From Markdown to LaTeX

A focused week on the writing side of the RAG study.

RAG saturation

I ported the manuscript from Markdown to LaTeX and drafted an outline for a reproducibility section: which configs produce which table rows, which seeds were used, and how to regenerate the figures.

The porting itself was mechanical. The useful part was that typesetting every claim next to its table made the unsupported ones easier to spot. A couple of sentences carried over from the May draft still described results I hadn't actually produced. I flagged them for the audit that eventually removed them.

Reflection

A paper's reproducibility section is a good forcing function. If I can't say which command produced a number, I shouldn't be reporting that number.