Anthropic’s new Fable and Mythos models push the frontier toward stronger coding and knowledge work. The question is increasingly how such capability is evaluated in practice.