Computers Beat Humans at Chess 30 Years Ago. Chess.com Still Has 250 Million Players

In 1997, IBM's Deep Blue defeated Garry Kasparov, and a common verdict followed: if machines can win, why would people keep playing? Nearly 30 years later, the opposite happened. Chess.com co-founder and CEO Erik Allebest tells Sarah Guo on No Priors that the company now has about 10 million daily active users, 40 to 50 million monthly actives, more than 250 million registered members, and expected revenue a little over $200 million this year. About 650 people work there, fully remote, around a mission Allebest boils down to making members happy, growing the game, and being the best possible place to work.

The origin is almost comically small. In 2005, while Allebest was at Stanford Business School, the chess.com domain came up in a bankruptcy auction. The previous owner had been building chess software, raised a little money, and failed. Allebest and a friend bought the name for $56,000, hoping to build a MySpace-like community where chess players could express themselves, learn, and share. Sand Hill Road was next door, so he took the meetings. Most investors told him the idea was uninvestable, that he should get a real job and stop wasting his time. Classmates he asked to join said they were going to do something serious.

Nobody funded them, so they built it themselves: the friend as CTO, Allebest on product. He put in money from earlier businesses he had sold, borrowed a little from his friend's mom, and paid it back later. They launched in 2007 as he graduated. About 18 months later they started charging for online chess learning memberships and became profitable quickly. From there the company grew at the speed of cash: more revenue meant one more hire. At 50 people they swore they would never go past 50. At 100 they swore the same. They eventually stopped making that promise, because chess kept surprising them. The headcount is now 650.

The advice he later gives founders is mostly a recap of that path. Today's curse, he says, is too much founder advice. Twenty years ago the standard script was hire a Stanford technical co-founder, raise a pile of money, get an office, chase a huge market, and spend on customer acquisition. Chess.com did the opposite: a friend from San Jose State, remote work, no fundraising, no paid acquisition, and a market everyone called too small. Figure out what you want to exist in the world, use the minimum resources to prove it, work with great people, and execute. Listen to customers and pivot if it is not working. Stop listening to people telling you what to do.

Before Covid and *The Queen's Gambit*, Chess.com had about a million daily active users, and the team was already shocked. Then the pandemic, the Netflix series, and a run of news cycles sent usage through the roof. Internally the fear was familiar: this might be treadmills and sourdough bread, a lockdown fad that would fade. Usage did come down from the 2022 peak, but it stayed far above the old baseline.

2023 produced another spike. Short-form content pulled kids into school chess clubs. Chess.com's Mittens bot — a cat that pretends to be terrible and then wrecks you — went viral. A high-profile cheating scandal kept the game in the news. When that wave receded, the floor was higher again. Only in 2024 did Allebest really conclude that the growth was sticking around and compounding. "If your business is shrinking, it's scary. If it's flat, it's scary. If it's going up, it's scary," he says. For years, product and tech scalability problems arrived one after another. The last year has been calmer: the team is in place, growth is more predictable, and they have finally gotten their hands around what it means to run a 650-person fully remote company.

The ambition scaled with the numbers. If 250 million people have already registered, he asks, why aren't there 50 million daily actives? A billion people should be playing chess. Culture shifted around the same time. Allebest did not start playing until he was 18, heading to college, and in his world chess was a secret nerd hobby — he was a skater and snowboarder who kept it quiet. Now his kids' friends play, jocks play, and Louis Vuitton centers fashion shoots on chessboards. Lewis Hamilton plays on his phone on a private jet. Guo, who grew up as the high-school nerd at the chessboard, says about 20 percent of kids in her Bay Area town do after-school chess, on a par with soccer.

Allebest thinks Chess.com helped manufacture that change, not just ride it. Chess can live in your pocket; the content is funny and relevant; and the company rewrote who counts as a player. The old club rule was that you were not a real chess player until you were rated 2000. Chess.com celebrates hanging your queen. A 500 rating still makes you a chess player, still part of the community, still on a journey. The line he returns to is the one the game itself teaches: you never really lose if you learn.

