Comment by Silagi

That's wild to say when comparing to the OpenAI family of models. I've never seen so many tests that validate tests that validate tests.