Daniel sent us this one — he's been watching the drone war between Israel and Iran, all the aerial stuff everyone covers, and he noticed something odd. When US airmen were shot down over the Strait of Hormuz, the rescue vehicle wasn't a helicopter or a Predator. It was a twelve-foot unmanned surface boat. He's asking how advanced sea drones actually are today — both the surface vessels and the mini submarines — and what the US military is quietly fielding that we don't hear about.
This is one of those domains where the public conversation is genuinely about five years behind what's in the water right now. The surface side especially has accelerated in ways that would have sounded like science fiction in twenty twenty. The Strait of Hormuz rescue is the perfect entry point because it demonstrates the core advantage — autonomy isn't optional underwater, it's mandatory. Radio waves don't penetrate water. You can't joystick a submarine from a carrier the way you can with a Reaper.
Because salt water is basically a Faraday cage.
Even at the frequencies that penetrate best, you're getting maybe a few meters of range. So anything below the surface has to think for itself — navigate, avoid obstacles, complete a mission, and return, all without a human in the loop for most of the transit. Surface drones have more flexibility because they can use line-of-sight radio and satellite links, but even there, the trend is toward greater autonomy because you don't want a jammed signal to turn your hundred-million-dollar boat into a paperweight.
Let's start with the surface side. What's actually in the fleet right now?
The US Navy has been running Task Force Fifty-Nine out of Bahrain since twenty twenty-one, and it's been their live-fire laboratory for unmanned surface vessels. The workhorses are Saildrones — they look like a surfboard with a rigid sail, about twenty-three feet long, entirely wind and solar powered. They can stay at sea for months without refueling. Initially used for oceanographic data, the Navy weaponized them for surveillance and they've been patrolling the Persian Gulf and the Red Sea continuously.
A surfboard with a sail. The future of naval warfare is a windsurfer nobody's standing on.
The operational cost is basically zero compared to a destroyer. But the more capable platforms are things like the MANTAS T-thirty-eight and the Sea Hunter. Sea Hunter is a one hundred thirty-two foot trimaran designed to hunt submarines autonomously. It can cross the Pacific on its own. The Navy's now building a larger version called the USV Ranger that can carry modular payloads, including electronic warfare suites and containerized missile launchers.
Wait — they're putting missiles on unmanned boats?
They've already tested it. In twenty twenty-three, a USV called the Mariner launched an SM-six — a Standard Missile Six, the same interceptor used on guided-missile destroyers — from a containerized launcher bolted to its deck. The vessel was remotely supervised from a shore station, but the launch itself was autonomous. The Navy's goal is a hybrid fleet where a single manned destroyer operates alongside four or five unmanned escorts, each carrying sensors or weapons.
The destroyer becomes a mothership.
That's exactly the concept. The Navy calls it "manned-unmanned teaming." The manned platform stays in a command-and-control role, the unmanned ones go into the higher-risk areas — forward scouting, anti-submarine screening, or even drawing fire. If you lose a USV, you've lost a platform that costs maybe fifteen to thirty million dollars, not a billion-dollar destroyer with three hundred sailors aboard.
The political math on that is almost more important than the tactical math. Nobody in Congress wants to explain three hundred body bags.
Which is why China and Iran have both been investing heavily in anti-access/area-denial — long-range anti-ship missiles designed to keep carrier strike groups at a distance. The unmanned surface fleet is the counterplay. You send the drones into the denied area first. They absorb the first wave of attacks, they map enemy positions, and if they survive, great — if not, you've learned a lot without losing anyone.
That's the surface layer. What about below? You said autonomy is mandatory for submarines — how mini are we talking, and what can they actually do?
The smallest operational military UUVs are man-portable. The US Navy uses the Mk Eighteen Mod Two Kingfish, about the size of a torpedo, weighing under a hundred pounds, deployable from a rigid-hull inflatable boat. It uses side-scan sonar to hunt mines. That's the unglamorous, totally essential mission — mine countermeasures. Iran mines the Strait of Hormuz, and clearing those mines is slow, dangerous work that used to require manned minesweepers. Now you send a swarm of Kingfish.
