This was extremely impressive to me. AtCoder has the hardest problems these days, usually the human onsite final round contestants can't solve more than 2 or 3 problems. This year the problem setter sets the round in a way that maximizes humans chance of winning. Then OpenAI just comes in and solves all problems..

No point for them to even go to IOI or ICPC this year anymore, those are all much easier than this Atcoder contest. And given the mathematical nature of AtCoder problems, not sure if there's any value to do IMO either, except for publicity of course.