Forty standalone articles: three review guides, eight KataGo and service-engineering articles, seventeen research features on AI and joseki history, and twelve deeply read games from the Edo period to the AlphaGo era. Each answers one question with steps, sources, and a focused visual.
On August 30, a game with unsupported komi first triggered a KataGo parameter error. The same investigation then uncovered an oversized stream completion, a registration runtime mismatch, repeated Token creation, and old-client sync failures. This is the timeline, measured impact, recovery evidence, and unfinished work.
LoGos mixes structured Go data, long reasoning, and reinforcement learning and reports professional-level move play. Yet fluent commentary can contain atomic claims unsupported by the board or engine. Validate claims, not style.
An ICML adversarial policy beat KataGo at superhuman settings in more than 97% of games—not by playing stronger Go, but by inducing systematic mistakes. Average strength and worst-case robustness are separate properties.
A vanished recommendation was not necessarily refuted. Timing may have changed, search may have been unstable, or training may have repaired a blind spot. A museum preserves why an answer once made sense and why it later failed.
HumanSL predicts rank-conditioned human moves, but maximum probability, sampling, and target strength are not the same. Teaching should locate the next reachable judgment, not the person most similar to AI.
Komi moves the winning threshold; area, territory, ko, suicide, and tax rules change the final ledger. The useful question is not how much win rate moved, but where candidate order actually crosses.
In a public professional-game sample, early bare-4-4 3-3 rose from 9.72% in 2017 to 40.06% in 2018-2020. Adoption rates show that more players used it; whether they also learned to exchange less, tenuki, and return has to be read from the games themselves.
Change the network, engine, search budget, or symmetry and the first candidate can move. This article shows how to read the full analysis conditions before deciding whether an old result still holds, needs a note, or should be retired.
Early 3-3 changed the definition of completion, not merely one corner answer. Omitting three moves, preserving cuts, and taking another corner can be more efficient than cleaning locally—if you keep managing the unfinished aji.
New Fuseki broke the habit of completing one corner at a time in 1933. Chinese and Kobayashi formations went further: stones already on a side began pricing approaches, pincers, and tenuki before the joseki started.
The Shusaku diagonal was never an answer outside history. Restore no-komi play, Black's incentives, house research, and later komi, and the same coordinate moves from steady to slow—then returns under AI for a different reason.
Across the same 50 professional games and 500 positions, a 9d Human SL policy covered 82.7% of the AI one-point good set at the median, versus 77.5% for 5d. The difference is real, but it is not a ladder that rises in every position.
Five hundred professional positions, two strong models, and eight board symmetries show that more search usually reduces median drift but can expose a split first. Stable coordinates, stable values, and trustworthy teaching labels are different claims.
From Jowa's ghost moves and Shusaku's ear-reddening move to AlphaGo 37 and Lee Sedol 78, twelve variation trees repeat one boundary: objective value, practical difficulty, and historical courage are not one metric.
Fifty professional games, 500 positions, 397,561 actual search visits, 10,000 whole-game bootstrap replicates, and targeted audits up to 8,192 visits across all eight board symmetries. The question is no longer only which move is best, but why people miss it, whether the label is stable, and what deserves to be taught.
A six-stage professional review workflow, large-scale studies, and 60 Read19-analyzed games explain five elite styles and the real relationship among match rate, wins, and memorization.
Once open models are far beyond professional strength, win rate, score lead, and one top move compress the most useful evidence. This article joins optimality, human discoverability, and model uncertainty so review can become a learning instrument instead of an answer machine.
Reconstruct the 2016–2026 frontier from papers, released weights, and direct matches. Compare width, depth, parameter count, training, search, and why 5,185, 14,554, and relative +300 Elo do not share a scale.
Desktop review, guided cloud study, professional databases, online play, and problem training are different products. Choose a primary workflow first.
Models shape speed, strength, and style; search controls stability. Read win rate, score lead, visits, principal variation, and ownership together.
Sign in once with the native CLI, then choose GTP or Analysis Engine JSONL. Local clients do not manage a model or GPU.
Both use standard input and output, but GTP v2 and Analysis Engine JSONL differ in state, framing, and intended clients.
Understand a Read19 encrypted WebSocket session: every text frame carries standard KataGo Analysis Engine JSON, with updates while search continues.
Separate the engine, neural-network model, search, protocol, and interface before deciding what to download, connect, or debug.
A deep review routine that avoids chasing every number: reconstruct your thinking, find turning points, compare why the played line and three KataGo routes differ, and keep lessons for your next game.
Start with FF[4] source text, learn nodes, properties, coordinates, and variations, export from OGS, KaTrain, Sabaki, or SmartGo, then analyze the record with Read19 and KataGo.
Keep the KaTrain interface on your computer while Read19 runs KataGo in the cloud, without configuring a local GPU.
Ten years ago, professionals still asked whether a joseki had been completed. Today the sharper question is which exchange must happen now, and which can wait ten moves. A 96,058-SGF archive, two tournament records, and node-by-node KataGo analysis trace that change.
The legend promises divine inspiration and a defeated prodigy coughing blood. The board is harsher: Jowa did recover and win, but modern KataGo sharply disagrees with the traditional praise of the first ghost move.
Black 127 landed near tengen. A doctor supposedly read the moment not from the board but from Gennan's reddening ears. KataGo now offers a useful objection: magnificent direction, questionable urgency.
In 1933 the board's greatest institutional authority sat as White. Nineteen-year-old Go Seigen answered with territory, speed, and then a stone at the exact center.
When Minoru Kitani put Black 121 into an envelope, sealed moves entered a major Japanese title game. A fairer clock also became a weapon, and Yasunari Kawabata turned the collision into The Master of Go.
The Hiroshima bomb exploded while White 106 was being replayed after adjournment. The shock wave scattered the room; people helped one another, restored order, and resumed. The game matters only if disaster is not romanticized.
Fujisawa Shuko needed one win for a seventh straight Kisei. Cho won four in a row and game seven by 1.5 points. The useful drama begins at Black 143, where a leader gradually returns the initiative.
The first win-and-stay team match reached its two captains. In Beijing, Nie beat six-time Kisei Fujisawa by 3.5 points. The public story became enormous; the board's useful divide is White 50 and its whole-board direction.
Korea sent only Cho Hunhyun to the new world event. He recovered from 1-2 down and beat Nie Weiping as Black under Ing rules. The Korean Baduk Association identifies the early White 28 as slow: world titles can begin leaking in the opening.
Rui became the first woman and first foreign player to win Korea's Kuksu title. The harder board story begins when White 60 tenukis and Black attacks two groups not by chasing one to death, but by making each group's escape burden the other.
Textbooks say not to play a bad ladder. From Black 67 to 97 against Hong Chang-sik, Lee deliberately pursued stones that could escape. Each turn collected a forcing profit elsewhere, turning the board into a diagonal ledger.
P10 was not magic. It pressed White low on the right while preserving freedom toward the center. What felt alien was a priority humans rarely chose, expressed through perfectly legible Go logic.
Modern KataGo still gives AlphaGo about 98.7% after L11. The move deserves its name because it creates a corridor of concrete questions where one machine error can rewrite the match.