John Falcone, Head of RLHF

We’ve all seen the recycled themes and quotes. Allusions to “flash war”, “the Terminator conundrum”, and probably the most cited, “whoever becomes the leader in [AI] will rule the world.” These phrases are consistently pulled from their original context and contorted to align with the author’s thesis. 

What they miss, however, and what the original speakers understood, is that warfare remains a distinctly human endeavor. And the systems we build must reflect this reality. 

Which is why, after nearly a decade in the defense tech ecosystem, I’m excited to join Smack Technologies as Head of Reinforcement Learning from Human Feedback. 

For AI and other emerging technologies to secure America’s warfighting edge, they must be grounded in the “profession of arms”. At Smack, I’m scaling the premier network of warfare domain experts and integrating their knowledge throughout our reinforcement learning pipeline to capture the art and science of warfighting. 

To do so successfully, I’ll be drawing on lessons from a career that has spanned military service, policy, and defense technology. This includes serving as a Navy Surface Warfare Officer in diverse theaters from the Mediterranean to the Western Pacific and authoring fleet-wide processes to align human operators with newly integrated technologies. It also includes experience in the United Nations Department of Peacekeeping Operations and the Naval War College’s Humanitarian Response Program, which influenced thought leadership pieces on AI and nuclear decision-making. Most recently, my work at DIU helped spearhead  AI integration across the maritime services and highlighted the challenges we face to achieve decision dominance where it matters.

These experiences helped lay bare a similar, underlying problem facing military decision-makers: how do you build a system that reflects the way experienced operators think and adapt, when the picture is incomplete and conditions are changing in real time? That question is a large part of why I joined Smack.

During my time at DIU, I had the opportunity to review hundreds of white papers and, through them, develop a broad understanding of what much of the defense AI ecosystem is building. The pattern tends to look familiar: train LLMs on formalized doctrine, written tactics, and whatever data can be collected at scale. Doctrine matters, of course, but anyone who has served knows it’s only the foundation. What comes next is the harder problem to solve: ensuring a plan can both survive contact with reality and be repaired under time and physical pressure.

That’s where Smack’s approach is truly one of one. 

The company has gone the difficult but necessary route of training their models in an environment designed to capture the judgment of experienced operators planning and making decisions under pressure. It’s the difference between a system that learns from real domain expertise and one that delivers output that is polished, plausible, but ultimately shallow.

Our domain experts, collectively known as the Smack Den, are a force multiplier. Their expertise and judgment have been honed in conflict, at the table with national leaders, and through academic and research inquiry. And they want to work with and contribute to Smack precisely because we were warfighters and deeply understand the need to deliver the best tools for the next generation. 

My decision to join the team goes beyond both the technical approach and the success of the founding team to date. I was drawn to the culture. From engineers to the C-suite, this is an organization where people are willing to wrestle honestly with both hard problems and, quite literally, each other. And there is no question about our mission: deliver frontier AI decision-making platforms for the warfighter wherever the need to exert decision dominance exists.

Warfare is a human endeavor. As Head of RLHF, I aim to ensure Smack’s systems reflect the realities of war as understood by the people who’ve lived it. If we do that well, the result will be a decisive decision advantage grounded in human judgment, operational relevance, and trust.