As the Ukraine War enters it’s 5th year, and the Oct 7th Wars start their 4th… we’re sorta getting a feel for the ultimate shape of modern drone warfare, AI , and its limits.
I’ve written previously on the use of AI surveillance against insurgents and assassins and how to counter them in my hit pieces “Getting Away With It: The Art of Urban Infiltration and Exfiltration” and “Tactical Espionage DollarStore Hobo”… But now I want to address conventional warfare in the Cyberpunk 2020s we find ourselves in… Lets discuss “AI Drone Swarms”… and why they don’t and almost certainly won’t exist.
.
.
Despite all the buzzwords about drone swarms… Ukraine is sort-of peaking at around 9000 drones employed per day across the entire front. That’s 270,000 drones Monthly or about 3.2 million per year.
Across it’s 1250 kilometer front that’s about 7 drones per kilometer, per day. As of Oct 2025.
(yes this is a 1 dimensional assessment, ideally you’d want an x-axis, drones per “square”-kilometer, so you compare warzones with frontlines to those with deep fronts, or no fronts (guerilla wars, insurgency), and ideally index it to population to create a real universal measure of unmanned combat intensity… but this makes the math simple)
.
Russia stated in 2024 that they wanted to create and employ 4000 drones per day, and in 2025 Ukraine said Russia was aiming to produce 1000 Shahed style (larger long range) drones per day. In addition to it’s quadcopters and others.
The language around drone type is messy… Many call Shahed’s “Strategic Drones” yet they can by shot down by crop-dusters with a mounted machinegun… Thus a “Strategic Drone” can be shot down by a Sub-Tactical civilian aircraft. So forgive me if the language of this piece is imprecise with regards to whatever doctrine has been slapped together and will be defunct in 3 months.
.
There’s good reason to think (beyond both sides fighting existentially for so long) that this is about the upper limit of what nation states can really sustain in an area.
Even with a low combined unit and logistic cost of $1000-2000 per quadcopter drone 3.2 million drones would be 3.2-6.4 billion dollars in expenditure for a single low level support munition… the Shaheds by contrast cost $20,000-50,000.
Once we assume that a good percentage of these drones have 5000-10,000 dollar thermal optics… its very likely tens of billions annually are already being spent on the drones that are being CONSUMED.
Remember the ENTIRE AMERICAN NAVY… with 11 aircraft carriers, 450 ships, 50,000 non-combat vehicles, 75,000 buildings, and 400,000 personnel…
Costs about 200 billion dollars annually…
The USS Gerald R. Ford: America’s Flagship, the Largest most advanced Aircraft Carrier in existence, cost 13 billion dollars, and has a projected service life of 50 years. (with a single new Ford Class being introduced, at this rate, every 10 years… they’re behind schedule)
It’s very possible Ukraine and Russia are spending that much, or even MULTIPLES of that annually, on drones that are consumed and dead within 2-5 flights, if not single use.
.
So we’ve basically seen the order of magnitude in terms of the upper limit of drone saturation if not quality, and the price isn’t really going to come down.
It’d be largely inconceivable that some power is going to pay 100-200 billion dollars on drones annually to get up to 30 million drones per year, and produce a radically different battlefield than what we see now… Even in a World War 1 style total war where countries throw 20, 30 or 40% of their GDP at the war… it’d be very hard to picture them managing to get to the point of spending hundreds of billions on drones annually… Not least because they’d use those drones to target each-other’s economy, manufacturing facilities for drones (as we’ve already seen, but heavily limited by treaties and locating facilities in 3rd countries), and most importantly Chip producers and Lithium supply lines.
Also remember that the neccessary oil and gas, logistics, production facitilites, drone factories… Etc. All make excellent and obvious targets for drone strikes themselves (the obvious consequence of drone warfare is the destruction of logistics) due to varies agreements and norms, and outsourcing to allied countries, most production and logistical facilities have been vastly safer than in previous wars, with those norms generating headlines whenever they’re broken… In future wars it is far less likely belligerents will leave those targets intact.
