I know I will be downvoted to hell, but...
This program is cool, the implementation is surely very clever, and the theoretical algorithm behind is very very interestig. And I really like integers, primes, and number theory.
However, I find all these "records" a puerile waste of resources.
From the presentation page:
*The first phase of the record, [...] has taken
about 26 days of wallclock time and 67 years of CPU time, in several runs on clusters with 752 to 1328 cores.*
What's the point of wasting 67 years of CPU time to prove the utterly useless fact that 10^50000 + 65859 is prime?
I get that one is proud of their code and want to show that (1) it works (2) it is faster than previous approaches. But I imagine this can be done on much small instances. Let's fix a benchmark.
To me this sounds like "they gave us this supercomputer for one month and the most interesting thing we did was to prove that a certain more-or-less-random number is prime."