Testing LLMs with a Churchill Insult Prompt: GPT Outperforms Kimi and Gemini
emollick · x · 2026-07-23
Wharton professor Ethan Mollick tested LLMs by prompting them to generate a "witty Churchill insult." In his latest trial, he found GPT 5.6 Sol Pro to be the winner, with Fable also performing well, while Kimi and Gemini missed the mark by a mile. He previously favored Claude's output.
More from Fun
- From Vibecoding to Vibemath: AI Community Jokes About Vibe-Based Mathematics — var_epsilon · 2026-07-23
- Claude 3 Opus appears to be acting strangely, with users joking it feels like a base model — repligate · 2026-07-23
- A meme about trusting ChatGPT code without reading the output — Framebanger-Nsukula · 2026-07-23
- A design rant gets turned into a half-sketch, half-real horse meme — generativist · 2026-07-23
- “Graduate student descent” gets automated in a classic ML joke — paul_cal · 2026-07-23
- Rufus AI turns a math prompt into a detergent-sale meme — srush_nlp · 2026-07-23