Chunking the Notes

splitting notes into study-sized pieces

Part of: Flashcard Maker

Real notes aren't one tidy sentence. They're a wall of text: a whole lecture, a textbook chapter, pages of bullet points. If you paste all of it into one call and ask for cards, two things go wrong: the model skims and misses facts, and long inputs cost more tokens per call. The fix is chunking : break the notes into smaller pieces and feed them in one at a time. Why chunk at all A model reading a short, focused chunk pays attention to every line. A model reading ten pages tends to summarize the highlights and drop the details, exactly the small facts you want on flashcards. Smaller inputs also mean each call is cheaper and easier to retry when something fails. Chunking is the single change that most improves flashcard coverage . A chunk is just a slice of text under a size limit Pick a maximum size, say a few hundred characters, and pack text into chunks that stay under it without ever splitting in the middle of a sentence. Cutting on sentence boundaries keeps each card's context intact: This is a greedy packer : keep adding sentences to the current chunk until the next one would blow the limit, then start a fresh chunk. Each chunk is a self-contained bundle of a few sentences. Ho

Challenge: Pack Notes into Chunks