Every Card System Led Me Back to the Tree

The Graph Lives Inside, Not on a Screen

Knowledge may be a web. But a web is not a view.

Inside my head, ideas leap freely. A conversation about cooking reminds me of a thermodynamics lecture. A song lyric unlocks a paragraph I have been stuck on for a week. This associative movement is how thinking actually works, and no tool on a screen needs to replicate it — because it already runs silently in the brain.

The trouble starts when we try to project that internal graph onto a flat surface. Nodes overlap. Edges tangle. Zoom in and you lose the whole. Zoom out and you lose every part. The network is the wrong abstraction for display, for navigation, and above all for the daily work of reorganizing what you know.

Card-based note systems ask you to build this network externally. They promise that once enough cards exist and enough links cross between them, structure will emerge on its own. I believed this for a long time. What emerged, more often, was clutter.

Three Paths, One Destination

Nearly every card-based tool I have used organizes its cards through some combination of three methods: folders, tags, and bidirectional links. Each one promises something different. Each one arrives at the same place.

Folders. Folders give you nesting — files inside folders inside folders. This is, by definition, a tree. A folder system is an outline wearing different clothes. The moment a card tool leans on folders for organization, it has quietly conceded the argument.

But folders are a worse version of the outline they imitate. A file can live in only one folder at a time. Every placement is a forced choice: does this note about cognitive load in interface design belong in the Design folder or the Psychology folder? Pick one. Lose the other. In a true outliner, the node exists where the thought exists, and if the thought migrates, the node moves with a single gesture.

Tags. Tags seem to solve the single-folder problem. One note, many tags. No forced classification.

It does not hold at scale. Within weeks you have dozens of tags sitting over unsorted piles. #productivity, #design, #reading, #ideas — flat labels with no internal order. Tags without hierarchy are just categories. And flat categories cannot organize a large, evolving body of thought.

Every serious tag system eventually discovers the same fix: nested tags. Obsidian supports #design/ui and #design/systems. Bear organizes its entire sidebar around slash-delimited tag hierarchies. But a nested tag — parent slash child — is a tree. It is a folder expressed in tag syntax. The moment you nest a tag, you have re-invented the hierarchy you tried to escape.

The reason is fundamental: without nesting, you cannot meaningfully organize many ideas. And the moment you introduce nesting, you are building a tree.

Bidirectional links. This is the purest expression of the graph philosophy. Any note can link to any other. Links run both ways. Over time your notes form an organic web — no hierarchy imposed, no classification forced, just ideas pointing at ideas.

I was drawn to this deeply. I still think there is something philosophically honest about it. But in practice, I kept running into the same wall: a pure graph has no center. There is no starting point. No overview. No way to say "here is the big picture" without creating a summary node that manually imposes a hierarchy — which is, once again, a tree trying to rescue a graph.

Lost Inside a Brilliant Man's Notes

Andy Matuschak's public notes are among the most respected examples of networked thinking on the internet. His sliding-pane interface is genuinely brilliant — click a link and a new column stacks horizontally beside the previous one, so you can trace the path of your exploration. The first time I encountered it, I was struck by the elegance of how thought could unfold across a screen.

Then I tried to learn from it.

I wanted to understand his thinking on spaced repetition, on evergreen notes, on why books are a weak medium for serious learning. And I could not find an entry point. Each note linked to five or ten others. Each of those linked outward again. The individual cards were sharp and clearly written. But there was no map. I could not answer a basic question: what does this person believe, and how do his ideas relate to each other?

His notes are built for his thinking — for the associative, private, leaping process of generating new ideas. And for that purpose, they may work beautifully. But they are nearly impossible to read as a coherent body of thought. The network is a thinking tool for its author. It is not a communication tool for anyone else.

This is the tension that graph-based systems never resolve.

And then there are the graph views. Roam Research, Obsidian, Logseq — they all offer a visual map of your knowledge. Nodes as dots, edges as lines, clusters blooming in color. People share screenshots of these graphs like trophies. Two years of notes, rendered as a glowing constellation. It looks like neural pathways. It looks like a galaxy. It looks, for a moment, like the diagram of a mind made visible.

When was the last time you used a graph view to actually find something? To restructure an argument? To decide what to work on next?

I never have. After years of these tools existing, the community has slowly arrived at the admission most users quietly knew: the graph view is decoration. It proves that connections exist, but it says nothing about what those connections mean. It is too dense to read, too tangled to navigate, too abstract to act on. It is a poster on a wall. It is not a tool for thought.

Compression Is Not Loss — It Is Thought

If knowledge is a graph, but graphs fail as display, as navigation, and as organizational interface — what do we actually need?

We need compression.

A tree is a compression algorithm for human attention. It takes a web of ideas and flattens it into a path you can walk. It forces each idea under a single parent. It discards lateral connections. It loses information — deliberately.

And this is exactly why it works.

When I organize my notes into an outline, I make explicit decisions about what matters right now. I say: in this context, this idea belongs under this parent, at this level of abstraction. I let go of every other valid placement in favor of one navigable path. The outline does not pretend to be the territory. It is a map, and maps are useful precisely because they leave things out.

I have written before about what this kind of compression meant during my hardest season of work. What I want to say here is something broader: the constraint of tree structure does not merely compress — it generates.

When I drag a node from one branch to another, I am not tidying. I am rethinking a relationship. When I promote a detail to a top-level heading, I am recognizing that my understanding has shifted. When I collapse three separate notes into one because I finally see their common root, I am not losing nuance — I am gaining insight.

This is entropy reduction. I take a disordered cloud of loosely associated thoughts and impose structure on it. The structure is lossy. But the act of imposing it — of choosing hierarchy, of deciding what is central and what is peripheral — forces me to think harder about my own ideas than free-form linking ever has.

Restructuring a tree is thinking. In a way that restructuring a graph simply is not.

What the Tree Gives That Nothing Else Can

Let me be concrete about what outlines do better than any system I have tried.

Structure and readability. An outline gives me instant visual hierarchy. I see the shape of my thinking at a glance. Main claims sit at the top level. Evidence nests beneath. I can collapse branches I do not need and open the ones I do. Compare this to a card system: open a card, follow a link, open another. How many cards until you see the argument? You cannot know. There is no zoom-out that reveals the whole. An outline gives you the whole before you ask.

Refactoring. This is where outlines leave everything else behind. Knowledge is not static. Understanding evolves, and the structure holding it must evolve too. In an outliner, reorganization is physical and immediate: drag a node to move it, press a key to promote or demote it, merge nodes that say the same thing, split nodes that carry too much. Every operation is one gesture. Every gesture is reversible.

In a card system, moving a card between folders is clunky. Changing the level of a linked note is a concept that does not exist. Merging two cards means finding and updating every link that pointed to either one.

Context compression. When I work inside a single branch, everything else disappears. I see this node, its siblings, its children — and nothing more. The rest of my knowledge base is folded away, not demanding attention, not pulling me sideways. The tree shows me what I need now and hides the rest.

In a time where every surface competes for attention, the ability to not see is the most valuable thing a tool can offer.

I do not claim the outline is the most truthful model of knowledge. It is not. Knowledge is a web, and the web is beautiful, and the people who built tools to honor its shape were right about the underlying nature of ideas.

But a note-taking system is not a philosophical statement. It is a daily negotiation between your mind and your material. And in that negotiation — in the real work of structuring, navigating, refactoring, and communicating what you know — the tree remains unmatched.

The graph is how knowledge exists. The tree is how I can hold it.

I stopped trying to make my tools mirror the shape of truth. I let them match the shape of my attention instead. And my attention, like yours, moves one branch at a time.