I'm sad that it's crowding some of those things out of existence, not just out of being talked about.

Not limited to here, of course. Net-new publications to ArXiv for some (most?) CS subcategories are >=90% about models, transformers, training, quantization, or some other directly related field, or how to apply these towards a different specialty.

This is completely normal when a new thing is in the third section of the technology adoption curve.[0] AI will either go away (unlikely) or become a footnote in posts about what people are doing with it in the next stage.

[0] https://en.wikipedia.org/wiki/Technology_adoption_life_cycle

Alternately: the trough of disillusionment.

https://en.wikipedia.org/wiki/Gartner_hype_cycle