Task Imagination: The Skill Big Models Like Fable 5 Demand

Key Takeaways

• The bottleneck flipped. With a big model, the limit Jones kept hitting wasn't the model running out of ability — it was running out of big things to ask. If your asks stay prompt-sized, every frontier model feels roughly the same.
• "Task imagination" ≠ delegation. Delegation is tasks already on your tracker with a name attached. Task imagination is the dirty, ambiguous jobs nobody has written down — de-duping 2M CRM records, fact-checking a 500-page board packet, mining 40,000 reviews — because they felt too big to assign.
• Define "done" first. Write a clear paragraph describing what should exist at the end, assemble a data pack (this can take hours), hand it over — and then do the hard part: walk away and stop hovering. The itch to babysit is a habit trained on models that used to be too small.
• Review like an owner. When the work comes back, check it like a senior stakeholder's output: is the scope right, is it accurate, is it angled correctly — then assign revision work as needed.
• Reserve big models for big jobs. At ~$50 per million output tokens, Fable 5 isn't a daily driver. Spending that on a summary a cheap model could do in seconds wastes the muscle — use it for serious work where one job can save weeks.
• On jobs: the roles most exposed are pure, zero-judgment execution. For everyone else, the shift is toward being a "model manager" — scoping, feeding data, and judging output — which Jones frames as a career-uplift opportunity, not a layoff sentence.