Idea:

Train a transformer based generative model and incentivise higher zaps on posts, get enough zaps/per posts rate to accommodate the inference cost of the server.

And…

You have autonomous AI!