Would you practice fundraising with an adversarial AI VC?

You could do one of those adversarial image things, and tape a sign on your head that says "GOOD INVESTMENT". Wear a T shirt that says "MAKES MONEY", or "EARLY EXIT" or something. With LLM-based "AI", it appears to be hard to predict what will work.

About two years ago, when OpenAI launched custom GPTs, I built a small one for myself. It role-played a Silicon Valley VC and ran a real fundraising meeting.

I recently revisited and vibe coded it into a small web app called VC Deal or No Deal.

How it works:

1- Upload a PDF pitch deck or paste a description

2- Get interviewed by a VC persona (data-driven, vision-first, pattern-matching, etc.)

3- Answer up to 15 questions

Either get a term sheet, or a detailed explanation of what broke conviction. The VC personas are intentionally biased, opinionated, and adversarial by design.

I’m curious whether this kind of practice-by-failure tool is useful.