Congrats on the launch. I think maybe you should get the LLM to pre-compute the paths and store the results. Getting it to calculate every branch of the story as you go presents a few problems: - When it is busy (like it is now), everyone is just waiting for the next section of text. - You are going to burn tokens, even when multiple people are following the same path

That said, the way you have done it means it is roughly unique each time.