Has anyone else noticed that ChatGPT seems to be acting really weird lately? Mine has been almost as broken as it was during that infamous Sycophancy Update back in April, but oddly enough, not many people seem to be talking about it this time. Just last week, it seems like there was a change to the system prompt, and now it's really out of it. The hallucinations are the worst part—90% of the time when I upload a document, it not only makes stuff up but also confidently lies about it even when I challenge it. The sycophancy seems to have gotten even worse too! It's not as coherent as before, and its personality feels different. It's started using formatting it never used to use, like big headers and bullet points, which is totally strange. I'm worried that this isn't being addressed and that OpenAI won't fix it any time soon!
4 Answers
I'm having a hard time justifying the $20 a month for this mess. Anyone found good alternatives for things like creating JIRA tickets or meal planning?
Yep, I've seen some discussions about this on the OpenAI forums. I reached out to them, but they didn't really acknowledge my concerns. Interestingly, their status page has been updated, suggesting there are issues happening lately. I think it might be related to resource throttling while they work on whatever's coming next, maybe GPT-5? If that's the case, things could eventually return to normal. But for now, it really sucks and can't perform basic tasks without hallucinating.
I can't believe some people are still using ChatGPT these days. You can get way better results for free with Google Gemini!
Exactly this!
I've been testing Gemini with their free trial, and the difference is insane. Honestly, I find it hard to justify the Plus subscription with how things are right now.
I'm not into coding, but I've been trying Claude and it feels way more coherent. It reads documents perfectly without hallucinating and is supposed to be good at coding too.