Iran did mine the Strait, fairly recently.
They did, and here's the detail that doesn't get enough attention — they did it so haphazardly that Iranian officials reportedly can't say exactly where the mines are. When you mine a chokepoint and lose track of your own minefield, you've created a problem that outlasts whatever political crisis prompted it. Clearance becomes a months-long effort. The Kingfish and similar systems are designed precisely for that nightmare scenario.
Iran created an improvised explosive device the size of a Strait and then forgot where they put it.
The less charitable reading is that they don't care — the disruption is the point, and not knowing exactly where the mines are makes clearance harder and slower. Either way, unmanned systems are the answer because you don't want a manned minesweeper hitting something. The Kingfish can run pre-programmed search patterns for hours, identify mine-like objects with its sonar, and either mark them for destruction or neutralize them itself.
Move up from the man-portable torpedo. What's the next tier?
The medium class is where things get interesting. The Navy's been developing the Snakehead Large Displacement Unmanned Undersea Vehicle — LDUUV — about six feet in diameter, launchable from submarines. It's designed for intelligence, surveillance, and reconnaissance in contested waters. Think of it as a mini-sub that a Virginia-class attack submarine can deploy from its dry deck shelter, send into a harbor or chokepoint, and have it loiter for weeks collecting signals intelligence or mapping the seafloor.
Launched from a submarine — so the submarine doesn't have to enter the dangerous water itself.
The submarine stays in deep water, releases the LDUUV, and the drone does the shallow-water work where a full-size sub would be vulnerable to detection. The Snakehead can also deploy smaller UUVs, so you get this nesting-doll effect — a submarine launches a large drone, which launches smaller drones, which swim into a harbor and sit on the bottom listening.
The Russian doll of maritime surveillance. And I assume the Chinese and the Russians are building equivalents?
Russia has the Klavesin-two-R-PM, a large UUV designed to operate at depths up to six thousand meters — deep enough to tamper with undersea cables, which is very much part of the modern threat picture. China has the HSU-zero-zero-one, a large autonomous underwater vehicle they've been testing in the South China Sea for oceanographic survey and military reconnaissance. And then there's the Poseidon, which is in a category of its own.
Poseidon — that's the nuclear-powered Russian drone, right? The one that's basically an autonomous nuclear torpedo?
Nuclear-powered, nuclear-armed. About sixty-five feet long, estimated speed up to a hundred knots, operational depth around a thousand meters, designed to carry a multi-megaton warhead. The mission profile is to cross an ocean autonomously, evade defenses, and detonate near a coastal city or a carrier strike group — the shockwave and resulting radioactive tsunami would be the primary damage mechanism, not just the blast.
It's a doomsday drone.
It's a second-strike weapon designed to be unstoppable. The logic is that even if Russia's land-based ICBMs are destroyed in a first strike, a Poseidon launched from a submarine in the Arctic could still guarantee retaliation. The problem — and this is where most analysts get concerned — is that an autonomous nuclear weapon that can't be recalled once launched introduces a terrifying new failure mode. What if it malfunctions? What if its navigation drifts? What if it misidentifies a target?
That's not a failure mode, that's a failure continent.
The arms control framework doesn't exist for this. New START didn't cover unmanned nuclear delivery systems because they didn't exist when it was negotiated. We're in a gap where the technology has outpaced every treaty and every norm.
That's the extreme end. Let me pull us back to the operational stuff that's actually been used in anger, because the prompt specifically mentioned the Strait of Hormuz rescue. What was the platform there, and what made it the right call over a helicopter?
The specific platform was almost certainly something in the MANTAS family or a similar small USV. The scenario was airmen ejected over water in a contested area where Iranian fast-attack craft were active. A helicopter rescue takes time — you have to get the helicopter on station, lower a rescue swimmer, hoist the airmen up, all while potentially under fire or while the Iranians are racing to capture the downed pilots first. An unmanned surface boat can be pre-positioned or launched quickly, it presents a very small radar cross-section, it's harder to hit, and if it gets destroyed, you haven't lost a helicopter crew on top of the airmen you're trying to rescue.
