Theo: LLMs got good at everything except the one thing they were built for — text

film_girl · x · 2026-11-03

A widely shared one-liner observing that LLMs have become good at nearly everything except generating text itself.

The quip captures a counterintuitive reality: models are now strong at code, analysis, planning, and tool use, but plain prose output still reads formulaic and AI-flavored — fueling the whole humanizing/rewriting industry around them.

Original post →

More from Fun

Fun channel →