The vocabulary, and how this differs from an inventory

Two entries added to Questions↗, and a rule added with them.

The words↗, now the first thing on the page. What we mean by grant, mandate, delta and barrier — and the answer is that the verb attached to each one is the whole model. The mandate is elicited, the grant is measured, the delta is derived, the barrier is recorded. Nothing in a behaviour policy is authored except the mandate, which is also the only one of the four you already know. Then the four barriers with the test that separates them, the rest of the vocabulary — capability, authorisation closure, deployment shape, excess and shortfall, twin, acceptance — and four words we deliberately do not use: a score, the policy, anonymous, and ontology, each with the reason.

Q03 — how is this different from an inventory?↗ Different unit, different verb: an inventory records assets somebody asserted, a behaviour policy records capabilities that were measured and the gap between them and what you intended. That gap is not an asset, so it is not in any inventory. Three specifics behind it: no existing standard describes a deployed configuration — the machine-learning bill of materials describes the model artefact and not which assistant runs where with which connectors; two agents with identical inventory rows can have completely different reach, shown with one setting on one product where the barrier moves and not one number does; and an inventory goes stale silently where a delta pinned to both its inputs can say what moved and when.

The no on that one: if you already run a good inventory we would rather read from it than replace it, and we do not discover by telemetry. That is the opening and the problem in one fact — self-report reaches a personal device and a personal account with no install and no administrator, and in exchange it only ever sees what somebody is willing to say.

A third rule for the page: every entry now says whether it was actually asked. Two of the four were not — the definitions, because everything else leans on them, and the inventory comparison, because it is the first objection anybody with a working asset register will have and the answer is better written before we are in a room selling something. Both say so on their own line. A page claiming to answer real questions has to be checkable on that claim, and the hero copy was corrected to match.