My Primes API was previously very slow, since the way it worked required that the server calculate the entire array of primes from scratch and then return it. Today, I finished a major update: a database-driven design that takes the result array from a pre-calculated set of prime numbers (all of them under 500,000) and then returns it. It has the same interface as the original design, but should be much faster, with the tradeoff being that the database takes up a small amount of server space (about 1.5 MiB).