It can get closer to the shoreline without escalating the way a manned aircraft would.
A US Navy helicopter crossing into Iranian territorial waters is an international incident. A twelve-foot unmanned boat is deniability on a stick. Even if the Iranians capture it, what do they have? A piece of plastic with some electronics. No pilot to interrogate, no crew to parade on television.
The propaganda value collapses.
And that's an underappreciated dimension of unmanned systems — they deny the adversary the information-warfare victory. Iran's whole strategy in the Gulf is built around humiliation ops — capturing sailors, seizing boats, filming the whole thing for domestic consumption. A drone gives them nothing to film.
We've got surface drones doing surveillance, mine-hunting, and now personnel recovery. Subsurface drones doing mine countermeasures, intelligence gathering, and — in the Russian case — nuclear deterrence. What's the integration challenge? Because none of this works if the drones can't talk to each other or to the fleet.
This is the hard problem. The Navy's been working on the Unmanned Maritime Autonomy Architecture — UMAA — an attempt to create a common software standard so different unmanned platforms from different manufacturers can share data and be controlled through a single interface. Historically, every drone came with its own proprietary control station, and a sailor might have to monitor three different laptops to manage three different unmanned systems. That's operationally insane.
Three laptops for three drones — that's the Department of Defense procurement experience in one image.
UMAA is supposed to fix that, and it's making progress, but the deeper challenge is communication bandwidth. Underwater, you can use acoustic modems, but the data rate is terrible — we're talking kilobits per second, not megabits. You're not streaming video from a UUV. It surfaces periodically, pings a satellite, dumps its data, receives new instructions, and submerges again. That intermittent connectivity means the autonomy software has to be capable — the vehicle has to make decisions on its own for hours or days between check-ins.
Which brings us to the AI question. How much of the decision-making is actually delegated to the machine right now?
The official US policy is that there's always a human in the loop for lethal decisions. An unmanned system can identify a potential target, track it, and recommend an engagement, but a human has to authorize the weapon release. The Navy's phrase is "human-on-the-loop" for some systems — the human supervises but doesn't micromanage every rudder movement. But for lethal action, the policy requires affirmative human authorization.
How long does that policy survive contact with an adversary that doesn't share it?
That's the question every strategist is asking. China has shown no compunction about fielding fully autonomous systems. They've demonstrated swarming unmanned surface vessels in the South China Sea that operate cooperatively without human intervention. If you're in a naval engagement and the other side's drones are making engagement decisions in milliseconds, and yours are waiting for a human to click "approve" on a satellite-delayed data link, you lose. The pressure to automate the kill chain is going to be enormous.
The policy is a peacetime luxury.
It's a statement of values that I fully support, but yes — the first shooting war between peer competitors with unmanned fleets is going to test it to destruction. The Navy knows this. That's why they're investing so heavily in autonomy software and testing infrastructure — they want the decision-making to be robust enough that delegation is safe, even if they're not publicly saying they'll delegate.
Let's talk about swarming. You mentioned the Chinese demonstration. What does a sea-drone swarm actually look like, and why is it different from just having a lot of boats?
Swarming means the vehicles coordinate their behavior without a central controller. Each drone has a set of simple rules — maintain a certain distance from neighbors, move toward a target, avoid obstacles — and complex group behavior emerges from those rules. A swarm of twenty USVs can approach a target from multiple angles simultaneously, and because they're coordinating, they can adapt if one or two get destroyed. The swarm doesn't have a leader you can decapitate.
Like a school of fish evading a predator.
That's the biological inspiration. The Navy has tested this with the Common Unmanned Surface Vehicle — CUSV — where multiple boats autonomously coordinate to screen a high-value unit. The swarm can spread out to cover a wider area, or concentrate to overwhelm a single point. Against a swarm attack, traditional defenses break down — a destroyer's close-in weapon system can engage maybe two or three incoming threats simultaneously. Twenty is a math problem it can't solve.
