In last week’s post on ML pitfalls, I made a passing remark about the potential for using large language models (LLMs) to help us check for errors in machine learning pipelines.