Abstract Most systems describe decisions after the fact. A smaller class can prove decisions. An even smaller class can refuse execution of decisions that cannot be proven. This paper defines that stack. Enterprise workflows increasingly depend on computationally produced decisions whose consequences are financial, legal, medical, or regulatory. The standard observability stack—logs, traces, dashboards, and audit narratives—supports debugging and operational awareness but does not produce proof. When a specific decision is challenged, these systems yield a plausible reconstruction rather than a byte-exact reproduction, and downstream systems lack a mechanism to condition execution on verifiable identity. This paper defines deterministic decision verification infrastructure as a distinct category. The category is characterized by four interdependent mechanisms: (1) deterministic computation producing byte-exact identical output from identical input under a declared version context; (2) canonical identity formation, in which each decision is captured as a committed artifact with stable cryptographic identity; (3) replay-exact verification, in which the artifact can be re-executed to produce identical output; and (4) proof-gated execution, in which downstream systems condition action execution on independent verification of the artifact and refuse actions whose identity cannot be established. The paper distinguishes this category from logs, signatures, timestamping, audit narratives, governance tooling, and model explanation systems. It introduces the artifact as the primary operational object, defines proof-gated execution as the control-layer consequence of identity, and provides a glossary to stabilize category vocabulary.Note:This work introduces a category definition for deterministic decision verification infrastructure and proof-gated execution systems, with implementation alignment in production-grade verification stacks.
Devin Bostick (Thu,) studied this question.