Random numbers that can't be reproduced, ways to collect noise, and circuit diagrams

Random numbers generators that don't produce the same numbers given the same starting conditions are typically based on noise. In principle anything can be measured with enough precision that it's state will be random. Note that these kinds of generators are frequently biased- producing more ones than zeros or more zeros than ones. However, there are many methods of removing bias, the key is to get something that isn't reproducible and post-process the data.

(someday, there will be program examples, and circut diagrams here)