~99.995% of the computing resources used on this are from somebody else's servers, running the LLM model.