Nice kata! Yet I was wondering about the mysterious test results, too. So, I would like to ask whether you could ease them by the following.

• Could you please keep the ordering of the solution triples as x, y, z? Reordering them would require guessing what is x, y, z for each triple. What should this be good for?

• Could you please let us hand in the solution triples with largest z as a set rather than a list. Why artificially order them? Plus, your test results are not sorted the way you claim they are.

Sorry for nagging, first of all thanks for your work:)

Thanks! Now it passed:)

The second 'no hardcoding' test says I am cheating. But where? What I do is storing the sequence members computed and the numbers below the maximum of these members which have not occured yet in a cache. That's what I understood by 'no extra computations'. Is that forbidden?

By the way, on my middle-aged computer at home my code needs about 3 seconds for computing 20100 in one shot, but when I put CHALLENGE = True, it times out. Why that?

I would be glad about any hint:)

Hi Mr President:) I think I have a solution in Python (not the one where only the clues are given), but I do not know C++. So, I would like to know whether you plan to publish a Python version some day?

Uses that any power of the base matrix is determined by its first line, thus avoiding full matrix multiplication. Saves time for larger n.

Oh, no problem at all. Nice to meet a sensei:) By the way, what's your opinion about this solution? I'm rather new to this programming stuff and often struggle with runtime.

Best regards, Matthias

Actually not, at least as far as I can remember:) May I ask why you ask?

The "someone else" mentioned is da_big_fella, who in turn gave credits to the algorithm wrote in C by g964. This is so much better than the solution I had put together before. Thank you both for the input (whenever you may see this).

Sorry, this does not create a new list as required but rather modifies the input. I have corrected this.