The Ouroboros Thesis — A Machine-Verified Substrate for Governed AI (v1–v23 unified)

Peer-style thesis family establishing bounded recursion as a system primitive for auditable AI, the Lutar invariant Λ as a governance trust aggregator, and a machine-verified proof substrate. v23 unifies v1–v22. Honesty doctrine: Λ is Conjecture 1 (unconditional uniqueness is machine-checked false); conditional uniqueness is proven in Lean under a declared block-consistency axiom. Locked kernel: 749 declarations / 14 unique axioms / 163 sorries. Only kernel-verified results are stated as proven.