News for hackers. Stuff that matters.

Thoughts on ChatGPT-5? Especially Around Accuracy, Critical Thinking, and Memory

For those who’ve been using ChatGPT-5, I’m curious how you feel about it—especially regarding three areas that seem to challenge current LLMs:

Hallucinations & Transparency – Still seeing confident but wrong or outdated answers? Does it better admit “I don’t know” or distinguish facts from speculation? How are citations and real-time knowledge?

Critical Thinking – Does it still feel overly agreeable, validating contradictions or avoiding pushback? Or does it challenge flawed ideas and act more objectively?

Context & Memory – Any improvement in holding long conversations, keeping project context, or indicating context limits? Does it handle long, iterative tasks without quality drop-offs?

Would love to hear your experiences—good, bad, or mixed.

( 1 comment | on HN )
Re:Thoughts on ChatGPT-5? Especially Around Accuracy, Critic... by gnabgib on Saturday August 09, 2025 @06:57PM (#44851141)
Have you looked in these two recent posts:

ChatGPT 5 is slow and no better than 4 (39 points, 5 hours ago, 25 comments) https://news.ycombinator.com/item?id=44848431

ChatGPT-5 Can't Do Basic Math (14 points, 1 day ago, 11 comments) https://news.ycombinator.com/item?id=44840532