I'm not a math expert but I think it should take 20,825 times longer (583=195,112)*. There is also the fact that the underlying algorithm will also be different that lacks most of the pre-computation that the current algorithm for known missing places has so it will be slightly slower. However, as I said before I have no idea how to generalize this so I hardcoded the heck out of it (only for this special case: missing 3 chars, probably will add a couple more smaller cases later), the benefit of it is that I know how to run it in parallel so the more threads your CPU has the faster it would run and I don't think 4 billion is going to take that long to finish.