A truck parks near an airfield
On June 1, 2025, wooden cabins on the backs of ordinary trucks opened their roofs at points scattered across Russia. Small quadcopters lifted out and flew short hops to nearby air bases. Some of the aircraft they struck were strategic bombers that Russia no longer builds and cannot replace. Ukraine had smuggled the whole operation across the border over eighteen months. The drones themselves were the cheap part.
Analysts at CSIS put the cost of an FPV drone of that class at roughly $1,000. A Tu-160 bomber runs around $250 million. That is a ratio of 250,000 to 1, and it is the number this post is about. The raid was clever. But what matters more is that it was ordinary. It used commodity airframes, commodity radios, and planning done by people who did not have a defense budget. The expensive thing on that airfield was the target.
Operation Spiderweb was reported everywhere for about a week. Then the attention moved on. I do not think the implications have moved anywhere at all. They are sitting in plain view, mostly unprocessed.
The inversion is not an anecdote
The same ratio shows up wherever cheap systems meet expensive ones. In the Red Sea, warships have spent nearly two years firing interceptors at Houthi drones and missiles. The interceptors routinely cost one hundred to two hundred times what they destroy. The defense works. The economics do not. Every engagement is a small victory and a small bankruptcy at once.
Expensive precision has its own problem. The Washington Post reported in May 2024 that Russian jamming was degrading U.S.-supplied guided munitions in Ukraine, including Excalibur shells and guided rockets whose accuracy collapsed when their guidance links were denied. A weapon that costs six figures and depends on a signal is a weapon that costs six figures until someone turns on a jammer. The cheap FPV drone with a human or an onboard brain flying it kept working.
So the pattern by late 2025 is clear. Destruction got cheap while defense got expensive, and anything that operates without a fragile link got more valuable fast. Militaries have started to absorb this on the hardware side. Attritable drones are now a funded category in most Western programs. But almost nobody has followed the logic into software, and that is where I think the bigger inversion is hiding.
The most fielded computer in history
The phone in an operator's pocket carries more compute than a data center rack did in 2015. It has a GPU, a neural accelerator, radios, cameras, and precise positioning. It is ruggedized by the consumer market at a scale no defense program could buy. And it is already fielded. Every soldier in every Western military walks around with one. No procurement cycle put it there.
Ukraine noticed. CSIS documented in November 2024 how volunteer-built applications spread through the Ukrainian military from the bottom up. A single artillery app built outside any official program reached tens of thousands of users because it made gun crews faster and it ran on hardware they already owned. Chatham House made the broader point in March 2025 that Ukraine's real innovation is not any one device. It is the speed at which software moves from an idea to the front line, measured in weeks.
Even the Pentagon has conceded the direction. In March 2025 the Secretary of Defense signed a memo ordering the department to buy software the way software is actually built, with rapid pathways and commercial solutions as the default. The policy now points the right way. The software it is pointing at mostly does not exist. The device is treated as a browser for systems that live somewhere else, on a server, behind a link, exactly the architecture the jamming reporting tells us to distrust.
The question this blog will keep working
Put the two halves together. A $1,000 airframe can now destroy a $250 million aircraft, so mass and cheapness win the hardware argument. And the cheapest capable computer ever fielded is already in every pocket, running software that treats it as a phone. The platform arrived before its mission software did. That has happened before in computing, and the gap never lasts. Someone always writes the software eventually.
I do not know yet what that software looks like. I have suspicions. It probably assumes the network is down rather than up. It probably runs its intelligence on the device instead of calling home for it. It probably gets fielded the way the Ukrainian artillery app did, by being useful, not by being mandated. But those are guesses, and guesses are what a first post is for.
So here is the question I intend to keep pulling on. What would mission software look like if it were built, from the first line of code, for the hardware operators already carry? The next war will be fought with computers that are already fielded. Somebody is going to write the software for them.