First of all, I'm terribly sorry for what happened today; Problem C was solved by an AI. It is legal to use AI in AGC contests. However, this rule was adopted under the assumption that we would not provide problems that are solvable solely by AI. What makes the situation worse is that, as far as I know, only paid plans are able to solve this problem. This is especially serious, and I fully accept your frustration. Of course, I tested this problem (and other problems) with GPT5-Pro until I reached the query limit, and it didn't succeed. I should also note that the initial version of C was asking for a slightly different thing, and I tested the current statement only twice, but I don't think that's the root cause of this incident. I should have paid more attention to the development of AI. I knew that some hard CF/IOI/ICPC/PE/etc. problems were solved by AI. Many people asked me about whether (or when) to ban AI in AGC. I had several chances to reconsider the rules. However, I was overconfident and thought, "Hmm, there do exist good problems that aren't solvable by AI, so I can just choose them." As you know, I didn't have the ability to identify such problems. As I said, the current rules of AGC don't restrict the use of AI, so making the contest unrated would feel like applying an ex post facto law. Therefore, my current plan is as follows: - Keep this contest rated (and count GP30 scores). - Ban AI for future AGC contests. - Hold more AGCs to mitigate the effects of this incident. To realize the third point, I have to reconsider the style of AGC. I'm not exactly sure about what changes to make, but one possible idea is to make a contest that is somewhere between ARC Div.1 and the current AGC. Again, this is entirely my mistake and I sincerely apologize to all participants. I would also like to express my deepest gratitude to everyone who participated in this contest under such circumstances. AGC admin, maroonrk