Why do I read? The many ways vary depending on context. Reading purely for entertainment means lazing in a hammock while reading some fiction cover to cover. Studying for a class means leaning over a textbook, maybe with reference material on the side. Poetry gets read aloud, savoring each word, for an audience (real or imagined). I gained something different from each way. Still, I kept feeling that there was something more to be done. Reading takes time and effort, and I wasn’t sure I was getting as much as I wanted from it. It’s been nearly a decade since I first started thinking about this, and my approach has changed a lot.
My default way of reading was to choose something that sounds interesting and start reading from page 1. I’d read one word after the next until I reached the end, then put the book down, and that would usually be the end of it. I’d be left with a warm fuzzy about whether it was a “good book” and some of the most memorable parts.
I decided to record the books I read, with a goal of reading at least 1 book per month. This kept me accountable to my goal, and also helped me understand which topics most interested me, and to decide what to read next. It’s not a complete picture: many books didn’t make it in, and no textbooks or articles. This system persisted for a long time - 9 years! It has 119 entries and so has served me well in meeting that goal.
One thing I noticed when I reached a double-digit number of books was how few details I actually remembered. I sometimes found myself recommending books, then stumbling when asked what specifically made it worthwhile. It became clear that just reading more books wasn’t the right goal. I wanted to remember the things I read about and be able to use them in my life. I informally tried a few strategies to consolidate my memory: reading the Wikipedia page when finishing a book; taking already-read books off the shelf and opening to a random page to check how familiar it was and whether I could still follow the narrative; finding a willing listener and trying to explain the gist in a way that felt satisfying.
I soon learned that what I was doing informally was Spaced Repetition. My first introduction to spaced repetition software was Anki. I spent weeks using it, but it never stuck. At first, my interest was in memorizing vocabulary in different languages, programming libraries, or classification systems I thought were interesting, such as for clouds or geology. Downloading someone’s pre-made deck or creating my own felt empowering; actually making it part of my routine quickly became a chore. After spending 20-30 minutes doing flash cards, I felt a sense of accomplishment for gritting through it, but also tired and zoned out. My recall seemed to improve as long as I kept doing it, but I felt aimless. I was remembering facts, but not learning skills. I definitely wasn’t having fun.
I realized a different approach was possible: to do spaced repetition on the things I read, not just on a list of facts. I spent lots of effort to highlight my books, whether on kindle, pdf, or paper. My goal was to extract these highlights, centralize them in one place, and refine them more using spaced repetition or progressive summarization. If I am honest with myself, this totally failed - my bookshelf is a graveyard of highlights that will never resurface in a useful way. The closest I came was in using a paid service called Readwise, which automates much of this work (it’s a new product and still needs polish). However, what convinced me to stop was not that it was too much effort, but that it was the wrong strategy altogether.
The key insight I had about Anki was that knowledge is something I use, not something I have. Anki doesn’t help me relate the knowledge in a deck of flashcards to the things I already know; it doesn’t help me apply it. Learning with Anki is rote memorization based purely on frequency of exposure, which is an important factor in memory, but not the whole story. I lacked appropriate cues. Recall was best when practicing a deck, but the facts didn’t feel accessible when I needed them. They weren’t integrated in my “latticework of models” (as Charlie Munger calls it). I realized this when I revisited using spaced repetition, but this time in Roam using roam toolkit.
My way of exploring this new learning method is to learn a constructed language called Toki Pona. I chose this for several reasons: 1) intrinsic interest 2) it’s small enough that it doesn’t need to be a lifelong endeavor 3) it’s complex enough that the skill can’t be rote memorized. The crucial difference between spaced repetition before and after is that I’m not simply trying to remember facts - I’m trying to perform a particular skill (speak and write a language). While the notion of reading a book and retaining its contents is still foundational, the method changes in a subtle and important way. The vocabulary, as well as sentences that illustrate grammar by example, became my flashcards. I remember one in particular, the word selo, which describes an outer shell, or skin, or wrapper. I frequently failed this card and saw it almost every day for a couple weeks. I noticed myself failing at this memory task, yet still could not remember the answer. Noticing that the current strategy was failing, I tried something else. I used a picture of a seashell as the cue:
My ability to remember this word instantly improved. I started applying a similar strategy to other problem words, using language of origin, movie scenes, mnemonics, or anything else that had emotional content. It is much easier for me to remember words by using a visceral and colorful image, rather than looking at the same words on a flashcard again and again.
A corollary of differentiating between facts versus skills is that memorization is only the start of learning. For example, in learning a language, vocabulary and grammar are given (a priori; axioms) and meaning is created by composing them together. It suggests to me that the dichotomy between facts and ideas is false. This led to designing questions that would help me "climb the pyramid". I created a flashcard whenever I detected a gap or saw an insight, for example to explain differences between confusing synonyms or homonyms, or to explain themes in the logographic symbols used for writing. As I move up the taxonomy, I might for example ask myself not to just translate, but make multiple translations and justify why one might be better than another. This effort is still in progress and my success here feels limited - it’s the idea of bootstrapping myself into an expert when no expert is available: how do I create a curriculum for myself? I might write more about this in the future.
