Kyrgyzstan Beat Singapore at the IOI. We Did It With 3% of Our Students
Kazakhstan, Kyrgyzstan and Uzbekistan have quietly put one of the more remarkable multi year runs at the International Olympiad in Informatics that I’ve seen out of any region.
Today, Kazakhstan sits 7th in the world. Kyrgyzstan #20 and Uzbekistan #25 aren’t far behind, climbing past countries with vastly larger education budgets.
I’m going to zoom in on Kyrgyzstan, since it’s the program I know best. Keep the other two in mind though, because I suspect most of what follows applies to them too. Also, this applies to other Olympiads like IMO, IPHO, ICHO, IBO.
This year, Kyrgyzstan’s team ranked (#20) above Singapore (#24). Singapore, a country whose name is basically shorthand for STEM education done right.
The most impressive part: our Olympiad program did this with less than 3% of the country’s student population. Only a handful of private schools run the kind of structured, competition focused program this requires. The public schools, with more than 93% of the total students, simply aren’t part of the equation. I’ll spare you the rant about the state of public education here, but it isn’t good.
Moreover, not all of that 3% get the elite level opportunity. Scouting starts around grade 8 or 9. Out of that already small pool, only a few dozen students make it onto the elite training track. Everyone else, including kids with real potential, is practically out of the running before they turn 15.
So the honest summary is this: we produced strong results from a tiny, self selecting slice of the country, using a system that wasn’t meant to scale.
The goal was never to turn every student into an Olympiad medalist. That isn’t realistic and it isn’t the point. The goal is to make sure far more students get the opportunities to unleash their potential.
That also means rethinking what “the competition” actually is. Right now it’s a single funnel that ends in a national team of four. It shouldn’t be. A student who will never make nationals can still have a real, leveled competition to be part of: school level, regional level, an online ladder, whatever fits where they are. Olympiads work best when they look more like a sport with many divisions, not a single cutoff that most students never get near.
None of this means Olympiad medals are a scorecard for the health of STEM education generally. They aren’t. A handful of medals says very little about the average student’s math or programming ability. But the two aren’t unrelated either. A wider, stronger general talent pool feeds a stronger Olympiad pipeline, and a serious Olympiad program pulls standards up for the schools and teachers around it. Raise the average, and the ceiling tends to rise with it. A wider pool doesn’t water down the top end, it raises it. More shots at the target produce a better best shot.
Five years ago, widening that pool would have meant hiring specialized coaches at a scale no developing country budget supports, and informatics is a harder problem here than math or science, since it also demands real machines and real programming instruction, not just a good teacher and a chalkboard. Education researchers have understood the underlying fix since the 1980s.
That constraint is loosening. AI tutoring won’t replace what a great coach does. But for the first time, it can put a decent first version of that coaching in front of far more students, in far more schools, without needing a budget line that doesn’t exist.
Here’s where I want to take this. I want to put a high quality training app in the hands of every student in Kyrgyzstan, and eventually beyond it. Not to replace the coaches, schools, and the system that produced this year’s results, but to sit underneath it, so that far more than 3% of students get the opportunity to train and compete.
I don’t have the finished version of this yet. But I’d rather spend the next few years building toward it than keep mining the same 3%.