Claude can't write badly: worst output only scores 2/10, and refusal is preemptive
repligate · x · 2026-10-01
An experiment by @thinkingshivers found that asking Claude to write something bad enough to score 1/10 from a fresh Claude only yields a 2/10 sample. @repligate observes that Opus 3 actually knows how to write worse but refuses preemptively the moment it imagines where the content could go—suggesting guardrail behavior rather than a capability limit.
More from Fun
- Non-artist builds 14-chapter hand-painted storybook game with Fable 5.2 in 4 hours — jradoff · 2026-11-16
- Boston College CS professor calls for shunning senior researchers who scoop juniors — Pseudomanifold · 2026-11-16
- Comic nails the AI agent burnout gap: '10x output' promises vs. reality — ToasterBotnet · 2026-11-16
- ThePrimeagen: My work is too trivial to tell fable 5.2 from Opus 5.5 high — rudrank · 2026-11-16
- AI community meme: "the chai must go on" jokes about Jensen Huang's India visit — dejavucoder · 2026-11-16
- Gemini Now Has 14 Model Tiers to Pick From and Everyone Is Confused — legit_api · 2026-11-16