I have been trying to solve the puzzle but every time I tried I got an headache. I dont know where to start and I keep on researching until I end up with nothing. I even ask some of my friends who are good with this things but I cannot understand fully their explanation. I guess I will stop doing the challenge and will focus on the signature campaign. Well good luck to you guys in solving the puzzle its hopeless for me.

Hmm.. your both messages look like you are asking to increase the bounties again  For example from 0.64BTC up to 6.4BTC for wallet #64 and so on

That is a very complex problem and addressed to experts in programming, I will be watching to see who sends the solution.

It is not clear what you are doing there and what result you get?
Write me in PM.

How many private key can be generated per second using this tool ? I am not sure but when I watch th video its look like take more time + don't forget that using bitcrack and gpu can create more than 100Mkey if you have good resources, so this game should take lot and lot of time than bitcrack.
and at all its just chance, you can get it fast if you have chance

So you just confirm that the key is random! Your example like "that the same bit could not repeat more than 8 times" just prove that the key is random.

How do you think the real random works? The random number is not selected within the eide range, no. The random is applied to every bit. So, the random entropy was used to create each bit in random way, which is absolutely 50%/50% for every bit.
That's why there is no pattern. That's why it is so hard to guess and find any connections. When you apply 50% probability for every single bit, so you do not guess and imagine how is the whole key will look, it is really not important.

I beleive that all the keys for this puzzle transactions were generated randomly on the lowest level - on bit level.

I do not argue that the keys from the puzzle were obtained randomly. What was this generator, bitwise, ekdska, super random, etc. irrelevant.
But here we go into the theory of probability. For example, if you get a random number from 1 to 100, then theoretically one kind of number, for example 3, can fall 10 times in a row. But it is theoretically. If in practice this happens, then it will be more like a generator failure. Another example. We create a million random numbers from 1 to 100. In theory, some numbers will turn out much more, and some much less. But in practice, the quantitative discrepancy will be no more than 10 percent on all numbers. That is, not entirely by accident.
From this we see that the numbers found have some rules or conditions. The number of set bits can be from 30 to 70 percent, a sequence of up to 8 identical bits in a row, the same rotation no more than 10 times, etc. The same rules can be chosen for the 8-decimal version and the 16-decimal representation. Repeating the same numbers, ladder, etc.
And applying these rules, we can discard most of the keys so as not to check them for correctness. Yes, there may be exceptions to the rules, but most should be subject to these rules.
This increases the speed of search. And you can start not with boundary values, but let's say averaged ones.

Sort, less than 50 and more than 50.

Count 35 Average 60,77%

Count 30 Average 41.43%

Count 5 Average 50% not used in calculation.

That is, the search can begin with a range from 41 to 61 percent.
And if we search for those already found with this range, then we will find 58 keys out of 71. That is more than 81 percent of the locations.
In this case, the search is performed in only 20 percent of the possible options, that is, 80 percent of the options recline !!!
If you have not found, increase the range, say up to 31 - 71, while also excluding options from 41 to 61 percent. Etc. Or we proceed to search for the next key.

Thank you so much for explaining ,and sure he or she will not give the gpu code until he finish and find the key.
and about gmpy2 ,i tried to use it but can't, its problem with installing gmp and the wheel is damaged.
So now we can just go to #64 and let gpu users doing it 

Yeah I tried that too but still nothing,


if i use ubuntu i get that need gmp .if i tried windows i get this problem . i readed also that gmpy add 8% fast calcul.

Thanks