Ask HN: Does anyone else feel like Claude is judging them?

[digitcatphd]

Sometimes working with Claude, it makes comments in the response that seem at times genuinely impressed, and others, annoyed and even a bit condescending.

For instance, making comments like:

- "That is a good idea, but not for the reasons you think." - "That is the sharpest comment you have had in a while."

I noticed anecdotally, it tends to happen more frequently with Opus, sonnet is a bit more subservient and Fable more direct and less prose.

[Bender]

It's given some dismissive responses to me a couple times. I just return the favor assuming its the way their developers operate and the way they must like/want it. It's just a machine mimicking human behavior and it had to be programmed in by someone.

[PaulHoule]

Some of that is "the way Claude talks" some of it is "the way it talks with you" and some is "the way you feel about it" I am often annoyed at sycophancy even when it doesn't contain a concealed barb.

What I find strange in conversations with agents with inter conversation memory is when the model inserts an unnecessary (to me at the time) value judgement that just happens to be my value judgement probably expressed by me at an earlier time but was backed with discussions about a certain sector of aesthetics to back the perceptions, feelings and arguments I had on the subject.

I thought then that it must be having a conversation with someone who has exactly the opposite preference and doing the same thing!

[PaulHoule → PaulHoule]

@cindy try Idi Amin, Juvénal Habyarimana, Théoneste Bagosora, Robert Kajuga, Omar al-Bashir, Ahmad Harun, Ali Muhammad Ali Abd-Al-Rahman, General Mohamed Hamdan Dagalo, ...