Did anyone else think it was strange, in an article about runtime efficiency, for them to say that problem-specific optimisations promise speedups of "several orders of magnitude"? Because, well, that doesn't change the order of the algorithm...