Distill To A Short List
Cut 50+ business-model ideas to 3–5 candidates worth pressure-testing — without flattening every winner into 7/10 mush.
The Five-Gate Funnel · 50 Ideas → 1 Bet
Each Gate Kills The Wrong Ideas — Or You Kill It Later, More Expensively.
Filters are sequential, not weighted-averaged. A candidate that fails any single gate falls out — no 7/10 average can rescue a 3 on founder-fit or a 2 on defensibility.
Does the model match how this ICP actually buys? Kills self-serve dreams in committee-only markets.
Binary, not numerical. Insider insight on the model — distribution, monetisation, technical asset. Two of three or drop.
What's the moat at month 24, when ten copies have shipped on Cursor and Claude Code? Software is no longer the moat.
Time-to-evidence under 6 weeks? Margin clean once compute and Stripe fees are backed out? Bootstrap-survivable.
Gut written down before the rubric total is revealed. Would you still build this with no audience? That's the bet.
Walk The Day · Five-Year Commit Test
Imagine Building This Specific Model Every Day For Five Years.
Annie Duke (Quit, 2022): gut data is real, but degrades fast once contaminated by the rubric. Write each row before un-hiding the totals.
What does the first day of building this actually look like — calls, code, contracts?
Six weeks deep — is a falsifying loop firing, or are you staring at a sales pipeline that won't move?
Who are the first ten customers — and would you take that exact call every week for the next year?
No audience, no funding round to chase — would you still want this specific business in three years?