About a year ago I read How to Take Smart Notes¹ by Sönke Ahrens. Since then my study methodology has gone through a Renaissance of sorts. Before this pivotal point I did read books, but most information tended to disappear shortly thereafter. I recall reading topics that I very much enjoyed in the moment, but had a difficulty recalling key points of. The effect was very noticable when reading dense text books on new subjects I wanted to learn. This effect has been subdued by a great degree (I feel like, not having a way to measure the before and after) by adopting the methodology that Ahrens shines a light upon. This magical method — this panacea of a technique, pegged as such by the subtitle — is indeed simple: write your thoughts down.

Introduction

To expand a bit on the idea, Ahrens advocates for taking three types of notes, illustrating the note-taking system designed by Niklas Luhmann (another German who likes books) some decades ago: you should take

• Reference notes: When you read a book and find something interesting in it, write a short note about the general idea, and on what page it can be found on.
• Fleeting notes: When you think of something interesting during the day, write the thought down such that you can revisit it later. This frees up your mind to perform other work than just remembering stuff.
• Permanent notes: Looking over your recently written reference and fleeting notes, expand on their ideas: lift them out and apply them to other contexts; push an idea to its limit; following a train of thought as deep as you want. Do these thoughts remind you of anything else? Can you associate this note with another one?

Luhmann took his reference notes and neatly sorted them into a reference library. He processed these and his fleeting notes into permanent ones. On these small pieces of paper (known as a Zettel) he wrote some alphanumerics in the top corner in order to associate them to related notes. He then put them in a slip-box (kasten; thus etymologically, Zettelkasten). What this effectively resulted in was external mental scaffolding. Or in popular parlence: a second brain. This scaffolding was written in his own words and formed from his own thought: he added a creative angle to the whole ordeal of learning, and in the process, impressed the new-found knowledge better in his long-term memory.

Now, picking up any note from this box would result in two things:

1. it allowed him to revisit ideas he'd previously had (retrace mental steps), and offer a mechanism to expand upon them with new knowledge (by simply adding a new note); but perhaps most importantly,
2. it would reveal a thread of ideas (linked by the ordering metadata) that could take some surprising turns, revealing new relations or inspire new thoughts.

Let's consider an example. In fact: let's pull a random note from my own slip-box, verbatim:

Fuel/oxidizer ratio is not always set stoichiometrically

Ideally we'd burn all fuel and oxygen to 100%, but limits and requirements of

• specific impulse,
• tank volumes and weight, etc;

forces a design that leaves some fuel or some oxygen unburned.

This is a reference note I took while watching a video on rocket engine cycles. Let's see if the two things enumerated above can be evoked:

1. The note says that all fuel in a rocket engine isn't burned. We abstract the statement to encompass combustion engines in general. What, other than rockets, have a combustion engine? A car, of course. Now, having recently learned some new things about radiometry,² I now know that exhaust gases stand out a lot in the infrared spectrum due to heat. I subsequently think about an idling vehicle standing in a forest, and an IR camera that is looking at this tranquil scene. Aha! I make a new association:

   Non-stoichiometric combustion increase platform signature

   The sooty exhaust gases offer a contrast against a forest scene at ambient temperature. Consider also the vibration modes of the exhaust; molecule doppler effects.

   I have added more value to the slip-box by peeking into it and realized a connection between radiometry and rocket propulsion. Is this added note correct? From all angles? Will a sooty exhaust increase the infrared signature? Maybe. Depends. But if not, then probably in some other frequency band. Absolute truth is not important in the slip-box: the accuracy of this new note is sufficient to activate my neurons and recollect the connection between the founding ideas. I find personal value in this Zettel and that is all that matters.

   Note also the information density in the second line: not everyone knows what vibration modes are, or what molecule doppler effects could mean, but again, these notes are personal. The only person that needs to understand them is you.

