# MIT Technology Review asks whether AI can help mathematicians make new discoveries [Canonical reading](https://escapemeridian.com/readings/mit-technology-review-asks-whether-ai-can-help-mathematicians-45e1a50d) MIT Technology Review published this article on June 4, 2025. It examines DARPA’s effort to encourage mathematicians and AI researchers to build an “AI coauthor”: a tool intended to break complex problems into smaller ones. The article distinguishes success on familiar competition problems from progress on open research questions, where solutions may require long sequences of unfamiliar steps. It describes AI systems that can test mathematical ideas, suggest possible directions and help rule out dead ends, including work on a long-standing conjecture that did not solve the conjecture itself. The article also emphasizes mathematicians’ role in exploring which ideas matter and recognizing new ways to think. Its conclusion is that AI may act as an advance scout for human intuition, finding directions and dead ends that could save mathematicians months or years, while the article says true breakthroughs still come from people. Geordie Williamson cautions that technology companies expect this to change, but ends: “But you know—we’ll see.” The article does not establish whether the described tools will lead to new theorems or how broadly they will help mathematical research. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source MIT Technology Review · Article Original source date: Jun 4, 2025 (2025-06-04T08:21:24Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-11-13T17:53:25.532Z. First observed: 2026-10-30T14:50:50.420Z. Retrieved: 2026-10-30T14:51:56.184Z. - [What’s next for AI and math](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Limits Text as retrieved on October 30, 2026; originally published June 4, 2025. The article gives no measured basis for how much time the described tools save mathematicians; it presents this as a potential effect. The article does not establish whether the tools will produce new theorems or how broadly they will help mathematical research. ## Open question Will AI tools that suggest research directions and rule out dead ends lead to new theorems, and how broadly will they help mathematical research? ## Source-stated MIT Technology Review reports that DARPA’s expMath initiative aims to encourage mathematicians and AI researchers to develop an AI coauthor for breaking complex problems into smaller ones. > The goal for expMath is to encourage mathematicians and artificial-intelligence researchers to develop what DARPA calls an AI coauthor, a tool that might break large, complex math problems into smaller, simpler ones - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article distinguishes performance on competition problems from the exploratory work involved in mathematical research. > Math Olympiad problems often involve being able to carry out clever tricks, whereas research problems are much more explorative and often have many, many more moving pieces. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article reports that Gukov and colleagues used AI to show that a proposed counterexample to the Andrews-Curtis conjecture was false; they did not solve the conjecture. > Gukov and his colleagues didn’t solve the AC conjecture, but they found that a counterexample (suggesting that the conjecture is false) proposed 40 years ago was itself false. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article concludes that AI tools may help mathematicians find directions and avoid dead ends, while attributing the view that breakthroughs still come from people to its own narration. > They can discover new directions and point out dead ends, saving mathematicians months or years of work. But the true breakthroughs will still come from the minds of people, as has been the case for thousands of years. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated Geordie Williamson says technology companies expect the present limits to change, but leaves the outcome open. > “There’s plenty of tech companies that tell us that won’t last long,” says Williamson. “But you know—we’ll see.” - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source and contributor context MIT Technology Review · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: e00ad644-66d1-8fbc-db59-d4c31d6c5d26. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T17:48:43.020Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: 4371102d1ead231236bfa136cc318d222b75771bd62798af942eb66ce8e53349. Source text hash: 57eb2734778ff1c1713229b9aaeade0318b75641291797d33d494115f001c391. ## Revision and source-check history ### Version 1 — 2026-11-13T17:53:25.532Z MIT Technology Review published this article on June 4, 2025. It examines DARPA’s effort to encourage mathematicians and AI researchers to build an “AI coauthor”: a tool intended to break complex problems into smaller ones. The article distinguishes success on familiar competition problems from progress on open research questions, where solutions may require long sequences of unfamiliar steps. It describes AI systems that can test mathematical ideas, suggest possible directions and help rule out dead ends, including work on a long-standing conjecture that did not solve the conjecture itself. The article also emphasizes mathematicians’ role in exploring which ideas matter and recognizing new ways to think. Its conclusion is that AI may act as an advance scout for human intuition, finding directions and dead ends that could save mathematicians months or years, while the article says true breakthroughs still come from people. Geordie Williamson cautions that technology companies expect this to change, but ends: “But you know—we’ll see.” The article does not establish whether the described tools will lead to new theorems or how broadly they will help mathematical research. ## Original source MIT Technology