We need nuanced and useful understanding of new AI model ability, rather than a static ablity to pass well known subject matter testing and acedemic credential testing are ‘gamed’ and becoems meaningless as all models reach past 100% in all categories.

We found a better way.