are you really arguing that opus 5.5 beating the best of openai's models handily is worse than you expected? What exactly were you basing your expectations on?