← sources

How do you stop your agents from using outdated information?

source: reddit (reddit:AI_Agents)published 2026-10-02 14:53ingested 2026-10-03 02:35

retained text

When a source document gets edited or deleted, how do you handle its old chunks, embeddings, and any memories the agent created from it? Rebuild everything, update incrementally, fetch fresh data at query time or something else?

I'm interested in any concrete examples: what went wrong, and what actually worked?

Disclosure: I’m a co-founder of Antfly, working on retrieval infrastructure. This is a problem we’re designing around, and I’d like to understand how others handle it.

document fetched 2026-10-03 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1wvvv5e/how_do_you_stop_your_agents_from_using_outdated