What happens if you take the output of a mainstream LLM and send it through this app? Would that solve the issue of the original article?