The Witnessed Difference Programme - From Finite Distinction to Witnessed Measurement

The Witnessed Difference Programme: From Finite Distinction to Witnessed Measurement develops a finite obstruction calculus that makes the assumptions behind mathematical and operational constructions explicit. Starting from an exhibited distinction, it examines what additional structure is required for retained witnesses, succession, duration, finite probability, observers, measurement, and weighted histories. The mathematical core includes finite symmetry obstructions, cochain complexes, quotient defects, closure diagnostics, and exact optimization certificates. Matching repair and dual witnesses certify an optimum under declared hypotheses. Presentation and adapter theorems characterize when compressed reports preserve specified operations, relations, and queries; counterexamples identify information that compression loses. These results provide concrete criteria for checking whether a proposed representation preserves its intended meaning. A layered observer architecture distinguishes presentation, memory, evaluation, and policy. Observation requires a registered, witnessed, re-accessible interaction accepted by a declared evaluator. Measurement adds a specified measurand, reported-value map, calibration, uncertainty, provenance, and any claimed metrological traceability. Permission and authority require their own declared rules. Version 0.5.0 consolidates the primitive-axiomatics and Genesis-alpha sources and develops additional bridges to the actual Genesis raising and marked-result constructions. Complete quotient keys retain directed assertions, loops, and joint incidence, while ordered raw histories preserve the information needed to replay declared costs and transitions. Nineteen historical targets receive finite or conditional resolutions through constructions, exact criteria, or counterexamples; the wider Genesis reconstruction obligations remain separately registered. A further extension connects Meron orbit cancellation to finite character sums and gives precise conditions for preserving observable numerators. The CODA-CS/Q branch develops conditional interval enclosures and refinement interfaces. Exact finite transfer traces and rational extrapolation illustrate the distinction between arithmetic error and model error: an enclosure for a finite interpolation calculation bounds a thermodynamic limit only when justified scaling-remainder bounds are supplied. The accompanying evidence architecture connects statements to implementations, tests, selected Lean 4 formalizations, source provenance, and independently checked receipts. Formal proofs, computational demonstrations, and physical interpretations retain their separate scopes. The contribution is an explicit framework for locating the assumptions, preservation conditions, and missing interfaces in a derivation. Quantum-capable representations are developed under declared inputs; unrestricted quantum reconstruction, Born-rule uniqueness, continuum physics, and the proposed higher-dimensional field-theory targets remain open.