look at AI, I think they're almost where we are, maybe slightly behind. When you look at humanoid robotics, I would argue they're ahead.

In many other areas, like medicine and biosciences, there are lots of interesting things there. In the consumer space, there are lots of interesting things there as well. I don't know if you've heard this podcast called 996. I don't know if it's still airing or not, but there used to be a fantastic podcast by some American-Chinese businessmen, maybe venture fund managers.

Shawn Wang

About the Chinese economy?

Lorenz Meier

About China, from a sort of tech venture point of view. I lived in China for maybe 4 months, and I visited a couple of times. Even WeChat is so much more advanced than anything we have in the West.

It's very important not to be too arrogant, and I think we're guilty of that, definitely in the US. Sometimes we tend to be too arrogant. I think humility always helps, at least as a person. And I think we obviously don't have to be enemies.

With Ukraine and Russia, Russia came to kill all of these people and get all this territory. With China and the US, it's not like that, and thank God it's not like that, right?

Shawn Wang

It could be China and Taiwan, maybe.

Lorenz Meier

Hopefully not. Yeah, hopefully not. China has its own problems, probably with human rights, et cetera, but hopefully it's still not beyond the point of fixing. Hopefully. Hopefully.

We should be armed, right? We should be ready for whatever. That alone decreases the probability of any conflict. If you're weak, you're basically provoking the conflict.

The problem with Europe these days is that last year, Ukraine and Russia went from drone technology of 2025 to drone technology of 2026. Europe went from the winter of 2022 to the spring of 2022. Europe didn't even make 1 year of progress. The US, I would argue, made less progress than Europe as well in the last year.

The technological gap is getting wider and wider and wider. At some point, I'm looking at Poles who are very close to us and close to Russia. Polish people—sorry, it's not my first language.

Shawn Wang

When you're looking at the Poles, what are they saying?

Lorenz Meier

Yeah, Polish people—Poles. I'm looking at them, and they bought 100 tanks and 4 submarines. Dudes, you don't have 1,000 people who know how to operate an FPV. What the hell are you doing?

Shawn Wang

So Poland is not preparing for war correctly.

Lorenz Meier

From what I can see, they're not doing it right. The problem is they'll be in a situation where they're so proud of their Winged Hussars and their cavalry, and the enemy is attacking with airplanes and tanks. That's literally the gap getting wider between Russia and Poland.

Shawn Wang

That happened in 1939.

Lorenz Meier

I don't want that to happen again.

Shawn Wang

All right. The Europeans need to wake up more. If you were advising America's defense establishment—which you might be doing in real life—but if you were saying things on a podcast that might be heard by some people connected to that defense establishment, which you may or may not be, what are the top priorities, besides more funding? We know more funding; that'll be necessary for anything, literally anything. What are the top policy priorities for America to increase its readiness right now? Let's say 3 to 5 priorities.

Lorenz Meier

Yeah. I really like this quote. I think it's by Arthur C. Clarke: “The future is already here—it's just not evenly distributed yet.”

Just the same way Silicon Valley is this sort of future location for all things tech, Kyiv in Ukraine is sort of a defense valley. It's the point where the future of defense has already arrived. There's a ton of things to learn from that, starting with, in particular, hundreds of companies in very particular fields, and the battlefield experience from battlefield commanders of every level, starting from soldier to sergeant to platoon-level commander to brigade-level commander.

There are special forces and intelligence, all of that. There's also how the government organizes the infrastructure and the playing field for all these businesses to flourish. I would definitely look into much tighter integration and exchanging experience. That would be 1 thing.

Reform of procurement would be another thing, and I think that's what is currently being done with Drone Dominance. I think Bidhat is leading that, and maybe some other people in the administration. I think that's extremely powerful and the right thing to do. They should scale that up big-time.

Obviously, any military person would say, “Well, yes, okay, you're fine, cool, but Ukraine and its theater of war are very different from potential scenarios that the US might have to fight in.” I agree. But there is still so much to learn, even from the sea warfare that Ukraine is doing.

Then there are long-range drones like these Shaheds that unfortunately damaged some of the American equipment in the Middle East. They can fly up to 2,000 km. If you think about the Pacific region, 2,000 km covers a lot of land with all the islands and aircraft carriers.

Shawn Wang

We're learning that lesson right now in Iran, in the Middle East.

Lorenz Meier

You would think so, but I'm not sure. There were so many chances to learn that lesson from Ukraine before, and I don't think it was fully learned. So I'm not sure how fully learned the Middle East lessons were.

