Preventative Architecture: Structural Constraints on Admissible Construction

Under conditions of operational closure, not all formally specifiable constructions are admissible as primitives. Treating downstream objects—such as semantics, optimization criteria, explanatory frameworks, or values—as generative starting points introduces structural incoherence rather than error. This paper introduces preventative architecture as a distinct design category defined by structural constraints on admissible construction. Preventative architectures operate by rendering entire classes of illegitimate constructions unrepresentable at the level of grammar or structure, eliminating failure a priori rather than detecting, correcting, or regulating it post hoc. We formalize the defining characteristics of preventative architecture, distinguish it from corrective, normative, and verification-based approaches, and analyze near-miss systems that restrict behavior without constraining expressibility. The category is shown to be instantiated across multiple domains, including programming language design, cryptographic protocol construction, engineering safety, and epistemic systems operating under closure. The contribution is architectural rather than prescriptive: preventative architecture constrains admissibility itself, ensuring that illegitimate moves cannot arise as objects of construction. This distinction explains its necessity under closure and its robustness under adversarial conditions.