Paste your idea. Watch it run a knockout gate, get scored in confidence-weighted ranges, survive its own strongest kill case, and reach a make / hold / kill verdict — live.
A strong score on one axis quietly cancels a disqualifying one on another. MakeOrKillIt separates fatal from merely weak — instead of summing them.
One number per dimension, collapsed to a mean. A 9 on market hides a 1 on legality.
The fatal flaw is caught first. Survivors are scored in confidence-weighted ranges.
Five must-meet criteria. Any clear failure kills the idea before it's ever scored.
Each dimension gets a range whose width is set by confidence. You review low-confidence calls.
Where the aggregate band sits — and how wide it is — maps to make, hold, or kill.
Guess a market size instead of reading it, and confidence drops — so the band widens and the verdict leans toward hold. Uncertainty is priced in, not hidden.
Weights sum to one — a starting point the calibration loop retunes over time.
Any clear failure is an instant kill. Genuinely unclear? The gate asks you.
Buildable lawfully and ethically at all.
A meaningful floor of users to serve.
The core build is reasonably possible.
Fits the stated mission and strategy.
You can reach the minimum resources.
It builds the strongest case for killing your idea.
For each dimension the model names the single strongest disconfirming evidence; for the whole idea it writes the most convincing kill case it can — whether it believes it or not. Your best defense against your own optimism.
low ≥ 65 — the worst realistic case is still good enough.
Wide → reduce uncertainty. Narrow but middling → reshape.
high ≤ 40 — the best realistic case still isn't worth it.
Paste the problem, the user, and how it earns. Watch the verdict resolve live.