a16z's @stuffyokodraws built a benchmark for LLMs playing Tetris. It revealed distinctive personalities in each model, and the differences between human and AI strategies. Here's what happened: (a16z)