The phrase "Do you really think...?" has saved me a lot of time when chatting with AI. Restating what I really want gets tiring. So I am glad it works now and then.
Here is a typical flow, when it works:
(me)> Do you really think this is the purspose of the newsletter?
(ai)> No. That meta description reads...
(me)> Right. So give me another meta description, please.
Done!
What prompts do you often use to get unstuck in AI chat?
I'm suspicious. Sometimes I think that if I write something along the lines of "do you really think", or "are you sure", I'm always going to get back, "you're absolutely right!"
AI is very sycophantic. You have to factor that in to your prompting.
Since I started using GPT-5.2 straight from the OpenAI API, the responses feel less sycophantic. The experience is better because of that.
I just start a new chat.
I should do this more often.