Reclaiming re-reading and the Leitner box system

From my other posts, you can see I have a particular interest with squeezing every drop of value out of spaced repetition for tackling long problems beyond ordinary flashcards. Flashcards are traditionally built for atomic information or 1:1 facts that minimize mental load. My previous post goes further into optimizing this approach, which is a new way to handle long chains of reasoning without overloading each individual card with too much mental effort, which is a real road block when you're reviewing a deck of thousands of cards daily.

We usually apply a spacing effect to atomic definitions, or concepts compact enough for a single card, and now even to large problems. But what about your actual notes?

I often write a lot of notes, hundreds of them accumulating in an Obsidian vault or an org directory, usually buried away, never to be opened again. To be completely honest, I'm not a devoted practitioner of the Zettelkasten method. I prefer long, linear notes that capture everything I want to say about a subject, that's just how my brain prefers it, splitting them only when it feels truly necessary. Now, expecting anyone to actively recall a thousand-word note from memory is unrealistic. Yet I do want to reread them. I want to refresh my memory on what I've written, and demanding 90% retrieval accuracy for that is pointless.

The studies are unambiguous in this case, and there is no room for argument: re-reading is inferior to the testing effect, which is objectively more efficient in virtually all types of recall. But I've found the rare exception that breaks this rule. Re-reading is an effective strategy when two conditions hold: the notes are in your own words, and they're long enough to make active recall genuinely impractical. And that's the whole reason behind why I built leitner.el.

The Leitner box system is one of the oldest spaced-repetition methods for fighting against the forgetting curve. It's so old that most people disregard it as obsolete next to modern algorithms like FSRS and the newer SM-2 variants which are not open source. Having implemented both myself (a post for another day), I still chose Leitner. Leitner does indeed become inefficient at massive scale, but for smaller quantity long-form notes, it actually works beautifully. More importantly, an optimal spacing interval matters far less when you aren't applying a testing effect at all, in fact it's irrelevant for re-reading.

So leitner.el repurposes this old system for a use case it was never designed for, but which I found it fits surprisingly well: resurfacing your old notes so they don't rot away in your filesystem, buried away for years. Moreover, most people never look at their notes again. They sit in folders, untouched, for years. It isn't even a question of whether re-reading with Leitner outperforms active recall with a better algorithm, it's that most people take one extreme or the other, and their notes are never resurfaced at all. Any system that gets you to read them again is better than none.

This creates a clean separation of concerns. Use active recall and FSRS for material that demands retrieval. Use Leitner + re-reading(not recalling) for material you simply want to refresh your mind on.

The classic Leitner system is a ladder of boxes with fixed review intervals. leitner.el adds to this foundation with a couple features designed specifically for the re-reading workflow.