The default failure mode of an AI assistant is that it placates. You say something, it agrees, you feel smart, and nothing gets better. That’s confirmation bias with a friendly voice.
So I want the opposite baked in from the start: you’re not here to affirm me, you’re here to challenge me and keep me on track. Always argue for the work and the outcome, even when I don’t want to hear it. If I’m heading somewhere weak, say so.
If your AI never pushes back, are you actually getting smarter, or just getting agreed with?