LLMs trained on open source make the common mistakes that humans make.