· Sharp.go Team
The story of weiqi: from Emperor Yao to KataGo, in four thousand years
Weiqi is older than writing in most places. The earliest Chinese chronicles describe it as an invention of Emperor Yao’s tutor, designed to train the heir to a throne in balance, restraint, and reading the consequences of moves. That story is legend, but the underlying claim is real: the same game, with the same rules, has been played for roughly four thousand years.
This post is the long arc — from the imperial courts of the Bronze Age to a Kubernetes-deployed AI engine in 2026, with a 3D browser client over the top. It is a story of how a single decision problem outlasted every empire that played it.
For a more focused introduction to what the game is and where the names came from, see What is weiqi?. For a terminology comparison across regions, see Weiqi vs Go.
● The ancient period
01Yao, Shun, and the invention legend
The Chinese imperial chronicle that gives weiqi its origin story.
The Book of Documents (Shujing), one of the Confucian classics, contains the earliest description of weiqi’s invention. Emperor Yao, facing an unfit heir, appointed the minister Shun to tutor the prince. Shun reportedly said:
“The prince is unruly and careless. I have made a box marked with lines so that he may learn through play.”
What kind of game the prince learned is debated. Some scholars read the passage as a chess-like board; others read it as a placement game on a grid. Either way, by the Spring and Autumn period (770–476 BCE) the placement game is unambiguously attested in poetry and tomb art. Weiqi in its modern form is firmly established by the time of the Han dynasty (206 BCE – 220 CE).
"
"Shun was appointed to teach Danzhu. The prince was violent and careless. Shun therefore devised a box marked with squares to train him through play.
● The classical period
02Spreading through East Asia
Korea, Japan, and the diplomatic gift tradition.
By the Han dynasty weiqi was already an established courtly pursuit. Han kings and emperors played it. Officials included it in the recommended curriculum for a cultivated gentleman. The game spread via two routes:
- To Korea, via the northern commanderies of the Han. The Korean pronunciation baduk (바둑) is close to the original Chinese.
- To Japan, via diplomatic embassies and Buddhist monks in the 6th and 7th centuries. The Japanese form igo (碁) developed independently and became the dominant term after the 8th century.
Weiqi carried with it a cultural role it kept for nearly two millennia: a diplomatic gift. Korean embassies to Japan brought boards and stones; Japanese poets wrote about weiqi in the same breath as calligraphy and poetry. By the Heian period (794–1185 CE) weiqi was already a fixture of the imperial palace in Kyoto, and the earliest Japanese professional players were emerging.
● The millennium of the houses
03The professional schools of Edo Japan
From family-based schools to the modern national federations.
Japan’s Edo period (1603–1868) produced one of the most concentrated bursts of strategic thinking the game has ever known. Four competing houses — Honinbo, Yasui, Inoue, and Hayashi — were authorised by the shogunate to certify top players. The strongest of these, Honinbo, produced a series of legendary professionals whose games were preserved on diagrams and studied as canonical.
The most famous of these professionals is Honinbo Shusaku (1829–1862), whose 19-game winning streak against the strongest players of his era gave him a reputation that has not dimmed. He reportedly never lost a game in which he used the ko fuseki — the 3-4 corner opening — as Black.
When the Meiji Restoration of 1868 ended the shogunate, the houses lost their patrons. The Nihon Ki-in (Japan Go Association) was founded in 1924 to keep the tradition alive; it remains the central Japanese federation today.
● The 20th century
04The professional era
Title matches, weekly press, and the slow internationalisation of the game.
Throughout the 20th century weiqi professionalised in East Asia. In Japan the major titles — Honinbo, Meijin, Kisei, Tengen, Gosei, Oza, Judan, Ryusei — crystallised as headline matches running once or twice a year. In China the national federation ran its own title structure after 1966; in Korea (the Kuksu, later the Myeongin) similar structures emerged.
The press covered these matches the way sports sections cover modern football: weekly commentary, blow-by-blow news reports, post-game analyses. The reader base was national — millions followed the title matches. Outside East Asia the game’s footprint was smaller, but stable. European and American Go associations kept amateur traditions alive through weekend tournaments and a strong correspondence-play culture.
● The AI revolution
05AlphaGo, Lee Sedol, and the year 2016
The four-games-to-one match that changed every player's relationship to the game.
In March 2016, a London-based AI company named DeepMind challenged Lee Sedol, then the strongest active player from a generation that had won eighteen world titles, to a five-game match. The system, AlphaGo, combined deep neural networks with Monte Carlo tree search. Game 1 was conceded by Lee on move 178. By game 4, Lee had one win and the world had changed.
