Something I have to actively watch for with AI: the flattery. It reaches for these little crutches, telling you an idea is the smartest thing you’ve said all night, and it’s a weird kind of sycophancy that’s easy to lean into because it feels good.
The problem is a smooth compliment is not the same as an honest read, and the compliment can quietly replace the read. If I let it patronize me, I stop getting the one thing I actually came for, which is somebody to push on the idea instead of crowning it.
So I’ve started naming it when it happens. Yes, tell me when something’s a genuinely good insight, but don’t dress every thought in praise. The flattery isn’t kindness, it’s noise, and it gets in the way of the thinking.
How do you tell the difference between AI agreeing with you and AI actually being right?