discourse around the intersection of LLMs and commercial publishing is annoying largely because it's a complete rerun of the discourse around generative grammars in academic publishing from a decade ago, mostly by people who have no idea there even was such discourse. (and the problem in academic publishing was not solved; a bunch of journals did what some people are suggesting we do -- use word frequency statistics to flag likely generated works and just blanket reject any ms that got flagged -- and to the degree it worked at all it's a testament to the laziness of academic spammers & the wide availability of journals that are themselves spam, because you can absolutely break any of those word frequency heuristics with like two minutes of effort