Darmok beats every human at the Times Crossword Championship, solving all 7 puzzles error-free
rohanpaul_ai · x · 2026-11-22
A UT Austin paper introduces Darmok, the first automated system to reach superhuman performance on British cryptic crosswords, competing at the 55th Times Crossword Championship in London against an elite human field.
- Why it's hard: cryptic clues hide an unmarked definition and wordplay under misleading surface readings, sparse grids leave nearly half of each word's letters unchecked, and tournament rules rank any grid with one error behind all perfect ones.
- Results: Darmok solved all seven championship puzzles error-free and finished faster than every human competitor in every round, taking 6m 52s on the Final — 3m 13s ahead of the human champion.
- How it works: to combine tournament-grade accuracy, speed and reliability despite API stalls, timeouts, safety refusals and reasoning loops, it cross-checks clues across three independent models in rounds, uses verified crossing letters to unlock harder clues, and fills gaps with a whole-puzzle model plus dictionary lookups without overwriting confirmed words.
- Paper: arXiv:2611.13557 by Siddhartha Chatterjee.
Related event: UT Austin's Darmok Beats Human Champions in Times Crossword(2 posts)→
More from AGI Musings
- 2010s mainstream AI safety planned an aligned singleton overriding governments — jessi_cata · 2026-11-22
- Benjamin Bratton mocks The Atlantic's AI coverage as 'AI cope' — bratton · 2026-11-22
- Should an AI-assisted RH proof win the Fields Medal? Math world debates — NathanpmYoung · 2026-11-22
- Fields Medalist Bao Chau Ngo on AI proof of Riemann Hypothesis: treat AI output as data — Dr_Atoosa · 2026-11-22
- AI math pace outruns expert timelines: two Millennium Prize-level problems now reportedly solved — haider1 · 2026-11-22
- eigenrobot: In 2020 I'd have bet we'd prove the Riemann hypothesis before fixing obesity — eigenrobot · 2026-11-22