I get the idea but I don't think it can work; say you make a list from 0 - 10 million; now that's only 23-24 bits of entropy, but the problem is, even if that list was big enough to be useful, and even if you were examining the raw 'entropy input', (and say the hashing output from that input follows...