Open-source BYOK playground forces LLMs to output legal moves across five board games

wilsonye · reddit · 2026-09-08

The author built the opposite of typical "play chess against an LLM" demos: models must emit a parseable legal move applied by a rules engine, and three illegal or unparseable outputs in one match means a loss.

The author wants data on which models stay legal for a full game, whether listing legal moves changes strength or just syntax, and where Xiangqi/Go collapse compared to Chess.

Original post →

More from coding & agent

coding & agent channel →