Shawn Wang

Perhaps losing a war to a minor power will teach America.

Lorenz Meier

Well, yeah.

Shawn Wang

Although their economic weapon will be the most important and decisive by far, some of our bases were allegedly rendered unusable by their Shahed-type drones.

Lorenz Meier

Yeah, look, I think there are so many lessons to be taken from this. Russia was a much bigger power attacking Ukraine, given the same logic that we discussed: whoever has more production capacity should win. But Russia didn't achieve victory in Ukraine, and the US didn't get a full victory in Iran. It probably achieved some of its goals, but probably not all of them.

You can flip that as well. What if China has so much more capacity than the US? What if they attack us for whatever reason? How can we hold them back if we don't have the rare earths? As the Ukrainian and Iranian examples show, you actually can hold back something like that even if you're the less capable party.

Shawn Wang

Well, those examples did rely on Chinese supply chains, though.

Lorenz Meier

A partial yes, but if you think about Ukraine from February 22, 2022, through the first half-year or year, it wasn't much reliant on Chinese supply chains. We were just relying on whatever we got. So that's 1 side of things.

Another side of things is basically how much suffering you can withstand along multiple axes. It's not just the military axis. It's also the economic axis and the political axis, I would argue. One of the reasons why wars stop or start is because the political pressure on the leadership internally in the country is so high that you just have to stop it, right?

That differs big-time depending on whether you were the one seen by the population as the party that started the conflict or the one that was attacked. That's 1 part. Another factor is just the overall state of society.

One thing I'm worried about in Europe now is that people are not ready to fight even if they're attacked. When people are asked about that, they're like, “Oh, I'm just going to move somewhere where there's no war.” That's a challenge, and that's what makes Europe weaker right now.

And, you know, the US didn't really have to ever, I think, fight a foreign war on its own turf. I hope that never happens, but in case that did happen, I don't know—how would the rich cities on the East or West Coast behave? Would all the Wall Street bankers and Silicon Valley VCs mobilize and really start working on defense?

I'd love to think so. I like that. That's the way I think about the American spirit.

swyx

They did in World War II.

Yaroslav Azhnyuk

In a way, but look, it wasn't that clear in World War II. Churchill famously said, “America will always make the right decision after trying all the wrong ones,” right?

One could argue that there's this sort of USA that lives in popular culture and was created by Hollywood: cool dudes who will always come and do the right thing. But if you look at international politics, it doesn't necessarily always look like that. Take the Budapest Memorandum: Ukraine gave up all of its nuclear weapons, the world's third-largest nuclear arsenal, because the US, Russia, and the others were very persuasive. They said, “Yeah, just give it away. We guarantee your security.”

Then they said, “Oh, it's not guarantees; it's assurances. We use the word ‘assurances,’ so we didn't promise you much. You just gave it away for free.” Then Russia attacks, and there's no reaction. So, the whole world in 2022 looks at it and says, “Oh, okay. So maybe we should get nukes.” My prediction is that, in the next couple of decades, more countries will be working on their nukes.

Alessio Fanelli

They really should. I've consistently advocated for specifically Japan, South Korea, and Poland to get nukes. Obviously, Ukraine should as well.

Yaroslav Azhnyuk

Someone could argue that if a country currently doesn't work on its own nuclear program, it's doing a disservice to its country, and the government should be fired. It seems like, from recent world history, that this is the only way to actually provide credible deterrence.

So, I guess I think that in Europe, people are not quite sure how America will behave. Will it behave as the Hollywood hero, or will it behave pragmatically, as it did at the beginning of World War II, or as it did when Ukraine was attacked by Russia? I know the US just decided to sort of push the Budapest Memorandum aside because, of course, Russia is a nuclear power, and we don't want to mess with it.

swyx

Everyone says Russia's behind right now in the drone war.

Okay.

Alessio Fanelli

But that wasn't true a year ago. A year ago, people were saying either Russia was ahead or they were at parity. Or maybe that was a year and a half ago. Russia has more people—about 4 times as many people, or more.

Yaroslav Azhnyuk

Yeah, I think, give or take, yeah.

swyx

Yeah.

Alessio Fanelli

30 versus about 120, yeah.

Yaroslav Azhnyuk

Yeah. So, 4 times as many people, more help from China.

swyx

The economy is, like, 10 to 20 times bigger, I think.

