Prethermal Plateau Unification: Pad Geometry Plus MAP Translation

This critical manuscript addresses a persistent failure mode in prethermal plateau and stability claims: the gap between demonstrating a long-lived dynamical regime and certifying that the regime is usable under fixed, replayable rules. Recent platform results show that structured random driving can suppress heating and extend prethermal lifetimes on many-qubit processors. While these results establish physical existence and controllable lifetime, they do not by themselves certify operational utility or guard against boundary narrative repair. This manuscript extends the earlier pad and triad-of-triads program by applying MAP as a unifying translation layer to prethermal plateau claims. Prior work established the signed-triad ledger, Strong-to-Weak downgrades, STLF refusal discipline, and tokens-only replay as independent enforcement tools. Here, those tools are bound together by MAP and explicitly externalized from the engine, allowing plateau lifetime claims to be evaluated as contract objects rather than as narrative curves. The result is a carry-forward certification spine that preserves earlier triad logic while adding a thermodynamic interpretation of acceptance leakage and a regime-transition marker (ZFlip) suitable for plateau-style datasets. The central contribution of this work is Mapping and Amplification Protocol (MAP), a pad-centered translation layer that externalizes certification logic from the engine and forces enforcement tools to act coherently under a shared coordinate meaning. MAP is not a new physical mechanism and not an optimizer. It is a deterministic translator from tokens to enforcement actions that binds Strong-to-Weak, STLF, Knob, and Sum-of-Zeros to the same declared pad and corridor. By doing so, MAP removes silent degrees of freedom that otherwise allow filtering, tightening, or drift to masquerade as stability. Under MAP, plateau behavior is no longer evaluated by a single flat curve or preferred projection. Instead, claims are promoted, downgraded, or refused through a triad-of-triads contract that separates signed state, windowed pattern fingerprints, and meta-promotion logic. Two independent redundancy cages (view square and ladder square) are enforced via double Four-Solves-Three auditing, blocking majority narratives and forcing disagreement to remain visible. Acceptance recovery is explicitly bounded: any recovered utility must be paid for through declared downgrades rather than repaired into Strong. Within this framework, the manuscript introduces a contract-level thermodynamic separation: what platform studies describe as “heating” is treated here as acceptance leakage in time under a fixed corridor. The quantity TzTzTz is operationalized as a burn-rate proxy that tracks how quickly tension is consumed as trajectories approach refusal, while ZFlipZFlipZFlip marks the first persistent regime transition where the effective plateau description fails. Plateau lifetime is therefore treated as necessary but not sufficient; the resource question is whether the plateau survives under fixed accounting until the flip. The intended outcome is a reusable, platform-independent certification object. MAP and the surrounding contract infrastructure are designed to carry forward across qubit, magnonic, photonic, and hybrid platforms, allowing future plateau datasets to be evaluated without modifying the rule spine. The result is not a critique of platform physics, but a durable method for turning plateau observations into bounded, replayable, and falsifiable objects of record.