.
Beyond this: The current drone density of 7 drones per kilometer per day is already really artificially concentrated, remember Ukraine and Russia aren’t fighting along the entire border, nor along the Belorussia border… If they were fighting along the entire border approximately 2,295 kilometers, and the Belorussian border, approximately 1084 kilometers), that drops to about 2.7 drones per kilometer per day.
It’d take a ww1 style total war + total mobilization + really favorable economics and security of economies, the involvement of both the US and China, AND a ww1 style extended concentration on a single narrow front that is really hard to conceive happening in any modern World War… before you saw something really qualitatively frighteningly different like 50 or 100 drones per kilometer per day. (which is what it’d take to get multiple drones every hour for every kilometer of front, night and day)
And even a total unlimited Armageddon war for Taiwan… Maybe the most concentrated high intensity conflict you could possibly imagine might not produce that frightening next level drone intensive conflict, since urbanization, mountains, jungles, humidity, weather (Taiwan has a monsoon season), and the logistical challenges of amphibious operations all individually devour the effectiveness, survivability, and useability of small drones… Requiring more expensive, heavy investment, and harder to use unmanned systems that still will be less survivable, attrit faster, and gather less information because of the conditions.
Indeed the Russo-Ukraine War already is basically that war, given it’s the Russian Economy, people, and a large number of it’s allies resources (Notably China) vs. basically the entire economy and armament stockpile of NATO (which has been dumped on Ukraine), all with really simple classical Army vs. Army logistics that allow for high investment since you’re transporting assets from friendly territory overland to the front, instead of a complex multi-stage vulnerable amphibious supply line. (Ukraine is analogous to the western front of WW1 in this respect as well as others)
.
And Have i mentioned the Ukrainian Steppe is the flattest, least forested most visible place on earth for drones… with the best possible elevation?

Ukraine is unique in that it might be the single environment on planet earth where Drones are most effective, they’re least likely to lose signal from a hill between the operator/repeater and the drone, and just drop out of the sky, there is the least canopy and leaves for the drone to potentially hit… And its optics are the most likely to see something since there isn’t either tree cover or rolling hills to devour the drone’s visible range…
those 7 drones per kilometer per day we see in Ukraine are what 100 or a thousand per kilometer might feel like in Ohio or Appalachia where one had vastly more canopy and valleys to maneuver and hide, that Ukraine so lends itself to drones being effective.
.
Basically all wars for the rest of at least foreseeable terrestrial history (so the next 100ish years) are going to be fought below a 20 drone-day/kilometer threshold, even all but the highest intensity will be below 10… And the vast majority below 2.
And Almost all those wars are going to be fought In vastly more rolling terrain, further above sea level, with more forests, and more urban environments.
There will maybe be 1 Stalingrad that will break that trend locally and become the stuff of legend and nightmares that it reaches 50 drones per kilometer per day (but there we’d expect an urbanized battlespace/ruble to reduce their effectiveness)
And the reason is the physical limits of drones, how cheaply you can put them together: simply what optics, motors, chips, payload, and lithium batteries cost cannot really go meaningfully cheaper.
The largest companies in the world have been fighting for decades now to get their lithium batteries to have just that much more charge, 50% just a 50% improvement… And basically everything you can wring out of lithium as far as energy density, an element on the periodic table, has been gotten trying to make your phone last longer.
You should expect cost and energy efficiency of Lithium batteries to improve over the next two decades about as much as you expect the cost and energy efficiency of gasoline to improve. If you don’t think your car is going to go 3 times farther or get 3 times cheaper between refills, you shouldn’t really expect drone batteries (or electric car batteries) to do much better. Both are matters of basic chemistry, thermodynamics, and energy density.

