The bit-fishing-table is the core element in AHS-RNG. The first bit-mixing tools are cooperating for the creation of a random value to serve as address of a bit from the BFT. In this table exists an equal number of zeros and ones. For a table of 65536 bit (8K, as used in this demo) we distribute 32768 "0" and 32768 "1" in a good random pattern. Every element of one bit in the table has his own address, from 0 to 65535. In binary that means from 0000000000000000 to 1111111111111111. The last four bits indicate the position in a line of 16 bit. In the demo we display only the concerned 16-bit line. You are able of course to open in a separate window the complete BFT (the supplementary information on 'BFT' on the start message) and check yourself the correctness of the bit in the whole table at the indicated address.
The number of possible different patterns for a BFT of 8K is 6.244451173... * 10exp19725, as we can easily calculate in less than 5 minutes: 65536! / (32768! * 32768!). Every different pattern will generate completely different streams of random numbers. We tested a change of "01" to "10" at an arbitrary position in the table, and after a warm-up period we got two different streams. We checked the quality of the difference between the two streams by XORing the two streams and creating a new stream with the differences. This new stream passed the TestU01 bigcrush without any failure.
The fact that we include in the calculation of a new "FeedBack Modifier" the last 16 bit from the 32-bit random number just created is at the origin of this surprising effect. If you enter the 'step-by-step' mode just before the completion of one of the 32-bit random numbers, you can follow this feedback algorithm. Appendix D in the white paper from 2006 (page 29) gives supplementary explanations.
This endless variability and the statistical results exceeding those from quantum random number generators is the reason why we don't accept that the AHS-RNG is considered as Pseudo Random Number Generator. If one insists on this attribute, he must prove the differences in the result against true random numbers!
The (false) declaration that a digital computer cannot generate random numbers comes from the fact that nearly the complete research on random number generation was explored in the past by mathematicians without deep enough understanding of modern digital computer, their programming and their operating systems. Our AHS-RNG is able to produce deterministic true random numbers or secret and non-reproducible true random numbers.
As we proved that a digital computer is able to produce true random numbers and not only pseudo random numbers, we hope that mathematicians start now to discuss seriously the question if there may exist infinite solutions to produce random numbers by digital computers. Anyway it is not good science to spread opinions as facts without being able to prove them. We are ready to prove our claims.