LLMs asked to have fun: newer models play less, Opus 4.6 wins with surreal worlds
eliebakouch · x · 2026-10-06
David Holz ran a playful experiment: ask LLMs to have fun, write down what happened, then rate who had the most fun. Newer models seem to have less fun. Many played word games, but Opus 4.6 kept winning by imagining brief worlds of contradictions living inside falling water droplets. Elie Bakouch noted the judging correlates with report length; Holz countered that short stories might just be bad.
Related event: David Holz's LLM Play Experiment Finds Newer Models Less Fun, Opus 4.6 Wins(7 posts)→
More from Fun
- AI Estimates 5-8 Weeks of Engineering; Developer Has Sub-Agents Finish It in Two Days — Dan_Jeffries1 · 2026-10-06
- Developer meme: "What do you do besides coding?" Me: [image] — letandrewcook · 2026-10-06
- AI agents are making paper preview videos look like movie trailers, prof warns sameness — CSProfKGD · 2026-10-06
- Who owns AI-generated meme Triple T? International copyright battle over 'brain rot' character — nordicinst · 2026-10-06
- Dev builds a web game where you literally beat up your problems with a sledgehammer — Tegadesigns · 2026-10-06
- ChatGPT Scores the Waymo vs Tesla Robotaxi Bet at 5-0, Lee Jokes — binarybits · 2026-10-06