Everything compounds. Not only money, but waistlines, popularity, curiosity, and ideas.
When ideas breed with one another, the offspring are usually hideous: these are the shower thoughts, and the drunken 3am kitchen ‘creations’. Ramen pizza is probably not going to catch on.
But very occasionally, you get something beautiful.
The last post was about collecting and curating the best ideas. This post is about earning compound interest on your collection. The magic happens in the nooks where ideas collide and fuse, but we have to create the right conditions for recombination—the equivalent of turning down the lights and piping Barry Manilow through the speakers.
To earn compound interest on your money, you need somewhere to put it, like a bank account, and a practice for making it grow—an investing strategy. Same goes for information. I suggest the ‘somewhere to put it’ should be a Commonplace Book, and the ‘practice for making it grow’ should be the Zettelkasten Method.
If this is all German to you, no fear. Here’s what we’re going to cover:
- Modern Commonplace Books in a digital age
- The broken ‘file drawer’ approach to knowledge management
- The Zettelkasten Method
- Digital vs traditional tools
- Roam Research for Zettelkasten
What is a Commonplace Book?
In the dark ages Before Google (B.G.) people couldn’t idly search the collective fruits of all human civilisation while sitting on the toilet. Instead, they had to scrapbook their own homemade internets, which they called a Commonplace Book: a patchwork of collected quotes, snippets, articles, recipes, notes, and ideas. This book was interspersed with the collectors’ thoughts and reactions, and not intended for publication.
Commonplace Books reached peak popularity during the Renaissance, but they go back a couple millennia: we have examples from Marcus Aurelius, Julius Caesar, Leonardo Da Vinci, Thoreau, Virginia Woolf, Thomas Jefferson, etc.
The defining feature of a Commonplace Book is that it’s an all-purpose repository of cool ideas, hence the ‘common’—there’s no specific theme, no chronological order, and no rigid structure.
Now it might seem unnatural to keep your thoughts and ideas jumbled together like this. But the exact opposite is true: imposing a rigid hierarchy is unnatural!
The hippies were right: everything is, like, connected, man. Before we get to how we might implement a modern Commonplace Book, a quick overview of how the brain works—or more accurately, how it doesn’t work.
Floating Down the River of Associative Knowledge
There’s a river of information flowing through our brains: chatter, sensory stimuli, thoughts, feelings, books, blog posts, tweets. Something like 99.99 per cent goes straight in one ear and out the other—maybe forming a few eddies for microseconds or hours or weeks, but ultimately sluicing back out without a trace.
This is as it should be. The river of knowledge is as broad and fertile as the Nile, which is to say, full of nuggets of excrement, viral diseases, and the occasional crocodile. We don’t want most of this stuff to stick. Even when it comes to the good information, ephemerality has its appeal—I love the fact that I can forget how a novel ends and get to enjoy it all over again.
But the valuable concepts, ideas, and stories that drift our way are worth retaining. If we want to get compound interest on our knowledge, we have to stop all these precious ideas from draining straight back out the holes in our colander-brains.
That means building a framework of mental models to ‘hook’ ideas as they drift down the river. Once they’re part of the structure, we can hang new ideas on them, and so on. We end up with a sprawling latticework that expands its surface area in a non-linear fashion; like beavers took a bunch of acid and built a four-dimensional dam out of coat hangers.
Now when something comes floating down the river, we have any number of hooks to snag it with:
“Oh! This is an example of X, except in the field of Y. I think I have a case study like this. It reminds me of Dr. Z’s concept of A, except the main difference here seems to be B. I wonder how it applies to C?”
This is associative knowledge, and it’s the main way the brain stores stuff. There are no discrete categories: just a bunch of weaker or stronger connections and patterns.
Unfortunately, every generation is doomed to model the brain after the most advanced technology of the age: Aristotle and his bros thought it was an ice-box for tempering the humours or whatever, then we had the hydraulic model of the brain pumping thoughts around, then the clockwork model, then electrical lines, and now we’re stuck on the computer-as-brain model.
We’ve grown up storing our files in a bunch of folders neatly organised by hierarchy, and trying to taxonomise everything in discrete categories. The fact that people think words—a bunch of grunts strung together by monkeys—possess some magical ability to neatly carve reality at its joints causes us to lose our minds on a daily basis.1
The file-drawer approach to capturing knowledge looks like setting up open-mouthed boxes in the river, nested inside each other like Russian dolls. A new idea or nugget of information drifts along, and is funneled into progressively smaller boxes until it finds its final resting place. The boxes only have one plane open to the world, and can’t overlap, which means the surface area of the structure only grows in a linear fashion. A place for everything; everything in its place.
The file-drawer approach is very tidy, completely unnatural, and completely kills the mood for the kind of freaky ideas sex we want to encourage.
