August 26, 2022

The Graph, the Gardener and the Architect

We shape our tools and they shape us back.

Paul Rony

Founder of Kosmik

Networked computers by Mel Furukawa for BYTE (May 1985).


George R.R. Martin said there are two types of writers: gardeners and architects:

"The architects plan everything ahead of time, like an architect building a house. They know how many rooms are going to be in the house, what kind of roof they're going to have, where the wires are going to run, what kind of plumbing there's going to be. They have the whole thing designed and blueprinted out before they even nail the first board up. The gardeners dig a hole, drop in a seed and water it. They kind of know what seed it is, they know if planted a fantasy seed or mystery seed or whatever. But as the plant comes up and they water it, they don't know how many branches it's going to have, they find out as it grows."

In knowledge management, we can see a similar pattern in how people consume, save and create knowledge. Architects plan and organize their work before creating, they are comfortable with hierarchical file structures, and thrive within constraints. Structured tools like Notion or Obsidian work great for them. 

Think Tank an “Idea Processor” based on an outliner editor from the late 80’s.

Gardeners are wanderers, they explore multiple paths, play with different hypothesis and have only a rough idea of where they want to go. They need a different type of tool to suit their way of working.

You can tell a gardener from an architect by looking at their desktops. Gardeners organize files by position, color, icon. Architects prefer to use names, tags and other “meta labels”. The distinction is clearer if you express it in terms of “filers” vs “pilers”. Filers, as in "I put in a folder/file" (like architects do) vs piler, as in "I put in a pile" (more unstructured, like gardeners).

Graphical hypertext systems can combine the two to propose a structure that offers the best of both worlds. But we are not yet there.

The first tools for gardeners and architects

In the 80’s, the hypertext field was already divided between people advocating for systems like ThinkTank that relied on an outliner (like Roam or Obsidian today) and those that relied on usually visual “index card” metaphors (HyperCard, Notecards). Looking back, this divide was an early manifestation of the architect vs gardener debate. While hypertext is non linear in essence, the UI sitting on top of its link forest can be very structured. The problem of hypertext systems is creating and enforcing a structure without breaking the flow and serendipity that comes from using such systems. 

The Macintosh was a revolution for gardeners because the finder was spatial (really spatial, unlike now - you can learn more about it here). Spatial here does not mean “zoomable” or “infinite”. It means that an object (icon, window, menu) will keep its position and place in the user interface. The position adds context and gives cues about other items’ positions.

The user can find its way around the system by using visual cues combined with a robust hierarchical structure. The file structure is rigid but its representation remains fluid, allowing for changes, updates and serendipity. 

Despite the advances, creating a tool that truly suits the unstructured workflow of gardeners is difficult. So while there are a few tools in the market that serve the architects well, we couldn’t find a tool made for the gardeners amongst us. So we decided to build it.

We shape our tools and they shape us back

Why so much insistence on creating “the perfect tool”? Just as we shape our tools, they also shape our thinking. So having a tool that works for us can push forward human thought. We think that’s a pretty good aim.

Some people use just one program for their note-taking/knowledge management needs. But there is a clue that they aren't completely at ease and that we still have a lot of work to do to make digital tools easy, adaptable, and fluid: they carry a small notebook with them. In the past, these notebooks were called “commonplace books”.


Everything starts with commonplace books 

Until the 20th century people used “commonplace books” to copy quotes, poems, recipes and thoughts in one place. H.P Lovecraft also used a commonplace book between 1919 and 1934, the opening phrase of which is a good summary of the commonplace book ethos:

This book consists of ideas, images, & quotations hastily jotted down for possible future use in weird fiction. Very few are actually developed plots—for the most part they are merely suggestions or random impressions designed to set the memory or imagination working. Their sources are various—dreams, things read, casual incidents, idle conceptions, & so on. —H. P. Lovecraft


Commonplace books are fluid, adaptable and except for the chronology imposed by the bounded pages, completely non-linear. They are a repository of ideas and material. In essence the notebook gives a contextual structure to fragments that floats on its surface.

And the fluidity of those analog tools gave birth to the idea of a “file system” that could adapt and house unstructured data. This idea was explored by Ted Nelson,  in his famous talk A file structure for the complex, the changing, and the intermediate in 1965, where he used the term hypertext for the first time.

The file system presented by Ted Nelson relied on a pool of unstructured data (text and other types of content) that could be called by the user in a graphical interface where he would be able combine them into lists or “enfilades”. The enfilade is a data structure with the sole purpose of maintaining the order of the items used in it and save “contextual transformations” (insertions, comments, deletions) added by the user. By making the edits contextual, the raw data stored in the data pool remains unchanged but the user can explore alternatives and compare several options.

Such a fluid structure allows users to encapsulate a full writing workflow inside the system. The user can easily switch between:

  • Note taking

  • Organising and structuring

  • Outlining

  • Maintaining references and bibliographies.

Because the system is so fluid and can switch between a “pile” mode where fragments are displayed on a spatial canvas and a “file” mode where enfilades are used to give a proper structure (and store contextual data like tags for examples), it could finally reconcile gardeners and architects.

Kosmik, the adaptable software

With Kosmik and its underlying database/file system we tried to stay as close as possible to Ted Nelsons’s original paper, to build a tool that would work for gardeners and architects alike.

Remember, gardeners usually combine material until their final product - be it a document, note or presentation -  begins to emerge. This is the curation phase. To aid it, we build in transclusion and are working on backlinks to place the breadcrumbs that will bring them insights.

When we look at how people work in Kosmik we generally recognize three steps:

1. Brain dump (or capture) phase

The user imports items, and creates the basic blocks that will be re-used to later build a more linear product.

2. Curation phase

People sort the content or add context to the documents by adding annotations, connectors or post-it notes.

3. Sense-making phase

People connect the elements visually and build structure on the canvas. The beauty of Kosmik is that it allows you to continue to brain dump if you find connections or relevant things to add to your document. You can interrupt and change course at any time if you want to explore a new idea. With transclusion you can work on the same content anywhere on the canvas.

This is how we hope to finally help both gardeners and architects!


P.S. This post is an expanded and annotated version of this twitter thread: The graph, the gardener and the architect 🧵

Have a great weekend and thank you for reading! Subscribe for free to receive new posts every month ✨