We asked @mikeknoop (Co-founder, @arcprize) about continual learning and the evolution of AI reasoning benchmarks: "ARC V1 was introduced back in 2019. It was designed to challenge deep learning as a paradigm, before language models really took off." "V2 challenges a new paradigm of AI reasoning systems. Even though the puzzles look similar to V1, V2 generally requires longer reasoning chains, which makes it harder." "Now, with V3, we’re defining what we’re calling an interactive reasoning benchmark; to evaluate and challenge the new generation of frontier AI agent systems."