OpenAI counterexample solves the millennium problem in a Yes/No manner - we do know in general that singularity could arise, but it still isn't solved in terms of "Why", ideally we could take an arbitrary physical problem and decide if Navier-Stokes equation can be applied to it.

This is similar to general solvability of the quintic equations - Abel provided a proof first but only with the advent of Gallois theory we could basically understand it in full and decide for any quinitic if it's solvable by radicals or no.