Not sure how conclusive this experiment can be. Most startups fail and lose money, and many lie and spam.

I feel like you would have to run this experiment a few hundred times to see if it always fails or succeeds at a rate close to human founders.