2. Going back to the original note, it contains some metadata (partially shown in italics) that links the idea to the specific impulse of rockets and the oxygen molecule. Following the former a few steps I happen to end up on a note on the topic of dynamic range, which redirects me to the use of decoupling capacitors in low-power systems. Does this note relate to rocket propulsion? Not directly, but the relation is unexpected, even is disconnected by a few degrees. Following this particular thread did not spawn any new ideas, but so it goes: you can't strike gold every time.

The point I'm trying to show here, and one that is lifted multiple times in Ahrens' book is this: relating notes to each other creates additional value. The whole of the Zettelkasten is worth more than the sum of its Zettels.

So, does it work?

For me, yes. Absolutely. I find myself absorbing more of the information I read and watch, and realizing the worth of cross-pollinating ideas.

When I find a passage or a video snippet that interests me, I stop and write down whatever thoughts that spawn. This transitions my brain from passive to active as I have to actually perform work to create something,³ rather than nod along and say "yeah, that makes sense," and immediately flush the context window for the next new idea. Pausing and doing work commits the concepts better to long term memory and contributes to the value of the slip-box.

I find myself

• branching out into topics I would never before consider dipping my toes into;
• contributing better to discussions; and
• having more fun reading books.⁴

One context in which the Zettelkasten method surprised me with its effectivity was when a question dropped into my email inbox. The question happened to relate to something I had read not very long ago. I looked up the subject in my slip-box and found a summary of the topic in my own words, along with a reference to the book and page where I had read it. I composed the mail in a few minutes and went back to work.

I see further great potential of this method.

---

Will Zettelkasten offer you the same metamorphosis? Only you can know the answer. If the topic sounds interesting, I recommend getting your hands on the book.

On quality

As I write notes on a daily basis I do not always find reason to dive deep into a thought. Sometimes I've just finished a book that really wasn't worth the time and just want to commit the few key points of interest to the box. These notes are minimum effort. Some notes I don't even link to another, simply because no relation pops into my head at the time of writing. I've found that trying to maintain a set quality level is difficult and not really worth the effort: the dataset will invariantly be heterogenous.

But one rule is constant: the entropy⁵ of the note must be sufficiently low that you can be reminded of the mental context you were in, while writing it. You should be able to stand on the scaffolding you've built and observe the painting on the wall. The more context and fleshed out the note is, the stronger the structure, and the clearer the painting. If you encounter a note of low quality you may not understand it: there is no painting on the wall or the scaffoling wont hold you weight. The thought you considered important enough to commit to paper wasn't elaborated upon enough, and has now passed away with time. Sad.

Weapons of choice

In my setup (org-roam in Emacs) all notes live in a single directory,⁶ ~/org/roam/*.org, and are indexed by a unique identifier (which I don't see directly) and the title of the note. This lack of a "proper structure" removes the option paralysis of trying to figure out where a note belongs. I can capture one from any context I am in, connect any notes I feel are related in the moment, and attach a tag or two if I feel like it.

In practise I don't interrupt my usual work for a few minutes of writing, but the interface is mirrored when looking into the box: as I'm writing this paragraph I press <C-c n f how to take smart RET> and access all my reference notes for Ahrens' book I wrote in November last year. Let's assume that I don't know what the rest of this section should cover. So: I copy, transform, and expand upon some notes I have already written. Sourced from page 129:

As you write about what you read you read you will become proficient at finding problems in the text:

• "Wait, the authors states X over here, but the equation lacks this variable I expected."
• "This sentence took me five passes to interpret. I would instead have written it like…"
• "This analogy is weak, what if the idea is instead explained via…"

and from page 83:

You can easily fool yourself into thinking that you understand a subject, while it really is acquaintance with the subject. Writing about the subject we're learning allows us to differentiate the two.

The above method is also highlighted in the book. As you have already written a bunch of notes on a subject, you will never again start with a blank page when you have a report to write.

---

But the spirit is weak: I sometimes leave the comfort of Emacs and pick up a pen and paper. In fact, sometimes I even create a voice recording. The most effective tool to extract a thought from your brain, I've found, is not always the same. In some moments I can barely push a paragraph to paper, but I can utter a sufficiently comprehensive monologue that I can decipher tomorrow. I recommend experimentation, and not locking yourself to a static set of tools.

