AI work method
Why ChatGPT can seem to forget what you told it (and a note that survives it)
You told it three chats ago. Now it asks again, or it suggests something you already ruled out.
A new chat does not automatically include the earlier chat. If Memory is on, ChatGPT may carry over some details, but OpenAI says Memory “does not retain every detail from every conversation”. ChatGPT manages a shared context window, and part of it is used for system instructions and, if enabled, memories (pricing page footnote). In my experience, answers in a long chat can stop following early instructions. Written notes you control are the dependable part.
Sources, checked 2026-11-02: Memory in ChatGPT; Plans and Pricing (ChatGPT; footnote on the shared context window). Features change, so check these pages for the current wording.
Three things that can look like forgetting
| What you see | What is going on | What to write down |
|---|---|---|
| A new chat knows nothing about last week’s work | A new chat does not automatically include the earlier chat. Memory, if it is on, may carry some details, not all (Memory in ChatGPT). | The goal, what was decided, and the next step |
| A long chat starts contradicting its early answers | In my experience this can happen in a long chat; I don’t state a cause. For background, OpenAI says the context window is shared with system instructions and memories (if enabled) (Plans and Pricing). | The decisions so far, before the chat gets long |
| An instruction you gave stops being followed | In my experience this happens; I don’t state a cause. | The instruction itself, word for word, to paste again |
What to write down
Paste this five-line note at the start of a new chat. Keep it short enough to read in a minute.
Illustrative example (a made-up project):
Goal: a two-page plan for a small community garden's spring season.
Decided: raised beds only; no pesticide; budget stays under the grant.
Do not redo: the planting calendar (finished on Tuesday).
Open question: who handles watering in May?
Next step: draft the volunteer rota.Paste it back and ask the AI to say it in its own words
After the note, add one line:
Say this back to me in your own words, then list anything you are unsure of.If the answer gets a decision wrong, you are more likely to find out now, not three answers later.
Doesn’t Memory fix this?
OpenAI’s own page (checked 2026-11-02) says: “Memory does not retain every detail from every conversation. ChatGPT decides which available information is relevant to a response, and Memory can change as your context changes.” (Memory in ChatGPT). A note you write is a separate thing: you decide what is in it.
What this does not fix
A note does not make the AI correct. It can still misread the note or make mistakes, so check the output, and ask an AI for evidence of “done” before you rely on it.
If your project spans many chats
End each session with a seven-line handoff note, and start the next one with the five lines above. Mothex, the written method, is how I keep that up across a long project.
Get the free Handoff Kit (six plain-text files, no email)
ChatGPT is OpenAI’s product. Lizvo is independent and not affiliated with OpenAI.
— Lizvo
General information only, not professional advice. See the disclaimers.