Sections

AI Agents

Software that decides what to do next, rather than waiting to be told. An agent chooses which tool to call, reads what comes back and decides again, which is what separates it from a pipeline and what makes it hard to test.

Articles in this section cover the parts that hold up in production: how a model reaches your systems, where the boundary between the model's judgement and the application's control should sit, what an agent costs when it is allowed to loop, and how to measure behaviour that is not the same twice.

We write about mechanisms rather than frameworks. A framework is a season; the question of who decides what gets called, and on whose data, outlives it.