AI Made Chess Boring, Then Made It Better
In the Stockfish era, players imitated a perfect engine and games turned dull; with neural-network engines playing aggressive, unconventional chess, human skill got pushed higher.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Only in 2024 did he confirm this was a big business
Erik says he didn't truly believe this was a big business until 2024. Before that chess.com had only about a million daily actives, and even they were shocked. COVID and Queen's Gambit brought the first wave, and at the time they judged it might be a pandemic flash in the pan like Pelotons and sourdough; in 2023, because of short videos, the mittens bot, and the cheating scandal, schoolkids started playing again, and after the second spike faded the baseline was far higher than before. His conclusion: this isn't a flash in the pan, it's compounding. He also admits the other side of growth is fear — shrinking, flat, and rising all have their own terrors.
— Erik AllebestPrivate equity came in but gave not one dollar of growth capital
General Atlantic became the first PE to believe in them in 2020 or 2021, and CVC came in later. Erik stresses both were secondaries — buying shares from existing shareholders, with no capital injected. He describes a recurring pattern: investors first say "I'll hold forever," then a few years later say "I have to sell," and the company is forced to find a larger buyer. He admits the PE round process was unpleasant, that most of the people he met focused on things he considered wrong, until he met people who believed in the mission. Counterintuitively, he says the company got better after PE came in: they point to specific areas to look at rather than just saying "do better," and they bring in forecasting and operating discipline.
— Erik AllebestAI first made chess boring, then made it better
This is the most valuable mechanism in the whole piece. After Deep Blue beat Kasparov many predicted chess was finished, but Erik says the real trough came from engines like Stockfish: it was too perfect, players tried to imitate the optimal line, grinding out dull endgames, and matches became boring for a while. The turning point was neural-network engines — Leela Chess Zero, trained by self-reinforcement learning on different positions, started beating Stockfish in unexpected ways, playing aggressively and unconventionally, pushing the game forward. So computers first made chess boring, then made it more exciting than before. Human players' level rose along with it, and AI coaching, review, and personalized puzzles made the game more enjoyable for humans.
— Erik AllebestEqual information doesn't make everyone equally strong
The host asks: if everyone has the same access to information, why isn't everyone equally good at chess? Erik's answer is repetition. He offers a multiplication framework: take whatever you do on any given day and multiply it by a thousand, and that's your life — five chess puzzles a day times a thousand is five thousand, a hundred a day times a thousand is another order of magnitude. He thinks what AI is best at is letting people dig deep into a topic at extreme speed, get ideas, and polish ideas — like putting a professor's office in your pocket; but he also admits AI is biased, tells you what you want to hear, and inflates your ego. He says plainly: there's no shortcut to practice, tools can make it faster, but the repetition has to get into your brain.
— Erik AllebestCheating can't be stopped, but it can be caught
Erik says chess has faced cheating far longer than AI, and many once thought it was an existential threat. chess.com has two independent workflows: one for general detection, one for professional events with prize money. He declines to reveal specific methods, on the grounds that "I don't want to give cheaters ideas," but says they track a lot of data and have statistical and machine-learning models, because they know both how humans play and how computers play. His judgment is restrained: cheating can't be fully prevented, you can't put a camera in everyone's home, but you can do it very well — catch them, ban them, and try to educate.
— Erik AllebestIn an age of infinite content, the classic is worth more
Erik uses music as an analogy: once content is infinite, people go back to the Beatles and Led Zeppelin, because that's the classic everyone can share. Same with chess — no loot boxes, no new skins, no rule changes, no new hero to learn, the simplest rules, perfect information, no luck, playable on a device or by pulling a board out of a drawer, the same shared experience forever. It's complex enough to astonish you and simple enough to learn. He says this is why chess became the "king of games" through evolution, and what people want to hold onto in today's content-explosion world.
— Erik AllebestFounder advice is itself a kind of curse
Erik says one of the biggest curses of being a founder today is that there's too much founder advice out there. He recalls the playbook from twenty years ago: recruit a technical founder from Stanford, raise a big round, rent an office, target a big market, spend on acquisition. They did the exact opposite — found a friend from San Jose State, worked remotely, didn't raise, didn't spend on acquisition, picked a tiny market. His advice is to listen to your own heart, figure out what you want to exist in the world, then relentlessly do it, rather than listening to what the market says. He also admits you have to listen to customers and pivot, but the core questions are: what's your vision, what's the minimum resources needed to prove it, and who do you want to do it with.
— Erik AllebestPoker ratings must make people care about points like money
Erik says the poker rating algorithm is still evolving, involving the ratings of others at the table, how many chips you won, how many hands you played, and they'll keep adjusting. He points to a difficulty: behavior differs a lot between a $100 game and a $1,000 game, and when playing for M&Ms someone will go all in constantly and keep refilling. But he's betting on a psychological shift — when people start caring about their poker rating, they'll care about it like money, because the rating reflects their skill and value as a player. His own example is concrete: losing a hundred dollars is nothing, but dropping a hundred rating points really hurts. He thinks this would be an interesting innovation for poker.
— Erik AllebestIn their own words · checked verbatim
Yeah, I think 2024 was when we thought this would be a big business. That was not that long ago.
Erik Allebest9:31
And that chess computer like actually made chess pretty boring for a little while because it was so perfect.
Erik Allebest21:01
And so what was interesting is that like, computers at first made chess a little bit more boring, but then made it so much more exciting.
Erik Allebest21:01
And one thing that I really love is the concept that if you just take any given day in your life and multiply that by a thousand, that's kind of what your life looks like.
Erik Allebest25:09
It's the same as when you train a neural net, right? Like you give it a bunch of data.
Erik Allebest28:10
I think one of the curses of being a founder today is just how much like founder advice there is out there.
Erik Allebest35:26
I think people are going to care about it as much as money because it's a reflection of like the value and their skill as a player.
Erik Allebest44:48
Figures
| Chess.com daily active users | about 10 million | 2:11 |
| Chess.com registered members | over 250 million | 2:11 |
| Chess.com annual revenue | slightly over $200 million this year | 2:11 |
| Chess.com team size | 650 people, fully remote | 2:11 |
| chess.com domain purchase price | $56,000 | 3:11 |
Glossary
- Stockfish
- A traditional chess engine, extremely strong but considered too perfect and dull in style.
- Leela Chess Zero
- A neural-network chess engine trained by self-reinforcement learning, playing aggressively and unconventionally.
- mittens bot
- A cat-shaped bot that went viral on Chess.com in 2023, driving a large wave of new users to play.
- ELO
- The rating system commonly used for chess and other competitive games.
- secondary
- An investor buying shares from existing shareholders; the money doesn't go into the company and isn't used for growth.
How to listen
Founders and investors building consumer products, community platforms, or games going global — especially anyone interested in how AI changes the value of human skill, and how a company that never raised grew to $200 million in revenue.
The cold open and guest introduction from 0:00-2:11 can be skipped.