> extreme tells It's not a "tell" if it has no predictive value. Most things popular in output are that way because they were already popular--or possibly hack

> extreme tells It's not a "tell" if it has no predictive value. Most things popular in output are that way because they were already popular--or possibly hackneyed--in inputs. I'm pretty cynical when it comes to LLM overhyping, but mimicking forms and styles written text is one of the things they actually do well. Even if you see something odd like "adumbration" (vague foreshadowing) you're still stuck distinguishing between (A) a human trying to sound a certain way and reaching for a thesaurus vs (B) someone setting up an LLM prompt to try to mimic the first kind of person.