a) 50% result is not solving the problem. Especially when the implementation is brute forcing the problem and is against the spirit of ARC.

b) He is not being overly negative of LLMs. In fact he believes they will play a role in any AGI system.

c) OpenAI CTO has publicly said that ChatGPT 5 will not be significantly better than existing models. So the rate of improvements you believe in simply doesn't match reality.