Definition
A memory architecture for AI assistants that adjudicates which stored facts are still valid when new information arrives.
A long-term-memory framework that does write-time conflict adjudication with an LLM judge, topology-triggered propagation across dependent attributes, and constrained readout that blocks stale memories from grounding answers.