A chess engine learning from zero — trained by everyone here, live.
Leaderboard
| # | contributor | games | positions | last seen |
|---|---|---|---|---|
| Loading… | ||||
Champion showcase
What makes it different
Never shown a single human game or a hand-written evaluation. It only ever learns whether it won, lost, or drew — every bit of chess skill is self-taught from nothing.
The entire engine is written in a custom programming language, from the ground up — not a fork of Stockfish, Leela, or any other engine. No borrowed or stolen code.
It learns purely from self-play like AlphaZero, but searches with classic alpha-beta instead of MCTS — a combination almost nobody has tried.
There's no big lab behind this — every game the community plays makes the champion stronger. The board above is the swarm at work, live.
Download the worker, run it, and your PC quietly plays games for the current champion in the background — every game makes the engine stronger. It auto-pauses while you're gaming.
Questions or want updates? Add Maswabe on Discord.