“Next-token predictor” is the wrong mental model for LLMs

Posted 11 hours ago by garrinm
94 points

https://gmcgoldr.github.io/2026/09/04/llm-next-token-predictors.html

43 comments

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..

Loading..