I think it's extremely important to distinguish being friendly (perhaps overly so), and agreeing with the user when they're wrong
The first case is just preference, the second case is materially damaging
From my experience, ChatGPT does push back more than it used to