When Chess.com entered, the field already had FICS, ICC, and Yahoo Chess. Allebest's strategy was stubbornly personal: build the product he wanted as a player — free, in the browser, and a pleasure to use. Over time, most rivals fell away. Lichess remains, and he treats it generously, calling it the Linux of chess: open source, important for the community, serving a different part of the market. Chess.com funds growth from revenue; Lichess serves people who want the non-commercial version.

Chess is not patentable. The app stores hold thousands, maybe tens of thousands, of chess apps. What wins, he says, is product, community, and content, plus a relentless focus on experience. Because Chess.com is pure software and content, and because no major tech company was bearing down on them, they could take the slower 20-year road, build culture, and skip outside capital. Other markets, he grants, really do need speed, hardware, or inventory money. This one did not.

Outside capital still found them, awkwardly. The first external investor insisted he just wanted to be part of it and would hold forever; five years later he needed out. A second investor made the same forever promise and was selling three years later. At that scale they had to shop for a different class of investor. Allebest did the private-equity rounds and disliked most of the process and most of the people, who he felt were focused on the wrong things. General Atlantic, around 2020 or 2021, was the exception: Tanzine and the firm believed that if Chess.com focused on its mission, the business would follow. They bought secondary shares. No capital was infused.

When General Atlantic later needed to return capital to partners, CVC came in, again buying secondary, with deep experience in gaming, sports, and community businesses. General Atlantic reinvested. Allebest stresses that the company still does not need outside money to grow. Against the private-equity stereotype, neither firm has barked at them to hit short-term numbers. They share industry pattern recognition, get into the weeds on product and community, and pushed operational forecasting and reporting that made Chess.com stronger as an operator, not just as a reporting machine. The job of CEO, he adds, is a different landscape at hundreds of people and a multi-billion valuation than it was with a handful of teammates, and the investors helped him grow into it.

Guo frames chess as perhaps humanity's longest-running experiment in what happens to a human activity after machines become superhuman at it. Deep Blue is 30 years in the rearview, and people care about chess more, not less. Allebest's explanation is almost anthropological: humans want to do human things — compete with other humans, make things, solve problems. In chess, machines pushed people to get better at the human-versus-human game rather than abandoning it.

The plot has three acts. First, world champions used early computers as idea generators, then took those ideas into over-the-board play. Second, engines such as Stockfish became so perfect that chess got boring: players imitated the machine and ground out dull endgames. Third, neural nets arrived. After Leela-style self-play transformed Go, the chess community trained Leela Chess Zero, which started beating Stockfish in unexpected, aggressive, unconventional ways. Computers first made chess a little more boring, then made it more exciting than it had ever been.

For ordinary players, AI is already in every loop: a coach that reviews your game, puzzles generated for your weaknesses, a computer always ready to play. Allebest's observation is that AI has not replaced human-versus-human chess; it has raised the human floor. Seven-year-olds now give grandmasters a run. He mentions an English girl who is crushing everyone and declines to say her name for fear of butchering it. The same pattern shows up at home: his son, now 18, taught himself video-game creation on YouTube and uses AI for research, not for the creative core, and is better at it than Allebest ever would have been.

Inside the company, AI is equally pervasive. Support automations answer common questions while human agents handle the cases that need a person — a "great symbiosis," he says. Analysts query data with LLMs. CTO Josh built an internal system called GNS years ago, an authentication and knowledge layer with permissions across the whole organization; it is now infrastructure everyone depends on. Development is getting more agentic, shrinking the time from seeing a problem to shipping a fix. On the product side they are prototyping an AI coach in your pocket that looks at last week's games, offers feedback and training, and will chat if you want. It is not a replacement for a human coach, Allebest is careful to add. If you really want to get systematically better, hire a good one.

If everyone has access to superhuman chess knowledge, Guo asks, why isn't everyone equally good? Allebest's answer is almost disappointing: you still have to do the repetitions that put knowledge into your brain. Chess.com sits on what may be the largest dataset of humans learning a skill — 250 million users, mistakes and improvement at every rating — and the conclusion is the same. Do puzzles. Review your mistakes. Train the openings you want to play. Tools can have shortcuts. The process cannot.

