As the article says, the previous one was a brute force implementation. And the 100x number is kind of nonsensical since the real speedup depends on the size of

As the article says, the previous one was a brute force implementation. And the 100x number is kind of nonsensical since the real speedup depends on the size of the input (the asymptotic complexity of the algorithm went from O(nmk) to something like O(n+mk) I believe).