Gergely Orosz Tests New Model Writing Capabilities with Multiple Personal Interviews, Including Fable and GPT-5.6, All Resulting in Failure