undefined | Better HN

0 pointscjbgkagh1mo ago0 comments

I find the LLMs target their language to the audience, so instead you could say, “I am Dutch so give it to me straight.”

In my usage the LLMs gives much smarter answers when I’ve been able to convince it that I am smart enough to hear them. It doesn’t take my word for it, it seems to require evidence. I have to warm it up with some exercises where I can impress the AI.

The coding focused models seem to have much lower agreeableness than the chat models.