Most database vendors offer users a large corpus of information on database tuning, including documentation, blog posts, and research papers. In this paper, Immanuel Trummer explores how this content (targeted at human DBA) might be used to augment automated tuning tools. He presents an approach wherein pre-trained language models are used to extract tuning hints from text documents, and then mined hints are applied directly for configuration or as input for further optimization. He presents the results of a simple prototype system that can improve the performance of MySQL and Postgres on TPC-H.