from Hacker News

Do you find LLMs to be judgmental?

Have you fixed with custom instructions?

Inspired by something I was told re claude vs chat - "[claude is] too judgmental, i literally don’t go to claude because it feels judgmental of me, whereas chat feels like it’s on my team and wants to help me" "claude feels like a weird librarian or teacher that’s too wordy and doesn’t wanna help me for a bunch of things"

I personally use Claude a bit more these days, but I've also set up custom instructions to make it less judgy and do recall it feeling judgy before that, sometimes to the point of feeling insulted by it a couple times actually (strange feeling to have a computer throw such shade at you)... But yeah, custom instructions seems to have solved that for me.