The website was updated today. Unfortunately, I may have to N/A this market.

@BraydonDymm I don't think this needs to N/A. The statistics as they stood when the site stopped showing them are preserved in Terence Tao's community database, which mirrors the site's problem statuses. It moved its headline counts off the front page on 6 October in line with the site, but the full history is still published here:

erdosproblems/data/statistics_history.csv at main · teorth/erdosproblems

(breakdown at erdosproblems/STATISTICS.md at main · teorth/erdosproblems )

Comparing the database to itself: its solved count was 440 on 1 January and 579 when the counts came down, so 139 problems marked solved so far this year. (The site's own counter read 451 on 1 January; the database counts slightly differently, so database against database is the fair comparison.)

On top of that, the erdosproblems.com front page now notes significant claims relevant to 25 problems from the OpenAI release. 23 of those are problems not currently marked solved (one of them, #508, is only a partial result). If they hold up, that takes the year to over 160, already past 150, with more than two and a half months still to go before the market closes.

erdosproblems/STATISTICS.md at main · teorth/erdosproblems

A community database for the problems on the erdosproblems.com site - erdosproblems/STATISTICS.md at main · teorth/erdosproblems

One question of understanding: Does "Manifold" in the name of the site is meant in the sense of "Riemannian Manifold"? (RM has a fixed - positive - meaning in the math community.)

@KeGuThueringer I do not know the origin of this platform’s name

Is it allowed to bet here and at the same time to secretly solve lots of open Erdos problems - and decide near the end of the year how many of these solutions to disclose? To be open: The intention is to improve my betting chances here.

@KeGuThueringer yes

Does an existing solution found by AI literature review count as "solved during 2026"?

@ZongshuWu The market will be resolved by the difference between the N in “N have been solved” as stated on the website at the end of the year and 451. Right now, the website says “475 (41%) have been solved” so the difference today is 24.

@BraydonDymm I take that as a "yes" to my question

Make sure you do not measure this by how the "total solved" changes, since this will include old solutions also. You have to manually count which of those solutions happened during 2026.

@ThomaslZZ0 The market will be resolved by the difference between the N in “N have been solved” as stated on the website at the end of the year and 451. Right now, the website says “466 (41%) have been solved” so the difference today is 15.

@BraydonDymm You should reword "solved by the end of 2026" as "solved during 2026". And you may want to specify a timezone.

@vee updated

On 1 January 2025 there were 292 solved out of 895 total

@duck_master I’m using the website as the official source. Do you have a different official source?

@BraydonDymm That website is fantastic. However, the statistics it keeps, is unreliable. - old-bielefelder on erdosproblems.com

@KeGuThueringer the link you shared is the same one I’m using above. I’m not aware of any more reliable statistics.