A war run out of group chats
An artillery crew in eastern Ukraine receives a fire mission the way most of us receive weekend plans. Someone posts a target in a chat thread. The gunner pulls up a mapping app that volunteers built and gave away for free, checks the coordinates and fires. The whole loop runs on commercial phones and tablets that soldiers bought themselves or crowdfunded. CSIS documented this in late 2024, and the detail that stayed with me was not the improvisation. It was the reach. Kropyva, the volunteer artillery app, spread from unit to unit until it was closer to standard issue than most official programs ever get.
I keep coming back to that report because of what it says about demand. Soldiers do not write software for fun. They write it when the thing they need does not exist and nobody sells it. A volunteer app outcompeting an official program is a market signal in a uniform. And the need sits below headquarters, down at the level of the team and the battery, where the mission gets planned in the morning and argued over at night.
Platforms for the top, pipes for the bottom
Look at where the serious money in defense software is going. Anduril's Lattice and the Maven line of work are turning into the operating pictures of commands. They fuse feeds from hundreds of sensors and put a coherent battle on a screen in front of a general. I want to be clear that this is good software. The enterprise C2 push is well funded and well built, and headquarters needed it badly.
At the other end of the stack sits the plumbing. TAK gives everyone a shared map and a common language for position data. Radios and mesh networks move bits through terrain that hates bits. This layer is unglamorous and it mostly works, which is the highest praise infrastructure gets.
Between the two is where a mission actually lives. A team plans a patrol, runs it, then sits down afterward to figure out what happened and what to change. Walk through that middle today and you find spreadsheets, chat threads and a laptop someone configured by hand. The Pentagon knows the way it buys software is broken. The March 2025 software acquisition memo and the November 2025 acquisition transformation strategy both push toward buying software the way software is actually made. But that reform energy aims at how to buy. The catalog it buys from still has an empty shelf in the middle.
What would have to live in the middle
I do not have a good name for this layer yet. Call it the mission layer for now, the software a small team runs for itself while it plans, executes and reviews its own mission. Naming it matters less than sketching it, so here is my rough sketch.
It would run on hardware the team already owns, the laptops and tablets and small servers a unit can buy this month rather than the terminal a program delivers in four years. It would treat disconnection as the normal state of the world. Sync when a link appears, keep working when it does not. Ukraine's digital tools grew up under constant jamming, and Chatham House argues that this pressure is exactly what made them work. Any AI in this layer would have to live on the device for the same reason. A model that phones home is a model that dies with the link.
Mission data would need an ending. A team operating near the enemy cannot drag an ever-growing archive around, and a captured device should hold less the closer it gets to the fight. So the software would need a way to close a mission out, keep what the team decides to keep and destroy the rest on purpose. And it would answer to the team. The enterprise can get its reports, but the team owns the plan, the record and the machine it all sits on. That inversion feels strange to write down. I suspect it is the whole point.
I keep looking for it
Every few months I convince myself that someone must have built this already. The pattern is too obvious and the demand signal from Ukraine is too loud. So I go looking and I find workarounds. Enterprise platforms that assume a cloud and a headquarters. TAK plugins that solve one slice. Soldiers gluing the rest together with group chats and shared drives, the same way the Ukrainians did, only with less urgency behind it. The product is missing.
Maybe I am wrong about the shape of it. This post is me thinking out loud more than declaring anything, and the sketch above will probably look naive in a year. But the gap itself feels real. Two well-built layers, a hollow middle and soldiers filling the hollow themselves. I am going to keep pulling on this thread in the next few posts, starting with what it would cost to build and who would actually buy it. If someone has already built the thing, I would genuinely like to see it.