> there is no shortage of so-called studies about LLMs that inexplicably use comically old and bad models like gpt-4o to reach a strong conclusion about the ineffectiveness of LLMs in general

> one cannot help to suspect that this specific kind of academic malpractice does in fact have a point which is to undermine the technology as a whole out of fear, to keep the academic "consensus" in favor of tenured academics

Serious studies often take a lot of time to do. Also, models outdate very fast. So, of course nearly every serious study on such a topic will be based on models that are considered "outdated" by LLM aficionados.

So, no need to invent conspiracy theories that such studies are made to keep the academic "consensus" in favor of tenured academics.