A developer has created Sameshi, a chess engine that fits within just 2KB of code while achieving a 1170 Elo rating. The project uses a 120-cell mailbox board representation and implements Negamax search with alpha-beta pruning. Built as a tribute to the demoscene computer art subculture, the engine demonstrates how sophisticated game AI can be compressed into remarkably small footprints.

For move evaluation, Sameshi relies on material-only assessment with capture-first move ordering and full legal move validation including checkmate and stalemate detection. The developer measured performance through 240 automated games against Stockfish at various Elo levels (1320-1600) using fixed depth-5 searches and constrained rules. These tests established the engine's rating with a 95% confidence interval of 1110-1225 Elo.

The project joins a tradition of size-constrained programming challenges where developers push boundaries of what's possible in minimal code. While Sameshi doesn't implement advanced chess features like castling, en passant, or promotion, its achievement demonstrates the elegance of algorithmic optimization. The source code, written primarily in C, is available on GitHub with 107 stars and continues to attract interest from the programming community.