Or that opus and mythos are training on the data somehow such that there solutions are incorrectly right. Or that openai is lying/wrong. Or that all of these companies are cheating so much it doesn't really matter and never did.