Exploring LLM Free Association: Deliberate Typos and Omissions to Spark Ideas

yungcontent · x · 2026-08-24

A user shared a favorite AI use case: being intentionally obtuse and leaving out huge swaths of phrases or completions to see what dots the LLM tries to connect next. Noting that some users avoid even the smallest typos to save tokens, this user maximizes typos to induce "'07 Weezy levels of free association" from the model.

Original post →

More from Fun

Fun channel →