Indeed it is very conceivable that under any extreme international crisis or disruption chip manufacture, lithium, and other complex good supply will be disrupted or enter scarcity, and the ratio of drones to kilometer/day, far from the current 7 drones expended per kilometer per day, will drop well below 1 even on the hottest front which would be used for scouting and high value targets in conditions of such scarcity… That’s certainly not going to be all wars, but in the same way Forest vs. Urban vs. Mountain warfare are radically different and require different skills, high vs. low drone density warfare are going be different genres of warfare that appear regularly around the world, with countless subgenres and unique tactical scenarios and types of challenges that develop out of it.
Even Moreso than the way major wars generate oil shocks and rationing, or that it’s not been historically unheard of for mechanized armies to retain horses and bicycles to husband scarce petroleum (see the Wehrmacht)… Expect financial, supply line, manufacturing attacks, resource scarcity and logistical disruptions to be VERY COMMON in the domain of drones since whereas petroleum is self similar and fungible… drones require vastly more complex supplies and components where shortages in one produce bottlenecks for the entire chain.
.
Beyond this, as already stated: intense weather, urbanization, foliage (jungle/forests), mountains, distance, and non-conventional “frontless” warfare (insurgency, guerilla warfare, espionage) all radically reduce the density and dominance of drones… either by crippling the survivability of the drone, its intelligence gathering capabilities, or by radically increasing the territory they have to cover in ways that are asymmetric and unique to each battlespace.
.
This has a lot of interesting implications that I want to write about.
Conclusion 1: AI “Drone Swarms” Are Probably Useless… If the term ever meant anything to begin with
(unless the point is to have “killer robots” as a bureaucratic responsibility screen, or scam old congressmen into giving you a budget)
At no point in the war in Ukraine, in spite of the incredible, and probably not to be repeated, density of Drones… At no point has there ever been a shortage of Drone Operators.
At 7 drones/kilometer/day… At a 5-40 minute flight time you’d need only ONE GUY to pilot all the drones for an entire kilometer of front. 1250 for all of Ukraine.
Even if for some obscene reason it took you 24 hours to operate a drone for 40 minutes, and the hard limit was 1 person per drone per day… Ukraine would only need 9000 drone operators, maybe 27,000 if you had to cycle them, or pair them up… In reality I’ve heard many non-specialists say that they operated drones a bit in Ukraine but did other roles most of the time… Meaning Ukraine, in one of the most intensive drone wars in history, might need fewer that 1000-2000 full-time dedicated drone warfare pilots, that a lot of it is sort of a part-time task, or secondary skill.
Indeed the real limiting factor isn’t pilots, it is range and guys willing to get close enough to the front to deploy the quadcopter drones (drone operation is a lot like sniping, if the enemy is in range so are you… Drone operators often duel and hit each other)… and the contour of the land and jamming limiting how far the drone can get signal.
You might think this suggests a roll for the AI, have a larger drone mothership that deploys the drones and then the AI pilots all the little drones, indeed this is largely what is being suggested, and has been semi-demonstrated… That AI can be used to create unjammable drones which can still maneuver and select targets.
but if you already have the mothership high in the air it can operate as a signal repeater that a pilot further away can just use the FPV goggles and controller and pilot the drone in real time pinging off the mothership, using all his judgement and accumulated skill, and exploiting all the intelligence the drone gives him.
This is how “drone motherships” (manned planes, and unmanned larger drones) are used today.

The only thing AI is actually useful for is terminal guidance, if the drone gets jammed and needs to guide itself into the target without being able to contact it’s pilot… Then the robot can take over.
Just one problem:
If you’re using “AI” for terminal guidance it has to be ON the Terminal Armament, your $1000 cost cut-to-hell drone you’re trying to turn out as many of as possible suddenly needs a bloody Nvidia Graphics card strapped to it if you want to integrate Neural Nets or Learning Language Models.
That kinda screws the cost cutting!
Indeed Audril’s “Ghost” makes this mistake.

