The reality, however, is that the CNS algorithm is not an efficient way to find collisions. In fact, it’s so inefficient that it’s beaten by the standard non-quantum method of finding collisions, namely “parallel rho” collision search from 1997 van Oorschot–Wiener. Let’s take an in-depth look at what’s going on here.

Besides the beatdown, some good notes on finding collisions and performance assessment. Not always as simple as big O.