Did OpenAI solve the wrong Navier-Stokes problem?

OpenAI’s proof seems eligible for a million-dollar prize—but only by using a controversial loophole