A Human Beats Go's Top AI Engine — With a Two-Stone Handicap

In a three-game exhibition series played July 17, 19 and 21, 2026 in Seoul, Shin Jin-seo, the 26-year-old world number one Go player, defeated KataGo, the strongest Go-playing engine, two games to one — but only by starting each game with a two-stone handicap, extra stones placed on the board before play begins. The match, called the Ssen Math-Hankyung Gisin Championship and organized by The Korea Economic Daily and Truebook Sinsago, marked the tenth anniversary of the 2016 series between Lee Sedol and Google DeepMind's AlphaGo, the match that first showed a computer could beat the strongest human players at Go.

Unlike that 2016 series, which Lee Sedol played on even terms and lost 4-1, organizers built the 2026 rematch around a head start, because a decade of engine development has pushed the gap the other way. As one commentator on the match put it, if the 2016-era AlphaGo played KataGo 10,000 times today, KataGo would win all 10,000 games. Two stones has become close to the accepted boundary of what a human can still overcome against a modern engine.

Shin lost the opener on July 17, unable to hold his position through a midgame fight. He won game two on July 19 by 4.5 points, becoming the first professional to beat KataGo in an official two-stone-handicap game. He closed out the series on July 21, winning by 11.5 points over 221 moves playing black, to take the series 2-1.

Shin said afterward that he had never defeated KataGo at this handicap before the event, and that treating the extra stones as "less an advantage than a necessity" shaped how he played: rather than trying to match the engine's calculation, he built solid positions, neutralized its attacks before they could develop, and avoided the complications where a search advantage tells most. "The beauty of human Go is that players can make bold or unexpected moves depending on the situation," he said.

Truebook Sinsago, the event's co-sponsor, paid Shin a 150 million won appearance fee plus a 50 million won bonus per win, for 100 million won in victory bonuses, plus a car — a package the company valued at about 250 million won, roughly $108,000, in total. Its chief executive, Hong Beom-jun, framed the result as being about method rather than outcome: "The significance lies not in win/loss, but in how humans continually adapt their approach."

In the decade since Lee Sedol's loss, elite human players have shifted from teaching engines to studying them, and pros now measure their own play by how many points per move they lose against current AI rather than by matches won outright. Shin's 2-1 record does not close that gap — by his own account, the two stones were doing real work — but it marks how much of a head start now separates the strongest human from the strongest machine at a game AI mastered a decade ago.