I am not sure on what I'm going to say, but I think that LLMs are a technological dead end. They might get some use now, but eventually the industry will shift towards better models for machine text generation. And, if those models rely on a tiny corpus of hand-reviewed data, instead of shoving down as much text as possible into the model (the first "L" in "LLM" is "large"), then Reddit posts/comments will become outright useless.
In other words: Reddit is degrading further the trust of its userbase, and it might not even get much in return.