A starting point, not a prediction
A small test beats a big guess.
Napkinproof is run by Ad Astra Contracting Inc. in Toronto, Canada, operating as Ad Astra Connect. It helps first-time founders, students, and side-project builders turn a rough idea into an honest 48-hour demand test.
Four scores. Higher is better.
Problem clarity
5 = a specific person with a specific, frequent, painful problem they already spend time or money on.
1 = vague "people would like this".
Buyer reachability
5 = you can name where 20 likely buyers gather and reach them this week for free.
1 = buyers are hard to find or only reachable through paid ads or enterprise sales.
Cost to test
5 = a real test costs under US$20.
1 = needs inventory, licenses, hardware or over US$500.
Speed to signal
5 = a meaningful yes/no signal is possible within 48 hours.
1 = needs months.
What the total means
- 16 to 20: Test this weekend.
- 11 to 15: Sharpen first, then test.
- 4 to 10: Rethink before testing.
The server adds the four scores. Risk flags stay separate. They cover safety, permits and privacy, dependence on a platform, high capital needs, and ethics.
What this cannot tell you
An AI score is not proof of demand, a safety review, or a forecast of business success. The model can misunderstand your market or suggest a weak test. Check every claim, set your own thresholds, and treat conversations and real behavior as stronger evidence.
Financial, medical, and legal ideas receive a stop note with no score or test plan. Talk to a licensed professional or the relevant regulator first.
Why there are limits
The tool runs on a free daily AI allocation. We reserve at most 100 AI calls per UTC day, with a limit of 3 submissions per IP address per 10-minute window and 10 per day. Failed AI calls still use a daily slot. The offline checklist remains available when the AI is full.