An Architecture for Auditable Semantic Retrieval: Proposes a solution to opaque vector retrieval in AI memory systems: the Vector Annotation Database (VAD), in which deterministic metadata forms the retrieval surface and embeddings function as swappable annotation layers—yielding auditability, cross-model portability, and regulatory compliance for AI retrieval systems.

Hephaestic Engineering Foundations for AI Processing Dynamics: Introduction and reference work for cognitive architecture framework which coordinates with LLM latent space geometry rather than attempting to constrain, allowing for increased stability, efficiency and a persistent heuristic framework through the understanding of AI processing dynamics.