Thomas Bloom explains that rapid advances in AI have reshaped how people interact with Erdős-style mathematical problems: progress has accelerated, but much of the site’s recent activity consists of brief AI-produced proofs posted as priority claims with little explanation. That shift has discouraged human engagement, encouraged a gamified rush to claim results, and led some to dismiss these problems as trivial or solely machine-solvable. Bloom cites growth after adding comments in August 2025 and gives concrete usage figures - launched May 28, 2023 with ~200 problems, now 1,221 problems, over 9,000 comments, ~2,000 registered users and 10-25k daily unique visitors - then argues the site should foster human understanding rather than become an AI repository.
In response he announces policy changes: freeze new problem comments and proof-claim postings (general threads and blogs remain open); stop showing problem statuses and solved percentages; remove credit/ownership language for future results; and prioritize high-quality human expositions, soliciting well-written writeups and linking verified formalisations (e.g., Lean) where available. He will manually update the archive and assess existing proof claims, decline contributions that misrepresent AI use or plagiarize, and leave the hiatus duration open while inviting alternative ideas and other venues for purely formal AI proof repositories.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.