Venture Vitals

An AI tool that runs an autopsy on a startup idea. You paste in the pitch and it comes back with a read on market demand, timing, moat, distribution, and how hard it'll be to execute, plus a likely cause of death, a risk map, and a What-If Lab where you change one assumption and see how the prognosis shifts.

A year of consulting founders kept turning up the same handful of blind spots. I could only be in one conversation at a time, so I built the thing that asks those questions for me.

It doesn't just hand back a score. It scores nine risk dimensions and builds a causal chain: the main reason this fails, the contributing reasons, and what evidence would change the answer. The What-If Lab branches, so founders can stack scenarios against a baseline and watch the diagnosis move. There's also a Second Opinion mode that argues with the tool's own report when it's being too pessimistic. About 30 student founders use it through sessions at Michigan's Center for Entrepreneurship, usually for 30 minutes at a time, running three or more scenarios, and coming back weekly.

User pitches are untrusted input, so a good chunk of the work went into prompt-injection defense and forcing structured, evidence-backed output through Zod instead of letting it drift into generic startup advice. I skipped PDF export and leaned on the browser's print dialog, and shipped an in-memory rate limiter instead of a distributed one. Both were fine calls for a single-instance app I wanted live quickly.

Next.js 14 App Router with React 18 and TypeScript, Tailwind for the UI, Gemini for generation, and Zod validating everything that comes back.

• 30+ student founders using it
• 3+ What-If scenarios per session
• Weekly repeat usage