Pushing objectively false answers should render an AI doodad bad for the job, no?
But that's why I said "if it's good for the job it's good for the job"
If there's something that Grok *positively* does better than other LLMs, why wouldn't you want to use it, because, _boohoo_ Musk bad.