A frontier AI lab for national security, built by people who lived the lag it's trying to erase.
In an office park in El Segundo, twenty-six people are building artificial intelligence for the worst day imaginable. Smack Technologies is a defense AI lab. Not a chatbot company, not an analytics dashboard with a flag on it. The pitch is narrower and stranger than that: the next major conflict will be decided less by who has more steel and more by who decides faster. Smack exists to make sure that side is ours.
The company calls the goal "Decision Dominance." Stripped of the capital letters, it means something concrete - taking the messy flood of sensor data that pours into a command post and turning it into a decision a commander can actually act on, in seconds instead of staff meetings. Two products carry the work. Omega thinks at the command level. Alpha thinks at the edge, out where the radio cuts out. Between them sits a thesis about modern war that is either obvious or audacious, depending on how much time you've spent inside one.
Military doctrine has a famous loop: Observe, Orient, Decide, Act. The OODA loop. Most technology of the last twenty years poured money into Observe - more satellites, more drones, more sensors feeding more pixels into more screens. The result is a strange kind of abundance. Commanders today drown in observation. What they lack is time to orient and decide.
That middle stretch - turning data into a plan - still runs at the speed of human staff work. Briefings. Whiteboards. People in a tent comparing options while the situation moves. The founders watched this firsthand, which is the uncomfortable detail that makes Smack hard to dismiss. They weren't reading about the lag in a white paper. They were living inside it, downrange, waiting for a decision while the window closed.
The skeptic's question writes itself: hasn't every defense contractor promised to fix exactly this? Fair. The difference Smack bets on is where it points the AI - not at watching more, but at the part nobody could speed up.
Decades of spending bought better eyes. The gap moved to the brain - the orient-and-decide stretch of the loop, still bound to the pace of human staff work.
Months → Minutes. That's the entire pitch, compressed.
Above: the unglamorous truth of modern conflict - the slowest part of the loop is the part you can't photograph.
Andy Markoff and Clint Alanis met the problem the hard way. Both came out of Marine Special Operations Command - the Raiders - with more than two decades of combined combat experience between them. Markoff later did a stint in operations at Palantir, which is roughly where defense software people go to learn what software can and can't do for a warfighter. In 2024 they founded Smack.
Their bet was contrarian in a useful way. Instead of building yet another platform to display data, they would build domain-specific AI models trained to reason about combat itself. And rather than wait for real-world data - scarce, classified, and grim to collect - they would generate it. Smack built a proprietary, physics-based simulation environment that spins up millions of synthetic engagements, then trained models inside it using deep reinforcement learning, including multi-agent algorithms like MADDPG.
Former MARSOC officer. Did operations at Palantir before starting Smack. The one who says the quiet part - 2027 - out loud.
20-plus years in Marine special operations. Brings the part of the product spec you can only get by having been there.
The founders, in short: people who waited for a decision under fire and decided to go fix the waiting.
Smack ships two stacks because war happens at two speeds at once. There is the campaign - the months and days of planning - and there is the right now. The company's models are built to plan across all three horizons simultaneously: one to twelve months, one to four days, and the immediate moment.
The strategist. Omega fuses multimodal data and converts a commander's intent into an executable plan in minutes, using multi-agent reinforcement learning and proprietary knowledge graphs. Application suites like Fires and Logistics sit on top of the core models.
The operator. Lightweight models plus proprietary hardware that keep working at the tactical edge - in degraded, low-bandwidth conditions where the connection to the cloud is the first casualty. Built for autonomous systems and platform decision support.
Under the hood it's less science fiction and more careful engineering: graph networks, diffusion models, and transformers stitched together for tactical inference, with commercially available language models handling the human-machine interface. The clever part isn't any single component. It's the synthetic battlefield they're all trained inside.
Two products, one idea: the cloud is a luxury, and luxuries are the first thing a war takes away.
Talk is cheap in defense tech, where demos outnumber deployments. So here is what's verifiable. Two of the harder customers in the building - the Joint Fires Network and the Marine Corps Warfighting Lab - are paying Smack seven-figure sums. The cap table reads like a defense-and-frontier-tech who's who: Geodesic Capital and Costanoa Ventures leading, alongside Point72 Ventures, Felicis, First In, Scribble Ventures, Bloomberg Beta, Washington Harbour Partners and more.
Smack's central claim is about collapsing the orient-and-decide phase of the loop. Illustrative, not to scale - the point is the order of magnitude, not the exact bar.
Source: Smack Technologies' own framing of the OODA decision timeline. Treat as directional.
A chart that's really just one sentence wearing a suit: the slow part used to be the brain, and Smack wants to fix the brain.
Here's the part where a defense company usually waves its hands. Smack's mission is more specific than most: deliver Decision Dominance to the U.S. Department of Defense, its allies, and partners. The deeper logic is almost pacifist in shape, even if it arrives wearing camouflage. If your side can decide faster than an adversary can, the adversary's best move may be not to start. Speed becomes deterrence.
It's a big claim, and the honest reader should hold it at arm's length. Whether software can actually compress a kill chain the way a slide deck promises is something only deployment proves. But the framing is coherent, and the customers signing checks seem to find it credible enough to fund.
"Delivering Decision Dominance to the U.S. Department of Defense, its allies, and partners."
The premise: in a contest of equals, the side that orients and decides faster sets the tempo. Tempo wins.
The mission, de-jargoned: be fast enough that the other side reconsiders. Then never have to prove it.
Return to that office park. Twenty-six people, a simulator humming through its millionth synthetic battle, two product names on the whiteboard. Two years ago this was an idea two veterans had about a problem they couldn't stop thinking about. Today it's $32 million, a handful of government contracts, and a roadmap to scale.
The tension that runs through everything Smack does hasn't been resolved - it's been picked up. Modern militaries can see almost everything and still struggle to decide in time. Whether Omega and Alpha actually close that gap at the scale of a real conflict is the open question, and a serious one. Smack is honest enough about the stakes to put a date on it. The rest of us get to watch whether the bet pays off before we find out the hard way why it mattered.
What the company offers, in the end, is not a weapon. It's time - the one resource a commander can never manufacture and never get back. If Smack is right, the most important thing it builds won't be a better way to fight. It'll be a reason not to.