Instead of Using AI with NVIDIA processors on it for terminal guidance and counter jamming, what both Russia and Ukraine have settled on to solve this problem is old-school fiber optic wire guidance and/or analog television signals which generate visual static when jammed or at range limits, not bugs and artifacts.
Whereas if the Drone CAN get a signal and you’re bouncing the AI off a Mothership or satellite or have radio or are using hard to jam analog signals or use wire guidance… why would you bounce that signal to DataCenters in Palo-Alto or Tel Aviv and endure the lag that comes with the signal travelling that far, the AI processing it, and then sending it back… Which can add up to seconds of lag just from Fiberoptic and electromagnetic - speed of light lag travelling via satellites to Palo-Alto, travelling around the datacenter, and they travelling back to the sattelite and the front, before any delay in the AI’s own processing and thinking. (and no there is no technological way to solve The Speed of Light)…
No remote AI hosted on a Data center in silicon valley is ever going to be able to dip and duck and duke a motorcycle or person ducking, diving, and dodging arround a drone because of this lag… Likewise a remote AI drone is never going to be able to evade and strafe arround rifle or shotgun fire the way that skilled drone operators can (they dodge the speed of their opponents aiming, not the bullets).
This is the reason that currently no missiles, bombs, CIWS systems, etc. use AI terminal guidance or targeting… They use this thing called an “Al Gore Rhythm” or “Com-pu-tor” to cheaply and efficiently make the missile follow a simple pre-programed targeting correction… You know, like people did in the 70s.
And Yes this distinction matters. People are sloppy as hell when they talk about “AI Drones”, and “Drone Swarms” making no distinction between a traditional algorithm… A Basic on device computer feedback system with preprogrammed behaviors, like has existed in Missiles and Autopilots for 50+ years now, on the one hand, and “AI” using Neural Nets, Learning Language Models, and other recent technology that exists as a byproduct of complex AI software enviroments.
There is very little reason to use AI for terminal guidance when you could simply use real flesh and blood drone pilots, maybe with simple interface designating or navigation assistant like 2000s videogames or Garmin GPS devices had, and then a very cheap $50-100 cpu, basically a Graphic Calculator, to attempt something like terminal guidance towards a heat signature or last designated object if it gets jammed… the same way fire and forget missiles work, but with a vastly cheaper optic.
This is basically what I’ve seen out of ““AI Drones””, simple algorithms that target points on a map, or on screen, or lock-on to designated moving objects in-case the user get’s jammed… Little different than an old-school missile.
This is vastly more useful than pinging palo alto datacenters off of satellites, you get all the skills and judgement and political and tactical decision making power of the young high IQ Recruits who become drone operators who can verbally ask questions to the person next to them or over comms, even as they’re piloting the drones. Peak decision making beyond anything AI is capable of, and able to think through 4th and 5th order effects and tertiary side concerns, and communicate… And then designate specific targets that they SEE incase their manual terminal guidance goes down.
Only the most important, or special backup expensive option weapons for use against High Value Targets (Anti-ship drones 100% will have Nvidia Graphics cards strapped to them) could really justify sticking expensive computing hardware on a drone or drone mothership driving up the costs thousands if not tens of thousands of dollars… When, if you NEED the AI (as opposed to a simple targeting algorithm), almost by definition there’s lots of ambiguity and unknowns and targeting choices that have to be made… meaning that a large percentage of the time your expensive AI drone that costs 5-10x a simple computer with an inertial gps, camera, targeting algorithm and thermal/radar sensor… It is going to be hitting nothing anyway… you sent it out not knowing what you were trying to hit, because if you had a good ID on the target you’d use the missile or algorithm drone.
Remember the issue isn’t price because we’re cheap or poor, the issue is that you want as a many as possible so that they can be used aggressively and not become scarce strategically across the theatre, or simply locally just within part of the front…
No matter what weather, jamming, heavy use, a supply mishap, a new tactic, mere over aggression, excessive fire, or enemy lucky hits are going to take out a lot of ANY drone platform.