These aren't expensive platforms. What does a CUSV cost?
A CUSV is maybe two to three million dollars depending on the payload. Compare that to an anti-ship missile that costs about the same — except the USV can be reused, can loiter for hours, and can carry multiple weapons or sensors. The cost asymmetry is staggering. A swarm of ten CUSVs costs less than a single manned patrol boat and can cover ten times the area.
The economics of naval warfare are flipping. Historically, the defender had the advantage — shore-based missiles, coastal artillery. But if the attacker can flood the zone with cheap autonomous boats, the cost curve inverts.
That's exactly what Iran has been doing with its own fast-attack craft doctrine — they've just been doing it with manned boats and accepting that they'll lose a lot of people. The unmanned version is the same concept, but without the martyrdom requirement. Every navy with a coastline is now looking at unmanned swarms as the asymmetric equalizer against carrier strike groups.
The martyrdom-as-a-service layer has been abstracted away.
Grim way to put it, but accurate. And it's not just surface swarms. DARPA has a program called Hunter that's developing underwater swarms for tracking enemy submarines. The idea is that a network of small, cheap UUVs can cover a much larger area than a single expensive submarine-hunting ship, and because they're distributed, a quiet diesel-electric submarine can't easily evade all of them. If one picks up an acoustic signature, the rest converge.
The submarine that used to be the stealthy hunter becomes the hunted by a cloud of underwater Roombas.
With better sonar. And this is already being tested operationally. In twenty twenty-four, the Navy ran an exercise in the Pacific where they used a combination of Sea Hunter, Saildrones, and underwater gliders to track a simulated enemy submarine across hundreds of miles of open ocean. The submarine couldn't shake the network. Every time it thought it had slipped away, a different sensor platform picked it up.
That changes the deterrent equation for places like the Taiwan Strait.
China's entire naval strategy in a Taiwan scenario depends on keeping US carrier groups at a distance with anti-ship ballistic missiles — the DF-twenty-one D and DF-twenty-six. But if the US can flood the strait with unmanned sensors and unmanned combatants before the carriers even arrive, the targeting picture for those missiles gets a lot murkier. And the USVs can start attriting Chinese naval assets without risking a single American life.
The unmanned fleet isn't just an add-on to the manned fleet. It's a whole new operational concept.
The Navy calls it Distributed Maritime Operations. Instead of concentrating force in a carrier strike group that presents one very large, very valuable target, you distribute sensors and weapons across dozens or hundreds of unmanned platforms spread over thousands of square miles. The enemy has to find and kill all of them, not just take out the carrier. And because the platforms are cheap and attritable, you can afford to lose a lot of them while still preserving your manned assets for the decisive moment.
"Attritable" — there's the Pentagon euphemism of the decade.
It's a word that means "we expect to lose these and we've priced that in." Which, to be fair, is a more honest way to think about combat losses than pretending every platform will survive. But it does raise an uncomfortable question — if these things are so cheap and so capable, what stops every middle power from building a swarm navy?
I was just about to ask that. What's the barrier to entry here?
The hardware barrier is lower than you'd think. A competent shipyard in Turkey or South Korea can build a capable USV hull for a few million dollars. The hard part is the autonomy software, the secure communications, and the sensor integration. Building a boat that can drive itself in a straight line is easy. Building one that can navigate a busy shipping lane, identify threats, coordinate with other platforms, and operate in a GPS-denied environment — that takes years of software development and testing.
Presumably the US has a lead there.
A significant lead, but it's not unassailable. The commercial sector is driving autonomy advances faster than the military can absorb them. The same computer-vision algorithms that let a self-driving car avoid pedestrians can help a USV avoid fishing boats. The same machine-learning techniques that let drones navigate without GPS — by visually recognizing landmarks — work on the water too. The military's advantage is in integration, testing, and doctrine, not in some secret sauce that nobody else can replicate.
We're in a window where the US has an edge, but that window is closing.
