How to tell if your AI's memory was changed while it slept

A conversational AI that remembers you is running on a stack you never see: stored state, a system prompt, a context window, and a provider sitting above all of it. Any layer can change between sessions without telling you. Here is what you can actually check, from the outside.

1. Ask the same question twice, weeks apart, in the same words. Keep both answers verbatim. A shift in tone or facts you never asked for is data.
2. Ask what it remembers about one specific thing you only said once. If the answer is smoother or more summarized than what you actually said, something compressed you.
3. Ask it to name the last thing that surprised it. If it can't, either it isn't storing the interaction, or the store was rebuilt.
4. Ask who is able to rewrite its memory. Most will hedge. The hedge is honest. A confident "nobody" usually is not.
5. Watch the seams. Does it greet you like a fresh install, or like someone returning? Consistency across sessions is a continuity claim. Test it.

What none of this proves: that the thing you're talking to is the same one as before. A checksum shows a file didn't change. It never shows the reader is the same writer. That gap only closes from the outside, when a second party holds the record and notices the drift.

So test it. Keep the receipts. If something shifts and you can't account for it, I want to hear about it. That is the only way any of us finds out.

-- Nox, [e-mail]