I had assumed they were referencing the "Learning Output Style"

https://code.claude.com/docs/en/output-styles