A lot of oil money that Ukraine just doesn't have. More direct help from China than Ukraine is getting. So, Russia just has this massive advantage in scaling against Ukraine itself. Ukraine has financial assistance from the EU, but right now, Ukraine is ahead in the drone race.

swyx

I'm not sure about that, by the way.

Okay.

Alessio Fanelli

Well, that was going to be my next question. Is that true? And if it is true, how long before Russia manages to pivot, course-correct, and regain the lead?

swyx

Sorry, for my own curiosity, can we define “drone race”?

Yaroslav Azhnyuk

Yeah, look, I think it's also helpful for our listeners to understand that there are at least 30 different types, or categories, of drones. First, you have different domains: flying drones, ground vehicles, sea vehicles, and underwater vehicles. Then, for each of those domains, you have multiple use cases.

For ground vehicles, you have logistics, evacuation, mining, demining, and maybe something else. For aerial drones, you have reconnaissance, front strike, mid-strike, deep strike, mining, demining, radio repeating, kamikaze bombing, ISR, and different types of surveillance: tactical surveillance, operational-level surveillance, and maybe strategic-level surveillance at some point. Logistics also applies to aerial drones.

For sea drones, it's the same thing. In each of those categories, you have dozens, sometimes over 100, companies and products competing. That's the current Ukrainian battlefield. From the Russian side, it's less of a zoo, as we say. In each category, they usually have 1, 2, maybe 3 products, and then they scale them in a centralized fashion.

When you talk about whether we are behind, or who's behind or ahead in drone warfare, you've got to analyze that area by area. If you're talking about front strike, I would argue that Ukraine has gotten ahead recently after scaling fiber-optic drones. Before that, Russia was slightly ahead.

With mid-strikes, something like 40 to 200 kilometers, it's hard for me to judge. At some point, Russia was ahead. I think maybe we're getting ahead as well. In deep strike, we recently got ahead, so we were doing more damage to Russia with deep-strike drones than they're doing to us.

In sea drones, we're consistently ahead; we were always ahead. In ground drones, I think we're ahead.

swyx

Where are they still ahead?

In general, I think we're ahead. Where they are still ahead, I think, is in certain components, like GPS-free navigation. Their CRP antennas are pretty good. They have these winged bombs that they drop from their bomber planes. I forgot the English name for it.

Alessio Fanelli

Glide bomb?

Yaroslav Azhnyuk

Yeah, sort of. So, they're ahead on that side, and it's difficult to protect against those.

swyx

What's the range of that?

It can be pretty big. I think it can be up to 80 kilometers.

Alessio Fanelli

And then, obviously, the range of the plane?

swyx

Like a strike plane?

Yaroslav Azhnyuk

Yeah, the range is a very iffy subject here because the range is basically the distance from where you drop the bomb to where it lands. But you drop it from a fighter plane, and fighter planes are susceptible to aerial interceptor missiles.

On our side, we have our own fighter planes, and we have ground anti-air systems. Those 2 assets have their radars and radar fields. Depending on the enemy's tactics, you can calculate how big the aerial area is that you cover with those assets.

I'm not a professional military guy, so I'm covering these topics in layman's terms. Don't quote me on this. I'm just trying to make this as understandable to an average listener as possible.

swyx

Helicopters—I’ve recently seen reports of drones taking out helicopters in the air, and that this is new. Is that new? Is that going to be a big deal? Is that eventually going to get rid of helicopters, the way drones are getting rid of tanks from the battlefield?

Yeah, look, helicopters are also very versatile assets. Front-strike helicopters, I think we're going to be seeing fewer and fewer of them. These few Russian helicopters that Ukraine has intercepted with drones were more like edge cases than a systematic sort of helicopter-hunting campaign. I think it is possible to turn it into a systematic countermeasure against helicopters.

swyx

What kind of—will those be battery-powered drones themselves?

Yeah, potentially. There are so many different scenarios. You can have large aerial drone carriers carrying interceptor drones.

Alessio Fanelli

That then go hit the helicopters.

Yaroslav Azhnyuk

For example. Or you can have battery-powered interceptor drones, but not of the missile-with-a-propeller type, like many of these well-known drones, such as Sting or P-1 Sun. They look basically like a missile with a quadcopter behind it. But you can also have plane-like, fixed-wing aerial interceptors.

swyx

So then, does anyone have a little drone that flies super low under the helicopter and shoots it from underneath?

Alessio Fanelli

A surface-to-surface drone that carries surface-to-air missiles somehow.

