Discovery at Machine Speed, Admissibility at Human Trust: A Multitime Receipt Standard for Millennium-Class Problems

The acceleration of scientific discovery via artificial intelligence and high-performance computing is creating a structural admissibility gap: candidate generation increasingly exceeds the capacity to produce independently replayable, auditable, and signable evidence under bounded resources. This paper proposes the Multitime Receipt Standard (MRS), a normative interface that decouples high-cost discovery from bounded replay verification by requiring a hierarchical receipt bundle with pinned artifacts, deterministic replay, contestability traces, and scale-ledger instrumentation. MRS standardizes uncontrolled artifact growth and unresolved ambiguity as formal admissibility failures, enforced by mechanical scale guards and recursive contestability through discriminator sub-receipts. We instantiate the protocol through two archetypes: discrete certified evidence (SAT-class receipts) and continuous envelope evidence (validated-numerics receipts). The contribution is not a solution claim for Millennium problems, but a testable governance standard that localizes admissibility bottlenecks and makes “what is signable” mechanically checkable.