Cloud AI made a bet that data would flow to the model. In military operating environments that bet fails, for two distinct reasons. This post explains why, and closes a three-part series on the design principle that runs through every layer of defence AI.
The problem in defence software delivery was never really about distance. It was always about how much of the real problem survives the journey from operator to engineer. LLMs change what fidelity is achievable when physical proximity is not possible.
Defence software delivery consistently fails because of the structural distance between the engineer who builds and the operator who uses. The case for forward deployment, why it works, and why the commercial model is the hardest barrier to fix.
AI platforms have shifted from productivity tools to enterprise operating systems. This post examines what that means for competitive strategy, workflow redesign, and how product teams should think about building and scaling AI agents.
A framework for thinking about what makes a deftech AI company genuinely hard to compete with over time. From compliance and accreditation infrastructure to knowledge architecture and cleared human capital, the most durable moats in defence AI are not the most obvious ones.
Silicon Valley spent a decade building an ideology about escaping state power. What it actually built was the most capable surveillance and military contracting infrastructure in history. This post asks what that contradiction means for the founders now being recruited into the defence tech wave.