Brainstorm

Generate a set of concrete product ideas from a problem space using structured ideation techniques.

Instructions

1. Parse the Prompt

$ARGUMENTS is a problem space, goal, or domain. Not a solution.

Good inputs:

• reduce time from signup to first value in B2B SaaS
• help developers understand production errors faster
• make team retrospectives actually useful
• authentication for the next 5 years

If $ARGUMENTS sounds like a solution ("build a better Jira"), reframe it: "I'll rephrase that as a problem space to get better ideas: 'help software teams track work with less friction'. That will generate more diverse ideas."

2. Context Scan (2 min)

Quickly read the existing project (if in one) to understand:

• What does the product already do?
• Who are the users?
• What's already been built / evaluated / rejected?

Check docs/ideas/, docs/evals/ for existing ideas to avoid duplicates.

3. Apply Ideation Frameworks

Run all 4 frameworks. Each produces different kinds of ideas.

Framework 1: Jobs-to-be-Done (JTBD)

Ask: "What job is the user hiring a product to do in this space?"

Format: "When [situation], I want to [motivation] so I can [expected outcome]."

Generate 3-5 distinct jobs. For each job, ask: "What's the minimum product that would nail this job?"

Framework 2: How Might We (HMW)

Reframe the problem space as a set of "How Might We" questions. Each HMW opens a solution direction.

Problem: developers struggle to understand production errors
HMW... make the error context self-explanatory?
HMW... bring relevant code directly to the developer instead of the reverse?
HMW... make errors impossible to ignore until resolved?
HMW... let the system diagnose itself before paging a human?
HMW... make fixing a repeated error take one click?

Generate 5-7 HMW questions. For each: sketch one concrete product response.

Framework 3: Analogy Thinking

"How do other domains/industries solve an analogous problem?"

Find 3-5 analogous problems in different industries. For each, extract the core mechanism and apply it.

Problem: too many alerts, alert fatigue
Analogous domain: Email → Gmail Priority Inbox (ML-ranked importance)
Applied: Smart alert triage that learns which alerts actually need action
→ Idea: ML-prioritized alert routing

Analogous domain: News → Morning Brew (curated digest instead of firehose)
Applied: Daily digest of most important alerts instead of real-time noise
→ Idea: Async alert digest

Framework 4: Constraint Reversal

Remove a constraint that everyone assumes is fixed. What becomes possible?

Problem space: B2B onboarding
Assumed constraints:
- "Onboarding must be done by a human sales rep" → What if it was fully automated?
- "Onboarding takes weeks" → What if first value in under 10 minutes?
- "Users need training" → What if the product trained itself to the user?
- "Pricing is decided before sign-up" → What if pricing emerged from actual usage?

For each reversal: one concrete product idea that exploits the removed constraint.

4. Score and Filter

From all frameworks (15-25 raw ideas), select the top 7 by:

• Novelty: meaningfully different from existing solutions?
• JTBD fit: solves a real, specific job?
• Feasibility: achievable without 10-person team?
• Market signal: any evidence users want this? (even anecdotally)

Remove ideas that are:

• Minor features of an existing product (not product ideas)
• Already well-solved (high-quality existing solution exists)
• Exactly the same as something in docs/ideas/ or docs/evals/

5. Create Idea Seeds

For each of the top 7, create a brief idea seed with:

• The core insight (one sentence: the non-obvious thing)
• The job it solves (JTBD format)
• What's different from existing solutions
• The riskiest assumption to validate

Save to docs/ideas/discovered/YYYY-MM-DD-brainstorm-<name>.md.

6. Output

BRAINSTORM: <Problem Space>
════════════════════════════

Frameworks applied: JTBD, How Might We, Analogy, Constraint Reversal
Raw ideas generated: <N>
Top ideas selected:  7

TOP IDEAS
──────────────────────────────────────────────────────

1. <Idea Name>
   Core insight: <the non-obvious thing>
   Job solved:   "When [X], I want to [Y] so I can [Z]"
   Different:    <what no existing solution does>
   Risky assumption: <what must be true for this to work>
   → /idea <kebab-name>

2. [...]

[...]

Idea seeds saved: docs/ideas/discovered/ (7 files)

Suggested next step:
  Run /evaluate <name> for each idea you find compelling.
  Or: /discover <most promising idea's domain> to validate with market research.

Arguments

$ARGUMENTS examples:

• reduce developer on-call toil — pain-focused
• B2B SaaS billing that doesn't suck — outcome-focused
• authentication post-password — trend-focused
• make code review faster and higher quality — process-focused
• (empty) — ask for the problem space

After This

• /idea — capture the best idea for evaluation
• /evaluate — evaluate the idea for Go/No-Go