i've been throwing the same hard question at a few different models (or the same models) for a while now and honestly i've stopped caring where they agree. when they all land on the same answer it usually just means the question was easy, or they all grabbed the same standard take from overlapping training data, so agreement is often just a shared blind spot. the useful part is always where one of them breaks from the pack, that gap tends to be land right on the thing i was glossing over. i got ...