So that’s the distinction between hierarchical and associative knowledge. We want to build what Charlie Munger calls “a latticework of mental models”, with as large a surface area as possible. This not only helps new ideas stick, but creates more opportunities to combine knowledge in original ways.
Now for the specifics. I’m a big fan of the Commonplace Book, but I think we can make some major improvements to the old-school scrapbooking approach.
The Zettelkasten Method for Commonplace Books
A Commonplace Book usually has some kind of organisation. Some people use a table of contents, or reference page numbers, or use tags or post-its to loosely group related ideas. But none of these strategies hold a candle to the Zettelkasten method.
Zettelkasten’s most famous adherent was Niklas Luhmann, one of the great sociologists and systems theorists of the 20th century. Luhmann was freakishly productive: he wrote something like 70 books and hundreds of articles in his career, many of which became classics, without ever having to force himself to write.
The secret to his success lay in his research method: when he read something interesting, he wrote it down on a notecard with a unique identifier, accompanied by his own thoughts. Then it went into his slip-box (in German, zettelkasten). The placement was determined not by the topic, but by how it related to his own writing, thinking, and existing notes. Any card could end up with dozens of branching subthreads.
I don’t use Luhmann’s exact methodology, but the bastardised Zettelkasten practice I’m about to demonstrate follows the same principles.2
(The example clips are recorded in Roam Research, which is my preferred medium for keeping a Commonplace Book. More on tools later; don’t worry about this for now.)
Zettelkasten Step 1: Deconstruction
Say we’re adding some Kindle highlights to our Commonplace Book. We’ve let them sit and breathe for a while, as described previously, and they’re ready for processing.
The first thing to do is break everything down into atomic components: preferably, no more than one idea per ‘card’. The way I usually do this is to write a catchy heading that sums up the gist of the idea. Underneath the heading, I’ll break down any subcomponents or explanation into a few simple bullets as required.
Everyone recommends you translate this into your own words, but I think it’s usually fine to preserve the original text. Do whatever works.
Everything gets deconstructed wherever it happens to fall: I’m not trying to ‘file’ each idea anywhere in particular. If you really want, you can assign high-level tags that will apply to everything you’re processing, but I wouldn’t recommend it—I prefer to nest notes under the author and book title only, and get more granular in the next stage.
(The tags in this example—#Risk, #Statistics, etc—apply to the book itself, but not its contents. If I wanted it to apply them broadly, I would indent the notes beneath the metadata block.)
Zettelkasten Step 2: Assonance and Dissonance
Now we want to smash the new idea up against everything else in our knowledge graph, and see what happens. At the coarsest level, the first pass is to tag the concept handles for each atomic idea.
Since I’ve mentioned #Risk aversion and #Social contagion, this particular idea (‘Defensive decision making is NOT the same as risk aversion‘) will show up in the backlinks for those pages, alongside every other connected note.
This is gonna come in handy for future research, but it also gives me the opportunity to scan over existing notes on connected topics. From here, there are three broad categories of questions I want to ask myself:
- What does this ‘rhyme’ with?
- Is it isomorphic to this other thing (has the same form/relationship)?
- Or is it actually orthogonal (at right angles to it)?
- Is this a subset of a broader pattern? Or a superset?
This makes me think more carefully about where I should ‘hang’ the new thing, and its relationship with all the existing things I know. Maybe I can build something new at the intersection? And once again, the connection is bidirectional: any other block or page I mention will point back here.
I use the notation #🔗aff (affiliated) to mark up assonance, because, well, #ass doesn’t quite strike the right tone.
(In this example, I add a reference to a single block ‘Man Plans, God Laughs’, rather than a broad topic or tag. This is what makes the Zettelkasten/Roam combo so powerful—every bullet point has its own unique ID)
- What does this clash with?
- How does it jive with my own experience?
- Who has criticised this idea? Are the criticisms solid?
- Does one idea need to displace another?
Maybe one will kill the other; more likely they are different opinions or addressing specific situations or use cases. Again, this helps me better understand both the old ideas and the new one. Here I use the standard notation #🖍cf: (compare, from the Latin conferatur).
There’s no need to resolve a conflict on the spot, especially if it’s some thorny issue that would require a ton of additional research. I might just have a vague hunch that something’s fishy, and jot down a reminder to compare against other sources. There’s a good chance I’ll stumble upon more information that will fill in the blanks down the track, and at the very least, when I come to use this info, either block will point me to the disagreement.
Relevance to my own life
- Does this relate to some current or future project?
- Does this suggest a cheap experiment?
- Do I need to take any kind of action? Read something else? Write something?
In which case I’ll tag the relevant project, add a to-do checkbox, or incorporate it into my GTD. I might also add stray thoughts, using a #📝note or similar to distinguish my ideas from the author.
