Ben Ford served as a Royal Marine Commando and now works as a software engineer, with a background in Haskell and category theory. Mark “Moose” McGrath and I asked him how John Boyd’s real OODA loop shows up in agentic AI.
Ben argues that much of the AI field still works from the linear cartoon of the OODA loop. In his view, a large language model takes text in and puts text out, with no destruction and creation and no update to its model in real time. Active inference agents do both.
Orientation at the center
Ben makes the case that orientation is the whole loop, drawing on neuroscience and category theory. He describes unfolding circumstances as affordances in space and time, and he connects that to Wardley Maps and the spatial web. He links Boyd’s Destruction and Creation to active inference, the free energy principle and Markov blankets, the boundary that separates an agent from its world.
Exploration and the edge
We talk about knowledge of a thing and knowledge about it, and why Ben thinks exploration beats rigid process. He also gives his forecast: agents built on the free energy principle running on edge devices, using far less energy than today’s large models.
Orientation sits at the center of Boyd’s real OODA loop. Ben’s work shows what happens when engineers build from that center.
What would your systems look like if they could rebuild their own models in real time?
Listen and watch:
YouTube:
Apple Podcasts:
Spotify:
No Way Out on YouTube:
https://youtube.com/@nowayoutpodcast