> I was amazed by how fast he found that bug in code he had never seen before. I'm not sure if he read carefully through the original paper or came up with the

> I was amazed by how fast he found that bug in code he had never seen before. I'm not sure if he read carefully through the original paper or came up with the clever debug strategy of "since this is a port, let's compare to the original, looking for typos, with a particular focus on the loops". The latter is exactly what I would have done. If the algorithm did not match the expected complexity, the most likely reason is a bug in the implementation, and the quickest way to find that bug is to compare it with a known-good implementation, such as the one that the code was ported from.