Tech bros optimized war… and it’s working

0:00 Yesterday, the US Department of War announced it's going allin on a new primary operating system for the battlefield. A tool that's proven to be so effective at blowing people up that it's rolling out to every branch of the military. The Army, Navy, Marines, Air Force, and even the Space Force are going to be powered by the Maven Smart System, an AI platform that shortens the kill chain for kinetic operations. You heard that right. The same AI models that can't spell strawberry are now being used to turn people into a fine mist faster than ever. That might sound inhumane and terrifying, but don't worry, there is still a human in the loop who needs to click the accept all cookies button before the missiles can be launched. The Peter Teal's Palanteer is the main company behind Maven, but all the big hyperscalers and AI labs are cashing in on the taxpayer funded US war

0:45 machine. In today's video, we'll take a look at the future of war and find out how prompt and destroy slop ops actually work at a low level. It is March 24th, 2026, and you're watching the Code Report. In Modern Warfare, you can't just carpet bomb an entire city. Instead, you need to first identify a target, then verify it, then verify it again. Otherwise, you might end up killing a school full of innocent children. >> Very bad intelligence. >> I'm sorry. >> And that's where the Maven Smart System comes in. It's an AI platform that uses computer vision and sensor fusion to automatically analyze surveillance data like drone footage, then identify, track, and prioritize targets. There's still a human to push the kill button today, but eventually this process could become entirely autonomous. Now, before we look at the technical details behind the system, we first need to meet the

1:31 tech bros who created it. The core platform is provided by Palunteer. The current CEO is Alex Karp, and it provides the operating system that glues everything together. We've got AWS and Azure helping out with cloud infrastructure. And Google used to be involved, too, but they had to back out after their hippie employees started protesting. But the Maven system needs tons of real world data and it gets much of that data from Palmer Ly's Anderil which provides terrifying kill machines like the ghost drone, Amble Interceptor and Ghost Shark underwater drone. And then finally the system runs on multiple large language models until recently anthropics Claude was their champion. But then their soyo Daario started crying when he found out that his tech might be used to harm humans. War Chad Pete Hegsth, who can bench press 315 pounds, by the way, drank Daario's tears

2:17 and banned Anthropic as a national security threat from all government contracts. Luckily, Sam Alman was happy to step in for sloppy seconds. And the web of people here goes way deeper, but as a developer, I'm more interested in how Project Maven actually works under the hood. The exact tech stack is classified, but we have enough public data and leaks to piece together a similar system with open- source software. At the first layer, we need to ingest tons of data in different formats like video streams from our drones, ecoms from our special ops teams, GPS from our satellites, and so on. And to do that, we're going to use a tool like Apache Kofka. But basically, Kafka allows us to stream multiple data sources in one place, allowing this entire complex system to stay updated in real time. And now that we have incoming events, we can use a tool like Apache Spark to subscribe to a Kafka topic. So,

3:03 we can start transforming that data into something useful. Like, we might send drone footage to OpenCV to segment it and detect actual objects in those images. But now, here's where things get really interesting. In order for AI to blow people up, it needs to understand the relationships between all the different resources in our system. At Palunteer, their secret sauce is called the ontology, and the government is paying them billions of dollars a year to use it. But what is it? Well, basically it maps messy fragmented data from different sources into a shared structure while capturing the metadata and relationships between these objects. You can think of it like a digital clone of an entire organization, which might be a manufacturing plant, a hospital, or in this case, the military. At this point, we have data, but we don't understand the relationships between our data points. Ironically, we won't use a

3:49 relational database here, but instead a graph database like Neo4j, where people, vehicles, and bombs become nodes, and their movements turn into edges. And now our entire battlefield is mapped in a way that replicates the real world, where it can be queried and visualized by humans and AI. And now that our world's recreated, we need to set some ground rules before we start taking action. A tool like Open Policy Agent could help us do that by enforcing policies across the entire stack. It looks good to me. Now we can start dropping in AI agents with the model context protocol. From here you can grab your favorite open Chinese model like Kimmy or Quen. Then use Heretic to uncensor it. And now it can start performing actions based on this data. But from there you just need to wire up some tomahawk missiles and you'll be blowing people up on pure vibes in no time. And what's really crazy is that you don't need a trillion dollar defense

4:35 budget to build this thing thanks to tools like Tracer, the sponsor of today's video. It's a spec driven development tool that lets your whole team work together with agents to build real world software. Just start by telling it what you want to build, and Tracer's epic mode will ask you follow-up questions to create a series of specs and tickets that map out to your requirements. But from there, you can invite your teammates to the project and live edit the specs together. It then assign people to specific tickets. A tracer then passes all of that context to your favorite coding agent and validates the output along the way to make sure it actually meets your requirements without drifting. So if your team wants to use agents to ship software that actually works in production, then try Tracer for free at the link below. This has been the code report. Thanks for watching and I will see you in the next one.