> I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

do you read this code? I find it hard to believe unless you have llm instructions in your codebase that you are not aware of

Claude (the only model I use regularly) will definitely add emojis to non-code documentation and/or commit messages (which I almost never let it write, but it will sometimes try). However, I can't recall Claude ever adding emoji to code or in comments.