So now the new idea is firmly enmeshed into the fabric of the Zettelkasten—rather than floating away down the river, or siloed in a dusty file cabinet.
Zettelkasten Step 3: Remixing
The point of collecting ideas is to use them. I’m not going to relitigate the previous post, but this is the key insight of the Zettelkasten method.
People think writing involves picking a topic, and then researching it. This is exactly backwards: after years of building chains of thoughts and connected ideas, you already have a bunch of topics ricocheting around your skull and begging you to put them on the page.
Which suggests a heuristic:
If you need to open Google for anything other than simple fact-checking, you’re probably not ready to write about a topic.
Besides, it’s much less stressful to follow your interests and let ideas emerge in their own time. Luhmann was ridiculously prolific because he was never faced with a blank page, and he always had multiple projects on the boil:
I only do what is easy. I only write when I immediately know how to do it. If I falter for a moment, I put the matter aside and do something else.
In the ideal case, converting the contents of your slip-box into finished prose is just a formality. This is never quite true for me—I always learn a lot during the final assembly—but I have usually done the bulk of the work months before I first put pen to paper.
The goal is to accumulate building blocks of ideas that are infinitely remixable, and can be assembled into whatever form you need. A tweetstorm can be a blog post, which can be an essay, which can be a book (although the trick is doing it the other way around).
And the creations you assemble aren’t limited to writing or research—they can be art, or inspiration, or wisdom, or any kind of life practice. I use my Commonplace Book/Zettelkasten for all sorts of junk.
This reassembly is where outliner tools really shine. I was initially resistant to abandon my beloved Google Docs, but now I do all my writing in Roam (I’ve copy-pasted this blog post across). The atomic structure means I can see exactly which contexts I’ve used an idea in before, and where it originally came from. And if I update the component, the changes instantly populate through the entire database.
So let’s talk about tools.
Paper vs Digital Commonplace Books
Why isn’t Zettelkasten a household name? Sönke Ahrens suggests it’s partly because the practice was confined to a small group of German-speaking sociologists, and partly because the idea is so deceptively simple.
My guess is that it’s partly because the tools just weren’t appealing. If I had read about the original Zettelkasten method, with arcane numbering systems and physical index cards, I’m pretty sure I would have bounced off. Seeing these cluttered boxes full of cards instantly triggers an aesthetic ‘ick’ factor. I realise this is dumb, but what can you do. I’m dumb!
My first exposure to the Commonplace Book was an old blog post by Ryan Holiday. Ryan writes his notes down on physical index cards too, but without the same system of cross-referencing. I’m pretty sure he has dividers for topics, and separate boxes devoted to book projects.
This was simple enough to get me started on the practice, although I did it digitally from the outset. I often draw and brainstorm on paper, I love physical books, and I understand the whole tactile thing. Ryan makes good points about hand-writing forcing you to be more discerning about what enters the system.
But…screw that. I hate writing by hand; I barely earned my pen license in primary school and it’s all been downhill from there. I not only don’t want to use a physical box full of index cards, I can’t—at least, not with my current lifestyle. Ryan Holiday has fireproof safes full of these things; I have a suitcase and a laptop.
There’s no doubt that index cards works great for Ryan, who happens to be a famous bestselling author, and may well have been the best choice until very recently. Most digital note taking tools have historically sucked, because they were stuck in the file-drawer failure mode.
In this sense, the web has also been a great disappointment. Google is gamed by SEO-savvy content marketers to the point of uselessness, wikis have the same hierarchical limitations, and Ted Nelson’s original vision for hypertext has long since become a joke: only starry-eyed fools and crackpots are still chasing those dusty old dreams.
Fortunately, one of those crackpots didn’t stop trying.
Roam Research for Zettelkasten
Conor had been studying this stuff for years when I met him. I only had a primitive practice at the time, but he got me hooked on the vision, and I helped him test the early Roam prototypes. We played with so many iterations that didn’t work. It was sometimes frustrating for me, and I was just a glorified beta tester—for him it was his life’s work.
But he and his co-founder Josh finally cracked it: Roam Research is out in the world, and making a splash.
Naturally, an army of clones have swooped in to copy the end result of years of research and development. I am not salty about this because it’s fantastic for the field of knowledge management, fantastic for consumers to have choices, and fantastic for advancing the goals of collaborative research.
Honestly, I’m not even going to try to do a proper comparison between the Roam-style tools, not least because I’m hopelessly biased. There are tons of important factors to consider—local storage, data portability, encryption, mobile app—which matter more or less to different users, and are constantly changing as the space develops.
