Tag
The article reviews 16 Agent Memory papers from the past two weeks, pointing out that the trend has shifted from static GraphRAG retrieval to self-evolving, multi-signal routing, cognitive models, and low-cost memory systems, with per-paper overviews and links included.
This article proposes a 14-step roadmap from single agent to self-evolving system, emphasizing that base engineering (models, tools, permissions, context) is the key to determining the quality of loop output, and details practical methods for building an efficient base such as CLAUDE.md, sub-agents, skills, hooks, and state files.