Ask HN: Has ChatGPT suddenly become terrible at generating images?

It used to be that I could quickly whip up images, usually for demonstration or just silliness by saying something like, "A cartoon of a bunch of military aircraft at the bottom of the ocean with Poseidon rubbing his hands together and licking his lips while wearing aviator sunglasses."

And now it asks me 6 or 7 follow-up multiple choice questions. And then this time it said, "No more confirmations needed. Rendering the image will take a moment — I will deliver it in my next message." And 10 mins goes by. So I ask, "are you acutally going to?" And this triggers a "expectations reset" and it asks another four multiple choice questions.

The whole UX has just become worse than bad. It feels like they're meddling with trying to get more accuracy out of a system, and they're making the whole experience worse. Which I feel like is one of those classical times where some org obsesses over some KPIs and that obsession sends them entirely in the wrong direction. Perfect being the enemy of good and whatnot.