Escape Meridian

MIT Technology Review · Jul 25, 2024

MIT Technology Review reports a milestone for AI in mathematical problem-solving

Markdown · JSON · Original source

MIT Technology Review’s article, published July 25, 2024, reports that ’s AlphaProof and AlphaGeometry 2 solved four out of six problems from that year’s International Mathematical Olympiad (IMO), earning the equivalent of a silver medal. The systems had to solve the problems and prove their answers correct. AlphaProof uses to train itself to prove mathematical statements in , a formal programming language; a version of Gemini translates informal math problems into formal statements. AlphaGeometry 2 was optimized for geometry problems. The systems solved problems in algebra, number theory and geometry, but left two combinatorics problems unsolved. The article reports that AlphaProof generally performs better on algebra and number theory than on combinatorics, and says Google DeepMind is still working to understand why. It presents the result as a substantial advance in AI’s mathematical problem-solving, while leaving open how the systems will scale, whether they can be made faster, and whether they can extend to other kinds of mathematics. Cambridge researcher Katie Collins says more capable systems could help mathematicians solve and invent problems, and help people learn more about how humans tackle math. She also says much remains unknown about how humans solve complex mathematics problems.

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.

MIT Technology ReviewArticle · technologyreview.comOpen original source

Original source

MIT Technology Review · Article

Source details

Original source date: Jul 25, 2024 (2024-07-25T15:30:00Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-17T15:49:00.682Z. First observed: 2026-10-29T17:30:16.347Z. Retrieved: 2026-10-29T17:31:45.505Z.

Referenced by

Recorded citations in public readings. A reference alone does not establish agreement or influence.

No public references recorded here yet.

Reference data (JSON)

Limits

Text as retrieved on October 29, 2026; originally published July 25, 2024.

Open question

How will the systems scale, and can they be made faster or extended to other kinds of mathematics?

Source-stated

MIT Technology Review reports that Google DeepMind’s AlphaProof and AlphaGeometry 2 solved four of the IMO problems and earned the equivalent of a silver medal.

The systems—called AlphaProof and AlphaGeometry 2—worked together to successfully solve four out of six problems from this year’s International Mathematical Olympiad (IMO), a prestigious competition for high school students. They won the equivalent of a silver medal.

Source-stated

The article says AlphaProof trains itself to prove mathematical statements in Lean, with Gemini translating informal problems into formal statements.

AlphaProof, a reinforcement-learning-based system that trains itself to prove mathematical statements in the formal programming language Lean. The key is a version of DeepMind’s Gemini AI that’s fine-tuned to automatically translate math problems phrased in natural, informal language into formal

Source-stated

The article reports that AlphaProof generally performs better on algebra and number theory than on combinatorics, and that the team is working to understand why.

“Generally, AlphaProof performs much better on algebra and number theory than combinatorics,” says Alex Davies, a research engineer on the AlphaProof team. “We are still working to understand why this is, which will hopefully lead us to improve the system.”

Source-stated

The article reports Collins’s view that more capable systems could help mathematicians solve and invent problems, and says much remains unknown about how humans solve complex mathematics problems.

Creating AI systems that can solve more challenging mathematics problems could pave the way for exciting human-AI collaborations, helping mathematicians to both solve and invent new kinds of problems, says Collins. This in turn could help us learn more about how we humans tackle math.

Source-stated

The article reports Myers’s open questions about how the systems may scale, their speed, and whether they can extend to other mathematics.

“It will be interesting to see how things scale and whether they can be made faster, and whether it can extend to other sorts of mathematics,” he said.

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.

Reinforcement learning

Training by trial and error, in which a system is rewarded for the outcomes its designers want.

Google DeepMind

Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.

Lean

Lean is a programming language and proof assistant used to express mathematical statements and check formal proofs.

Editorial provenance

Version 1. Publication ID: 40d3d5a2-9c4d-805d-d938-8c6d83bfe930.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-17T15:45:34.895Z. 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: c1dac8d76155ce25e17a19f7c8d27900a8c3a7780c5a3e569a1eb8261964a1dc. Source text hash: 25e8edebc2c1d836e39c194cafd67cdc57d75b5a967ebd23c3d792c9f3233164.

Revision and source-check history

Version 1 · 2026-11-17T15:49:00.682Z

MIT Technology Review reports a milestone for AI in mathematical problem-solving

MIT Technology Review’s article, published July 25, 2024, reports that ’s AlphaProof and AlphaGeometry 2 solved four out of six problems from that year’s International Mathematical Olympiad (IMO), earning the equivalent of a silver medal. The systems had to solve the problems and prove their answers correct. AlphaProof uses to train itself to prove mathematical statements in , a formal programming language; a version of Gemini translates informal math problems into formal statements. AlphaGeometry 2 was optimized for geometry problems. The systems solved problems in algebra, number theory and geometry, but left two combinatorics problems unsolved. The article reports that AlphaProof generally performs better on algebra and number theory than on combinatorics, and says Google DeepMind is still working to understand why. It presents the result as a substantial advance in AI’s mathematical problem-solving, while leaving open how the systems will scale, whether they can be made faster, and whether they can extend to other kinds of mathematics. Cambridge researcher Katie Collins says more capable systems could help mathematicians solve and invent problems, and help people learn more about how humans tackle math. She also says much remains unknown about how humans solve complex mathematics problems.

