A Rupert’s hypothesis counterexample has sparked fierce debate in the mathematical community after a South Korean middle schooler published a paper claiming to solve the 300-year-old problem using artificial intelligence.
South Korean Middle Schooler Publishes AI-Aided Math Paper
According to the international research archive platform Zenodo and the developer platform GitHub on the 13th, a paper titled ‘The Rhombicuboctahedron is Not Rupert’ was pre-published on the 5th. The author of the paper is Jeong Gi-hyo (15), a third-year student at Gimpo Hangaram Middle School in Gyeonggi Province. In an interview with the JoongAng Ilbo, he stated, “To solve the mathematical challenge, I wrote code using Claude and verified it using AI to write the paper.” Jeong noted that he established the research topic and direction entirely on his own without the help of an advisor or other contributors. The calculations required for the proof reportedly took about 38 hours on a laptop. A post by Jeong sharing his paper on the American community platform Reddit reached number one on the Claude code sharing board on the 10th.
Challenging the 300-Year-Old Mathematical Assumption
The claim made by the middle school student is a counterexample to the ‘Rupert’s property,’ which posits that all convex polyhedra can pass through themselves. His research asserts through AI proof that a rhombicuboctahedron cannot pass a duplicate of itself of the same size. A rhombicuboctahedron is a solid figure composed of 20 equilateral triangles, 30 squares, and 12 regular pentagons.
Academic Community Reacts to AI-Driven Research
The middle schooler’s claim of achieving a complex mathematical proof using AI has drawn mixed reactions. Comments expressed astonishment that a student without higher education could carry out research in a professional academic field. Conversely, criticisms were equally strong, with some suggesting that “the author himself probably doesn’t understand the content.” Addressing this, Jeong remarked, “While it’s true I cannot independently verify every single detail of the paper, the conclusions are highly likely to be error-free because similar studies reached the same conclusion around the same time using different methods.”
For a paper to be recognized as a legitimate academic achievement, it must undergo verification such as peer review. Uhm Sang-il, a Chief Investigator (CI) at the Discrete Mathematics Group of the Institute for Basic Science (IBS), explained, “Uploading a paper to an archive platform simply certifies the publication date and does not guarantee its contents.” Suh In-suk, a professor of mathematical sciences at Seoul National University, pointed out, “The author of a paper must understand all the details and be able to answer questions from others. Even if AI was utilized in the research, the author must personally understand and reinforce the arguments for the work to be recognized as a complete achievement.”
The Changing Landscape of Academic Research
Assessments are also emerging that AI is transforming traditional research methodologies and the roles of scholars. Recently, as OpenAI announced it found a solution to one of the ‘Seven Millennium Prize Problems’—the ‘Navier-Stokes existence and smoothness’ problem—the mathematical community is paying close attention to how the contributions of AI and human researchers should be evaluated. Choi Kyu-dong, a professor of mathematical sciences at the Ulsan National Institute of Science and Technology (UNIST), noted, “Jeong’s paper is an interesting case demonstrating that an author’s lack of professional training and the academic value of a paper can be distinct. It will be worth watching how these changes impact the field of mathematical research.”>
