AtCoder is preparing to license its contest data directly to AI developers: 5,000+ problems with editorials and full judge test data, plus 30M+ verified human solutions across 112 languages — including failed attempts in submission order, with every submission labeled in-contest or post-contest for contamination-aware training and evaluation.
All data is first-party and consent-based, not scraped; opted-out submissions are excluded.
Learn more and join the waitlist at [AtCoder Datasets](https://datasets.atcoder.jp).