---

My most common methods with Zettelkasten are:

• while reading a book: creating almost-finished notes that I later process in a second pass, digitalizing it (in Emacs).
• attending a lecture or meeting: writing high-entropy notes that are processed in two additional passes: the second to almost-finished notes (within a day or two), and the third into Emacs (within a week).

For fleeting thoughts I either send the thoughts into a dedicated email folder or write it down in a A5-sized field notebook I always have close at hand. This is also part of my getting-things-done workflow, which is better dedicated to its own blog entry.

Things to try further on

The adoption of new workflows is a heavily transitory process. Small steps allow you to gradually improve the efficiency of your new tools, and fix (or adopt to) papercut issues. After a year I think I've bypassed (escaped from) the worst traps that come with trying to adopt too many tools too fast. As of writing, I have two key things I want to improve upon during the coming year:

• I need to look more into the box: I've done it a total of two times, and both within the last week. I'm thinking about creating some weekly task were I spend perhaps 15-30 minutes starting from a psuedo-random note. This will remind me of old thoughts with the hope of keeping them fresh, and allow me to add a note or two with new knowledge.
• Utilize tags: org-roam allows me to tag notes. I always add tags to new notes, but I have yet to exploit them when peeking into the box. Perhaps org-roam-ql and some fuzzy string matching can give me a leverage down the line.

Much further down the line is applying a large-language model to my dataset. Asking my slip-box a naturally formed question and have it answer in my own words is an enticing idea. But involving anything AI too early sounds like a detrimental step. If AI is to be used, it should augment my workflow, and not become a critical part.

Tips for getting started

Here I'll leave you with a summary of tips and warnings I'd like to communicate to anyone else that wants to try the Zettelkasten method out:

• Writing notes will take longer time than the actual reading: Don't be surprised if you end up spending two full weeks creating permanent notes for a book that took a week to read. Pausing your reading and creating permanent notes after each chapter can offset any eventual Zettelstress.

• Don't create a network of naked ideas: put some meat in there: If all your notes are "X happens because of Y", it is difficult to resurrect mental context. The painting on the wall isn't really worth the space it's taking up.

  A recommendation is writing down your stream of consciousness or relating the idea to personal experience. If you're writing a note on Ohm's law, perhaps you have encountered its effect in a lab that you can write about?

• Consider quality over quantity: An idea can easily take up a page or two of temporary notes, and look like they will result in a few permanent notes. But having more than a hundred unprocessed notes mocking you over the span of weeks is not fun. Prioritize your backlog, not writing the maximum number of permanent notes. If a permanent note ends up becoming five paragraphs, so be it. The flat structure and a good search function will be able to index it anyway.

• Don't worry too much about quality: Some days you just can't write things well. I have a lot of permanent notes that are just copies of temporary scribbles (no proper paragraphs or grammar; a haphazard drawing or two of a figure). As long as you can interpret the content, it is a good note.
• Start simple: Don't try to adapt too many tools or methods at once. A transition period is required. It is always better to start from zero, adding a few new things down the line, than trying to have everything at once. An overwhelming toolbox can easily push you away from something good, subconsciously. In short: avoid feature creep.

Get in touch

If you have any questions about the Zettelkasten method, tools (esp. those related to Emacs and git), or want me to expand on some topic in this blog post, feel free to shoot me an email.⁷

Happy note writing!

Footnotes

¹ subtitled Simple Technique to Boost Writing, Learning and Thinking. Available at the author's website.

² Techniques of measuring electromagnetic radiation.

³ Which, at times, can be surprisingly difficult.

⁴ At the cost of increased literature bills, and an every increasing to-read list…

⁵ Sparseness; a lack of information required to remember your state of mind when the note was written. A good Zettel has a low entropy, and can be read 10 years down the line without prompting questions like "what did I mean here?"

⁶ Version-controlled with git, of course (and git-annex for any binary files, like embedded images).

⁷ v@tmplt.dev.