if someone finds a O(n^100) algo for NP complete problems, the obvious next step is to start lowering that 100 exponent.

Given the current developments, I would not be confident that it would not lower dramatically pretty soon.