I always test AI with philosophical questions. All OpenAI models perform far behind Claude Sonnet in such tasks. Even o1. Even o1 is a long way from real thinking. So I don't have to worry about my job.