Mirror Passes, Roots Fail
OpenAI’s GPT-5 arrives as an on-demand expert with the knowledge and reasoning of a “PhD-level” scholar, folding two years of advances into a single adaptive model. It ends confusion over GPT-4o, o3, and o4 by modulating reasoning to match each query. GPT-5 prizes usability—faster, more intuitive, and personalized. I retested the ontological bias and still saw a tree without roots. I tested it with code development and, in my opinion, it performs better in this specific but important area. GPT-5 also passed my earlier “Strawberry Test”, counting letters in both “strawberry” and “mirror” without error. Used weekly by 700 million people, ChatGPT now aims to be your go-to assistant before true AGI. Read More →