Today after the div I decided to look at the leaderboard and I was surprised by the number of gray participants in the top 20 and the time it took them to complete the tasks. Some of them submitted problems D, E, F at intervals of 3-5 minutes. After looking at their code, I saw a clean GPT code style and a lot of obvious comments. A huge number of codes do not look human; in particular, their code contains long names of camel case variables. Sometimes they received errors when submitting them, but these were only compile errors and WA1, which is not entirely logical if you wrote the task yourself and tested it. 
A couple of examples that I noticed very easily: byteplay_studios and bdyby1009 had huge comments on problem F, which they solved in less than 5 minutes. You can also study people's codes nguyennhat182006 and WAIL-01. And now my two favorites: sun_jinwoo36 and cphelper127864: the first of them solved problem A 1:17, and all the following problems up to F took him less than 8 minutes each. The second one made parcels via E at the end of the round; his comment code contained half of the total volume. At the same time, his code passed first 0 tests and then 1, but was still unable to pass all of them.

Considering the rapid development of neural networks, it is also necessary to quickly develop decision control systems because such people can occupy about 500 places in the ranking and interfere with honest people