Turning Agent Memory Into Skills That Work — Will Lyon, Neo4j

Why watch

Will Lyon from Neo4j explains why retrieval-based agent memory falls short and demonstrates a graph-based approach using context graphs with short-term, long-term and reasoning memory. The talk covers entity extraction, shared ontologies, decision traces, and how to distill grounded, governed skills from agent memory traces. Includes a demo of building a patient-intake skill from healthcare conversations.