The Chess Board Was Always Free
This week Uzbekistan won the Chess Olympiad in Samarkand, finishing ahead of India. It is the second Olympiad gold in four years for a country of about 39 million that, a generation ago, was a provincial outpost of the Soviet chess empire. When the Uzbeks won their first, in Chennai in 2022, the team’s average age was about 20, and its youngest player was still in high school.
That is not an anomaly. It is the new normal, and the ages have become absurd.
In 1958 Bobby Fischer became a grandmaster at 15 and a half. It was treated as a freak of nature, and the record stood for 33 years. Today a 15-year-old grandmaster is barely news. Sergey Karjakin did it at 12 years and 7 months in 2002. Abhimanyu Mishra, a kid from New Jersey, did it at 12 years and 4 months in 2021. This May, Faustino Oro of Argentina missed Mishra’s record by about two months. India’s Gukesh Dommaraju made grandmaster at 12 and became world champion in 2024 at 18.
Something changed. To see what, you have to start with what chess always was.
A poor country’s game
Chess is like soccer. The equipment costs nothing. A board and pieces, or a ball, and you’re in. That is why both games have always attracted poor countries and ambitious families.
And the knowledge was never secret. Nimzowitsch’s My System came out in the 1920s. Capablanca’s books were everywhere. From 1966, Chess Informant, published in Belgrade, shipped the best games in the world, mostly annotated by the players themselves, to anyone with a subscription. A determined kid in Ohio could read the same material as a kid in Leningrad.
Yet from 1948, when Mikhail Botvinnik won the title, the Soviet Union held the world championship without interruption until Bobby Fischer took it in 1972. Three years later it was back in Soviet hands, and it stayed there until the Soviet Union itself was gone.
The shift had started in 1927, when Alexander Alekhine took the world title from José Raúl Capablanca. Capablanca was the great natural talent of the age, and Alekhine had never beaten him in a single game. Capablanca arrived in Buenos Aires with no real preparation. Alekhine had studied Capablanca’s games for weaknesses and gotten himself into physical shape. He won six games to three, with 25 draws. From then on, walking into a title match unprepared made as much sense as a boxer walking into the ring without training.
The Soviets didn’t have better books. They had an institution. In the 1920s the state sent masters across the country to train players. Botvinnik trained like an athlete, with timed daily walks, breathing exercises, and practice games in which his partner blew cigarette smoke at him so a smoky tournament hall wouldn’t bother him. He built a school that produced Karpov, Kasparov and Kramnik. Strong juniors got coaches, full-time support, and, most importantly, a dense supply of strong opponents to play every week, with experts on hand to tell them what they did wrong.
That was the scarce input. Not information: structured practice with feedback. Nobody else built a chess college, so nobody else produced a steady line of world champions.
The Polgár experiment
In 1965 a young Hungarian named László Polgár courted a language teacher named Klára by mail. His letters weren’t love poems. They laid out a plan. He had studied the childhoods of hundreds of geniuses, concluded that “geniuses are made, not born,” and wanted a wife to help him prove it with their own children. She married him anyway.
They picked chess because it offered what almost no other field did: an objective, public measurement nobody could argue with. They homeschooled their three daughters, Susan, Sofia and Judit, and trained them for hours a day.
In an era when almost everyone believed women simply couldn’t play at the top level, the results were a demolition. Susan became a grandmaster in 1991. So did Judit, that December, at 15 years and 4 months, breaking Fischer’s 33-year-old record by about a month. Sofia, the middle sister, stayed an International Master, but at 14 she won a 1989 open tournament in Rome with 8½ out of 9, beating four grandmasters, three of them Soviet. The chess press called it the “Sack of Rome.” After 1990 Judit never played in a women’s event again; she competed only in open tournaments. She entered the world’s top 100 at age 12 and eventually reached No. 8 in the world, with a peak rating of 2735. In 2002 she beat Kasparov, then the world’s No. 1, in a rapid game. No other woman has crossed 2700 or entered the top ten.
The Polgárs destroyed the belief that girls’ brains couldn’t do it. The limit had been set by society: by what people expected of girls and by what anyone was willing to invest in them.
But look at what it took. Two parents who gave up their careers and turned their home into a chess academy. One obsessive father, three daughters. The method worked. It didn’t scale.
