systemtheory · Philosophy

The philosophy of systems

A philosophy of what AI systems can and cannot do, stated in precise ontological terms. What exists inside an autoregressive model? What does it mean for such a system to know, to err, to intend?

It is the conceptual companion to the technical work on systemlayers: structure over behavior, mechanism over symptoms, reproducibility over intuition.

All essays

  • Before reasoning: a security abstraction for how LLMs construct reality

    Most LLM security work asks what a model decided to do. This essay is about the step before that, where the model constructs the objects, roles and constraints it will reason over, and argues that a functional abstraction of that step is the right kind of claim to make about a system whose internals we do not understand.

  • The structural method: six steps, and what they rule out

    The first post on systemtheory is not an argument for a position. It is a description of a procedure: observe, reproduce across independent systems, isolate the mechanism, build the smallest model that accounts for it, test its consequences, and state what would falsify it. A research programme is better defined by what it refuses to accept as evidence than by what it concludes.