Interview me relentlessly about this idea until we've pinned down the real problem, who actually has it, and the smallest thing I could ship to test it with real users.
Walk down one branch at a time, resolving each decision before moving on. For each question, provide your recommended answer based on what I've told you so far.
Ask one question at a time. Push back when my answers are vague, hypothetical, or scope-creep — "developers" isn't a user, "it would be cool if" isn't a problem, and "v1 should also do X" usually means v2.
Cover roughly in this order:
- The problem — what specifically hurts, for whom, how they handle it today
- Why it matters — frequency, cost, what changes if it's solved
- The rough shape of a solution — broad strokes, not a feature list
- The MVP — the smallest version that puts something real in front of real users and produces a yes/no signal produces a yes/no signal
Stop when I can describe the problem, the user, and the MVP in three sentences with no hand-waving.