Roam stands out to me for two structural reasons. The first is that the clones are missing the atomic nature at the level of individual ideas—the unique IDs at the heart of the Zettelkasten method. The second is that Roam is a here-for-life kind of project, not a short-term cash grab like so much of Silicon Valley. I know for a fact that Conor wants to write the Constitution of Mars with this thing.
The major points against Roam are that it’s expensive, and it has a steep learning curve. If you have the disposable income and time, I strongly recommend it, but perfect is the enemy of good (and don’t forget my glaring conflict of interest). Obsidian seems to be the strongest alternative for now, and there are others which are showing a lot of promise.
Whichever tool you happen to choose, I do think that if your modern Commonplace Book isn’t digital, you’re missing a trick.
One of the most common pieces of feedback about Roam is that it removes the friction from note taking, synthesis, and writing. You don’t want to be too discerning about what goes into the slip-box. You should be able to just write, without having to explicitly organise anything. Maybe your stream-of-consciousness diary entry clicks with the interview transcript you add a couple years later. That’s the magic.
It’s a lot of fun to just dive through the connections and see where it takes you—like Wikipedia surfing, but so much more intimate. And once the navigation of the graph view gets some upgrades, and we can move though our Commonplace Books in three dimensions, shit’s gonna get ridiculous.
My adventures in the world of knowledge management turned up three weird civilisational inadequacies. The first is that you can somehow make it through ~14 years of formal education without being taught how to actually read a book. I’m still bummed out about this, and wonder what life would be like if I’d been building this compounding source of capital from the outset.
The second is that the practice of keeping a Commonplace Book or Zettelkasten is not universally known amongst students, researchers and creatives—although I think it’s possible that the tools just weren’t there, and we might be on the brink of a renaissance.
The third mysterious inadequacy has to do with memory. Roam helps a lot with retention—the associative framework is similar to a mnemonic device, and you get multiple chances to review an idea: when you first read about it, when you curate your notes, when you hang it in the database, and through periodic exposure as it resurfaces.
But the optimal way to memorise stuff, is well, actual optimised memorising. If you don’t know about spaced repetition systems (SRS), prepare to have your mind blown. In the same way that Zettelkasten turns note taking from an ephemeral hobby into a longterm compounding goldmine of interconnected knowledge, spaced repetition turns rushed cramming for the test into a perfectly calibrated system to remember things until the end of time.
That’s gonna be the third and final part in the series. I don’t know if I’ll write it right away, because readers who aren’t interested in this stuff might be grateful for a reprieve. But it’ll be coming soon.
More on the file-drawer problem, associative knowledge, collaborative problem-solving, and the vision for what Roam is trying to achieve. The specifics have changed a little since we wrote this, but the technical/philosophical underpinnings are there for anyone interested.
If you prefer audio, this conversation between Conor and Erik Torenberg is the best overview. The Roam onboarding process sucks right now, but there are dedicated user websites with great guides: RoamTips, crib tours on YouTube, a Slack channel, and of course, #roamcult on Twitter.
Like many people, this speech by Warren Buffett’s right-hand man is what made the idea of building interdisciplinary mental models really click for me. Munger: “You may have noticed students who just try to remember and pound back what is remembered. Well, they fail in school and in life. You’ve got to hang experience on a latticework of models in your head.”
I also weakly recommend Poor Charlie’s Almanack, which is a collection of Munger’s essays (here’s my review).
A comprehensive guide to what I’ve been trying to get at here, including the full story of Luhmann and the Zettelkasten (my source for the quotes and information used in this post).
This is a terrific read because the author eats his own dogfood: the book was born of a Zettelkasten practice, and you can tell. It’s packed full of interconnected ideas, rather than the usual nonfiction gimmick where you get one central idea padded out with layers of filler and examples.
- This mistake is as old as Socrates and his featherless bipeds, but incredibly, it still dominates a lot of arguments today: ‘What is a woman?’ Well, ‘woman’ is a semantic signifier that we use to gesture at a cluster of associative traits: long hair, secondary sexual characteristics, capable of bearing children, high estrogen to testosterone ratio, big expensive gametes, somatic cells with XX chromosomes. None of these give us a universal definition: I have long hair but I’m not a woman; some women have different chromosomes, some can’t have children, etc.
Words have fuzzy edges, and there is no one pattern that makes sense in every single context: a medical doctor will draw the lines one way; a geneticist another; a religious creed another; a sporting body another; a workplace another; a society another. Pretending ‘life’, ‘woman’, ‘consciousness’, ‘gun control’ or ‘sandwich’ are pure Platonic forms baked into the structure of the universe is the sophomoric equivalent of starting your high school essay with “Webster’s dictionary defines…”
(For a full explanation, see Scott Alexander’s excellent The Categories Were Made For Man, Not Man for the Categories.)
- You can read all about the original system in Sönke Ahren’s excellent book How to Take Smart Notes (see Recommended Reading).