Erdős problems in the AI era
Explore recorded resolutions and formalized proofs over time, alongside AI model releases
Dates show database updates, not necessarily when the mathematics was done. Release markers provide context, not attribution.
Model releases
Loading historical snapshots…
Scroll sideways to explore the diagram, or read the data table
Problem status
Snapshot: September 28, 2026“Problems”: % of all problems. Lean columns: % within each status. A dash means no data or an empty status.
| Status | Problems | Lean-formalized | Not Lean-formalized |
|---|---|---|---|
| Proved | 336 (27.5%) | 183 (54.5%) | 153 (45.5%) |
| Disproved | 139 (11.4%) | 97 (69.8%) | 42 (30.2%) |
| Otherwise solved | 101 (8.3%) | 33 (32.7%) | 68 (67.3%) |
| Completely open | 593 (48.6%) | — | — |
| Decidable | 8 (0.7%) | — | — |
| Falsifiable | 25 (2.0%) | — | — |
| Verifiable | 7 (0.6%) | — | — |
| Not provable in ZFC | 3 (0.2%) | — | — |
| Not disprovable in ZFC | 6 (0.5%) | — | — |
| Independent of ZFC | 3 (0.2%) | — | — |
| All problems | 1,221 (100.0%) | — | — |