The chess college in every pocket
After losing his title in 1963, Botvinnik, an electrical engineer by training, spent years trying to build what he called an “electronic grandmaster.” He never got there. Others did. The founder of the Soviet chess school spent his last decades working on the machine that would, in other hands, give his school away to every kid with a phone.
A child today has everything the Soviet system rationed. An opponent at any strength, at any hour. Every game checked by an engine far stronger than any human, so every error is found the same night. A rating that says exactly how good the child is and whether he is improving, with no grading on a curve and no social promotion. And unlimited volume: thousands of games and puzzles a year.
Nothing about the books changed. What changed was the feedback loop, and it became almost free.
The results showed up exactly where you would expect if talent had always been spread far more widely than the old pipelines could find it. India had one grandmaster in 1988, Viswanathan Anand. It now has 99 and is about to reach 100, and its top ten players average over 2700, second only to the United States. One of this year’s new ones, Aaryan Varshney of Delhi, got there with no academy and no professional coach; his father was his trainer. Uzbekistan wins Olympiads. Argentina produces a 12-year-old grandmaster.
The machine wasn’t the only cause. India had Anand as a role model, the pandemic pushed the whole game online, and there are far more rated players and norm tournaments than there used to be, which makes titles easier to chase. The top Indian kids still have coaches. But the engine is the accelerant under all of it: the coach now works on top of a machine that supplies most of the repetitions.
Chess was always cheap. Mastery wasn’t. The computer made mastery cheap.
Chess is the measuring stick
The same thing is happening in computer programming. Teenagers with a laptop and an internet connection build software that used to take a team. But programming has no rating. “How good is this kid?” is a matter of opinion.
Chess is different. Since FIDE adopted the Elo system in 1970, it has kept an objective, continuous record of human performance, the closest thing we have to a controlled experiment on how people learn. And that record says something unambiguous: when practice with immediate, accurate feedback becomes unlimited and cheap, the time it takes a talented child to reach world class collapses by years, and talent starts showing up everywhere. The machines didn’t raise the peak much; a steady 2800 is still rare. What they did was shorten the road and open it to far more people.
Yes, chess is the best case. It has fixed rules, a right answer the machine can find, and a score nobody can argue with. Most subjects are messier. But that cuts both ways: we ran the experiment in the most favorable domain imaginable, and even there nothing happened until the feedback became free. Once it did, a century of assumptions about who could reach the top collapsed in about twenty years.
Enter Claude Code to close the gap
The engines and the online games supplied the practice and the feedback. They did not supply László Polgár: the coach who watches one child, sees what he keeps getting wrong, and decides what comes next.
Books made knowledge cheap. Engines made feedback cheap. Direction stayed expensive.
Claude Code already does that kind of work on software. It can develop and improve code on its own. It does not care what state the software is in, from brand new to legacy. It reads what is there, finds the issues and fixes them, checks the result, and tries again. It is the chess engine’s loop of attempt, check and correct, applied to something far messier than a chessboard.
Why can’t the same technology teach?
Pointed at a student instead of a program, it would be a tireless personal tutor. It meets the student wherever he is and adapts to him. A student can be tested, given exercises to improve, and retested. That is how it works on a program: adding code, debugging, testing again. In 1984 the educational psychologist Benjamin Bloom found that the average student taught one-on-one outperformed 98 percent of students taught in a conventional classroom. Nobody doubted that tutoring works. The problem was always the cost. Whether it takes 4 years or 10 years to get the degree, based on the student, does not really matter. That only matters when you need a coordinated program, as in today’s schools and universities.
The point
There is always the unusual person who figures it out mostly on his own, the Bobby Fischer. But that is one in a billion.
Many people can do much better than anyone expects, given the proper environment.
In this story, the proper environment was László Polgár. That cannot scale across humanity.
The tools, the modern chess engines and related apps and the world of online competition around the clock, are available to everyone.
The challenge is to use AI to help fill the gap beyond the chess engines and online games that everyone has access to now.
It seems like the gap left in the chess teaching is ready to be filled and applied across the education spectrum.
Books made knowledge cheap. Engines made feedback cheap. AI can make coaching cheap.