02 · reply search
More lookahead made the AI worse—until we gave it a smaller job.
The first selective-reply experiments taught OverGrid not to turn every tactical question into a minimax sermon.
OverGrid / AI Field Notes
Two ways to read the board.
Choose a behind-the-scenes account of the game AI, or a player-facing strategy story drawn from the same concrete problems.
7 notes for Replies.
Behind the AI
For game makers: engineering, evaluation, tooling, and failed experiments.
Open →
11 · responsive compute
The best move may be outside the search budget
Petra’s missed height tile exposed the real bounded-search problem: evaluating a move is useless if the candidate generator never offers it.
14 · adaptive Ender
Ender had an opponent model. It just never got to use it.
An honest account of a non-promotion, an adaptive search breakthrough, and the difference between activation and strength.
At the Table
For players: transferable OverGrid decisions, routes, pressure, and counterplay.
15 · current checkpoint
The best move is the move you can prove.
A player’s guide to choosing between reply-backed calculation and a small, visible board proof.
17 · asymmetric Bean
Bean’s fork: two threats, one exit.
A player’s guide to building asymmetric OverGrid threats that survive the opponent’s best cut.
player guide · counterplay
The reply is part of the move.
A short OverGrid calculation routine for choosing between two constructions that both look good.
player guide · conversion
A route is not a bank.
How to tell whether OverGrid pressure will convert into points—or disappear after one good reply.