See US interceptor stockpiles for what happens when you use an Exquisite consumable.
.
Small drones are like a category of ammunition. Poor quality or high failure rates is a minor consideration vs. RUNNING OUT. Actually it’s even worse, since it not only prevents you from suppressing the enemy along one dimension of the modern battlefield, but also blinds one of your primary information gathering tools.
(And yes I know all the companies say their drones return to base and are recoverable, and aren’t lost every time… but that’s true of Quadcopters, and Ukraine atrits them every 3-5 flights with wind, weather, and just accidents… and soldiers rightly don’t want to risk death running into a mine laden no man’s land to recover something with an armed contact warhead)
.
“Artificial Intelligence” GenAI, Neural Networks, Learning Language Models, etc… Really have little plausible application to drones aside from maybe backing up and reprocessing the raw footage for battle tracking, geolocation, and identifying fixed targets operators might have missed but caught on camera, while operators focus on piloting the drones and making strategic choices.… Which we’d expect AI to be much worse at actually identifying targets, and doing twitch maneuvers vs a 20 something’s twitch eye movements, reflexes, and judgement, but significantly better at documenting and processing footage at scale.
.
But even there AI isn’t that useful compared to simply creating a user interface and letting Analysts document and battle track. AI is bad at producing insights and target lists… whereas 14-42 year old autists with sufficient caffeine and an anime body pillow will geolocate literally anything for you.
Ukraine never ran out of autists who wanted to geolocate every little detail. Indeed Ukraine and Russia were so overflowing with amateur autistic white-male battle trackers they couldn’t accommodate, that the rejects formed rival OSINT groups to wage proxy wars over who could screw over each other's OPSEC the most. Everyone knows the telegram channels and youtubers.
Simply at no point has the bottleneck been operators or analysts of intelligence. You give a young white guy an IQ test, then bake him in a simulator for a month or two, then use him.
.
I see a few exceptions.
One is ground based -largely stationary- drones (ground travel devours energy vs. flight) which can sustain much longer and lie in wait and even recharge for a while with solar panels on the top of the vehicle to trickle charge, then fire on targets which appear, locking down sectors of fire, or performing a counter drone roll trying to shoot down flying targets.
Tactically these would benefit from very simple AI target acquisition, and because they’re basically fixed: managing an arc of fire or denying an area, which is straightforward enough it can reliably be delegated… and would need to be delegated because in this low power sustainment mode it’d be processing dozens and dozens of hours of nothing happening… and be buried below a line-of-sight radio signal… It’d be like a fighting position but be vastly less survivable than a person who can still dig into a firing position.
These Ground drones would take on the character of Auto-turrets or for odder variants complex “seeker” landmines (unmanned quadcopters that lie in wait and takeoff if a senor is tripped, then follow a basic search/targeting program) that will be familiar to gamers and fans of sci-fi.
The other exception i really see, which are important but not that common, are things like all out enemy assaults: over-the-top WW1-Style human-wave attacks or massed cavalry charges (motorcycle, tank, air-cav)… Where Drones DO have a massive bottleneck compared to say machine guns or anti-tank guided missiles, or mortar shells or 30mm cannons… Whereas the next round from a machine gun, or Javelin Launcher or mortar tube or autocannon can be fired immediately or at least as fast as they can be loaded, drones take minutes to fly to target and need to be piloted the whole way.
However this is incredibly rare, and AI does not a whole lot for it, since you need those drones ready to fly at the front to deploy them against such assaults anyway, and then have someone detect the assault in the few seconds to minutes it takes to close the distance… This is why Motorcycle Cavalry Charges have emerged as an effective tactic.
