Principle
AI can generate a hundred product ideas before lunch. That does not make ideation free — it makes choosing expensive. The failure mode shifts from “we couldn’t build it” to “we built the wrong thing beautifully.”
Problems worth building share three traits: recurring pain (not one-off annoyance), reachable users (you can observe behavior, not just opinions), and honest economics (someone will pay or repeatedly choose your solution over the default).
The decision
DEC_004
Curiosity (wide)
│
▼
AI synthesis ──► hypotheses (written, falsifiable)
│
▼
Human contact ──► behavior + counts (not vibes)
│
▼
Problem bet (narrow) ──► build or killSignals that matter
| Signal | Strong | Weak |
|---|---|---|
| Pain | ”I do this every week and hate step 3" | "That sounds cool” |
| Workaround | Spreadsheet, hack, paid alternative | Nothing — they don’t care |
| Commitment | Time, money, intro to a colleague | Polite interest |
| Urgency | Deadline, cost, risk they name | Generic “eventually” |
AI is excellent at turning messy notes into a table like this. You still own the column assignments.
Workflow
- Write the problem before the product — one sentence: who, when, what breaks. No feature list yet.
- Run an AI pass on public sources: competitors, forums, job posts, reviews. Output: map of how people already cope.
- Talk to five humans who match the who — Mom Test rules: past behavior, not future promises.
- Falsify — what would prove you wrong in two weeks? If you cannot name it, you are still brainstorming.
- Promote or kill — one problem bet goes to validation (next chapter). The rest go in a graveyard doc with what you learned.
Tooling
Perplexity, Claude, NotebookLM for synthesis. Calendly and a voice memo for interviews. A spreadsheet beats a Notion board for counts.
Common mistakes
- Letting the model pick the problem because its pitch was articulate.
- Confusing TAM slides with evidence — big markets hide weak wedges.
- Building for “people like me” without checking they share the pain on a schedule.
- Skipping discovery because vibe tools make building feel like progress.
Artifacts
- Problem one-pager: who · when · pain · workaround · falsifier · kill criteria.
- Graveyard list: ideas killed and why — revision fuel, not shame.
Further reading
- Chapter 01 — Mindset (product vs engineering risk)
- Chapter 04 — Validation before code (when the bet is ready to test)