AI Model Darmok Solves Cryptic Crossword Faster Than Human Champion
A computer model named Darmok recently solved the final puzzle of the Times’ annual cryptic crossword competition approximately three minutes faster than Liam Hughes, the 29-year-old human winner. This milestone marks the first time AI has outperformed humans at cryptic crosswords, a puzzle type long considered resistant to automation due to its reliance on complex wordplay. The achievement resolves a debate among puzzle editors about whether machines could master such intricate linguistic challenges.
Key points
- Dr. Sid Chatterjee, a computer scientist at the University of Texas at Austin, developed the AI model Darmok to demonstrate that cryptic crosswords are solvable by machines.
- Darmok solved the final puzzle of the Times’ annual cryptic crossword competition about three minutes faster than Liam Hughes, who became the contest’s youngest human winner at age 29.
- Cryptic crosswords are significantly more complex than American-style puzzles because they rely heavily on puns, wordplay, and double meanings, which previously hindered AI progress.
- Will Shortz, the New York Times puzzle editor, originally doubted AI could handle such wide-ranging linguistic complexity, a skepticism that persisted until recent breakthroughs.
- Earlier AI successes included WebCrow beating 75 human crossworders and Dr Fill winning the American crossword puzzle tournament in 2021, but cryptics remained the final frontier.
- The victory suggests that AI can now master the most linguistically intricate puzzle formats, potentially reshaping how puzzle competitions and human cognitive benchmarks are viewed.
A significant milestone in artificial intelligence was achieved earlier this month when a computer model named Darmok outperformed human solvers in the Times’ annual cryptic crossword competition. The AI program completed the final puzzle approximately three minutes faster than Liam Hughes, a 29-year-old who secured the title of the contest’s youngest human winner. This result marks the first definitive instance of a machine surpassing humans at cryptic crosswords, a puzzle format long regarded as the last major barrier for AI due to its heavy reliance on puns, wordplay, and complex linguistic double meanings.
The development was spearheaded by Dr. Sid Chatterjee, a computer scientist at the University of Texas at Austin, who began developing Darmok in 2025 to test the limits of machine learning in linguistic puzzles. For years, experts like Will Shortz, the puzzle editor for the New York Times, argued that while AI could handle structured logic, it would struggle with the wide-ranging and subjective nature of cryptic clues. Although previous AI models like WebCrow and Dr Fill had defeated humans at American-style crosswords, those puzzles rely more on general knowledge and allusion rather than the intricate wordplay characteristic of cryptics.
The success of Darmok resolves a long-standing debate in the puzzle community about whether cryptic crosswords would remain the final frontier for AI. By mastering this complex format, the model demonstrates that modern AI can navigate nuanced linguistic structures with speed and accuracy that exceed human capabilities. This breakthrough may influence future competitions and shift the focus of cognitive benchmarking toward even more abstract or creative problem-solving tasks.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.
