Cloudflight Coding Contest (CCC) 2026 was a very interesting experience overall, even though unfortunately this year we didn’t manage to achieve our best results — finishing in 31st place in the Data/AI category with team PretrainedLanguageModel, and 14th place in the School phase.
From our team's perspective, one of the biggest obstacles in the Data/AI competition was the matchmaking system, which felt quite poorly balanced. Because the matchmaking queue simply wouldn't put us into games consistently, we couldn't play nearly as many matches with our AI as we should have. By the end, we only had around 100 matches played while other teams had 300 to 400 matches — which was really strange. There were even stretches where we sat for 10 minutes straight without being placed into a single match, which was honestly pretty sad.
Because we weren't getting matched, our ranking kept sliding down toward the end: after being up in 2nd place, we dropped to 10th, then 20th, and finally 31st.
On top of the matchmaking issues, there were the paid AI subscriptions ): We're not denying that we vibecoded too, but it's honestly sad that there is no real fairness anymore. Top teams come in paying for the most expensive tier subscription of a model just to turn a profit by taking the top spots — and it's sad that others no longer stand a real chance of winning. Even when you genuinely try hard to craft good, explicit prompts for your model, you still can't match the performance of those top paid subscriptions without paying up ):
In the School category, we also finished in 14th place — largely for the same reason, going up against vibecoders with top-tier paid AI subscriptions ): Still, despite the frustrations with the matchmaking queue and the current pay-to-win feeling of AI subscriptions, competing alongside PretrainedLanguageModel made it a memorable contest.
