AI Agent Architecture — ReAct, Memory, Planning and Multi-Agent Systems · Agent Memory — From Context Window to Vector Store
LLM is stateless — memory is an architectural trick
Agent Memory — From Context Window to Vector Store
Introduction
An LLM remembers nothing between calls — each request is a blank slate. All agent "memory" is an illusion woven by the architect: a context window filled with carefully selected history, tool results, and retrieved memories. This lesson unpacks what that means technically, what limits follow, and why this is a deliberate design trade-off rather than a flaw.