Structure Not Content, and Trajectory Not Point

Governance is not control. Governing a system is about where its trajectory goes, and whether it can still get back, not about whether the single action in front of it happens to be allowed. That claim is argued elsewhere. On its own, though, it does not tell you what to build. This paper adds a perspective of its own, on the form runtime governance has to take, and it draws that form out in two distinctions. Together they make an ontology of governance: what it is about, and what it is made of. The first concerns what governance evaluates, the point, one action at a time, or the trajectory, across the whole sequence. The second concerns how governance is encoded, as content, which the operation can reinterpret, or as structure, which it cannot. Each rules out a familiar shortcut. Together, wherever a system must stay adaptive and the operation can reinterpret what it is told, they force one shape: a governance plane that stands apart from the operation, assesses the path rather than the step, and reaches it only through immutable infrastructure the operation cannot reframe. That shape is realised in a separate architecture; here the argument is why governance has to take this form, not how to build it. The treatment is governance-agnostic. It concerns how governance is encoded and what it watches, not which values it should hold. AI is the running illustration, for that is where the failure modes are currently live and documented. The two distinctions themselves are general.