At a Glance
Problem: Modern militaries see the battlefield in extraordinary detail and still need days to turn that picture into a plan. Against an adversary that decides in minutes, that lag loses the fight.
Solution: Smack Technologies is the first frontier AI lab for national security building domain-specific reinforcement-learning models grounded in both the physics of war and expert human priors. Their two products, Omega at the command level and Alpha at the tactical edge, compress the Orient and Decide stages of the decision loop from days to minutes and update plans in seconds when conditions change.
Validation: Smack already holds contracts with multiple branches of the U.S. military, including those awarded by the Joint Fires Network and the Marine Corps Warfighting Lab, and was pulled into accelerated Marine Corps production discussions as the Pentagon moved to diversify the AI vendors it relies on.
Why it matters: The administration’s 2027 budget requests $58.5 billion for AI and connected command and control, and as the Pentagon reorganizes around an “AI-first” doctrine, the reasoning layer that turns sensor data into decisions is where the defense AI stack concentrates its value.
The defining bottleneck of modern warfare is decision, not detection. Modern militaries field thousands of sensors, drones, and autonomous platforms, and the volume of what they see has outrun their ability to act on it. The decisive edge in the observe-orient-decide-act (OODA) loop sits in the middle. Whoever compresses Orient and Decide fastest holds Decision Dominance, and that compression is the organizing problem of American military modernization. Against a peer adversary in the Indo-Pacific or across simultaneous theaters, the side that decides first dictates the fight. The slower force can lose before it fires a shot.
The Pentagon has made this its top priority. In January 2026, Secretary of Defense Pete Hegseth ordered the department to become an “AI-first” warfighting institution, followed days later by an AI Acceleration Strategy. One of the priorities is to push AI into battlefield decision support, including the conversion of data into action as rapidly as possible.[1] The stated goal is to compress the time between observation and action across every service before the end of the decade.
Commercial AI models cannot solve this on their own. The large language models that dominate public attention are trained on the open internet and optimized for fluent text, not the physics of a contested battlespace, the structure of a campaign, or evolving military doctrine. Smack’s founders argue that commercial models can meaningfully serve less than 20% of military decision-making. They are too compute-heavy to be used at the edge, they were not built to reason over geospatial terrain, and they cannot repair a plan in real time as the battlefield shifts. There is also no training dataset for WWIII, because the scenarios that matter most have never been fought.
Smack is the first frontier AI lab built specifically for national security.
Rather than adapt a general model to defense, it builds domain-specific models powered by deep reinforcement learning, trained on both physics-based battlefield simulations in which thousands of AI agents learn to plan and adapt under realistic constraints, and the human priors of military domain experts.
Its two products work in concert. Omega, the command-level stack, fuses multimodal sensor data and generates multi-domain plans in minutes, then updates them in seconds as the situation changes. Alpha, the edge-level stack, runs locally at the front lines, so decisions keep flowing even when connectivity to the operations center is severed.[2]
“The future of war…will be more decentralized than any conflict that we’ve ever fought”
-Andrew Markoff, Co-Founder and CEO, Smack Technologies
Smack’s founders have lived the problem. Andrew Markoff and Clint Alanis are Marine Corps Special Operations veterans with more than 25 years of combined combat experience who learned firsthand that two decades of counterinsurgency left the United States underprepared for a high-intensity fight against a peer adversary.[3] They founded the company in 2024, pairing a research bench capable of building frontier models with their own operational ground truth, including experience coordinating and synchronizing effects in support of American strategic objectives: the problem Smack is built to solve.
Smack rides one of the fastest-growing lines in the defense budget. The Pentagon’s fiscal 2027 request includes $58.5 billion for AI and Combined Joint All-Domain Command and Control (CJADC2), the effort to connect every sensor and shooter into one command-and-control system.[4] As command and control moves to the budget’s center of gravity, commercial models are commoditizing the integration layer. The durable value is migrating to the specialized models that reason on top of it.
In March 2026, the Pentagon designated Anthropic a supply-chain risk and moved to wind down its use of Claude over the company’s limits on autonomous-weapons use, and within days began testing rival commercial models to replace it.[5] The lesson reinforced our thesis: even the best general-purpose models are substitutable at the integration layer, and the department will not depend on any one of them. It also opened the field to specialized vendors. Smack was pulled into multiple Marine Corps meetings on how fast it could reach production for combat use in 2026, an acceleration of more than a year.[6]
Smack’s early contracts ground the thesis. The company has already secured work across multiple armed-services branches, including the Joint Fires Network, the cross-service program for coordinating sensors and shooters, and the Marine Corps Warfighting Lab, which tests the concepts the Corps fields next.[7] These are footholds in the programs that will buy Decision Dominance at scale.
We invested in Smack because the reasoning layer is the most valuable position in defense AI and the hardest one to occupy, and because the people building it have lived the kill chain they are modeling. We are proud to support the team.
[1] W. Hartung, The Pentagon Is Going “AI First”, The Nation, 7 April 2026; S. Freedberg, Pentagon Rolls Out Major Reforms of R&D, AI, Breaking Defense, 13 January 2026.
[2] Smack Technologies News & Insights, Alpha and Omega: Smack’s Core Product Suite That’s Revolutionizing Modern Warfare, 21 August 2025.
[3] Smack Technologies press release, Smack Technologies Announces $32M in Funding to Build First Frontier AI Lab for National Security, Business Wire, 2 March 2026.
[4] D. Sennott and M. Lehrer, Understanding the President’s FY 2027 Budget Request for the Department of War, Greenberg Traurig, 11 May 2026.
[5] K. Manson, Pentagon Tests Rival AI Models in Race to Replace Anthropic, Bloomberg, 21 May 2026.
[6] M. Stone, Pentagon’s Ouster of Anthropic Opens Doors for Small AI Rivals, Defense News, 9 April 2026.
[7] Smack Technologies press release, Smack Technologies Announces $32M in Funding to Build First Frontier AI Lab for National Security, Business Wire, 2 March 2026.