The Structure Beneath Simulation: Phase-Gated Intelligence, Sign-Regime Drift, and the Governance of Persistence

Bounded observers cannot act from total reality. They therefore compress experience into signs, thresholds, proxies, metrics, beliefs, records, and institutional regimes. This paper argues that mediation itself is not the fundamental problem: signs can be constructed and nevertheless socially or operationally real. The characteristic failure is sign-regime drift, in which a sign continues to carry authority after the identity-bearing recurrence or governing regime it once tracked has changed. I develop this claim through an identity-persistence framework in which coherent judgments of sameness through change require declared conditions of identity, admissible transformation, invariance, observation, drift, and verdict. The framework is applied to belief and social coordination, proxy substitution and institutional metrics, scientific paradigms, and AI decision governance. On the resulting interpretation, a scientific paradigm is a bounded regime for inquiry, an anomaly is a newly operative distinction obligation, and paradigm change must be distinguished from ordinary refinement. Scientific discovery can then be understood partly through retrospective compression: once an appropriate regime has been found, distinctions that were previously independent become consequences of a smaller generative structure, making verification much cheaper than discovery. The paper also introduces an elementary retirement calculus for kernel-typed demands at frozen scope. Retiring a demand enlarges the safe quotient and hence the distinctions that may lawfully be forgotten. This permits exact distinctions among conservative retirement, transformative specification weakening, problem-class change, and regime change, while exposing a separate governance problem: mathematics can determine the semantic effect of withdrawing a demand without determining who is entitled to authorize its withdrawal. The resulting position is neither anti-symbolic nor relativist. Bounded systems cannot eliminate mediation; they can discipline it. The relevant question is not whether a sign is “real” or “simulated,” but what regime gives it force, what invariant it preserves, and when that persistence relation fails.