A short-lived test of Fable 5 becomes a useful lens on a bigger AI question: when several models, access tiers, and cost choices sit behind one interface, the prompt is only part of the result.