ParentFull threadjohn-h-k·> a next token predictor can figure it outDescribing LLMs as "next token predictors" is disingenuous and wrongView on HN