Your AI Forgets Everything Between Chats
Ten conversations in, it still asks who you are. Every new chat starts blank — the context you built up over an hour is gone the second you close the tab.
This is not a bug you can prompt your way out of. Each new session has no memory of the last one, no matter how much you explained last time. You are not imagining the deja vu: you really are re-explaining your project, your constraints, your tone, from zero, every single time.
The fix is not a better prompt. It is a saved one.
Stop trying to fix this by writing a longer message each time. Write it once, save it, and paste it first — before the actual question.
Here is the template:
``` Here's my standing context: [what you're working on], [your constraints or non-negotiables], [how you like answers formatted]. Treat every message in this conversation as if you already know this — don't ask me to repeat it. ```
Fill the three brackets once. Keep that block somewhere you can grab it in two seconds — a pinned note, a text file on your desktop, whatever you already open ten times a day.
Where to put it so you stop pasting it
Most tools now have a place for this that survives between chats, so you paste it once instead of every time:
- **ChatGPT**: Settings → Personalization → Custom Instructions, or start a Project and drop it in the Project's instructions.
- **Claude**: create a Project, put it in the Project's custom instructions — every chat inside that project inherits it automatically.
- **Gemini**: Saved Info, in Gemini's personal context settings.
If your tool has none of these, the manual paste still works — it just costs you three seconds instead of zero.
What this actually buys you
Not a smarter model. A model that stops making you re-teach it the same five facts every session. The quality of the first answer in a new chat goes up because it is not guessing at context you have not given it yet — it already has it.
Twenty minutes a week rewriting the same explanation adds up to sixteen hours a year. Write it once.
Want the calm version of AI news like this, once a week? Subscribe to the Sharp AI Hub newsletter →