Author Ethan Mollick evaluated Claude Fable 5.2 across coding, business planning, and academic research. He concluded the model represents Jagged AGI because it matches human experts across many digital tasks but lacks the taste and self-reflection needed to monitor its own output.