
OpenAI just spent three months fixing what should have worked in August: making ChatGPT write like a human instead of a committee. The writing quality upgrade matters, but the real insight is why they had to backtrack at all. The capability wars are over. The usability wars just started.
