News for hackers. Stuff that matters.
Re:Software Survival 3.0 by xyzsparetimexyz on Thursday January 29, 2026 @12:27PM (#46813326)
> I debated with Claude endlessly about this selection model, and Claude made me discard a bunch of interesting but less defensible claims. But in the end, I was able to convince Claude it’s a good model

This is not a good way to do anything. The models are sychophantic, all you need to do in order to get them to agree with you is keep prompting: https://www.sfgate.com/tech/article/calif-teen-chatgpt-drug-...

Re:Software Survival 3.0 by mrefish on Thursday January 29, 2026 @12:55PM (#46813744)
> Let me know what you think by complaining that AI sucks on HN!

At least you complied with the next sentence :)

EDIT: Whoa, I didn't check your link before I posted. That's terribly sad. While I agree that LLMs can be sycophantic, I don't think Yegge was debating with Claude about drug use in this situation. Other references might have worked better to support you claim like this first page result when I search for "papers on llm sycophancy": https://pmc.ncbi.nlm.nih.gov/articles/PMC12592531/

Re:Software Survival 3.0 by xyzsparetimexyz on Thursday January 29, 2026 @01:12PM (#46814042)
I don't think that AI sucks. I just think that it is something that is too maleable to debate with
Re:Software Survival 3.0 by doug_durham on Friday January 30, 2026 @04:26PM (#46830123)
You need to explicitly tell it to debate you. Also you cite and example of someone using ChatGPT to discuss personal issues. We are talking about technical discussions here.