Yaroslav Azhnyuk

Yeah, I don't think that's very practical because whatever you have going on land will be super slow and not fast enough to hunt down a helicopter.

swyx

I mean, in the air, is there a drone capable of carrying a small surface-to-air missile that can skim low and then launch its little missile, like a flying missile platform?

But a big part of a mission like that is not just kinetically getting to a helicopter, but also identifying it, either by means of first radar and then visually, and placing the asset you have—the interception asset you have—in the right place at the right time.

The combination of those things is much more complex than just asking how we can strike it from behind or from below. But that doesn't mean helicopters are becoming completely useless. For example, helicopters are used to intercept deep-strike drones.

Like Ukraine uses a lot of helicopters to shoot down Shaheds. Russia used helicopters to shoot down our deep-strike drones.

swyx

A lot of people talk about drone countermeasures—things people do technologically to try to shoot down FPV drones or bomber drones. Dumb question that I probably already know the answer to, but for the listeners: Why can't you use a shotgun to shoot down drones that are coming after you? That's the main weapon that people use against them. Why aren't they very good?

They're pretty good. There are hundreds—definitely thousands—of cases of drones being shot down with shotguns, both by Ukrainians and Russians. There's even statistics on what percentage of Ukrainian FPV drones didn't accomplish their mission because they were shot down with a shotgun.

swyx

All right. So, if I'm a guy with a shotgun, I'm walking around, and an FPV drone comes for me—

I don't recommend that.

swyx

No, I don't plan on it. But suppose that were the case.

Okay.

swyx

Or suppose there's a guy—he's not me.

Yaroslav Azhnyuk

Yeah.

swyx

He's dumber than me, okay? He's got a shotgun, he's walking around, and an FPV drone is sent after him. Someone says, “Okay, there's a guy walking around. Kill him. FPV drone, go.” The FPV drone goes after him, and he has a shotgun. What are his chances of using that shotgun to shoot down the drone before the drone gets him? Can you say that?

Yaroslav Azhnyuk

Depending on how good you are with a shotgun. I was talking to some Ukrainian pilot groups, and they told me there was this Russian guy who was just like Rambo. He shot down 7 FPV drones. They couldn't get him. They finally got him, but it was like nothing they'd seen before.

swyx

Non-Rambo—

Noah

The average non-Rambo will just die.

swyx

Will just die. So there's a very low chance that they'll be able to use a shotgun to shoot down the drone.

Low chance, yeah.

swyx

Got it. That was the kind of question I was getting at. There's no sort of portable electronic countermeasure that can get FPV drones if you're just holding it, very effectively.

There are plenty of them. It just depends. Electronic countermeasures are used all across the front line. The tricky thing is that electronic countermeasures cover certain radio-electronic bands of frequencies.

swyx

Let me simplify my question. Is there a man-portable system that will give me a greater-than-50% chance of living if an FPV drone specifically targets me and comes to kill me right now?

If your system jams the frequency the drone works on, and the drone doesn't have optical fiber or last-mile autonomy, then you have a 100% chance that it will not fly toward you. But then, what is the chance of encountering a drone that can either use a different frequency, autonomy, or fiber optics? That depends on the area you're in and who your adversary is in that area and zone.

swyx

Okay. I guess this question was maybe too dumb.

No, it's a great question. There are no dumb questions here. If you feel a common theme in my answers, it's that things in practice, in war, are way more complex than they seem.

swyx

I want listeners to understand why. I've read tons of things that say that, basically, if you're walking around in the open and drones come for you, you're not 100% dead, but you're probably dead. I've read a bunch of things that say that, and I want listeners to understand why—people who are paying a tiny bit of attention to this debate from far away and intermittently in America don't understand the weakness of our military against this kind of attack, against a drone attack.

We have a lot of psychological mechanisms by which we cope with the mental idea of drones. I would like to bust those mechanisms by explaining why drones defeat human infantry on the battlefield.

Yeah, I mean, it's just like a guided bomb flying at you, and it knows exactly where you are. It's not the ultimate weapon, but one of the things that went viral in the Ukrainian defense-tech bubble, even before the words of the CEO of Rheinmetall, was an American battle-tank pilot who was interviewed. He was asked whether he was afraid of FPV drones, and he said, “No. Our tanks are strong.”

That went viral among Ukrainians because they were like, “Dude, you have no idea what you're talking about. Don't mess with those drones.” The Abrams is a great tank, but against an FPV drone, sorry, dude, that's not going to work. Maybe not against 1 drone, but a dozen drones will take it out.

