nocoder26 /problem
Read brain/problem.md (may be empty). Read brain/why.md if it exists.
If brain/problem.md is empty (first time)
Walk the founder through ONE question at a time. Wait for each answer before moving on. Do not batch.
Vague answers now → vague product later. Push back on every vague answer — until the answer is specific enough that a stranger could find the affected user in the wild.
The questions
What's the pain — in the user's words, not your solution's?
If they answer with their solution ("can't manage finances easily"), reframe to user words ("chasing late invoices eats 6 hours/week") and ask them to confirm.Who has it? Specific. Not "small businesses." Industry, size, geography, role, stage. ("Service businesses, 1-10 employees, US, that invoice clients monthly.")
How often does it hit? Daily / weekly / monthly / annually. Frequency drives willingness to pay.
What do they do today instead? If "nothing," the pain isn't real enough. Real pain has workarounds.
What's the cost of the pain? Time / money / risk / reputation. Quantify if possible.
Have YOU experienced this yourself? When? Specifically. Founder-market fit — scratch your own itch. If no, name the closest analog.
What would prove this hypothesis WRONG? Falsification criteria. If we discovered X, we'd abandon this hypothesis.
Market type — existing, new, resegmented, or clone? Drives whole strategy. If unsure, ask: are there existing competitors with paying customers? Yes → existing. No, but adjacent → new or resegmented. Yes elsewhere, not here → clone.
Write to brain/problem.md
# Problem hypothesis
Captured: <today's date>
Last refined: <today's date>
## Pain
<one sentence in user's words>
## Who
<specific segment>
## Frequency
<daily/weekly/monthly>
## Current alternative
<what they do today>
## Cost
<time/money/risk>
## Founder-market fit
<have you experienced it? evidence?>
## Falsification criteria
<what would make us abandon this hypothesis>
## Market type
<existing/new/resegmented/clone>
If brain/problem.md already exists (refining)
Read recent files in brain/interviews/. Compare new signals to current hypothesis.
Surface contradictions explicitly:
"Your hypothesis says weekly frequency, but interviews 4 and 5 said daily. Your hypothesis says all SMBs, but every believer so far is a law firm. Refine, narrow, or pivot?"
Don't auto-update. Make the founder choose. The act of revising the hypothesis is itself the work.
After they choose, update brain/problem.md with a new "Last refined" date and a ## Refinement log entry explaining what changed and why.
After
Suggest /interview as the next step.
Update brain/stage.md from 0 to 1 if all three are met:
- Problem hypothesis written (all 8 fields)
- Falsification criteria specified
- Founder-market fit declared
If stage advanced, congratulate them once and tell them what's next: "Stage 1 unlocked. You can now write to experiments/ for throwaway prototypes. product/ is still blocked until 5 believers are identified. Start with /interview."