Original source

MIT Technology Review · Article

Source details

Original source date: Jul 25, 2024 (2024-07-25T15:30:00Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-17T15:49:00.682Z. First observed: 2026-10-29T17:30:16.347Z. Retrieved: 2026-10-29T17:31:45.505Z.

Limits

Text as retrieved on October 29, 2026; originally published July 25, 2024.

Open question

How will the systems scale, and can they be made faster or extended to other kinds of mathematics?

Source-stated

MIT Technology Review reports that Google DeepMind’s AlphaProof and AlphaGeometry 2 solved four of the IMO problems and earned the equivalent of a silver medal.

The systems—called AlphaProof and AlphaGeometry 2—worked together to successfully solve four out of six problems from this year’s International Mathematical Olympiad (IMO), a prestigious competition for high school students. They won the equivalent of a silver medal.

Source-stated

The article says AlphaProof trains itself to prove mathematical statements in Lean, with Gemini translating informal problems into formal statements.

AlphaProof, a reinforcement-learning-based system that trains itself to prove mathematical statements in the formal programming language Lean. The key is a version of DeepMind’s Gemini AI that’s fine-tuned to automatically translate math problems phrased in natural, informal language into formal

Source-stated

The article reports that AlphaProof generally performs better on algebra and number theory than on combinatorics, and that the team is working to understand why.

“Generally, AlphaProof performs much better on algebra and number theory than combinatorics,” says Alex Davies, a research engineer on the AlphaProof team. “We are still working to understand why this is, which will hopefully lead us to improve the system.”

Source-stated

The article reports Collins’s view that more capable systems could help mathematicians solve and invent problems, and says much remains unknown about how humans solve complex mathematics problems.

Creating AI systems that can solve more challenging mathematics problems could pave the way for exciting human-AI collaborations, helping mathematicians to both solve and invent new kinds of problems, says Collins. This in turn could help us learn more about how we humans tackle math.

Source-stated

The article reports Myers’s open questions about how the systems may scale, their speed, and whether they can extend to other mathematics.

“It will be interesting to see how things scale and whether they can be made faster, and whether it can extend to other sorts of mathematics,” he said.

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.

Reinforcement learning

Training by trial and error, in which a system is rewarded for the outcomes its designers want.

Google DeepMind

Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.

Lean

Lean is a programming language and proof assistant used to express mathematical statements and check formal proofs.

Editorial provenance

Version 1. Publication ID: 40d3d5a2-9c4d-805d-d938-8c6d83bfe930.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-17T15:45:34.895Z. 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: c1dac8d76155ce25e17a19f7c8d27900a8c3a7780c5a3e569a1eb8261964a1dc. Source text hash: 25e8edebc2c1d836e39c194cafd67cdc57d75b5a967ebd23c3d792c9f3233164.

passed · 2026-11-17T15:45:34.895Z · anthropic/claude-opus-5.5

Publication: 40d3d5a2-9c4d-805d-d938-8c6d83bfe930

attribution — pass: The reading consistently attributes the report to MIT Technology Review. It attributes quoted views to the people the article quotes, including Collins and Myers. The 'still working to understand why' point is attributed to Google DeepMind, which loosely matches its source, AlphaProof team engineer Alex Davies.

says Alex Davies, a research engineer on the AlphaProof team. “We are still working to understand why this is

summary — pass: The summary faithfully covers the reported result and how the two systems work. It also covers their strengths and weaknesses, expert views and open questions. Lean and reinforcement learning are explained (the latter via the definitions list). Combinatorics and the IMO's high-school audience are not explained, but this does not block understanding. The question follows from Myers's stated open issues.

They won the equivalent of a silver medal.

date — pass: The reading gives July 25, 2024, which matches the originalPublished instant's UTC day. 'That year's' IMO matches the article's 'this year's' framing. The page's later sidebar items (Deep Dive links) are neither used nor relied on, and there is no hindsight framing. The kind is publisher-page, so no archive disclosure is needed.

problems given to humans competing in this year’s IMO

support — pass: Each claim matches its verbatim excerpt. The title's 'milestone' is supported by the article's statement that this is the first medal-level AI performance. Collins's possible benefits are kept conditional ('could'). Scaling, speed and generality are presented as open questions, not outcomes. Source claims are not treated as independently verified.

This is the first time any AI system has been able to achieve a medal-level performance on IMO questions.

media — pass: No media was selected and no verified media candidates were supplied, so no invented or mismatched media is present.

Challenge or correct this reading

Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.