If an agent could forget most of its session, what should it be allowed to keep?
retained text
I've been looking at the recent push toward persistent AI memory, and the part I find more interesting than "give the agent more memory" is the write gate. A few current projects are experimenting with persistent, searchable or versioned memory rather than treating the entire conversation as permanent state. Anthropic's own agent-memory research also focuses on things like curation, deduplication, pruning, and selective injection as memory grows. That made me wonder what the actual gate should be. If Claude Code finishes a task and wants to write something into persistent memory, should it need to answer something like: \- Will this change future behavior? \- Is this a decision or just an observation? \- Is this still true? \- Is there already a memory saying the same thing? \- Would I want a future session to see this? I'm especially interested in what people who use Claude Code heavily actually keep in memory today. Do you save decisions? User preferences? Project architecture? Failed approaches? Commands? Or mostly let CLAUDE.md / project files carry the durable state? I'm not claiming external memory is universally better. I'm trying to figure out what deserves to survive a session in the first place.
document fetched 2026-11-19 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1xzf6g4/if_an_agent_could_forget_most_of_its_session_what