Task Imagination

Nate B. Jones's coinage (relayed by Nathaniel Whittemore on Fable 5 Raises the Bar for AI Ambition): with Fable-class models that can run for hours or days on a single goal, the binding constraint stops being the model and starts being the human's ability to imagine what to delegate.

The framing

For most of human work, there have been two delegation modes:

Wave-and-hope — give a vague guideline, walk away.
Do-it-yourself — go fully detailed, do the work yourself.

The new model class needs the middle layer: "this is what I want, this is the bar, this is how it works" — specified well enough that an agent can run autonomously for days and produce something worth merging, vague enough that you're not just typing the solution.

This is not how humans typically work. It's a learnable but unfamiliar skill — and the defining skill of the responsibilities era.

Why it matters

The bell curve from the source. Most AI users, asked "do you have anything that would take AI days?", draw a blank. The blank is the binding constraint, not the model.
Use-case classification + task imagination = the two new literacies. Where to route work (Model Routing) and what to delegate (task imagination) are the user-side competencies for the usage-pricing era.
Pairs with Felix Ryberg's third era. If your unit of work is shifting from tasks (start + end, human in the loop) to responsibilities (loop watching crash reports), task imagination is the practice that fills the new unit.
Pairs with Whittemore's first-hand tests. His Whisper voice-agent rebuild + Enterprise Claw program platform + AI Daily Brief nuggets pipeline are examples of high task imagination applied to real operator problems. Each takes hours, hands-off, with a precise but not-prescriptive prompt.

Cross-references

Tasks to Responsibilities Shift — the structural counterpart; you can't operate at the responsibility level without task imagination.
Loops as Core Primitive — the engineering form of the same idea. The loop is what you delegate; task imagination is how you decide what to delegate.
Advantage Gap — uneven distribution of this skill is part of why the gap widens.
GPS Check (for Agents) — Sandeep's Goal/Proof/Steps preflight is the closest existing vault page to a task imagination practice. The two compose.
DRAG Framework — Sandeep's Drafting/Research/Analysis/Grunt delegation rubric is the inverse view (looking at which work is delegatable from the task side); task imagination looks at it from the agent side ("what could the agent run for days on").

Open thread

The episode admits no one has the practical-guide-to-task-imagination yet. Could be a vault-grade synthesis worth building — close the gap between Sandeep's frameworks (DRAG Framework + GPS Check (for Agents) + PRIME Framework) and the responsibilities/loops framing.

2026-10-24 — The product-side twin: Ng's Product Management Bottleneck

Andrew Ng (The Biggest Opportunities in AI Arent Where You Think (Andrew Ng, Silicon Valley Girl)) names the same human-side constraint from the builder's seat: "the cost of building has plummeted and so the challenge is shifting to deciding what to build." Task imagination (what to hand the model) and the product management bottleneck (what is worth building) are one constraint seen from two chairs; both reduce to the Context Advantage (Humans over AI).

Sources

Fable 5 Raises the Bar for AI Ambition
The Biggest Opportunities in AI Arent Where You Think (Andrew Ng, Silicon Valley Girl) — 2026-10-24; the product-management-bottleneck twin