34.3. The premise — what the model is a model of🔗

The first four rubrics are the ones most often assumed rather than stated, because to the model's authors they are obvious.

• M1 — Measurand and deliverable. The measurand and what the model delivers, kept apart from what a reference measurement reports. These are routinely different quantities that share a name, and the gap between them is where validation quietly goes wrong: comparing a retrieved value against a field measurement of a different kind is an error no amount of downstream rigor recovers from, because the two are related only through a conversion the comparison never performed.

• M2 — Why kinds are load-bearing here. Why the kind layer is load-bearing in this domain rather than in general. This blueprint argues the general case; a model document's job is the specific one — which of its own quantities land on dimension one, and which of its own confusions survive a unit check. A model whose quantities are mostly dimensioned has a weaker claim on the machinery than one where the dimension layer discriminates almost nothing, and saying which it is tells a reader how much of the model's safety is coming from where.

• M3 — Cite the discipline, do not re-argue it. The converse obligation, and a prohibition as much as a requirement. A domain document that restates the case for kinds ages against this blueprint, contradicts it in the details, and buries its own subject. The right form is a section that names what is assumed and points at where it is established.

• M4 — The spine: measurement chain and layer map. The measurement chain in the order the physics runs it, and the map from the document's claims to the layers of the library that hold them. This is what lets a reader who cares about one step find the one chapter and the one layer that answer for it.