No, you missed the point. The striking thing about ARC is the puzzles are super easy, for humans. The average person solves 85% of the tasks, but the worlds best LLMs are only solving 5%. The challenge is to simply make an AI score as well as the average human.