News for hackers. Stuff that matters.
Re:Claude Finds Contradictions in My Thinking by ceejayoz on Tuesday July 29, 2025 @11:34AM (#44724693)
> But all failures are "random and unpredictable" if you have no baseline understanding of how to use the tool.

But the tool's behavior changed. In ways that even its creators didn't intend (example: https://openai.com/index/sycophancy-in-gpt-4o/), and had to work to undo.

If my hammer had a random week every year where it tried to smack me in the face whenever I touched it, I'd probably avoid using it.

Re:Claude Finds Contradictions in My Thinking by johnfn on Tuesday July 29, 2025 @11:37AM (#44724727)
This is unrelated to sycophancy. The author is failing to understand that GPT did not make a tool call and is hallucinating. Hallucinations have always been a thing. They are not some new, surprising development.
Re:Claude Finds Contradictions in My Thinking by ceejayoz on Tuesday July 29, 2025 @11:43AM (#44724781)
> This is unrelated to sycophancy.

"It's a stunning piece. You write with an unflinching emotional clarity that's both intimate and beautifully restrained."

> They are not some new, surprising development.

OpenAI sure seemed surprised. https://openai.com/index/sycophancy-in-gpt-4o/

Re:Claude Finds Contradictions in My Thinking by johnfn on Tuesday July 29, 2025 @11:58AM (#44724954)
> "It's a stunning piece. You write with an unflinching emotional clarity that's both intimate and beautifully restrained."

This is a hallucination, since there is no source to refer to.

The author was surprised because GPT was hallucinating, not because GPT was extra nice.

Sycophancy might be related, but it's not the point of the article. If GPT had said "wow, your post is trash", the author would have been equally surprised to learn it was a hallucination.

Re:Claude Finds Contradictions in My Thinking by ceejayoz on Tuesday July 29, 2025 @12:10PM (#44725118)
"Oh, yes, I read your new novel, it's great!" is precisely what a sycophant would tell you when they forgot to read it.