Listen and ask questions on NotebookLM: https://notebooklm.google.com/notebook/ba22d819-31a0-426f-98fc-ef3ea2f83e0e?authuser=1 This upload contains two connected works developing Trace Fidelity as both a theory of reasoning and a practical infrastructure model. Trace Fidelity: Semantic Graph Auditing, Orphan Nodes, and the Lineage of Reasoning introduces a method for evaluating whether claims remain connected to the evidence, object, constraints, and interpretive path that produced them. It defines semantic graphs, graph jumps, orphan nodes, constraint hierarchy failure, and semantic debt as tools for auditing human, institutional, and AI reasoning. Building Trace Fidelity Infrastructure extends the theory into an operational architecture for local-first reasoning spellcheck, claim-lineage capture, semantic debt scoring, historical archive scanning, and AI-guided revalidation workflows. Together, the papers argue that institutions do not merely need more data, search, or storage. They need systems that preserve why claims entered the record, whether their support still holds, and when unsupported relations should be repaired, downgraded, or retired.