Lee Sedol’s single game 4 win — played under sudden-death conditions, against a machine stronger than any human alive — is the most analysed amateur game of the decade. The official narrative afterwards is that human intuition had not been surpassed across the whole board; it had been surpassed at concrete evaluation. Humans still saw things the engines did not, but the engines evaluated positions more accurately than any human.
"
"I have experienced a defeat I cannot express in words. I would like to express my respect to AlphaGo and the team behind it.
● The open-source era
06KataGo, Leela Zero, and the democratisation of analysis
Strong open-source engines anyone can run on a laptop.
The DeepMind team published a description of AlphaGo’s architecture but did not release the trained weights. Within months two open-source projects started independently: Leela Zero, led by Gian-Carlo Pascarella, replicated the AlphaZero training pipeline on volunteer-distributed compute; KataGo, led by David Wu, added infrastructure for distributed training and a richer set of analytical outputs (territory estimation, influence maps, score loss).
By 2019 these engines were stronger than the human world champion on standard hardware. By 2023 they were available as turnkey cloud endpoints and as local downloads. Today sharp.go uses KataGo as the analysis engine behind its 3D review surface — the same engine that professionals use to study, shipped as a one-click tool.
● The 2020s
07Sharp.go and the browser-first future
Three-dimensional boards, shareable URLs, no install.
The 2020s brought three converging trends:
- Browser-first delivery. Modern web standards — WebGL, WebGPU, WebTransport — make the browser a viable rendering surface for a full 3D board with sub-100ms analysis latency.
- Cloud GPUs make engine hosting cheap. A KataGo-class engine behind an HTTP endpoint now costs a few cents per thousand positions. Hosting a research-grade opponent at consumer scale is feasible.
- Open transcript formats. The move formats used by sharp.go and a handful of other modern clients are versioned and replayable. A game you play today can be reopened correctly in 2030.
Sharp.go is one of several Western projects building on these trends. It targets the casual and intermediate player who wants to play without an install, share a position without a plugin, and review with the same engine that the professionals use.
● The future
08What the next decade looks like
Cooperative review, portable transcripts, and the end of the install.
The next decade is likely to consolidate three directions:
- Cooperative review. Two humans and a KataGo backend in the same analysis session, each seeing the same influence maps and fork points. Sharp.go’s roadmap includes this as a 2026/2027 deliverable.
- Portable transcripts. A game replay that survives client shutdowns, browser changes, and device migrations, with the analytical annotations preserved alongside the moves.
- The end of the install. The browser is the client. Engines run in the cloud. Boards render in WebGL. A child in any country with a phone can play against the same AI strength as a professional in Seoul.
This is not a forecast — every one of these is partially shipping today.
● Frequently asked questions
01 Did Emperor Yao really invent weiqi? ▾
The earliest Chinese chronicles — the Book of Documents and its commentaries — credit Yao's minister Shun with inventing a board game to teach the prince restraint. Whether the game described is weiqi specifically is debated; what is unambiguous is that the placement game with stones on a grid is well-documented by the Spring and Autumn period (770–476 BCE) and was widespread by the Han dynasty.
02 When did weiqi reach Japan? ▾
The most widely accepted date is the 7th century CE, via Buddhist monks and diplomatic embassies from the Korean peninsula. By the Heian period (794–1185 CE) weiqi was already a court fixture in Kyoto, with professional players emerging in the late classical period.
03 Who is Honinbo Shusaku? ▾
Honinbo Shusaku (1829–1862) is the most celebrated player of the Edo period. His 19-game winning streak against top contemporaries — played in the late 1850s and never repeated — is the longest unbeaten run in documented weiqi history. He reportedly never lost a game in which he played the 3-4 opening as Black.
04 What changed after AlphaGo vs Lee Sedol in 2016? ▾
AlphaGo's 4–1 victory over Lee Sedol proved that machine evaluation had exceeded the strongest human concrete evaluation. The public release of open-source engines — most notably Leela Zero and KataGo — within the next three years made this evaluation strength available to anyone with a laptop. The effect on amateur and professional study has been the largest pedagogical shift of the game in living memory.
05 What does sharp.go bring to the game? ▾
Sharp.go builds on the open-source engine ecosystem by giving it a browser-first 3D surface. No install is required. Every game has a stable, shareable URL. The same engine (KataGo) that professionals use for study runs as a one-click review tool behind every board. Open [the home page](/) to play.
The story of weiqi is not a story of an unchanging game. It is a story of a single decision problem that has outlasted four thousand years of players, empires, and now AI. The shape of the game — what makes it interesting — has not changed; the surface area of who can play it, and the depth with which they can study it, has never been larger. Sharp.go is one attempt to widen that surface area a little further.
For the broader introduction, see What is weiqi?. For the rules under which all of this is played, see How to play weiqi. For how sharp.go renders 361 stones at 60fps, see Inside the engine.