Show HN: Jevman – AI decision models play Pac-Man

Jevman is an open-source Pac-Man benchmark that tests AI decision models by having them play 100 games each against classic arcade ghosts, with results ranked on a public leaderboard. Any model behind an HTTP endpoint can be benchmarked by converting it to answer directional queries based on game state, with standardized rules including a 2-second response deadline and games capping at 5 minutes. The project enables researchers to compare AI model performance on a real-time decision-making task with reproducible, verifiable results.