# Models Begin Closing Long-Unsolved Problems

Frontier models with modest token budgets and minimal operator hints increasingly produce credible solutions to problems that resisted humans for decades or centuries, and third-party evaluation firms turn that backlog of famous unsolved problems into a recurring public capability yardstick, so expect repeated verified solves and disputes over how much operator guidance each required.

- Conviction: 36 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-09-14T00:00:00.000Z
- Last updated: 2026-09-14T14:04:09.672Z
- Canonical: https://polylog.news/ai/trends/ai-solves-long-open-problems
- Publisher: Polylog
- Affected regions: Global

## Recent evidence

- [confirms] Claude Fable 5.1 Produced a Reading of a 373-Year-Old Cipher in 44 Minutes (2026-09-14): Evaluation firm Vals AI reports Claude Fable 5.1 produced a reading of a 373-year-old cryptogram — ranked 28th on the standard list of famous unsolved messages — in 44 minutes using 176,000 tokens and minimal operator hints. The measurement is an independent evaluator's, not the lab's, and the token cost makes the solve cheap enough to repeat across a backlog.