Four rungs of a guild ladder, each honestly harder than the last.
A workshop in the old world had a ladder, and everyone on it knew where they stood: the novice swept the floor and watched, the apprentice was trusted with real work, the journeyman had earned the right to travel and be paid, and the artisan signed the piece. The computer opponents follow that ladder, because it says something true that “easy, medium, hard” does not — each rung is a different relationship to the craft, not just a bigger number.
Plays by reflex. It sees what is directly in front of it — a win to take, a four to block, a three that demands an answer — and beyond that it simply plays reasonable stones. It does not read ahead, which means it will miss the trap you set two moves out, and that is the point: Novice is where the game teaches you. If you are new to five-in-a-row and custodial capture, start here and stay until winning feels routine.
The first opponent that thinks. Apprentice looks a short way past the present move — far enough to punish a blunder, to see the reply to its reply, to make you pay for a carelessly placed pair. It is the rung for a player who has outgrown Novice but does not yet want every mistake collected in full. In our measurements it beats Novice decisively, and loses to Journeyman just as decisively, which is exactly what a middle rung is for.
A genuinely capable player. Journeyman reads several moves ahead and plays the whole game — shapes, captures, and the trades between them. It solves every one of the fifteen problems in the companion book’s final chapter, including one that defeated every earlier version of these opponents: the famous fork position where the only save is not a block at all, but a counter-threat against the pair that holds the fork together. If you can hold your own against Journeyman, you are no longer a beginner at this game, anywhere you play it.
The strongest opponent in the app. Artisan takes about four seconds over each move and spends them looking deeper than Journeyman can, under the same judgment with more time to apply it. It knows the oldest truth of the game — that a threat belongs to whoever moves next, that a four with the move simply wins, that an open three with the move is already a four — and it plays like it.
“Each level is harder than the last” is an easy thing to print and a harder thing to prove, so we prove it. Every rung is measured against its neighbours in head-to-head matches, under Classic and Keryo rules alike. And the whole ladder is scored against several hundred rated games by strong human players: the higher the rung, the more often it chooses the same move the expert chose — and the top rung is the one most likely to decline the moves weak players make. When a tester finds a position where an opponent plays badly, that position becomes a permanent part of our test bench, so a mistake once fixed stays fixed.
And none of the four is a script. Whenever several moves are genuinely equal — which is most of all true in the opening, on a nearly empty board — every opponent chooses among them, so no two games against the same rung need begin alike. Predictability is the one weakness they all refuse to share.
However far you climb, the hint button is always there — it lends you the computer’s eyes for a move, in your game, whenever you ask.