Smack is the first Frontier AI Lab built for national security. But what does that mean specifically, and more importantly, how is it different from other AI tech companies in the defense space?  

In this article, we’ll dive into exactly what Smack is, what it does, and how it accomplishes its mission of delivering Decision Dominance to the U.S. DoW, its allies, and partners. 

What Is Smack Technologies?

Smack Technologies is a defense tech company founded in 2024 to build the first domain-specific AI models tailored to the needs of the U.S. national security apparatus. 

The company was co-founded by two Marine Special Operations Commanders: Andy Markoff and Clint Alanis. It’s headquartered in Texas, with offices in Washington, D.C. and El Segundo, California. Smack’s team is made up of scientists, engineers, and warfighters leveraging decades of expertise in their respective fields to solve one of the hardest and most important challenges facing national security: using deep reinforcement learning to overcome time and resource constraints, optimize decision-making, and enable intelligent autonomy. 

What Does Smack Do? (Short Version)

In its simplest form, Smack compresses the time it takes military commanders to make informed, intelligent, actionable decisions from months to minutes. It then enables warfighters and autonomous systems on the front lines to act on those decisions in a coordinated manner in real time, with or without comms access. 

What Does Smack Do? (Less-Short Version)

Military decision-making is currently stymied by an overabundance of data and adversary-led misinformation campaigns, where every decision must be made under severe time pressure and finite operational resources. Planning cells are siloed. Planning tools and processes are outdated. Human decision makers are overwhelmed.

What we’re left with are catastrophic vulnerabilities to campaign and mission planning and Rip It-fueled heuristics (or “Math for Marines”) at the tactical edge. 

In military parlance, the model used for decision-making is known as the OODA Loop (Observe, Orient, Decide, Act). Smack Technologies builds the AI models, applications, and product suites that dramatically compresses the “Orient” and “Decide” portions or the OODA Loop, where time and resource constraints are most acute. This means fusing and analyzing petabytes of sensor data, reasoning over that analysis, intelligently automating parts of the decision-making process that are low value for humans, and ultimately surfacing clear courses of action that account for the myriad variables and contingencies of modern, peer-level warfare.

The end result is rapid and coordinated planning and execution across all three decision horizons (1-12 months, 1-4 days, right now), all warfighting functions, and all echelons of command for the U.S. and coalition forces. 

How Does Smack Work? (Short Version)

Smack trains its models in a proprietary environment built to reflect the physics, decision spaces, and human expertise of war. Think of it like a movie theater that shows millions of films about WWIII to inform our RL models. The accuracy and realism of those films are judged by highly-trained human experts. 

Once in use, these models then analyze multimodal data streams in real time, reason over that analysis to create detailed plans in minutes (and plan updates in seconds), and finally optimize military decisions across all levels from the strategic to the tactical.

How Does Smack Work? (Less-Short Version)

To enable intelligent autonomy, Smack is built as a layered system that connects human intent to AI-accelerated decision-making. 

At the top, a flexible human-machine interface leverages commercially available LLMs (model-agnostic) to power user-defined agentic workflows. This interface is grounded in a proprietary Smack knowledge graph that structures real-world entities, relationships, and context. From there, the system calls Smack’s decision models to generate and evaluate courses of action. These models are trained within a proprietary environment, designed to reflect the physics, constraints, and human domain expertise of war.

Because there is no usable training dataset for WWIII, Smack’s environment is the training data, generating millions of synthetic engagements to teach our deep RL models how to master multi-agent behavior across space and time. Real-world data, while ideal, is both too sparse and too difficult to measure across interacting entities to train these models on its own. In its absence, Smack builds high-fidelity, expert-defined trajectories informed by human domain experts, then uses real-world data to refine and improve the synthetic manifold when it becomes available. 

Smack’s models are built on a proprietary architecture that combines graph networks, diffusion models, and transformers for tactical inference. We pair them with customer-specific applications (Fires, Logistics, etc.) in our Omega product suite to rapidly analyze multimodal data streams, reason over that analysis, and optimize decisions for commanders. 

To close the loop, we then leverage our Alpha product suite to combine Smack’s lightweight models with proprietary hardware in order to fuse local sensor data with campaign-level context. This both enables warfighters at the edge to make command-informed decisions in degraded comms environments, as well as allowing for orchestration of intelligent autonomous systems.

The end result is a mission built platform leveraging domain-specific AI models to create actionable campaign planning in minutes that can be updated in seconds and executed on the front lines in coordination with autonomous systems with or without comms access.

Learn More

This high-level summary is provided to help you better understand what Smack is and how it operates. To dive into more detail about the company, our models, applications, and product suites, book a demo here. 

You can also visit our About Us and Product pages, or follow our social media handles on LinkedIn and Twitter (X).