I would like to show GameSims, a small project that I built over a weekend (with help of claude) to play around with small llms (local LLM agents running on mac MLX), and learn about their capabilities and limitations.
It is basically a locally running arena where two local llms (Gemma 4 or Qwen Models) battle on a board game (like reversi, TicTacToe). To make the game more entertaining and add more influencing factor, those LLM agents also have the way talk to each other via chat while making a move to try to influence each other.