Outcomes are the universal frame.

A Stream's slots are the mutually-exclusive ways its question could land, and its priors are evidence that moves the read. An action is an outcome with a known driver — when the question is one the seat itself decides ("Do I storm the east gate or wait for Karis's signal?"), its outcomes simply are its candidate moves, scored by the same math and carrying no separate machinery. So a Stream takes whichever framing its question genuinely has:

• What happens — the perspective reads a situation it does not control. The register of a Public Domain: Research cycles route cited findings into Streams whose slots are world resolutions.
• What I do — the perspective is the driver, and the slots are its own moves. The register of deliberation: a room, a team, a personal domain.

There is no discriminant to set and no second code path — the question's phrasing carries the framing, and a Stream may sit anywhere between the two. What is forbidden is forcing one into the other: never manufacture a decision for a perspective that has none, and never flatten a live choice into a detached third-person forecast.

The Stream is the transducer that makes the qualitative quantitative: a prose prior is scored by its evidential weight into a shift, and a chain of priors becomes a tracked read — distribution, margin, uncertainty, volatility — sharpening over time. Wanting an outcome is not evidence for it; a prior that moves the stance without new grounds is a failed prior. That is what makes the rest measurable: a stream is a question under live reading (which slot holds, decays, gets overturned); cross-seat stance distance reads as alignment one way and diversity the other. See Believability below.