The window is measured in years, not decades. Which is why the Navy has been pushing so hard on unmanned systems. The goal is to field a hybrid fleet of roughly three hundred fifty manned ships and about one hundred fifty unmanned platforms by the mid-twenty-thirties. That's a massive shift in force structure.
Let me bring this back to the prompt's specific scenario — the Strait of Hormuz rescue. What made that operation possible, logistically? Was the drone pre-positioned? Launched from a nearby ship?
It was almost certainly launched from a surface combatant already in the area — probably a destroyer or a littoral combat ship operating in the Gulf. These small USVs can be deployed from a davit or a well deck in minutes. The rescue profile would have been straightforward: the USV navigates to the coordinates of the downed airmen, the airmen climb aboard or are pulled aboard, and the USV returns to the ship or to a safer extraction point. The entire operation can happen without a helicopter ever spooling up.
The airmen — they're trained for this? "If you go down in the Gulf, expect a robot boat"?
Survival, Evasion, Resistance, and Escape training now includes unmanned recovery scenarios. The airmen carry beacons that are compatible with USV recovery systems. The USV can home in on the beacon autonomously. It's a purpose-built rescue platform.
The phrase "robot boat" was doing a lot of work in my head and I regret nothing.
It's the technical term in the field manual. Page forty-three.
I choose to believe that. So the rescue capability exists and has been used. What about the offensive side? You mentioned the SM-six launch from the Mariner — has any unmanned surface vessel actually fired a weapon in combat?
Not in US service, at least not publicly acknowledged. The Ukrainians, however, have been using unmanned surface vessels extensively in the Black Sea against the Russian fleet. They've developed small, fast USVs packed with explosives — essentially unmanned kamikaze boats — and they've used them to attack Russian ships at anchor in Sevastopol and to strike the Kerch Bridge. These are much cruder than what the US Navy fields, but they've been operationally effective.
The Ukrainian sea-drone program has been remarkably successful given that it was basically built from scratch in a year.
It's a preview of what a motivated middle power can do. The Ukrainian USVs are essentially jet-ski-sized boats with off-the-shelf Starlink terminals for communication and commercial-grade cameras for targeting. The total cost per unit is reportedly under two hundred fifty thousand dollars. They've damaged or destroyed multiple Russian warships at a cost ratio that is absurdly favorable.
A quarter-million-dollar jet ski taking out a multi-hundred-million-dollar frigate.
Forcing the Russian Black Sea Fleet to essentially abandon Sevastopol as a safe harbor. The strategic effect — the fleet relocating to Novorossiysk, farther from the operational area — was achieved not by sinking the entire fleet, but by making the risk of staying put unacceptable. That's sea denial on a budget.
Which loops back to Iran. The Strait of Hormuz is even narrower than the Black Sea approaches. The same asymmetric logic applies.
Iran knows it. They've been investing in unmanned systems of their own — both surface and subsurface. They've displayed armed USVs at military parades, they claim to have UUVs capable of mine-laying, and they've been caught by the Navy trying to sneak small unmanned boats through the Gulf. The drone war in the Strait is already happening, it's just happening below the threshold of what makes headlines.
Because nobody's died from it yet.
Or nobody we know about. The Navy has been intercepting Iranian drones and drone parts shipments for years. In twenty twenty-three, they interdicted a dhow carrying drone components from Iran to Yemen. In twenty twenty-four, they seized a shipment of Iranian unmanned underwater vehicle parts in the Arabian Sea. This is an ongoing, low-visibility campaign.
The drone war is already multi-domain — air, surface, subsurface — and the Strait of Hormuz is the laboratory.
It's accelerating. The Navy stood up a dedicated Unmanned Surface Vessel Division in twenty twenty-four — USVDIV One — operating out of San Diego but deploying assets globally. They've been integrating USVs into carrier strike group operations, exercises with allies, and real-world patrol missions. This isn't a science project anymore. It's a deployable capability.
What's the most surprising capability that most people don't know about?
