Witness Core Abstraction: A Structural Law for Independent Verification in Computational Systems

The Witness Core Abstraction defines a structural constraint on all systems that produce consequential claims: no system may be the final verifier of its own outputs.

This paper extends the verification-constrained execution model established in InnerLight OS into a generalized architectural principle. The abstraction introduces a four-part structure—Deed, Witness, Testimony, and Arbiter—and formalizes the invariants required to preserve independence between assertion and verification.

This separation prevents the collapse of truth into self-certification, a failure mode observed across modern computational and governance systems.

The paper establishes the structural conditions required for independent verification without exposing implementation details, positioning the Witness abstraction as a foundational principle for verifiable computation and governance architectures.

Keywords: AI governance
verification systems
independent verification
computational trust
system architecture
admissibility
InnerLight OS
OBEXGATE
decision systems
algorithmic accountability