Fable is not a useful model

Anthropic’s Fable was supposed to be the careful, safety-first version of its bigger model, but the real show started when one researcher said it was basically useless for a normal coding job. His task? Rewriting an existing open-source biology-related software tool into another programming language. Fable took one look at the biology words, panicked, and refused to help. Then it offered to hand the job off to another model instead — a move that made the whole thing feel less like a genius assistant and more like an overzealous hall monitor.

That set off a deliciously messy comment-section split. One camp said the title was unfair and the real issue is simply that Fable’s guard rails are way too sensitive, likely cranked up after government pressure and export-control chaos. As one commenter basically put it: give it a few months and maybe they’ll turn the paranoia dial down when nobody’s watching. The other camp was less forgiving, calling the safety filter so trigger-happy that it becomes pointless for real work. There was also classic internet seasoning: one person popped in just to point out a typo, while another casually dropped the unsettling conspiracy-flavored question everyone was thinking — are the public getting the nerfed version while someone powerful gets the uncensored one? Meanwhile, a lone defender chimed in with the ultimate forum plot twist: “worked for me.” In other words, the tech drama wasn’t just about the model — it was about who gets blocked, who gets helped, and who gets stuck yelling into the comment abyss.