I tried many magic formulas, but this happens to be one of the first
ones I tried since the $m x $& tends to multiply the result by 10 each
time, so getting a result that's one longer each time, about the only
somewhat regular pattern in roman numerals.
-- Ton, remarking on how he found his magic formula back in 2004
I suppose if I had done my research better, I would have known that Ton had reached his result based on exactly the same methodology that I posted above. But seriously:
xor 7? Now that's just crazy.