What is competitive vibe coding?
Competitive programming but you can vibe code your solutions.
Why is it superior?
The fun part of competitive programming remains. Competitive programming has long been criticised for not inculcating "good" programming practices. In reality, the hardest but most fun part about competitive programming is coming up with a solution. Implementing your solution is just a necessary evil for automating competitive programming contests and websites. Competitive vibe coding gets rid of this evil, since you can truly enjoy the heart of competitive programming (making observations to solve bugaboos) without writing out $$$680$$$ lines of implementation pain.
It's more relevant for the future. AI is somewhat capable of writing small pieces of code, and it is reasonable to expect AI to be able to write clean and understandable code in $$$5$$$ or $$$6$$$ years. At this point, implementation skills will become useless. The heart of competitive programming, understanding and toying with the bugaboo to make observations and find a solution, still remains essential in terms of solving other bugaboos in your life. Competitive vibe coding trains this without emphasising on machine-replaceable implementation skills.
Disclaimer 0: I may or may not have used ChatGPT to help me review my drafts.
DISCLAIMER 1: Practicing competitive vibe coding in competitive programming contests may get you disqualified.
Auto comment: topic has been updated by literalchild (previous revision, new revision, compare).
A company at my place recently did host such a contest (to promote and test their AI model). It turns out that being a good competitive programmer yourself is still required. Unless you burn your dollars on pro level models, for problems that matter you still need human in the loop.
sounds like an oxymoron to me
Unironically there should be a contest format where you're provided with some 2024-era model so you can just type the solution in English and have it implement it, but at the same time the model is too dumb to solve the problems itself. Then it's way more about finding the right algorithm.
Don't understand why people are downvoting, it's a good idea, I guess the only problem is verifying no one is using stronger models to ask for solutions. It could work offline-only I guess.
I thought/hoped this would be the future, so we'd get stronger data structure and complex system implementation problems that were not possible in the contest span before, bringing the peak of competitive programming closer to research level tasks. Instead it seems the community settled on the status quo.
agree...
Sometimes describing the solution in natural language clearly is much more difficult than actually writing the code oneself, at least for me.