Structured Brainstorming

Iron Law

DIVERGE BEFORE YOU CONVERGE

Phases

1. Diverge

Generate 10+ ideas without filtering. Quantity over quality at this stage.

• Write every idea down, no matter how wild
• No criticism, no evaluation, no "that won't work"
• Build on previous ideas — combine, extend, invert
• Include at least 2 unconventional or boundary-pushing ideas
• Goal: fill the possibility space before narrowing

2. Cluster

Group ideas by theme or approach.

• Identify natural groupings (e.g., "UI improvements", "backend optimizations", "new capabilities")
• Label each cluster with a descriptive theme name
• Note ideas that span multiple clusters — these are often the most interesting
• Look for gaps: are there obvious clusters with no ideas?

3. Evaluate

Score each idea on three dimensions (1-5 scale):

Dimension	1	5
Feasibility	Requires new infrastructure, unknown tech, or months of work	Can be built with current stack in days
Impact	Marginal improvement, few users affected	Transformative for core use case, affects all users
Effort	Large team, long timeline, high coordination	Solo developer, short timeline, self-contained

Calculate a composite score: (Feasibility + Impact) - Effort

Rank all ideas by composite score.

4. Converge

Select the top 3 ideas and for each provide:

• Summary: One-sentence description
• Why this one: What makes it high-value
• Risks: What could go wrong or block progress
• Next steps: 2-3 concrete actions to move forward
• Dependencies: What needs to exist before this can start

Red Flags — If You Catch Yourself Thinking:

Thought	Reality
"We only need 3-4 good ideas"	You need 10+ raw ideas to find 3 good ones. Diverge FIRST.
"That idea is unrealistic, skip it"	Write it down anyway. Unrealistic ideas inspire realistic ones.
"I can evaluate while brainstorming"	Evaluation kills creativity. Diverge and evaluate are SEPARATE phases.
"The first idea is obviously the best"	The first idea is the most obvious, not the best. Keep generating.

Rules

• Generate 10+ ideas before evaluating ANY of them
• No criticism or filtering during the Diverge phase
• Give every idea equal initial weight — do not pre-rank
• Output a ranked list with scores, risks, and next steps for the top 3