Confidential Narrative Intelligence on a Local AI Box: Epistemic Symmetry Theory with Algebraic Mixture-of-Experts Routing

Financial institutions want to read the narrative of markets with language models, but much of what matters most — internal research, positions, committee notes, client communications — cannot leave their perimeter. This report describes how Epistemic Symmetry Theory (EST), a framework that detects regime changes in market narratives and explains what changed, is designed to run entirely on a local AI box. The system is organized as a Mixture-of-Experts in which the experts are local language models assigned to distinct offces (measuring, embedding, answering, generating, verifying) and to distinct domain hyperspaces, and in which routing is decided by deterministic algebra rather than by a learned gating network: a small, fast recognizer chooses; a capable model answers; every composition of operations is checked before it runs. We describe the architecture, the confidentiality model — what stays on the box, what may leave and only by the user’s choice — the EST engine that runs on top of it, hardware sizing, and the limits of the current evidence. The report is conceptual: proprietary implementation details are not disclosed.