i really hope we find a new technology to run the models locally with less expenses