# @Sauers_ — 2025-09-09

♥134 ↻4 · https://x.com/Sauers_/status/1965474624910815332

The post is beginning to attract the people in the aforementioned subset https://t.co/czsYiBhrw3

(image G0bEnYFWEAAk_j7.jpg not yet published)

> transcription (photo):

# Tweet Transcription

**Sauers** ✓
@Sauers_

There's a subset of people who literally think LLMs just predict the most likely next token. I wonder if they ask themselves "if they're predicting the most likely next token, then why do they respond to me as a ChatBot Assistant™ instead of continuing the text, responding as me?"

tags: author:sauers_, has-image, kind:image, kind:tweet, thread-context, year:2025