When the IBM supercomputer Deep Blue beat Garry Kasparov at chess in 1997, it was an early victory for artificial intelligence over the human mind, and the culmination of decades of attempts. Almost 30 years later, a computer program has finally outdone humans at solving cryptic crosswords.<\/p>\n
Earlier this month, a model called Darmok solved the final puzzle of the Times\u2019s annual cryptic crossword competition about three minutes faster than Liam Hughes, 29, who became the contest\u2019s youngest (human) winner<\/a>.<\/p>\n After Deep Blue\u2019s victory, the New York Times\u2019s puzzle editor, Will Shortz, set engineers a challenge<\/a>. \u201cIs the computer going to be able to solve the clues involving puns and wordplay? I don\u2019t think so,\u201d he wrote. \u201cIn a much more wide-ranging and, frankly, complex game like crosswords, we humans still rate just fine.\u201d<\/p>\n Less than a decade later, a computer named WebCrow beat 75 human crossworders<\/a> at an AI conference, and by 2021 a program named Dr Fill \u201cwon\u201d the American crossword puzzle tournament<\/a>, the most prestigious of its type for American-style crosswords, which rely on a combination of general knowledge, allusion and double meanings.<\/p>\n But cryptics are much trickier, and as recently as 2024 my colleague Alan Connor was asking<\/a>: \u201cWill the cryptic crossword remain the last thing AI can\u2019t conquer?\u201d<\/p>\n Dr Sid Chatterjee, a computer scientist at the University of Texas in Austin, thought not, and in 2025 set out to prove it. He began work on Darmok, named after a Star Trek episode<\/a> in which the crew attempt to converse with a species that speaks only in allegories \u2013 an experience akin to solving a cryptic.<\/p>\n Sid Chatterjee and Darmok after solving the final puzzle in the Times\u2019s cryptic crossword championship.<\/p>\n Darmok\u2019s first tilt at the Times championship ended in failure when it gave up on the final puzzle after 18 minutes. Its triumph this year was partly as a result of successfully solving a clue that tripped up three of the five finalists, including 14-time champion Mark Goodliffe:<\/p>\n double quotation markWolf trap missing its lowest crosspiece? (5)\u201d<\/p>\n The letters filled in as part of other clues gave the pattern S_A_F. Both SCARF and SNARF mean \u201cto eat greedily\u201d (synonyms for \u201cwolf\u201d) but the rest of the wordplay relied on a visual element.<\/p>\n \u201cTrap\u201d indicates the synonym \u201cSNARE\u201d, and when the bottom line of the E is removed \u2013 \u201cmissing its lowest crosspiece\u201d \u2013 it becomes an F, making SNARF the answer.<\/p>\n While one might expect visual clues like these to be more difficult for an AI, Chatterjee said he couldn\u2019t detect a pattern in the types of clues Darmok struggles most with. It does find puzzles by John Henderson (known to Guardian readers as Enigmatist<\/a>) harder than most, though.<\/p>\n If you\u2019d like to challenge yourself against Darmok, it solved this Fed crossword<\/a> from last week in four minutes and 10 seconds. Good luck.<\/p>\n Darmok is based on four AI models: Gemini 3.1 Pro, Claude Fable 5, GPT-5 and GPT-6 Astra. The solving is primarily done by the first two, with the others joining in if needed. Chatterjee attributed Darmok\u2019s improvement this year to the extra power behind these frontier models, as well as some of his refinements, such as making Darmok wait until it had attempted all the clues before inserting answers.<\/p>\n The next frontier for AI and cryptic crosswords \u2013 setting a satisfying puzzle \u2013 will be an even bigger jump. \u201cIt is a very, very difficult problem because you have got to make a sentence that parses both in normal English and in cryptic English smoothly, and those two grammars are very, very different,\u201d said Chatterjee. \u201cIf [beating humans at cryptic crosswords] was the Mount Everest challenge, I call this next one the Olympus Mons challenge.\u201d<\/p>\n Clue of the week<\/p>\n double quotation markProportion of inspiration and perspiration? It\u2019s roughly 45% and 42% (5)<\/p>\n