The Architecture of Computational Closure: Scientific Priority under Extreme Compute Asymmetry

When a promising research direction becomes visible, the organization best equipped to complete it need not be the organization that developed it. This paper examines how concentrated computational capacity can amplify that separation. It develops a state-relative account of computational closure: the production of a checkable result from a specified, partially structured research state through computational search, synthesis, derivation, and verification. Closure is a function within discovery, not a judgment that the resulting work lacks originality. Building on research on scientific priority, collective discovery, and unequal access to AI infrastructure, the paper distinguishes conceptual genesis, roadmap formation, technical completion, and other evidentially separable contributions. It defines a matched-state last-mile compression ratio, derives a deliberately simplified research-race model, and proposes experiments that distinguish the effects of progress signals from those of actionable research guidance. The September 2026 OpenAI Navier–Stokes episode serves as a source-qualified illustration, not as a validation of the model or a determination of intellectual ownership. The analysis identifies computational hegemony as a possible institutional condition in which unequal completion capacity and unequal visibility jointly shape scientific credit. It proposes contribution records and proportionate provenance disclosure while retaining separate assessments of correctness, originality, and public recognition.