“The capability of large language models improves approximately every three months, and there is no sign of this trend slowing down.”