Does it fetch the whole database and filter the results on client side? Because it seems like it would download 54 sqlite shards to show the few entries available for Mersenne and Fermat primes. It costs 40+ seconds of time and ~172MB of bandwidth. A cheaper option might be to generate the prime numbers first and then just fetch the shards it needs.
Other than that, I thought it was an interesting way of sampling the news as the time delta grow further apart between each item.
Yes, I agree there's probably a better way for the Mersenne and Fermat. And even the Germain we probably don't need to scan every shard for just the first page of them. Good points! And good idea to invert the lookup.
Thank you for the comment on interestingness. I thought so to, but I didn't realize it in same way you did in your observation that it gets more selective over time - I like that, in a sense that's kind of natural as the environment is more competitive over time. Thank you for the interesting idea!
Other than that, I thought it was an interesting way of sampling the news as the time delta grow further apart between each item.