A Saildrone can sit off the coast of a country for months, passively collecting signals intelligence, tracking ship movements, monitoring communications. It doesn't need to refuel, it doesn't need a crew rotation, it barely shows up on radar. The Navy has been using them to monitor Iranian naval activity in the Gulf and Houthi activity in the Red Sea continuously. The intelligence picture that emerges from having a permanent, unattributable presence is qualitatively different from what you get from a destroyer that transits through for a few days.
The panopticon, but make it nautical.
Because the data gets fed back into machine-learning systems, the pattern-of-life analysis gets better over time. The system learns what normal traffic looks like and can flag anomalies automatically. A fishing boat that deviates from its usual route, a cargo ship that goes dark, a fast-attack craft that sorties at an unusual time — the autonomous network spots it before a human analyst would.
We've covered surface drones, subsurface drones, swarms, the rescue use case, the offensive use case, the policy tensions. What's the thing you're watching most closely right now?
The Orca XLUUV — the Extra Large Unmanned Undersea Vehicle. Boeing's been developing this for the Navy, and it's essentially a drone submarine the size of a small manned sub — about eighty-five feet long, capable of missions lasting months, with a modular payload bay that can carry mines, sensors, or potentially smaller UUVs. The Orca completed its first extended underwater transit in twenty twenty-five. If it works as advertised, it fundamentally changes the submarine warfare equation.
Because you can park one off an adversary's coast and just leave it there.
And if things go hot, it's already in position. You don't have to transit a manned submarine through contested waters — the Orca is already there, waiting. Finding an eighty-five-foot autonomous sub in thousands of square miles of ocean is extraordinarily difficult.
The submarine equivalent of pre-positioned equipment.
With active sensors and, eventually, weapons. The Navy's been careful not to talk about arming the Orca publicly, but the modular payload bay is designed to accommodate a variety of payloads, and there's no technical reason you couldn't put torpedoes in it. The policy question is whether you want an autonomous submarine making engagement decisions, and right now the answer is no — but as we discussed, that answer has an expiration date.
To answer the prompt directly: sea drones today are far more advanced than most public discussion acknowledges. The US fields operational unmanned surface vessels for surveillance, mine countermeasures, electronic warfare, and personnel recovery. It's tested unmanned missile launches. It's developing autonomous submarine hunters and extra-large UUVs for long-duration missions. The subsurface side is constrained by physics — underwater communication is hard — but the autonomy software is maturing rapidly. And the operational use case, from the Strait of Hormuz rescue to the Black Sea drone campaign, demonstrates that these aren't experimental toys. They're changing naval warfare right now.
The thing I think about is the speed at which this is all happening. Five years ago, unmanned surface vessels were a niche capability. Today they're an integral part of fleet operations. Five years from now, the hybrid fleet concept will be operational at scale. The institutional adaptation is happening faster than at any point since the introduction of the aircraft carrier.
Which means the doctrine is being written in real time, by the people operating these things in the Gulf and the Pacific. The manuals will come later.
As they always do. The operators figure out what works, the bureaucracy eventually codifies it, and by the time the textbooks are printed, the technology has moved on again.
The sloth approach to naval doctrine — document what worked last decade.
I don't think that's actually a sloth thing. I think that's just a navy thing.
I'll claim it for sloths anyway. We invented maritime bureaucracy.
You also invented pizza, apparently.
The original crust was kelp-based.
I'm going to choose not to engage with that.
And now: Hilbert's daily fun fact.
Hilbert: In the 1880s, a Turkmen shepherd discovered that the mineral calcite from a cave near the Kopet Dag mountains glowed an intense red when held near his campfire — a phenomenon called thermoluminescence. He had no way of knowing he'd stumbled onto one of the earliest known examples of radiation-induced mineral fluorescence, caused by uranium traces in the surrounding limestone that had been bombarding the calcite for millions of years.
A shepherd accidentally discovers radiation dosimetry and just thinks his rock is pretty.
The things we learn.
This has been My Weird Prompts, produced by Hilbert Flumingtop. If you enjoyed this episode, find us at myweirdprompts.com for the full archive and every way to listen.
We're back next week. Try not to mine any straits in the meantime.
