The U.S. military hit 1,000 targets in Iran in 24 hours using Claude and Palantir's Maven, and the real story isn't the AI, it's the decades of infrastructure that made it possible to use it.

The Signal

Claude, Anthropic's chatbot that most people use to write emails, just helped coordinate a major military operation. Combined with Palantir's Maven targeting system, it processed real-time intelligence for strikes across Iran and Venezuela. But here's what matters: this wasn't AI suddenly becoming combat-ready. This was decades of military investment in data infrastructure, personnel training, and system integration finally reaching a point where an LLM could slot in and be useful.

The distinction the source draws is critical. There are automated weapons (the sci-fi stuff, drones that pick targets) and decision support systems (software that helps humans make better calls faster). Claude is the latter. It's not deciding who dies. It's helping intelligence officers process more information than humanly possible and surface patterns that matter. Maven aggregates sensor data, Claude helps make sense of it, humans pull triggers.

This is the agent economy showing up in the place with the highest stakes and the most money. The military has been building the rails for AI agents since before anyone called them agents. Sensor networks, satellite feeds, classification systems, cross-platform data standards. They weren't waiting for transformers to be invented. They were ready when transformers arrived. That's the pattern to watch: AI doesn't create capability from nothing. It amplifies infrastructure that's already there.

The Implication

If you're building in the agent economy, the lesson is infrastructure first, models second. The organizations winning with AI aren't the ones with the fanciest models. They're the ones who spent years getting their data clean, their systems interoperable, their people trained. The military proved you can swap in a commercial LLM if everything else is ready. Your competitive moat isn't which model you use. It's whether you've built the rails for any model to run on.

Source: Fast Company Tech