945x is the actual measured difference, O(N) is the theoretical difference. Many times quadratic time complexity is ok because you are working on very small N.
The point is 945x is just the measurement for this particular test case. If the author made the test larger it would appear even faster.
This is the underlying problem with people who don’t understand what big-O complexity is saying.
I believe he means it's 945x faster in the current case, but O(N) in the general case.