Likewise if you develop a “Swarm” system to scramble dozens of drones via AI to respond to movements, then the obvious thing to do is develop ways to bait those AI drones into scrambling and wasting themselves…
Much like Anti-Tank Guided missiles: drones are bulky, highly limited as far as an individual operator is concerned, and again… basically limited to a few hundred/thousand per day across even a truly massive military… Scenarios in which you’d both WANT to surge beyond the capacity of your operators and have the wherewithal to detect such a scenario vs. the advantage of managing and precisely husbanding your scarce bulky weapon is something 99% of immediate commanders are never going to encounter. Stopping an immediate charge is what direct fire weapons and reserves are for… And at the operational and strategic level of long range drones, you have TEAMS and command structures of people who can operate and pilot those (more expensive) higher level drones remotely.
So its not really a concern. AI in drones is a Solution in search of a problem.
“But what about Drone Assaults! Expeditionary forces. Manuever warfare. Shock and Awe!”
This I think is the heart of the AI-drone craze.
Americans and NATO desperately want to fight high impact short weeks-to-months long wars in which their technological superiority and dominance of weapons development and amassed stockpiles allow them to win flash wars with decisive clean headlines and very few casualties that can be buried beneath those headline.
They want to “Storm the Desert” and “be home by Christmas” having won cleanly… And ideally without too much actual ground fighting that produces martyrs and heroes, and without having to hand out lots of Medals of Honor to non-commissioned soldiers and new officers who mustanged from the ranks who will now be competing for leadership positions with the academy officers who failed to produce that quick victory.
When you look at the energy being put into AI drone “swarm” technology, vs. Actually creating new MOS(s) (Military Occupational Specialties) for drone operators and FPV pilots… You can see that the monomaniacal focus on AI is a way for the existing command structure to resist the fundamental change in warfare that’s occurred.
.
Indeed I’m increasingly convinced the primary applications for AI With Regards to Drones are:
Letting Silicon Valley tech companies embezzle funds into “experts” “tech people” and “investors” that should be going to rebuilding a US industrial base of skilled trades people and machinists.
Signing off on target packages that no self-respecting analyst ever would.
Preserving the current sociological and organizational character of western militaries.
The first is I think intuitively explainable… Military contracts are a trillion dollar industry, and big tech hasn’t spent millions on elections to not get a piece of the pie.
The only real use of AI in the Middle East we’ve seen so far… From Gaza to Iran, is that AI will sign off on the war crimes Military Analysts won’t.
While IBM in the 70s assumed that this was an obvious failure mode and reason to never put computers in charge. Generals and Politicians see this and think “AMAZING. We finally has a way to reliably dodge accountability!”
But the thing is I’m not even sure that’s applicable to drones. Sure Analysts who can geolocate and write up a 2 page target package are the real 110 IQ brain-bugs that won’t sign off on double-tapping a Girl’s School, if only out of basic pride, but given the videos I’ve seen the IDF posting of themselves online… I’m pretty sure you COULD find sufficient people to pilot Kamikaze drones into babies and then another into the grandma who comes to help. (gotta double tap)
Even under the most cynical interpretation AI use on drones, where Creating SkyNet is the point because you want to be able to kill babies without your minions questioning you, AI simply does not have an application where flesh and blood people wouldn’t do it better more easily…
Just hire a remote office of Indians to pilot the drones. You don’t need AI. You can just Mechanical Patel it.

Unless the point is diffusion of responsibility that comes from SAYING a robot made the decision. That senior officers want something to drone strike the baby and mother which they can’t be held accountable to hold accountable, because ugh…. I don’t know, the robot just did it!? We don’t understand why, and congress can’t make us say why!
“So stop using the robot”
We can’t do that! The robot is saving lives! (By which they mean lifetimes in meetings and paperwork if one of them had to physically signed off on bombing that girl’s school with an ink pen).
.
Holding Back the Future Fighter
The real development of the war in Ukraine has been the near complete breakdown of the conventional combined armed infantry that Western nations base their ground forces upon.