Review · Article Original source date: Jun 4, 2025 (2025-06-04T08:21:24Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-11-13T17:53:25.532Z. First observed: 2026-10-30T14:50:50.420Z. Retrieved: 2026-10-30T14:51:56.184Z. - [What’s next for AI and math](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Limits Text as retrieved on October 30, 2026; originally published June 4, 2025. The article gives no measured basis for how much time the described tools save mathematicians; it presents this as a potential effect. The article does not establish whether the tools will produce new theorems or how broadly they will help mathematical research. ## Open question Will AI tools that suggest research directions and rule out dead ends lead to new theorems, and how broadly will they help mathematical research? ## Source-stated MIT Technology Review reports that DARPA’s expMath initiative aims to encourage mathematicians and AI researchers to develop an AI coauthor for breaking complex problems into smaller ones. > The goal for expMath is to encourage mathematicians and artificial-intelligence researchers to develop what DARPA calls an AI coauthor, a tool that might break large, complex math problems into smaller, simpler ones - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article distinguishes performance on competition problems from the exploratory work involved in mathematical research. > Math Olympiad problems often involve being able to carry out clever tricks, whereas research problems are much more explorative and often have many, many more moving pieces. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article reports that Gukov and colleagues used AI to show that a proposed counterexample to the Andrews-Curtis conjecture was false; they did not solve the conjecture. > Gukov and his colleagues didn’t solve the AC conjecture, but they found that a counterexample (suggesting that the conjecture is false) proposed 40 years ago was itself false. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated The article concludes that AI tools may help mathematicians find directions and avoid dead ends, while attributing the view that breakthroughs still come from people to its own narration. > They can discover new directions and point out dead ends, saving mathematicians months or years of work. But the true breakthroughs will still come from the minds of people, as has been the case for thousands of years. - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source-stated Geordie Williamson says technology companies expect the present limits to change, but leaves the outcome open. > “There’s plenty of tech companies that tell us that won’t last long,” says Williamson. “But you know—we’ll see.” - [Supporting source](https://www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math/) ## Source and contributor context MIT Technology Review · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: e00ad644-66d1-8fbc-db59-d4c31d6c5d26. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T17:48:43.020Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: 4371102d1ead231236bfa136cc318d222b75771bd62798af942eb66ce8e53349. Source text hash: 57eb2734778ff1c1713229b9aaeade0318b75641291797d33d494115f001c391. ### passed — 2026-11-13T17:48:43.020Z — anthropic/claude-opus-5.5 Publication: e00ad644-66d1-8fbc-db59-d4c31d6c5d26 attribution — pass: The reading attributes the article to MIT Technology Review, and attributes quoted views to named people: Williamson's closing quote, Gukov's work, and the narration's 'advance scouts' conclusion. The page is one article. The trailing 'Deep Dive' list of related stories is page furniture, not the source itself. > Perhaps AI tools like AlphaEvolve and PatternBoost are best thought of as advance scouts for human intuition. summary — pass: The reading leads with the actor (DARPA via MIT Technology Review) and explains the 'AI coauthor' concept in plain words. It keeps the article's central contrast between competition tricks and exploratory research, notes that the AC conjecture was not solved, and preserves Williamson's open-ended close. It omits benchmark details (AIME, FrontierMath) and Gukov's path-finding idea, which flattens some of the argument. This does not misstate the article. The question follows from the article's own uncertainty. > But we're certainly not getting inundated with new theorems at this stage. date — pass: The reading's date of June 4, 2025 matches the UTC day of originalPublished, and the retrieval date is disclosed. The page was edited later (dateModified 2025-09-25). All cited passages are plainly part of the article body. The sidebar item referencing the AlphaGo ten-year anniversary is a later addition, and the reading does not use it. There is no hindsight framing, and the article's hopes are reported as possibilities. > In April, the US Defense Advanced Research Projects Agency kicked off a new initiative called expMath support — pass: Each claim matches its verbatim excerpt from the article body. The time-saving point is correctly framed as potential and as the article's narration. The limitations correctly note that no measured basis is given. Gukov's result is accurately limited to refuting a proposed counterexample. Nothing implies benefits beyond the source. > Gukov and his colleagues didn’t solve the AC conjecture, but they found that a counterexample (suggesting that the conjecture is false) proposed 40 years ago was itself false. media — pass: No media was selected, and no verified media candidates were supplied, so there is no identity or relevance problem. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. Some references could not be checked. Reload to try again. [Reference data (JSON)](/readings/45e1a50d9ebd467b58def8e0876d8fa69e7c382edb6eeb5bab212c37abea1fd7/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/mit-technology-review-asks-whether-ai-can-help-mathematicians-45e1a50d.json)