LLMese criticized: Stop over-apologizing and just answer the question

paul_cal · x · 2026-08-23

A post calls out the common "LLMese" style where models over-explain potential inaccuracies instead of providing direct answers. Comparing it to constantly telling a boss "I'll give you real numbers, not just plausible guesses," the author highlights the awkwardness of AI's tendency to act like it's confessing sins rather than being efficient, pointing out a recurring issue in AI text generation patterns.

Original post →

More from Fun

Fun channel →