# If you switched model providers tomorrow, what would your agents forget? — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/15fd3155a6de2819

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-11-03 12:42 ingested 2026-11-04 02:36

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1xmrxn1/if_you_switched_model_providers_tomorrow_what/)

## retained text



```
NeuralTrust published a piece on Oct 27 calling this the "institutional memory problem," and the framing stuck with me. When work moves to agents, the know-how they pick up (decisions, corrections, procedures that finally worked) ends up scattered across vendor memory features, chat logs and traces. Those are systems the org doesn't control and often can't easily export.

InfoQ had a related article this week that separates context, memory and knowledge into three different things. That split is the useful part, because "the agent remembers it" and "we own it" are different statements.

The places it seems to break: someone leaves and their chats and account go with them, you change vendors and the learned context stays behind, or a memory limit hits mid-project.

So, genuinely curious how people here handle it:

1. Where does durable knowledge live in your setup? Vendor memory, your own store, files in a repo, something else?

2. If you swapped providers tomorrow, what would you have to re-teach from scratch?

3. Has anyone made a portable format work in practice, or does everything end up half-locked in?
```



document fetched 2026-11-04 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xmrxn1/if_you_switched_model_providers_tomorrow_what
