Why it was a question

QUESTION-252 asked it together with the leave-one-out design: which of the five conditions can fail while all the others hold, and which such configurations are impossible in which model classes. Working through that design, A4 and A5 of the exchange of 2026-10-20 found that the five were not coordinates of one kind. This part has an answer the record holds; the leave-one-out part does not, so they are filed apart (ADR-049).

The answer

Not of one kind (CLAIM-401):

  • Adequacy for a target. Predictive sufficiency and approximation fidelity hold only relative to a declared target, metric and horizon (CLAIM-152, CLAIM-348).
  • A restriction on operations. Tractability restricts the operations under which adequacy is assessed, and the same representation passes or fails with the budget (CLAIM-304).
  • Relations of the representation to the system. Dynamical independence and memory closure, the latter relative to a chosen projection (CLAIM-397, THEORY-752).

So a matrix over the five as independent binary axes conflates kinds. The answer does not say the five reduce to three, and it leaves the leave-one-out cells, once the regime (TERM-065) is fixed, to QUESTION-252.