To give some detail they didn’t, they designed an algorithm that’s as good as possible using the specific approach they used. I was all excited someone had finally designed a provably polynomial pivot rule. This is a more modest step forwards, and depends among other things on a small amount of random noise being added to the problem.