Hi HN! We built a game for fun where you answer What Beats Rock? And you can type whatever you want. An LLM decides the outcome. Highscores reset every week.
One fun finding: We tried a lot of models and we found that Llama-3 is not as good at linking concepts to emojis as GPT-4o. Ultimately, 4o had the best reasoning skills that made this game possible.
The generative part of language models can make for really fun "single-player" games where you're really competing with the inventiveness of the language model, so there's some sense that you're playing a game with infinite hidden complexity.