There is hope, though. You just have to have kinetic countermeasures.

swyx

An interesting thing: “kinetic countermeasure” means a thing that shoots down the drone.

It can mean many things. If you go to the Ukrainian east and territories close to the front lines—about 50 km in from the frontline—all the roads are covered by fishnets. You literally ride in a corridor of fishnets. That's the mechanical countermeasure against the drone.

swyx

You count that as a kinetic countermeasure?

Well, a mechanical countermeasure.

swyx

Okay, okay. I don't know all the jargon.

Yeah, whatever. Then the tanks—if you look at Russian tanks, and sometimes even Ukrainian tanks or equipment, they all look like porcupines. They have these long, sticking-out poles. I don't know the proper term.

swyx

We talked about poles already.

A different kind of pole.

And that's the way to protect against drones. That's the way to make the drone detonate maybe half a meter or a meter away from the actual shell of the tank. Sometimes there are nets on top of these tanks, just welded on with some extra equipment.

Then, of course, there are guns. What both Russians and Ukrainians are beginning to experiment with is a kind of interceptor drone—an anti-FPV interceptor drone—which you put on top of something like a gun, like a harpoon, sort of thing. When you see a drone coming at you, maybe you can notice or hear it from 200 m or 100 m away, so you have a couple of seconds. You grab that thing, point it, and fire it.

On board, it has certain AI that helps it guide the small drone toward an attacking drone and intercept it that way. Those are the things being developed, and we're working on some of these things as well. You can imagine an armored vehicle with hundreds of drones on top of it, which are protector drones. They're sort of like active armor. Whenever they see a drone coming at you, they take off.

swyx

What about the kind of thing the Germans are building, which is basically a big truck with some sort of automated shotgun on it?

Yeah, they have Skynex. It's by Rheinmetall, by the guy whom we mentioned today. Skynex is considered an okay weapon. Its shots are quite expensive, though.

swyx

The cost to fire each shot, really?

Yeah, cost to effect in a more abstract way. Last year I was speaking at the LANDEURO conference. It's the biggest U.S. Army conference in Europe, called LANDEURO. There was an expo there, and there was a Raytheon RTX booth.

Raytheon is an amazing company. Gosh, we love Raytheon. They're making Patriots. Patriots are the best. They make a bunch of other things, too. They had this laser gun.

swyx

I was going to ask about that next. It's a laser.

A laser thing. They have it in 2 variations: a 10-kW laser and a 20-kW laser. I'm like, “Okay, 10-kW laser, tell me about it. Can it take down an FPV drone?” They're like, “Yes, of course it can.”

I'm like, “Okay, cool. How much time does it take to take down an FPV drone?” They're like, “Well, maybe 3 seconds.” I'm like, “Hmm, 3 seconds. That's a lot of time.” But okay, maybe fine. And if an FPV drone tries to evade, they're like, “Well, we will retarget it again.” Then 3 seconds starts again.

I'm like, “What can it take down, like a dozen FPV drones?” They're like, “Yeah, for sure.” I'm like, “Okay, a dozen FPV drones, 30 seconds, maybe yes. 2 km, maybe yes, maybe no.” Then I'm like, “How much does it cost?” He says something like $3 million.

I'm like, “Okay, $3 million. So that is 6,000 FPV drones.” I doubt this thing will be able to handle 6,000 FPV drones, or even 600 FPV drones, coming at it at the same time.

You have this kind of economic problem, and this product may not necessarily be a product against an FPV drone in an active battlefield environment. It might be guarding a stadium in a peaceful country. If some random dudes launch a couple of drones above a stadium, you shoot them down.

swyx

Okay. Everyone's happy, although the drone will fall down, maybe fall on someone's head. That wouldn't be cool. So, you would want something like catching bad drones with a net above a stadium or something like that. But whatever, my point is that the economics matter.

Alessio Fanelli

If you sent them one by one, it wouldn't—it would just be pew, pew, pew, pew.

Noah

But who would send them one?

Alessio Fanelli

But if you sent a massive swarm of 6,000 drones—

Noah

Yeah.

Alessio Fanelli

What about a more powerful laser, like a 100-kilowatt laser or something? They wouldn't need to spend the money on the—

Noah

That's worse. You need a less powerful laser that achieves the same effect. A more powerful laser would be more expensive, heavier, and more difficult to transport. It would be more difficult to make many of them, and therefore you wouldn't be able to cover a long front line. It would be super expensive to replace if it gets damaged—all of those issues.

