OpenAI explains that the standard benchmark used to measure AI coding ability is 'no longer meaningful,' and that examining the problems that initially failed to solve revealed ...