DEV: Improve prompts and LLM iteraction through TEST set

[mmcky]

To make improvements to prompts and the quality of output, I think the best next steps are:

1. build a test set of before and after documents that can be used to build better prompts.

Test based engineering and development workflows seem to work best with recent LLM models, it can self iterate towards finding a better solution. Having a couple of high quality examples help to reduce over fitting the solutions.

• Develop before and after lectures that can be used in a test set.
• build a reference lecture that has "ideal styling"