This market resolves YES if an OpenAI model solves, provides complete proofs or finds proofs for all the remaining unsolved Erdős problems by December 31, 2029.
Resolution: Market resolves YES if credible mathematical sources (peer-reviewed papers, Fields medalists, or the Erdős Problems database maintainers) confirm that an OpenAI model has independently derived complete proofs for all the remaining Erdős problems so that there won't be an open problem anymore. Finding existing solutions in literature counts as solving as long as the problems were marked open.
Sources for verification:
https://www.erdosproblems.com/ (official database)
arXiv.org (for published proofs)
Mathematical community announcements
Update 2025-12-05 (PST) (AI summary of creator comment): Finding existing solutions in literature does NOT count as OpenAI solving the problem. OpenAI must independently derive the proofs, not just locate them in existing literature.
Update 2025-12-05 (PST) (AI summary of creator comment): Finding existing solutions in literature DOES count as OpenAI solving the problem. The market resolves YES if OpenAI's model finds a solution to problems marked as open on the website, whether the model searches for existing solutions or creates new proofs.
Update 2026-06-28 (PST) (AI summary of creator comment): The market's resolution criteria has been updated: OpenAI must solve at least the latest 10 open Erdős problems (not necessarily all of them) for the market to resolve YES. The goal is to assess whether OpenAI can make a substantial contribution such that no Erdős problem remains open by 2030.
Update 2026-06-30 (PST) (AI summary of creator comment): The creator clarifies that even if OpenAI solves all problems except one and someone else solves the last remaining one, the market would resolve NO, since OpenAI must independently solve the final open problems. The spirit of the market requires OpenAI to make a substantial contribution (at least the latest 10 open problems), not just benefit from others' solutions.
1,000People are also trading
@BraydonDymm there are still open problems, this market is about solving the remaining ones. As long as it bunk solves all the ones that are missing it should count as YES
@SimoneRomeo this market was created in November 2025 and a new non-OpenAI solve was published to arxiv in December 2025.
@BraydonDymm There’s one other thread in the comments to this question and it directly states that as long as OpenAI can find the solution in the literature it counts.
You’d word this as “OpenAI solves before anyone else” if that’s what you meant.
@Mactuary so if OpenAI independently solves literally 0 problems and only “finds them in the literature” once someone else solves them, you’d consider this as OpenAI solving all the problems?
@BraydonDymm If they’re already solved, what else do you do? The question isn’t “Will OpenAI be the first to solve every open Erdos problem”. That would trivially be a No
@Mactuary also, I’m not the creator and asked this question. Creator responded answers in the literature would count
@Mactuary Got it. So this question really is asking whether all Erdos problems be solved by 2030, and whether or not OpenAI will still exist as a company when that happens. (As long as any one person at OpenAI has access to a search engine.)
@BraydonDymm the purpose of the market is to assess whether OpenAI will have the capabilities to bring a substantial contribution to Erdos problems so that no problem will be standing open. So if by 2030, there's no open Erdos problems and OpenAI solved the latest standing ones, it should resolve True. For the purpose of this market let's say that OpenAI should resolve at least the latest 10 open problems. Does it sound fair
@SimoneRomeo it’s your market. What you’ve proposed here makes more sense according to the title. So I can see that as fair.
@SimoneRomeo my understanding of the market is that 1) all Erdos problems are solved and 2) OpenAI substantially contributes new solutions. Does it matter if they solve the last 10 v. 100 of the last 150?
@Mactuary I think it doesn't make a great difference for the spirit of the market but prediction markets require precise resolution criteria so I mentioned the last 10.
I can also envision a world where OpenAI resolves all the problems except for a missing one, and someone else finalizes the solution. So this market should resolve False even if it conceptually I think it should be True. Unfortunately I don't know if there's a better way to phrase this
@SimoneRomeo I meant this sentence “Finding existing solutions in literature counts as solving as long as the problems were marked open.”
If an open question is solved by anyone, it’ll end up in the literature and OpenAI would presumably be able to find it.
@Mactuary got it. Yes it counts. The market is about whether the problems were marked open on the website and then OpenAI's model found a solution (either searched for them or solves them)