Dendi Says He Could Have Stopped AI. A Former OpenAI Engineer Thinks He Has A Point

Dota 2 legend Dendi posted a joke in mid-October that landed harder than he probably planned. Sharing footage of his 2017 loss to an OpenAI bot, he wrote that this was apparently where the AI era began, and that he "had a chance to stop AI." The caption was a bit of fun. Then someone who actually built the bot replied.

Przemysław "Psyho" Dębiak, a former OpenAI engineer, answered on October 15 that the joke is closer to the truth than it sounds. If Dendi had won, he wrote, OpenAI "would've raised far less money afterward." He described the OpenAI of 2017 as a little-known research lab of roughly 50 people, and said the showmatch was the first time, after its initial launch, that it broke into mainstream news.

Some background on that night. It was August 11, 2017, at The International in Seattle, a tournament with a prize pool of about $24 million. Dendi, a former world champion and crowd favorite, played a 1v1 Shadow Fiend mirror against the bot and lost 2-0, according to GameRiv's recap. GeekWire reported that he was beaten badly in game one, gave up game two and declined to play a third. He called the bot scary while it happened.

CyberScore's write-up of Psyho's posts adds a detail I find funny and a little alarming. Just days before the showmatch, the bot could not yet beat pros consistently. Dota was chosen, he said, because of the prize pool, an open programming interface and Linux support. So the whole thing was less of a coronation than a bet that came in.

The feed that night did the rest. OpenAI's own account posted that its Dota 2 bot was undefeated against the best solo players in the world. Elon Musk, then one of OpenAI's best-known backers, called it the first AI to beat the world's best in competitive esports, thanked Microsoft for the Azure compute behind it, and then pivoted to a warning: "If you're not concerned about AI safety, you should be." By the next morning TechCrunch and Fortune had both run it.

Does the money trail back him up? Partly, on timing. OpenAI launched in 2015 with $1 billion in commitments from Musk, Sam Altman, Peter Thiel and others. In April 2019 its five-bot Dota system beat reigning champions OG 2-0. Three months later Microsoft announced a $1 billion investment and became OpenAI's exclusive cloud provider.

A sequence is not a cause, though, and nobody can rerun 2017 with a different winner. Most people date OpenAI's real money to the GPT models and ChatGPT, not to a Shadow Fiend mirror. I buy the narrower claim. A 50-person lab needed a spotlight, and a stage in Seattle gave it one.

The numbers now are a different universe. OpenAI closed a $122 billion round in March at an $852 billion valuation. Reuters reports it is racing to raise about $30 billion at more than $1.4 trillion, ahead of Anthropic's expected mid-November listing. OpenAI is telling investors its annualized revenue hit $50 billion and should reach $70 billion by year end. The round is still under negotiation.

Why does a crypto desk care about a Dota joke? Because AI tokens trade like OpenAI's mood ring. This summer, when reports said OpenAI was cooling on a 2026 IPO, TheStreet logged TAO down 3.94% to $208.45, RENDER down 5.56% and FET down 3.8%, with the whole AI and big data token sector at $17.70 billion. On October 6, CoinGecko data cited by Bitbase had TAO near $300 and about $3.4 billion in market cap.

Put that next to a $1.4 trillion private valuation and the sector is a rounding error. That gap is the part worth watching. OpenAI turned its attention into revenue. Most AI tokens have not, and one October analysis points out that token prices can climb without any rise in paid compute jobs. Crypto knows the loop well: a demo, a stage, a tweet, a bid. The difference is whether anything is billing on the other side.

Anthropic's listing is next to price the sector in public. Until then, Dendi's post is a decent reminder of how small the trigger can look from inside the moment.