Where have the International Math Olympiad Gold Medallists Ended up - Part ... Four of Three?
A new job for IMO gold medallists just dropped
I suppose we’re not finished yet
I had considered complete my three part series documenting where the International Math Olympiad gold medallists ended up. But recent news requires an update.
When I wrote the post, OpenAI had already made it into the top four and was by far the newest of the companies that were recruiting these candidates.
It seems like they have now built an LLM capable of reasoning through the very difficult and previously unseen IMO problems.
Notably, OpenAI’s Alex Wei, who ran the project that got the gold, has himself a gold medal in the IOI. He wouldn’t have been included in my original dataset but an IOI gold is still no mean feat.
They put the proofs of the answers in a public github repo for all to check.
I have nothing to add to the commentary on AI progress. It’s incredible that we have built machines that can talk to us and are probably smarter than many of us in many measurable ways. Will it continue? Will it be good for humanity? All these questions are beyond my pay grade.
And so up until this point I hadn’t considered this something worth writing about. Until I dug further into the comments on the HackerNews thread.
The new IMO medallist job - AI tutor
Per this comment it seems like AI companies have been hiring IMO medallists to teach AIs how to solve hard math problems.
Eg
And for some reason they’re only willing to pay up to $65/hour?
Reading through them it seems very possible that all three of these job descriptions is for the same job at xAI.
The original commenter however also writes:
And Outlier/Scale, which was bought by Meta (via Scale), had many IMO-required Math AI trainer jobs on LinkedIn. I can't find those historical ones though.
The job seems to be “write proofs of hard math problems so the AI can read them and learn from you”.
In other words, for $65/hour you can become an important part of our training data!
But there’s more! Your potential job is not just limited to creating training data for super intelligent AIs, but also grading them!
In his tweet thread, Alex Wei talks about having previous IMO medallists grade the AI’s results:
In our evaluation, the model solved 5 of the 6 problems on the 2025 IMO. For each problem, three former IMO medalists independently graded the model’s submitted proof, with scores finalized after unanimous consensus. The model earned 35/42 points in total, enough for gold!
It’s not clear to me why this job would be worthwhile for an IMO gold medallist. They tend to pick between tech jobs, finance jobs, and PhDs with prestigious internships, so $65/hour to unload their brain into the AI doesn’t seem super compelling to me.
Did they jump the gun on Google?
This is a very cool announcement and obviously an improvement in the state of the art LLM mathematical ability. However maybe this deleted tweet implies Google’s AI also got an IMO gold this year? And they hadn’t announced it yet?
The IMO haven’t officially confirmed any of this, in the official IMO closing statement they acknowledge that AI companies were working on this problem
Additionally, for the first time, a selection of AI companies were invited to join a fringe event at the IMO, in which their representatives presented their latest developments to students. These companies also privately tested closed-source AI models on this year’s problems and we are sure their results will be of great interest to mathematicians, technologists and the wider public.
Elsewhere IMO officials talk about waiting a week before announcing AI results, to allow the humans to celebrate their results. OpenAI did not do that and perhaps Google did.
We’ll find out in a week.