A Typed, Dynamic, No-Meta Theory of Autonomous Research Claim Certification and Release

This paper develops a first-principles, typed, dynamic theory for certifying and releasing research claims produced by autonomous research systems under finite verification capacity and public accountability constraints. The framework is built under two explicit restrictions: observable-only governance and no-meta governance. Decisions may depend only on public claim content, declared schemas, public histories, released artifacts, provenance records, replay outcomes, and declared inferential traces. Hidden evaluators, latent truth labels, and opaque semantic judgments are excluded unless they are externalized as declared public checks or admissible proof-carrying public maps. The paper introduces a measurable architecture based on primitive public state, a declared authority algebra, an authority profile, and a constructive typed transcript. Within this framework, it formalizes typed public checks, authority-eligibility rules, current-validity, repair and exclusion logic, fail-closed certification memory, and a separate release layer with versioned release units, lineage, certified tokens, shipment slots, and serialized support-ledger accounting. The main results show that the typed transcript is a lossless representation of the declared authority sigma-field; that certifiability has distinct structural and operational frontiers; that inferential validity can be sealed into a public certification pipeline via a program-level shell with predictable e-processes; and that finite verification service creates an intrinsic capacity limit for certified release. The theory also proves that transcript information alone does not guarantee a universal positive defect floor for non-admitted shipment, while calibrated public risk classes can yield conditional lower bounds on expected outgoing defect burden. The paper is intentionally narrow: it does not claim to solve unrestricted scientific truth, novelty, or impact. Instead, it provides a rigorous governance theory for what can be publicly certified, held, repaired, excluded, or released when autonomous research outputs must be evaluated under explicit no-meta and observable-only constraints. Archived on Arweave(Data Tx ID)TeX source: 6YWB9avaqGp8LOKQ0x6M6Z0Q2GfXce9tdDLrb-bcAo8PDF: XRAdpbIniq3TMkB9jsfT0hMOws7EYE9GR3Qd1ALiVHo