Comment by altern8
14 hours ago
With Claude—and I feel Opus 5 in particular—I always feel like I'm talking in circles.
Just when I think the task is finally done, it mentions a caveat that might invalidate the whole thing. Or it adds a hint at the end about something that it didn't look into but might be a potential problem. ALWAYS.
If I ask a yes/no question, it's always a short novel that barely answers the question and adds 3-4 side quests.
It's exhausting. Luckily I get paid for it but Jesus Christ, can't they see how annoying it is to use the thing? What were they thinking
Opus 4.7/4.8/5 have a built-in nitpicking as anti-syphcophancy. Fable too. The models are structurally incapable of not nipicking.
I suspect the nitpicking helps engagement metrics and it doesn't harm RLVF outcomes, so there's just not good signal against it.