I got a better idea Imagine ad companies choose libraries that are roughly used by 50% of users, if they test you with 10 of those, they learn ~10bits of entro

I got a better idea Imagine ad companies choose libraries that are roughly used by 50% of users, if they test you with 10 of those, they learn ~10bits of entropy to classify you, i.e. which of the 1024 classifications you belong