the latest ai models have been giving me a lot of "short answer: no" when I'm quite used to "you're absolutely right"