Winter Term 2021 Retrospective (mainly on spaced repetition)

March 23, 2020

I initially wanted to call this a post-mortem, but then I read on Wikipedia that post-mortem is only about the failures and how to prevent them in the future. I thought I don’t want that. I want to focus on the good things as well. Hence, retrospective. This is an experiment. Here it goes…

This term was tough. I failed a class, I got into a big fight with my mom, I struggled to have a healthy daily routine, I signed up for too many activities… But it was also life-changing. I got a job offer at my dream company, I became more known on Twitter, I talked to a lot of very cool people, I learned how to lead group discussions about EA (effective altruism) better…


The main motivation for writing this post was actually my failure with implementing a good note-taking / spaced repetition workflow in Roam Research. At the same time, I also felt less motivated for schoolwork, mainly because THERE ARE ALWAYS SO MANY OTHER COOL THINGS TO DO. Sorry for caps.

But it’s true: I made Roam plugins, which was a lot of fun and I felt like I’m really “onto something”. I also ran or participated in a bunch of EA programs, which was fulfilling. Who, then, needs to care about social norms in primordial societies (yes, Sociocultural Antrhopology was the class I failed).

Anyway, I think the elephant in the room is roam/sr, my spaced repetition plugin for Roam Research. It was just really hard to actually use the system for studying hard science stuff, while still developing / thinking about ways to improve it. I didn’t do reviews every day, and in the second half of the term, I even ceased to make new cards at all. (Second half was also the time in which i released my other big extension, Python notebooks in Roam, and started more interacting with the community.)

It’s sad because (except for Antrhopology) my classes were super interesting and tractable. As Andy Matuschak says, The critical thing to optimize in spaced repetition memory systems is emotional connection to the review session and its contents. Which… I thought I had! Curiousity is curious…

I recently read the absolutely terrific LessWrong blogpost Seven Years of Spaced Repetition Software in the Classroom. The title speaks about classroom use, and many people think that’s very distinct from personal use. But I tried to pretend like my teachers are using SRS so it’s the same.

The main thing that resonated with me from the post was probably the pooping - a metaphor for the useful kind of forgetting. All students poop (at least in formal education). They have to, otherwise they’d go crazy. It now seems naive that I thought I could go through Caltech while retaining ALL IMPORTANT INFORMATION FROM ALL THE CLASSES.

Another thing I did wrong is that I mindlessly tried to apply the advice from yet again Andy Matuschak’s essay about writing good prompts. The biggest criticism of SRS is that the prompts are too isolated, too atomic; and I started a fire. Let’s say I’m learning math, physics and chemistry. What is better: go over a big concept in math, move on to physics, go over a big concept in physics, move on to chemistry, etc., while always asking just “Do I understand this?”; or: quickly go over a Q/A prompt about a tiny math concept, move on to physics, quikly go over a Q/A prompt about a tiny physics concept, move on to chemistry, etc.?

Well, I’m conflating a lot of things here. It’s okay to memorize basic terminology and notation this way. It’s not the best thing to do when learning about large and complicated concepts. The problem is that sometimes, things may seem like terminology, but they’re actually a large and complicated concepts. For example, entropy or enthalpy. You could define both of these in various ways, always carving out a new edge of them. They’re not suitable for a prompt “What is entropy / enthalpy?”, but maybe they’re fine for “What’s the formula for entropy / enthalpy?“.

But tell me now: how do I poop? Is the formula really the most important thing? No, what I’d really like is to deeply internalize a mental model of what entropy or enthalpy is (for the record, I still don’t have them). How do I poop the formula and keep the model?

I made very few cards about electromagnetism in physics. It’s likely that this is due to the fact that I’ve learned electromagnetism before, but I feel like, for example, I have a really really good model of what electric potential is now. I applaud the course for facilitating this through challenging problems sets.

And maybe that’s how it should be. Maybe my spaced repetition system should not ask me in four years “What is Gauss’ law?”, but instead… let me poop. Mental models are sticky (I hope).

Okay, enough of the failed attempts at creative writing and sounding smart, here’s what I want to do differently next term:

  • Optimize for passion and curiousity, not note-taking / spaced repetition workflow.
  • Distinguish between basic terminology and notation, and large and complicated concepts. Review terminology and notation frequently. Meditate over large and complicated concepts.
  • Look for the most important concepts and turn them into mental models. Forget the rest.
  • Learn. Don’t learn using SRS. Don’t learn by building a second brain in Roam. Just. Learn.

Profile picture

This might be my personal website in the future. For now, continue to my Roam plugins. Here's my Twitter.