L3 — Rendered Artifact
Human-optimised views — committed or uncommitted
L3 denotes a rendered artifact: a human-optimised presentation — dashboards, reports, UI summaries — that may be committed or uncommitted. Renderings trade fidelity for legibility. They are appropriate for exploration, reporting, and standard operational UI when epistemic uncertainty is characterised and disclosed.
They are dangerous when they conceal epistemic uncertainty. A L3 dashboard that mixes live and cached data without distinguishing which is which, or combines streams of different vintages into a single authoritative-looking view, occupies the prohibited cell: L3 plus high uncertainty. The output appears ground truth; the underlying data may never have coexisted in reality.
L3 must never silently substitute for L0 or L1 in invariant enforcement. Consumers that cannot tell inference from decision, or fresh from stale, are not failing at presentation — they are failing at architecture. Renderings inform; they do not bind unless explicitly committed at a commitment boundary with full structure.