We have to work on the scaling. There is lots of web scraping and LLM calls that we need to make sure works under load. I'm sure there will be quality improvements as well.