80-minute breakthrough! 23-year-old amateur uses ChatGPT assistance to solve a 60-year-old math problem

A 23-year-old amateur enthusiast, with the assistance of GPT-5.4 Pro, unraveled the Erdős mathematical conjecture that had been unsolved for 60 years (#1196). The model connected integer structures through interdisciplinary Markov processes, proposing a proof path that humans had never attempted.

According to OpenAI’s official announcement on April 28 and an in-depth report by Scientific American on April 24, a 60-year-old Erdős mathematical conjecture (#1196) was solved with the help of the flagship reasoning model GPT-5.4 Pro. On the same day, OpenAI held an official podcast featuring researcher Sébastien Bubeck and Ernest Ryu in conversation with host Andrew Mayne, publicly explaining the event details and significance.

Main figure of the event: 23-year-old amateur Liam Price

Solver Liam Price, 23, with no advanced mathematics training, occasionally collaborates with Kevin Barreto, a second-year mathematics student at Cambridge University. Price stated: “I didn’t even know what this problem was—I just sometimes throw Erdős problems to AI to see what it comes up with.”

In the afternoon of a Monday in April 2026, Price submitted the Erdős #1196 thread for community review. Scientific American published an in-depth report on April 24, and OpenAI’s official podcast on April 28 was a week later, providing an external explanation.

Mathematical breakthrough: Connecting integer structures via Markov processes, Tao comments “the first step humans took was wrong”

Erdős #1196 falls within the research scope of “primitive sets”—a collection of integers where no element divides another. Erdős’s conjecture states: as the elements of such sets approach infinity, the maximum value of the “Erdős sum fraction” will drop precisely to 1.

GPT-5.4 Pro’s proof takes an approach “humans have never tried before”: linking the structure of integers (anatomy of integers) with Markov process theory. This interdisciplinary bridge was previously not part of anyone’s research path.

Fields Medalist and renowned mathematician Terence Tao has been widely quoted for his two comments on this event. He described “this problem is different from others—humans have indeed seen it, but collectively, the first step was wrong,” and added, “the significance of this contribution to the study of integer structures far exceeds solving this particular Erdős problem itself.”

Another mathematician from Stanford University, Jared Duker Lichtman, stated that AI’s approach verified his long-standing intuition: there exists “some kind of common unifying sense” among these problems.

OpenAI 4/28 Disclosure: Podcast discussion and subsequent verification

On April 28, OpenAI officially invited researchers Sébastien Bubeck and Ernest Ryu to discuss “AI’s role in mathematical research” with host Andrew Mayne on their podcast. OpenAI’s tweet read: “Earlier this month, a 60-year-old Erdős problem was solved with GPT-5.4 Pro. Now AI is good at math—what happens next?”

As of the time of writing, the proof submitted by Price remains in the community verification stage on the erdosproblems.com forum and has not yet passed formal peer review; TheDecoder reported on April 15 that “formal verification is still ongoing.” Today’s OpenAI podcast disclosure is a level of external communication and does not mean the full mathematical proof has been verified—readers interested in follow-up can monitor the Erdős Problems forum thread #1196.

• This article is reprinted with permission from: 《Chain News》
• Original title: “23-year-old amateur uses ChatGPT to solve 60-year-old math problem: cracked in 80 minutes”
• Original author: Elponcrab