What's interesting is your input: the weights that you applied to the software to make it generate the text it did. When the author gives us an article generated by an LLM, we have to sift for the input – what the author actually wanted to say before the LLM built up the cruft around it – to find the gold.