Moving on from behaviour trees to a system of loosely coupled OODA loops — observers, orienters, deciders and implementers — for autonomous software development that keeps agents productive without making me the bottleneck.

Using behaviour trees to model decision-making in autonomous coding agents, drawing from game AI techniques.

Can AI agents be creative? Exploring whether randomised conceptual lenses — like reading a random Wikipedia page — can help agents escape local maxima and find elegant solutions.

What if AI agents could do useful work autonomously overnight — fixing tech debt, building prototypes, and organising your backlog like helpful elves?

Exploring the idea that application specifications — not code — should be the source of truth, and how agents can help maintain them.