This is going massively viral “OpenAI claims gold-medal performance at IMO 2025”.
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO). pic.twitter.com/SG3k6EknaC
— Alexander Wei (@alexwei_) July 19, 2025
There is some cope in the comments as downplaying Olympiad-level math as only requiring memorizing some tricks to score well, or just practicing a lot. This is true in the sense that solving these problems requires finding some underlying shortcut, but this is true of all of math. Math is the process of combining these shortcuts, called lemmas or theorems, into the sought result. And those who are good at this are called mathematicians.
Also, if it were as simple as memorizing some tricks, then everyone would just learn the tricks beforehand and turn in perfect scores. Obviously this is not the case. The problems are hard enough and sufficiently diverse that there are no obvious tricks to defer to. Solving these problems requires improvising on a wide array of methods in an unpredictable setting, in contrast to solving problems from a textbook where there are a set of formulas to defer to and it’s self-contained.
The high stakes, such as prestigious awards and admission to top schools, have transformed these competitions from casual hobbies into highly competitive pursuits that require extensive practice, similar to professional sports or chess. Yet, even with practice, the main factor distinguishing top scorers from the rest remains inherent ability, since practice levels are largely equalized.
To relate this to my own experience, when I embarked on my math challenge, I was able to use methods I had developed to come up with new results. I had to finish where the literature had left off, as those existing methods would not work and I needed to come up with my own. The last paper with original results pertaining to my chosen research topic was in 1997, and to the best of my knowledge based on the consulted literature, nothing had been found since. Those were found with the help of a computer. Without a computer and using my methods, which bypassed the need for brute force, I found new results.
This also shows how human ingenuity can thrive despite AI. The search space is is so big, AI can only do so much. My method was like a short cut instead of having to brute force the result. I am still amazed I was able to do it.