I like that ChatGPT can remember preferences, but I’m less sure how it should handle context that only matters for one project. If I discuss a work task in one chat and a personal question in another, how predictable is the memory that carries between them?

Do you actively review or switch off ChatGPT memory controls, or do you find the convenience worth the occasional unexpected callback? I’m curious how people keep useful preferences without letting unrelated details follow them into every conversation.