AI Agent Architecture — ReAct, Memory, Planning and Multi-Agent Systems · Agent Memory — From Context Window to Vector Store
Reflection — higher-order memory synthesis
Agent Memory — From Context Window to Vector Store
Introduction
Reflection is a mechanism in which an agent "thinks about its memories" and generates new, higher-level entries: observations, insights, rules, and predictions. In Generative Agents, Park et al. (2023) trigger reflection periodically to produce new nodes in the memory stream that then participate in further retrieval. This lesson unpacks: how the trigger works, how reflection is generated, why reflections carry higher importance, and how agent knowledge is built hierarchically.