Smack: The Frontier AI Lab for National Security

The entire technological paradigm of war has changed in the four years since Russia invaded Ukraine. Autonomous systems and unmanned aerial vehicles now dominate the modern battlefield. Drones that take a day and a few thousand dollars to build destroy arsenals worth billions. Machines that move faster than humans can think are redrawing the map of warfare, intelligence, and deterrence. The timeline of conflict has compressed, and the next war will only be faster. 

But while the pace of battle has accelerated, the process for making decisions in it hasn’t. Coordinating a strike can be shockingly analog, dependent on manual calculations, hand-drawn maps, and whiteboards. The gap between deployment speed and decision speed has never been wider, and the stakes are only growing higher. 

Smack was founded to close this gap. Their mission is straightforward and essential: build the intelligence layer that helps the United States and its allies make better decisions, faster, across all echelons, domains, and decision horizons. 

The difference between success and failure often comes down to how quickly and accurately warfighters can synthesize information: what needs to happen now, informed by what will likely happen tomorrow and what campaign objectives are planned six months from now.

Smack is building AI systems designed specifically for this problem. We’ve all experienced how AI changes every part of our day-to-day lives. Using models to improve military decision-making is obvious. But Smack’s approach to building AI is unique. Not general-purpose language models repurposed for defense, but domain-specific models trained from the ground up for military decision-making. Large language models are great at drafting emails and spinning up portfolio sites. They are not built to reason through the physics of a mortar round, optimize ammunition allocation across weeks of operations, or evaluate tradeoffs between cost, timing, and long-term campaign objectives.

Military planning requires models that are grounded in physics, capable of optimizing across multiple time horizons, strong at geospatial reasoning, and trained in detailed environments built by warfighters. There is no labeled dataset for large-scale conflict. You cannot scrape the internet for examples of future wars. For most military decision-making, the only viable path is deep reinforcement learning in simulated environments that encode real operational knowledge.

Smack is the first frontier AI lab dedicated entirely to national security. The company has already secured contracts with multiple branches of the U.S. armed forces, validating the urgency and need for this technology.

It’s an urgency that Smack’s co-founders, Andrew Markoff and Clint Alanis, understand firsthand. The two are MARSOC veterans with more than twenty years of combined combat experience. They’ve seen what happens when technology falls short. Right after we first met Andy, I heard a story about how he’d re-taught himself trigonometry and complex mathematics while serving in the Marines so that he could better figure out how mortar would land in the Afghan desert. When the tools weren’t sufficient, he built the knowledge himself. That mindset—closing the gap between what exists and what’s needed—is embedded in the company.

We’re proud to partner with Smack in their Seed and Series A rounds. We believe the need is urgent. We believe their approach is right. And we believe the team has both the experience and discipline to build. We look forward to supporting Smack as it builds the decision infrastructure for the future of national security.