Claude Fable 5.2 evaluated for Jagged AGI

The author explores Anthropic’s Claude Fable 5.2 model to evaluate whether it represents an early form of Artificial General Intelligence, which he terms "Jagged AGI." Testing the model across challenging domains like hardware hacking, business launching, literary recreation, academic research, and 3D world building reveals impressive expert-level capabilities alongside persistent gaps in taste, judgment, and self-correction. Open article →