Molecular Memory Networks (MMN) v1.0 — Governed Knowledge Artifacts

This record contains the MMN v1.0 preprint and accompanying reference runtime demonstrations. Molecular Memory Networks (MMN) are proposed as an architectural framework in which knowledge is represented as governed, revisable artifacts (“molecules”) rather than as static parameters or unstructured memory. The preprint presents the core concepts, operational loop, and lifecycle protocols (application, revision, rollback, and dissociation) without claiming state-of-the-art performance. The accompanying artifacts include three minimal, fully reproducible runtime demonstrations:(A) a minimal operational sanity check validating signal dynamics (surprisal, adaptive level-of-detail),(B) a rollback scenario under regime shift illustrating safe withdrawal of an invalid hypothesis, and(C) a hypothesis ecology scenario demonstrating coexistence and contextual dominance of competing molecular hypotheses. The implementations are intentionally simple and not intended as benchmarks. Their purpose is to make the lifecycle of molecular knowledge artifacts observable and reproducible in a controlled setting.