The Hallucination-to-Insight Pipeline (HiP): A Source-Agnostic Institutional Discovery Architecture for Durable Hypothesis Custody, Retrospective Reassessment, and Prospective Discovery

This technical paper presents the Hallucination-to-Insight Pipeline (HiP), also interpretable as a Hypothesis Intelligence Platform, as a source-agnostic institutional discovery architecture for preserving selected non-winning hypotheses, designs, predictions, plans, and other candidate artifacts. HiP records not only the candidate, but also its provenance, evidence state, operational disposition, causal basis for closure, evaluation history, reopening conditions, viability environments, custody, and federation metadata. Its Retrospective Engine identifies changes in evidence, instrumentation, capabilities, costs, regulation, or evaluator availability that may weaken an earlier closure decision. Its Prospective Engine searches for enabling conditions, combinations, domains, institutions, or uses under which a preserved candidate could become feasible or valuable. The architecture applies to language and multimodal models, large tabular models, predictive world models, JEPA-style systems, scientific simulations, search and evolutionary systems, human-machine processes, and future computational paradigms. The paper provides a minimal public Governance Envelope, a worked Hypothesis Object based on continental drift, a reproducible 24-case historical replay, explicit refutation controls, an illustrative lifecycle cost model, and a proposed materials and process engineering pilot. The historical replay is presented as a retrospective test of the published closure and reopening rules, not as proof of prospective performance or an estimate of production accuracy. This record contains the public technical-paper release of HiP.