Turning Agent Memory Into Skills That Work — Will Lyon, Neo4j
Source
youtube.com
Author
AI Engineer
Date
Why it matters
Retrieval over embeddings returns facts but not usable know-how. Modeling memory as a typed graph and distilling skills with grounding and staleness checks gives agents knowledge they can act on.
Key takeaways · AI-distilled
Will Lyon argues that embeddingA list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.Full definition → retrieval gives agents recall but not actionable knowledge. Neo4j's answer is memory that is connected, typed and traceable rather than a pile of similar text chunks.
Neo4j models agent memory as one context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → graph with three layers, short-term, long-term and reasoning memory, built through entity extraction, entity resolution and a shared ontology.
Decision traces in the reasoning graph record how agents reached their choices, which Lyon presents as the way many agents can learn from each other through shared memory.
Citing Neo4j research that treats skills as typed execution graphs instead of prose, the talk distills skills from memory and checks them for groundingTying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.Full definition →, coverage and coherence, with alerts when a agent skillA reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.Full definition → goes stale.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
videoBrains vs Hands: How to Run AI Agents Safely in Production — Viren Baraiya
videoWe Mapped 115 Microservices for Our Coding Agents — Kamalakannan Nandagopal, Postman
videoWhy We Deleted Our MCP Server and Rebuilt It — Abhi Arya, Reducto
videoLet Anyone at Your Company Ship Internal Apps with AI — Garrett Galow, WorkOS
