Description Without Grounding: An Operational Model of Self-Grounding Limits for Finite Embedded Systems

This paper develops an operational model of the limits of self-grounding in finite embedded systems. It distinguishes a system’s capacity to describe its own rules from its capacity to certify that its actual realisation follows those rules, and separates certification from recorded evidence from certification carried out by the system’s own run. An evidence-fibre criterion characterises when a positive verdict is sound relative to a specified class of possible realisations. Within this framework, uninvoked transitions are not determined by past history under local freedom, while certifications that depend on a restrictive realisation assumption cannot establish that assumption from the very same evidence. For realised audits, masking constructions identify conditions under which an internal positive verdict is unsound, alongside a counterexample showing that model-relative positive self-certification is possible. Finite non-circular ultimate grounding is then shown not to close under either per-credential masking or an expressly non-primitive grounding semantics; the latter is a task assumption, not a consequence of finitude alone. The analysis also distinguishes diachronic audit from synchronic mutual checking and derives fault-tolerance bounds in terms of independent dependency classes rather than raw channel counts. The resulting framework clarifies both the uses and the stated limits of self-audit, formal verification and epistemic authority.