I saw this on reddit, thinking it’s a joke, but it’s not. GPT 5 mini (and maybe sometimes 5) gives this answer. You can check it yourself, or see somebody else’s similar conversation here: https://chatgpt.com/share/689a4f6f-18a4-8013-bc2e-84d11c763a99
While AI reasoning models will continue to improve, the data sets they pull from will continually get worse.
That's not my understanding based on this article I read yesterday.
https://minimaxir.com/2025/08/llm-blueberry/
That's a blog post, not an article.
Why don't you give this one a read: https://arstechnica.com/ai/2025/08/researchers-find-llms-are-bad-at-logical-inference-good-at-fluent-nonsense/
Or, even better, why not read the actual study: https://arxiv.org/pdf/2508.01191