logicallee 16 hours ago

At State of Utopia, we recently found this vulnerability in logical thinking that arises in ChatGPT 4.5 when asked not to show its thinking.