It’s iterative. I pasted the same prompt/code to gpt4, and said the following: „it removed polish characters from the text!”

It correctly figured out the problem and fixed it - removing only unicode emojis, not all the unicode characters.

Then I told him: „but it still didn’t remove things like :) :/ xD” (not specyfying that I mean non-ascii emoticons, not emojis really).

It replied by saying I meant emoticons and writing regexp for those and other ascii emoticons.