The Mom Test interview kit for vibe coders — validate the idea before you prompt the build

Vibe coding broke the old excuse. "Validating takes too long — I could just build it" used to be lazy; now it's almost true. You can ship an MVP in a weekend, so why interview anyone?

Because the cost of building was never the real cost. The real cost is the three months you spend marketing, polishing, and emotionally defending a thing nobody wanted — and that cost didn't drop. If anything, faster building means you reach the "why is nobody signing up?" stage faster.

The fix is a handful of conversations before you prompt the build. The method below is adapted from Rob Fitzpatrick's The Mom Test, compressed for indie builders.

The Mom Test rule

The book's premise: your mom will tell you your idea is great, because she loves you. So will most people, because lying is easier than arguing. The rule that fixes this:

Talk about their life, not your idea. Past behavior, specific events, money and time actually spent. Never opinions, hypotheticals, or compliments.

Bad questions (all opinions about a future that costs nothing to endorse):

• "Would you use an app that does X?"
• "Do you think this is a good idea?"
• "Would you pay $9/month for this?"

Good questions (all facts about a past that actually happened):

• "When did this problem last come up? Walk me through what you did."
• "What are you currently using to deal with it? What did that cost?"
• "Have you looked for a better way? What did you try?"

The brutal heuristic hiding in that last one: if they never searched for a solution, the problem isn't painful enough to make them switch to yours.

The 15-minute call script

OPEN (1 min) — "I'm researching how people handle [problem area]. Not
selling anything, and I haven't built anything yet — I just want to learn
how you work."  (Saying you have nothing to pitch is what unlocks honesty.)

STORY (8 min) — "Tell me about the last time you dealt with [problem]."
Follow-ups: What triggered it? What did you do, step by step? What tools?
How long did it take? What happened because it went badly?
→ If they can't recall a specific recent instance, that's your data:
  the problem is rare or mild. Note it and move on.

COST & WORKAROUND (4 min) — "What's it costing you — time, money, stress?"
"What have you tried? Why did you stop / why do you stick with it?"
→ A janky spreadsheet they built themselves is the strongest buy signal
  there is. Existing spend > claimed willingness to pay, every time.

CLOSE (2 min) — "Who else struggles with this more than you?" (referrals
to people with stronger pain) — and nothing else. Do not pitch. The urge
to pitch at minute 13 is overwhelming. Resist it; you'll contaminate
everything they said before.

The synthesis template

After each call, fill this in — one screen, five minutes, while it's fresh:

INTERVIEW #__ — [role / segment, date]
Specific instance described:        yes / no / vague
Current solution:                   [tool, workaround, or "nothing"]
Actual spend (time or money):       [be precise]
Sought alternatives on their own:   yes / no
Emotion when describing problem:    annoyed / resigned / on fire / neutral
Strongest quote (verbatim):         "..."
Surprise (something I didn't expect):
Verdict for my assumption list:     supports / contradicts / neutral

After five interviews, the verdict column reads like a chart. The decision rule:

• 4–5 contradicts → kill or pivot the idea. Cheapest "no" you'll ever get.
• Mixed → your segment is probably wrong, not the idea. Re-target, run three more.
• 4–5 supports, with real spend and self-sought alternatives → build. And now you have verbatim quotes for your landing page and a list of people waiting for the link.

Where AI actually helps

Not in having the conversations — those need to be real humans with real pasts. But before and after:

• Before: have an AI red-team your assumption list ("what must be true for this idea to work? rank by riskiest") and generate your question set, checking each question against the Mom Test rule — it's good at catching "would you..." leaks.
• After: paste your raw notes and have it extract patterns across interviews — recurring workarounds, segment splits, the quote that keeps repeating in different words.

Five conversations, one weekend, zero code. Then prompt the build — for something people have already told you they're paying for in time, money, or duct-taped spreadsheets.