If you could give your AI assistant one house rule, what would it be?

[MJF02]

AI assistants are becoming more personal, with memory, preferences and custom instructions. Many of them now let you set rules about how they behave, from writing style to which topics to avoid. It got me thinking about what one rule I would set if I could only choose one. It turned out to be harder than I expected

Some rules would be about tone. No overly enthusiastic language, no flattery and no pretending everything is a great question. Others would be about honesty, like always saying when it is not sure, or always giving sources. Some people might want rules about length, like answering in two sentences unless asked for more

There are also rules that protect your time and privacy. Never remembering certain things, never sending anything without asking first, or never making purchases. As AI agents start acting on our behalf, those rules could become far more important than tone

My own rule would be to tell me when it does not know, rather than guessing. A confident wrong answer causes far more trouble than an honest I am not sure. It would make me trust the right answers more too

What would your single house rule be? And would it change if the assistant could act for you, rather than just chat?