How To Take Smart Notes by Sönke Ahrens helped me find the process that I’m using now. I began the book by highlighting key excerpts, and then stopped, knowing it was likely the last highlighting I’ll ever do. It helped me realize that while highlighting feels productive and creates an artifact, it doesn’t actually help me think. When I read past highlights, it feels more like I am merely re-reading an abridged version of the book and having simpler, less sophisticated thinking.
Instead of highlighting, I write my own summary, focusing on the parts that are most personally salient. Benjamin Franklin took notes simultaneously with reading. Ahrens goes on to say that reading without pen in hand is about the same as no reading at all. Even though I read this in Franklin’s autobiography almost 10 years ago, it’s only now that I have internalized Old Ben’s wisdom.
The deliberate act of summarizing by writing in my own words forces me to confront gaps in understanding. This differs from speaking, where claims can be hand-waved, exact statements forgotten. It is also different from highlighting, where the familiarity of re-reading without critical analysis can falsely convince me of knowing.
Writing also creates an artifact that I use later. At some other time, I review my summaries and relate them to the things I already know, looking for distant, non-obvious, or insightful connections, and make them explicit by creating another note that links them. This collection of notes goes into my memex (which currently lives in Roam).
Through this process, it’s clear to me that writing is the medium of thought, not the outcome; rather than being the thing that follows from independent study, the act of writing is independent study. Writing isn’t how I explain what I know, it’s how I learn it in the first place.
My writing is an external part of my brain. I often feel resistance to accepting this because I want to feel independent and self sufficient, not needing of an external tool which can be destroyed or taken away from me. Yet I also recognize that memories in my brain are being destroyed or lost every day, and that the external medium is actually much more durable and reliable. Cuneiform tablets, hieroglyphics, papyrus scrolls, and Incan quipu ropes have all survived much longer than the brains of their creators and served future generations. Most importantly, they exist in the first place because they served a purpose to their creators while the creators still lived.
In How To Take Smart Notes, Sönke Ahrens introduces the shipping container analogy for notes. In the past, cargo ships were loaded by stevedores, who could efficiently pack a ship. When the notion of a shipping container was introduced, it wasn’t obviously better, and seemed less efficient. What need is served by introducing an extra box to put things in? Why is it better to load up many partly-filled boxes instead of neatly packing everything together, allowing more goods to be sent in a single trip? On its own, it isn’t better. The advantage only becomes obvious when the whole problem is taken into account. When a ship arrives at its destination, stevedores unload the cargo onto the dock, and their job is done. However, the ultimate goal of getting supplies to where they can be used is still not met. By looking only at the subproblem, the logistical pain of repeated loading and unloading (which is different for every kind of cargo) is overlooked. By contrast, shipping containers allow cargo to be attached to a truck and taken to the place it is useful using the exact same process every time. Expanding the scope to include the whole problem necessarily entails adding complexity to keep things manageable. Crucially, by adding complexity in one part, the process as a whole can be streamlined.
In the way that standardized containers simplified the process of moving goods to where they could be effectively used, standardized notes simplify the process of assembling a string of thoughts into a narrative. Conserved effort can be spent on the interesting parts, instead of the logistical parts. Writing, like shipping cargo, has many subtasks, which would be overwhelming to try to do all at once; no writing really starts from a blank page. The subtasks of writing are different skills and require different kinds of focus at varying levels of abstraction. The writing itself is a focused and effortful endeavor; making connections and reflecting are free-association exercises that call for relaxed and flowing thinking; editing, correcting, proof-reading are much more mechanical and procedural. By breaking these apart, they become tractable. The first draft of this essay was written by composing my existing notes together. Rather than starting with a blank page, I merely chose what order to present the ideas I’ve already written down. Through this process, I solidify what I have learned while reading.
To write is to consciously and explicitly engage with my own strange loop. The process is non-linear, and can be seen as a complex system because the process allows unpredictable conclusions to surface, even though each note is tractable and well-understood, because I wrote it. Being the originator of all the components in the system (notes in the memex) also means that its unpredictable conclusions are relevant and emotionally salient to me.
In the way life’s purpose is to make more life, the purpose of reading is to write. In the past, my goal was to read more. Now, it is to write something compelling. This holistically captures what I want to get out of reading: the deep understanding that comes from remixing ideas and creating something new. Concretely, this means closing the loop and producing writing that compels another brain to do the same, after which their new ideas can come into mine. By adopting and developing my own writing process, it feels like I’m finally finding my voice.