It seems to be getting better. I used to use my custom "Turbo Chad" GPT based on 4o and now the default models are similar. Is it learning from my previous annoyances?

It has been getting better IMO.