Data/ML/AIEmerging
Agent Memory
Also written as Long-Term Memory (AI), Conversation Memory, Agent State
How an AI agent keeps track of information beyond a single conversation or context window: past interactions, user preferences, progress on a long task. It is usually built from summaries, databases or vector stores that the agent reads and writes.
Think of it like
A consultant's case notes, so they don't have to ask the client the same questions at every meeting.
Junior or senior?
A fast-moving area with no settled approach.
Senior sounds like
Talks about deciding what to remember and what to forget, and about stale or wrong memories causing problems.
Ask them
“What did your agent remember between sessions, and what went wrong when a memory was out of date?”