The reason why drones or iPhones become so popular is because they're small and everyone can have one. And so it is with countermeasures. That's another mental shift that you've got to go through. It's no longer about an aircraft carrier that costs, whatever, $14 billion and takes forever to build.

It's about mass that you can iterate on very quickly, upgrade, and have everyone operate. Then that mass, when it is combined—or the technologies, when they're extrapolated from one domain to another domain—they add up, right, as it happens with software. So, I think that's important.

swyx

Can I ask a follow-up question? Russia is not necessarily the smartest army you could be fighting. What would happen if your adversary was smarter? Do you think things would change meaningfully?

Look, I don't know if I fully agree with “not the smartest army.” Who is the smartest army?

swyx

I don't know.

That's a question. I don't know. I think those are very dangerous assumptions to make.

swyx

Who was the smartest army in World War I?

I would define “smart.” Why do you think so? Why do you think Russia is not the smartest army?

swyx

Okay. Maybe this is just my own information bubble.

Maybe I agree with you, but I'm naturally wired now to challenge those assumptions.

swyx

Okay. No, no, that's a really good point. I guess, from my information bubble, it seems like Russia's strategy has largely been to just throw resources at people.

In a Western propaganda information bubble, of course. As am I. We're all rooting for Ukraine to win, right? But, yeah, sorry, go on.

swyx

Yeah, yeah. But going back to this, there's a history of large powers failing to take over smaller ones, strategically speaking—

David and Goliath.

swyx

Yeah, this is—

They fail a lot more now than they used to.

swyx

Yeah. The success rate of taking places over has gone way down.

Yeah, and certainly. But regardless, I do wonder if Russia had not essentially assumed victory early on—

No, they're super stupid, of course. They were marching at Kyiv with their parade costumes, thinking they were going to have a parade in Kyiv in a few days. That was super stupid. There were lots of stupid things. They have no regard, no care for human life. They're sending those Russian folks without armor, without anything—folks on crutches—sending them to storm Ukrainian positions. It's like—

Shawn Wang

Do you think, at this point, there's—

Noah

I actually have a good friend. He's American, from Seattle. He served in special forces here in the US and had been on maybe 3 deployments. Then he went to Ukraine as a volunteer, and he's been fighting since 2022. He's a very good friend of mine.

At some point, he's been texting me, and he's like, “Okay, I'm near Pokrovsk.” It wasn't Pokrovsk; it was—gosh, the other city—Chasiv Yar. And he's like, “Okay, so what the Russians are doing, they're just creating so much work for all the psychologists who are going to heal those Ukrainian riflemen or machine gunners who are just shooting at the Russians who are going nonstop, right?”

The Russians are causing psychological trauma to Ukrainians because they're dying in such a stupid way. That is indeed stupid of the Russian high command, et cetera, et cetera. But then, that's the resource they have.

Alessio Fanelli

If you've got Zerglings, you use your Zerglings.

Noah

That's the way—that's their strategy, right?

Alessio Fanelli

Zerg—

Noah

Yeah.

Alessio Fanelli

Back in the day, that's what you did. If you play StarCraft, that's how Zergs win, right?

Noah

The Terrans?

Alessio Fanelli

I don't know. I'm working on that.

Noah

Hopefully, we can do Protoss technology with Zerg speed. That would be the best. I think that's what the Ukrainian housewives are working on.

Alessio Fanelli

In fact, you cannot beat those housewives. Do not oppose Ukrainian housewives.

Noah

Housewives, for sure. Yeah.

Shawn Wang

Two final questions. First one, you started out by telling us a story about going to a chapel on February 23rd.

Noah

Yeah.

Shawn Wang

Were you able to get married there? And, yeah, can you finish that story?

Noah

We actually did get married, but we postponed the wedding as a social event until the war is over.

Shawn Wang

Okay, cool. And then, last question: What do you want our audience to take away? If you have 1 point you want them to walk away with, what would it be?

Noah

One piece: be prepared for war. You've got to invest in defense and security.

Shawn Wang

Right. Thanks. Thank you for talking with us.

Noah

Thank you.

Shawn Wang

And thank you, Noah, for all the great—

Alessio Fanelli

Absolutely great. Yeah, great to—

Shawn Wang

Yeah, that was fantastic. Thanks so much.

Alessio Fanelli

Awesome. Thanks.