ideate

Ideate

A four-phase process for producing ideas worth having, not just the first three that come to mind. The enemy here is premature convergence — anchoring on the obvious, judging while generating, and mistaking variations of one idea for a real range. The phases deliberately separate diverging from converging and put an adversarial coverage check between them. Work them in order.

Phase 1 — Frame (Socratic)

Before generating, understand what a good idea even means here. Ask sharp, batched questions:

• The real problem or opportunity. What's the actual job to be done? The stated brief is often a solution in disguise — find the need underneath it.
• What would make an idea valuable in this context — and what would make one useless? This is the filter for Phase 4; name it now.
• Constraints — and which to suspend. Note the hard limits, but identify which constraints are worth deliberately ignoring during divergence to avoid strangling ideas at birth.
• What's already been tried or considered, so you can push past it rather than re-surface it.

Confirm the framing in one line, then diverge.

Phase 2 — Diverge (generate wide)

Generate a lot, and generate varied. Defer all judgment here — no idea gets criticized in this phase. Force range by working through genuinely different lenses, e.g.:

• The obvious ones — get them out so you stop circling them.
• Different angles — solve it from the user's view, the cost view, the speed view, the opposite view.
• Borrowed solutions — how has an adjacent field, product, or era solved a structurally similar problem?
• Inversion — what would guarantee the opposite outcome? Invert it.
• Constraint flips — what if you had 10× the budget? Zero budget? One day? No existing system?
• Combinations — fuse two earlier ideas into a third.
• The deliberately wild — at least a couple that feel too bold. They're seeds, not commitments.

Aim for quantity and spread. Group loosely as you go so the landscape is visible, but don't prune yet.

Phase 3 — Challenge (adversarial coverage check)

Attack the set, not to kill ideas but to expose what's missing and what's fake range:

• Anchoring. Are most of these variations of a single underlying idea wearing different hats?
• Missing categories. What whole class of solution isn't represented at all? Name it and generate into the gap.
• Fake novelty. Which "ideas" are just the status quo rephrased? Mark them.
• The unexamined obvious-best. The idea that feels like the clear winner — why might it be wrong, or why has nobody done it already? Pressure it.

Generate another round to fill the gaps this surfaces. Good ideation usually has a second wind right when it feels tapped out.

Phase 4 — Converge & deliver

Now judge. Against the value criteria from Phase 1:

• Cluster the full set into themes.
• Surface the most promising few — say why each is strong, and what it'd take to test it cheaply.
• Keep the wildcards worth not losing, even if not chosen now.
• Name the tension, if the strongest ideas trade off against each other.

Deliver the shortlist plus the full set, and offer to save it (e.g. ideas-<slug>.md). The goal of convergence is a small number of ideas worth doing something about, not a ranked list of everything.