erdos-top10-problems.csvOpen problems sit on the right edge. Hover a point for resolver and notes.
On 2026-04-16, Thomas Bloom — the mathematician who built and maintains erdosproblems.com — published a list of what he considers the most important Erdős problems, solved and unsolved [@bloom2026top10]. The list has ten entries naming twelve of the site's problem numbers: 3, 139, 4, 20, 28, 52, 61, 67, 77, 90, 571 and 713. The progressions entry spans numbers 3 and 139, and the Turán-exponents entry spans 571 and 713, so this ledger scores twelve rows. Bloom states the list's basis:
"Naturally this is very subjective (and probably will change even for myself day to day), and others are welcome to post in the comments their own candidates." — Thomas Bloom, Top 10 Erdős Problems, erdosproblems.com forum, 2026-04-16 [@bloom2026top10]
A "discovery" in this series is a row with status resolved, dated by the year of the resolving work. Statuses and years are scored from the blog post's own discussion of each problem; the one resolution after the post's date (problem 90) is scored from the sources named in its register entry.
The collection-wide cumulative index redraws the ledger as rows remaining:
"Erdős' belief in the difficulty of this problem was well-founded - it was over 60 years since he asked this question till it was solved by Maynard [Ma16] and Ford, Green, Konyagin, and Tao [FGKT16]" — Thomas Bloom, Top 10 Erdős Problems, erdosproblems.com forum, 2026-04-16 [@bloom2026top10]
../math-erdos/'s independent dating"This was resolved in the affirmative by Tao [Ta16]" — Thomas Bloom, Top 10 Erdős Problems, erdosproblems.com forum, 2026-04-16 [@bloom2026top10]
At the post's date Bloom recorded the problem open:
"Erdős conjectured many times (and offered $500 for a solution) that this was best possible" — Thomas Bloom, Top 10 Erdős Problems, erdosproblems.com forum, 2026-04-16 [@bloom2026top10]
The resolution is the vendor announcement of 2026-05-20, titled
"Model disproves discrete geometry conjecture" — OpenAI, announcement title, 2026-05-20 [@openai2026discretegeometry]
with the argument digested by nine mathematicians into a human-verified
account (arXiv 2605.20695) and the explicit exponent, greater than 1.014,
stated by Sawin (arXiv 2605.20579). The catalogue's imputed-year dataset
records the same event: problem 90, disproved, 2026, dated by the AI wiki at
2026-05-20 (../math-erdos/,
erdos-solution-years.csv).
"the resolution of [139] was achieved by Szemerédi [Sz75] in 1975 using a very different, purely combinatorial, argument" — Thomas Bloom, Top 10 Erdős Problems, erdosproblems.com forum, 2026-04-16 [@bloom2026top10]
The rows are hand-scored from the blog post: each entry's discussion states
whether the problem is solved and by whom, and the resolution years for rows
4, 67 and 139 equal the imputed solution years the
math-erdos folder derives independently from the
catalogue's own pages. Row 90's status postdates the post and is scored from
the 2026-05-20 sources quoted in its register entry; its source column
names them. There is no fetch.py.
figure.py calls the shared problem_list_chart() in
../../lib/families.py with ai_problem="90",
counting rows with status resolved and a non-empty resolved_year by year;
the ai_caption argument sets the annotation to that row's verification
standing (an AI disproof with a human-verified account). The cumulative view
is the shared ledger_remaining_chart().
One row. Problem 90 (unit distances) is credited to an OpenAI model on 2026-05-20, per the vendor announcement and the nine-author verified account quoted in its register entry [@openai2026discretegeometry]. No other row's resolution or open status carries any AI credit in the blog post or on the catalogue's pages as of 2026-08-14. The catalogue-wide AI-contribution record is the frozen wiki [@erdosproblems2026wiki].