Today's concept: the Most Random Data.
Look here. This contains a small set of tests to determine the randomness of a bitstream.
Now: obviously every possible bitstream will have results in this test. We can define a set of weights defining how important each of these results is to us - or even a complex equation if we want. In the end, we'll get a single value as output - the Randomness Metric.
Since every possible bitstream has its own Randomness Metric, what does this imply? Well . . . we can choose the bitstream with the highest Randomness Metric. It might not be computationally feasible to do so, but we'll ignore that for now, since computational feasibility doesn't factor into this proof.
And now that we have the bitstream with the highest Randomness Metric, we can call it the Most Random Data, and use it as a random sequence for every random number generator in the world!
I find this highly amusing.