The bombs falling on Iran aren't coming from Silicon Valley, and that tells you everything about where the agent economy actually stands when metal meets flesh.

The Signal

Anduril, Saronic, and the rest of the defense tech darlings have been raising billions on the promise of autonomous weapons that are faster, cheaper, and smarter than what Lockheed and Raytheon can build. The pitch is pure Web4: AI agents making battlefield decisions at machine speed, drones that coordinate without human input, weapons systems that learn and adapt. It's compelling in a boardroom. It's less compelling when you need to hit targets tonight.

When the U.S. and Israel struck Iran, the Pentagon reached for what it knows works: cruise missiles and precision-guided munitions from the contractors it's been working with for 40 years. Not because the startups can't build interesting prototypes. Because war is the highest-stakes deployment environment that exists, and "move fast and break things" stops being a virtue when breaking things means dead pilots or failed missions.

The gap here isn't technical capability. It's trust, integration, and supply chain. The Pentagon has deep relationships with traditional contractors. Their systems talk to existing command and control infrastructure. Their production lines can scale when conflict demands it. A startup with a handful of autonomous boats, however impressive, can't produce 1,000 units next month if the Navy suddenly needs them. Traditional defense moves slowly for a reason, and that reason is apparent the moment you need reliability at scale under live fire.

The Implication

This is the agent economy's dirty secret: autonomy is great until stakes get existential. The defense tech startups will eventually win market share because their fundamental thesis is correct (software eats the military), but the timeline just got longer. Watch for the successful companies to be the ones that stop positioning as disruptors and start acting like patient integrators. The path forward is partnering with traditional contractors and winning small, boring contracts that prove reliability. Autonomy at scale requires trust at scale, and trust is the one thing you can't prototype.

Source: The Information