“Mark Chen of OpenAI stated on a podcast that the company had been thinking about the ARC v1 benchmark for years.”