Not sure how best to verify such a proof.
Perhaps somebody could independently make another computer algorithm (in case the first one had an error) for solving it and run it on an entirely different computer (in case there was a hardware fault on the first one) and then, if it gives a proof, make a third computer compare the two proofs to say if the two are equivalent?

Originally posted by humy http://phys.org/news/2016-07-longest-maths-proof-billion-years.html

Not sure how best to verify such a proof.
Perhaps somebody could independently make another computer algorithm (in case the first one had an error) for solving it and run it on an entirely different computer (in case there was a hardware fault on the first one) and then, if it gives a proof, make a third computer compare the two proofs to say if the two are equivalent?

I haven't looked at their paper, only the Wikipedia page [1], but if they've done what I think they have, then they'll have iteratively generated partitionings of the set of numbers, and whenever they find a partition with a Pythagorean Triple add it to the output and remove it from the list of surviving partitions. So you could probably check the output with a Perl script, which just checks that they haven't missed any cases. I think the data is probably fairly easy to analyse, it's just that there's stacks of it so any checking has to be automated.

The issue presumably applies to a large number of mathematical proofs / findings. For example if someone calculated the first few thousand places of pi, nobody would double check it with a pen and paper.