TheQwertiest
I think there are blunt solutions to the blocking mechanism that haven't really been considered here, but could be implemented with ease if anyone on the team had the will/ability.
The most blunt option is one I've seen on some imageboard sites where they use css-based blocking combined with dumping a list of categories and users into the tags- you could just have it hide images based on your blocklist. It would mess with your notification count, but at least you wouldn't have to see it.
An ai free front page would probably have to be a policy decision unless they really wanted to curate an ai and ai-free list. I don't know how they set that up but surely it's more a lack of will than a major issue. It seems like the site owners are thoroughly on board with ai image generation and don't much care about people who are put off by it.
To actually solve the notifications thing would be a bit harder, depending on the backend architecture. I haven't bothered to investigate what they're using, but my guess would be that it just sends out broad alerts based on the act of labelling an image with a tag, which fits into the way that tags are added now anyway. Depending on what kind of data store they use, querying to limit the users to those who allow specific kinds of tags would probably be a real hassle in terms of the speed and computing costs. I'm guessing this is mostly a shoestring operation.