The mental model he loves is: take any given day and multiply it by a thousand. That is roughly your life. Five puzzles a day times a thousand is 5,000 puzzles, and you will be much better. A hundred a day times a thousand is another magnitude. One sketch a day, one small run, a few pages of a book — the same logic applies in any field. What matters is the compounding of habits, not how hard you went on a single day. He compares it to training a neural net: models can ingest data in parallel across memory channels; human brains cannot. You have to put in the reps. Social media's immediacy, he thinks, makes people forget that practice cannot be skipped.

AI's place in that model is a professor's office in your pocket: walk in, ask questions, get exposure to ideas, refine them, go deep. He is not naive about the failure modes. Models tell you what you want to hear, inflate your ego, and carry biases. Research already suggests that in some settings, people who use AI stop thinking. Guardrails will have to be built, and society will have to learn which uses make people stronger and which make them weaker.

Cheating is the ugly twin of superhuman engines. Machines can produce superhuman output and pass it off as human play, and Chess.com has been fighting that for 20 years — longer than the current AI boom, because chess computers predated it. Allebest is openly disappointed in the slice of people who will steal rating points, prize money, or an ego boost. The company spends heavily on anti-cheating, with statistical and machine-learning models trained on vast data about how humans play versus how computers play. He will not describe the methods, to avoid giving cheaters ideas. Two workstreams run in parallel: general detection for everyday games, and a separate track for prize events. Cheating cannot be prevented, he says, because you cannot put a webcam in every home. What you can do is keep raising the odds of catching and banning people, and try to educate them.

In a world where you can prompt an AI and have 10,000 new board games tomorrow, why is chess the thing that dares a billion-player target? Allebest reaches for music. Streaming made supply infinite, and people went back to the Beatles and Led Zeppelin, the classics that offer a shared experience. Chess is that classic in games: no loot boxes, no new skins, no rule changes, no new heroes to memorize. Perfect information, no luck, the simplest rules, complexity that can occupy a lifetime, simplicity a child can learn in minutes. Pull the board out of a drawer and you are playing the same game as someone a thousand years ago. Chess.com, he says, is simply a chess company. Gaming, content, and social network are all fair labels because the game itself contains all of those things.

They have started copying the playbook. Gambit.com is a new poker site whose core idea is a rating: how good are you, really, not how many chips you can buy, whether you can run a bot, or whether you can bully the table. The algorithm looks at opponent strength, chips won, and hands played, and will keep being tuned for $100 games versus $1,000 games versus M&M games where someone is all-in every hand and just rebuys more candy. Allebest's lived sense of it is telling: losing $100 at a table does not bother him much; losing 100 poker rating points does. Guo's gloss is that it could be equalizing — finally you can tell whether someone is actually good, or just has more M&Ms and likes to bully.

On AGI and ASI, Allebest believes superhuman intelligence will arrive at a faster and faster pace and will be scary without guardrails. He is more optimistic than pessimistic, hoping it helps with climate, politics, and education, and worried it could do the opposite. Too much money, he says, is already flowing to too few, and AI may accelerate that. He treats this as a cultural problem more than a technology problem: humans have had nuclear weapons, guns, and economic systems for a long time, and outcomes depend on what culture does with the tools and who is in charge. He does not feel much confidence in the super-billionaires or politicians currently steering.

Guo's closing note is brighter, and Allebest agrees with it. Chess's boom is evidence that people still want to acquire human skills, and still want to watch exceptional human skill, even when a machine could do it better. In a future of surplus capability and abundance, people will spend a lot of time playing games. The claim that human skill becomes irrelevant just because of the technology itself, she says, is nonsense. Deep Blue's mourners could not have guessed that 30 years on, humans would be spending more time, money, and passion on chess than ever. The experiment is still running. The next domains machines surpass may yet give the same answer.