In My implementation of a personal mnemonic medium, I’ve noticed that working with prompts in prose format avoids some of the myopia characteristic of SRS authoring interfaces. But in one common situation, writing many prompts about a book I’m reading, the prose prompts still suffer for having been stripped of their source context.
Because they’re not associated with source material, it’s more difficult to evaluate whether I’ve “covered” what I meant to, or to revise prompts which are ambiguous or haven’t captured the right angle.
If I write many prompts about a single book, they’ll take the form of an enormous, visually undifferentiated text file. That’s awfully hard to navigate, visually and practically! So I’ll often separate conceptual sections with subheadings… but still. I find it’s often easiest to think about prompts in terms of the corresponding “position” in the text, but the prose-embedded prompts drop that spatiality.
See also my notes on Interface for writing spaced repetition prompts inline while reading web content for possibilities / alternatives.