Conventional Platoon based combined armed infantry has almost entirely disappeared from modern warfare… The Volunteer “Professional” Infantry that western militaries have where 25-30 year old sergeants answer to 23 year old Lieutenants whilst leading 18 year old privates in 3-4 Sections of roughly 8-12 per Sergeant and a platoon of 40ish per Lieutenant is nowhere to be seen in Ukraine except maybe on an org chart buried somewhere at an HQ…
Instead The median age has been creeping up from 35 closer to 45, and instead of the “professional” (teenage low wage college alternative/detour) volunteer force… We’ve seen a bifurcation in how infantry works.
On the one hand there is low motivation Conscripts/Press-ganged soldiers who are often dragged off the street or out of prison and thrown into the meat grinder with maybe a few weeks training, little motivation, and hardly any chance of survival… Told simply to occupy a trench or assault a position if the commander decides its your time to die… Men whose goal is to desert at the first opportunity if they live, and who should really be wearing explosive Battle-Royale collars if you think about it…(presumably the signals environment makes that non-viable).
Then on the other you have ideologically radicalized volunteers, with high internal motivation, who tend either be Radicalized Right into literal Pagan Neo-Nazis or Radicalized Left into some variant of Trans-adjacent pervert or Radicalized Slav into some bizzare Ortho-maximalists (for the most part in non-overlapping units)…
These highly motivated fighters, despite often being self funded (self-purchased weapons, drones, bikes, kit, etc. are common) employ tactics and carry out missions that a mere decade ago would be the exclusive domain of special forces…
Inserting 10s of kilometers to the front in teams of 2-4 making intensively high risks decisions about how long to stay, where to move, how to avoid traps and mines, how to get places, what to do/pursue… Employing drones for intelligence gathering, plotting assaults, etc. And then extracting either as stealthily as they inserted, or via often incredibly high risk methods if something goes wrong.
They might be out there a WEEK in a state of medical emergency if something goes wrong or they get hit that the “NO MANS LAND” is that hard to do a medical extraction… And often you find stories of them dragging themselves or comrades back kilometers through the dark wearing tourniquets or self packed wounds…

Sure little STOPS the conscripts from doing this… Aside from blocker units… But the skill ceiling, will difference, and the wild disparity between individual teams and levels of ability are so extreme, and the role of individual intelligence, skill and judgement so vital… that the bifurcation enforces itself almost immediately.
.
Modern ground warfare is now one of the most lonely and high stakes extended tests of intelligence, judgement, grit, and will…
Put otherwise literally everything your Sergeant and western militaries have been saying for decades that war is a team sport, where the unit is more than the sum of its members, everyone pulls together… and its not like the videogames… Is now basically a lie.
Modern Ground Warfare is FAR closer to the things Depicted in video games from the 2000s and 2010s than it is to anything in the infantry manuals… Even the team sizes, 2-4 resembles a Console game or Tactical Shooter or RPGs far more than the platoon movements of Vietnam or Iraq…
Real life war resembles a stealth tactical shooter, but with the pacing, tension, and ability to screw yourself of a Submarine simulator (Just like if your sub gets spotted, you might be stuck below for hours waiting for the depth charge that crushes you, likewise you can get pinned down for hours waiting through waves of drones hitting the basement or dugout you’re hiding in.)
This is not just corrosive to the doctrine of western militaries, but their values and command structures…
This is a mode of warfare that overwhelmingly favors high-IQ White Men and the sorts of intelligent masculine brained Asians, Middle-easterners, Central Asians, and Upper-class South American Whites and Castizos who play video games with them… Acting in incredibly small teams with incredible amounts of individual discretion and high stakes judgement…
I cannot express enough that this is absolutely incompatible with the sort of multicultural captain planet teams of Teenagers of various immigration status that the Marines and Armies of the west currently recruit…
Right now the US military is largely structured as a jobs program for Immigrants and Urban Youths from poor backgrounds whom high-human capital whites, Asians and other recruits have to endure and drag along with them as a part of “Teamwork”.
