The Six Tools

BIG Tools walks an idea through six stages. Each stage has a distinct job. Knowing what the tool is trying to do — and what it can't do — is the difference between a useful pass and a frustrating one.

1. NGT — Next Great Thing

Job: Generate a divergent set of ideas, fast.

Given your seed insight, NGT produces up to 5 candidate directions. They shouldn't all be coherent with each other — divergence is the point. Promote the ones that surprise you or that you want to argue with. Drop the obvious ones.

Working well: you're getting candidates that span the problem space — some conservative, some wild, some adjacent.

Not working: every candidate sounds the same. Try rerunning with a different LLM, or re-seed with a more specific observation.

See the recipe: Recipe: Next Great Thing (NGT)

2. Reframe

Job: Turn a promoted candidate into a How-Might-We question — a sharpened problem statement.

Good HMWs are specific enough to act on but open enough to invite multiple answers. "How might we help first-year students find their writing voice?" is a good HMW. "How might we improve education?" is not.

Working well: the HMWs scope the problem without prescribing the solution.

Not working: the HMWs read as solutions disguised as questions. Re-run and ask for more divergent phrasings.

See the recipe: Recipe: Reframe

3. Filter

Job: Pressure-test each HMW against your audience, context, and constraints.

Filter asks you to describe the people you're designing for (the Purpose Profile), then evaluates each HMW against what you said. HMWs that don't serve the people you're actually trying to help get flagged. The ones that survive are the ones worth carrying forward.

Working well: Filter is catching HMWs that sound good but fail the audience test.

Not working: everything passes. Your Purpose Profile might be too vague. Re-enter a sharper description of who you're designing for.

See the recipe: Recipe: Filter

4. Steal

Job: Borrow proven patterns from adjacent fields.

For each surviving HMW, Steal suggests patterns from domains where something similar has already been solved. Think: deadman switches from aviation, carbon copies from accounting, torque limits from engineering.

Working well: you're seeing patterns you wouldn't have reached for on your own.

Not working: the borrowed patterns feel forced. Drop them — Steal is suggesting, not insisting.

See the recipe: Recipe: Steal like an artist

5. Mixtape

Job: Curate the survivors into a coherent set.

By Mixtape, you have a scattered collection of HMWs, constraints, and borrowed patterns. Mixtape asks you to pick a unifying theme — a short name or metaphor that holds the survivors together. This isn't the pitch yet; it's the organizing principle.

Working well: the theme ties the surviving candidates together without flattening their differences.

Not working: the theme is generic ("design", "improvement"). Push for something more specific.

See the recipe: Recipe: Mixtape

6. Pitch

Job: Write the narrative you'd share with someone else.

Pitch takes the Mixtape and produces a summary statement — the crisp version of what you've found, why it matters, and what you'd do about it. This is what you share, defend, or carry into an implementation conversation.

Working well: someone who wasn't in the workshop can read the Pitch and understand the idea.

Not working: the Pitch hedges, restates the HMW, or reads as a list of features. Re-run with a different LLM, or edit the Mixtape and try again.

See the recipe: Recipe: Pitch

Also worth knowing: Recipe: Paper House — the non-LLM self-check you run on the final pitch to make sure it can stand on its own.

How the tools talk to each other

The tools aren't independent. A weak Reframe produces a weak Filter. A too-vague Purpose Profile in Filter produces a too-permissive Steal. If a late stage feels off, the problem is almost always upstream. Re-open the earlier stage and revise.