There is this thing known as the "enhanced games" which is basically olympics but they allow athletes to use performance enhancement drugs. Codeforces can also start something similar with contests which allow participants to use AI without any repercussions.
It would also level the playing field, if everyone is using AI then no one is. The rating changes as a result of these "special" contests can be kept separate from the original rating.
You can also make the problems like 10x harder so people cant just "claude solve this question for me make no mistakes"
Thoughts ?
PromtEngineeringForces
Auto comment: topic has been updated by oaxplyn (previous revision, new revision, compare).
it is just alternative of monkeytype
who copy and pastes the question to ai first
and tbh it will be a match between ai's, there is no point of a human participating in it
(also a bit of prompt engineering)
also, olympic enchanced players have to still use there skills, like one without any skill or experience cant win even with drugs
but in case of cp, i will solve every single problem for him and all he will have to do is live in delusion of "ohh i am so smart"
also i think it is just a "TRAINING CAMP FOR CHEATERS"
the rating for the ai assisted contests will be separate and not impact the main rating. i dont think so it will be a case of cp becoming monkeytype but instead a sort of arms race between AI agents.
We must keep in mind that for hard problems, even SOTA models take a lot of "thinking" time (TTC). Whoever would be best able to reduce this time, either through solving parts of the problem themselves, phrasing the problem in such a way that the LLM takes less time or any other technique would come out on top
it is clear why you want it to happen
you yourself have submissions like https://codeforces.me/contest/2116/submission/322282737 https://codeforces.me/contest/2116/submission/322265057 https://codeforces.me/contest/2109/submission/320063330 https://codeforces.me/contest/2102/submission/319281999 https://codeforces.me/contest/2108/submission/318004442 https://codeforces.me/contest/2104/submission/317647075
what the fuck u have not solved a single ac or wa by yourself your every single submission is ai
idk man the idea seemed pretty cool and interesting to me, i guess most people dont agree :(
"make a other platform for that", maybe i will
"what the fuck u have not solved a single ac or wa by yourself your every single submission is ai"
what makes you think the submissions you mentioned are ai ?
also i wont be replying any further because you clearly arent engaging in good faith. have a good day
it is just pure common sense, it is clearly ai anyone with eye can see that
i can only imagine how will hints in editorial look like!
have you tried gpt?
did you noticed problem require math so it must be deepseek
have you considered gambling and choosing the ai on pure luck
did you noticed tag is *2300 which mean you need to use paid pro max gpt claude version?
then what would even be the point of improving your own problem solving skills if all it turns into is ai vs ai? and also there is clear a difference between physical sports and mind sports. i absolutely disagree with your idea
Dude, If you want cheat then this place is not for you.
We are here to improve in CP while enjoying it, we are not here to improve our prompting skill!
i feel like it can be fun to participate this way, but i dont know what do you mean by "level the playing field" because whoever has a more expensive AI model with more quota and context will naturally have an advantage, making the competition P2W.
as well as that, div.1 contests are rare enough these days, and what you are asking for is essentially to make some div.1 contests count towards a different rating system, which is going to make a lot of >2100 people really unhappy.
if you really want to compete using AI you might wanna look into meta hacker cup AI track, but imo CF isn't the right place for this right now
maybe limiting models to only open weight ones (which can be freely accessed using API providers like Openrouter or Groq) can solve the pay2win aspect ?
yeah and I can see the unhappiness part as this post got nuked to hell lol
you can, but that doesnt change my argument about it being unfair. how are you going to check which AI has been used? in a regular contest you can at least have some methods to battle AI, here since everyone "cheating" you can't possibly deduct who is using AI fairly.
and sorry to say, but post probably got nuked because you are gray, not because people thought deeply about your post and figured that it's a bad idea