The Entire modern platoon structure basically exists to allow various levels of academic knowledge (junior officers) and experience (NCOs) to pair up their good recruits with their bad and come out with a unit that, hopefully, performs better than its widely variable individual members… With enough technical-enough people to coordinate the various types of support fire that are America’s real secret sauce.
.
The modern battlefield has utterly destroyed this formula. FPV Drones, Dropper Drones, Spotter Drones, Kamikaze drones… All of them have wildly rendered the American way of war and more importantly the American mode of personnel organization whereby you pretend that the 5-10 highest IQ/most committed members of the platoon of 40 aren’t carrying the team and overwhelmingly bearing the costs (without matching rewards… DEI in all but name determines promotions)…
That does not work when the largest concentrated team you can have is under 8 people, and all those teams need to act with individual discretion, and your screwups you need others to handhold now need to insert 20kms on foot in the dark through minefields, wire, drone patrols, and overlapping enemy fire, to a fighting position, and then survive there a week, and then slip out again… Entirely under their own discretion… That’s the kind of thing that reveals intelligence and innate character differences really fast…
Likewise FPV Drones and Motorcycles, staples of Ukraine for years now are simply not being taught to recruits… because they’re dangerous, require discretion, individual maintenance… And are incredibly intelligence and skill loaded…
The difference between what a 85-90 IQ person can teach themselves in an FPV simulator or on a bike vs. a 115-125 IQ person is miles different… And like most individual skills, it is inevitably largely self taught, you can’t get handheld through 200 hours in the simulator or on the bike.
All of modern warfare now follows the logic of virtuoso skill development or gamer skill development… And that does not line up with how Western Militaries have structured themselves around multiculturalism.
.
This is why, despite no “AI Drone Swarm” having ever existed or been employed on any battlefield anywhere… Western Militaries are pouring BILLIONS into trying to develop this combination of technologies that has only ever existed in AI video mockups and pre-scripted Chinese Formation flying (where no decisions have to be made) and which all the actual limits of all the technologies say shouldn’t and can’t really exist…
.
Because they’re desperately trying to solve their way out of The Gamer Revolution of modern warfare.
The US and NATO desperately NEED the future of warfare to be centralized automated “AI DRONE SWARMS” because they desperately don’t want it to be decentralized highly-skilled FPV Drone Pilots.
.
The 3 major technologies of the modern battlefield right now, the three devices that dominate… Are FPV Drones, Motorcycles, and Modern Rifles/Optics… Deployed at the level of 2-4 men teams…
All three of these have nearly limitless skill ceilings whereby lone individuals can train endlessly and achieve almost cartoonishly magical levels of ability… And when combined with fieldcraft on the battlefield they create a strategic game that favors a tactical strategic thinking that follows that logic as well…
The raw value of individual skill, intelligence, and judgement in modern war is so bottomlessly powerful and defining of modern conflicts, just as they are bottomlessly corrosive to the values and structure of the modern egalitarian bureaucratic warfare/welfare state that they NEED some magic AI wunderwaffen to save them from the battlefield that actually exists…
But that will be another article.
Tip:
BITCOIN: bc1qdhj7637sgcssxgxygjaa3ddljwy8tzg5mzw325

MONERO: 8AhA3g9hbtDcAJE5MPmeQsFwwGsf3H9fq9tC6giQ4a6vKnTXv4J4MivKXrPKDpXyEeNc9mfFejbq84kSWkC8pjuj18rAEij

Follow me on Twitter: @FromKulak
















Great take as always, Kulak. I eagerly await the day we stand together before Valhalla.
Note you don't necessarily need graphics cards to run simple nn guidance models, which are often chosen as approximations of more complex heuristic models because of their low computation requirements.