Writing Slowly

Thinking about atomic notes

Researcher Andy Matuschak talks about atomicity in notes, an idea also developed by the creators of the Archive note app, at zettelkasten.de.

To make a note ‘atomic’ is to emphasise a single idea rather than several. An atomic note is simplex rather than multiplex. And this form of simplicity relates to the idea of ‘separation of concerns’ in computer programming.

Back to the unit record principle

But the idea is much older than this. I found something very similar described in 1909, in The Story of Library Bureau.

Library Bureau was the company Melvil Dewey (of Dewey Decimal System fame) established in 1876 to sell library catalogue equipment. But in the late 1880s the company realised that library cataloguing principles could be adapted to the rapidly expanding business world, which would open up new markets for record-keeping equipment and furniture.

In The Story of Library Bureau, atomicity of information is called “the unit record principle in business”. This concept pre-dates the digital era and to a significant extent the digital era presupposes it.The unit record principle is simple, and these days it seems quite obvious. It’s the idea that all the information on an individual unit (whatever it is) can be stored on a single record. This is quite different from the previously dominant record-keeping practice of maintaining leger books. The unit record principle, it is claimed, is made possible by the system of index cards. This system, known commercially as the ‘business system’, was developed especially in the late 19th Century, in conjunction with the explosion of record-keeping requirements relating to the emergence of mass production.

The rise and fall of the wonder-working card system

Around the turn of the century office furniture suppliers expanded greatly. They both emulated the Library Bureau and competed with it. Each company - Rand Ledger, American Kardex, Clarke and Baker, and many others, developed its own range of office furniture and its own spin on what to do with it. To justify the wide variety of filing equipment they had begun to sell, they also marketed various record-keeping techniques, including the use of index cards rather than the more traditional ‘book-keeping’ methods. The efficiency of the unit record principle led directly to the development of electromechanical tabulating machines for processing data, which Herman Hollerith invented in 1889. Collectively this tabulating machinery was known as unit record equipment, as it relied on unit records, now in the form of punched cards. Hollerith’s company became CTR, then International Business Machines (IBM) in 1924. In 1927 Library Bureau merged into Remington Rand. These businesses and their paper-based legacies eventually led directly to the digital world we all know.

Writing today: between a rock and a hard place

These days, we’re stuck between two quite different approaches to writing.

The first approach to writing is the word processing principle, which no one talks about but which is hidden in plain sight. Microsoft Word, Google Docs, and similar word processing applications assume there is no discrete ‘unit’. It just goes on and on, and on. Even the page, a unit previously well understood, has been obliterated by the metaphor of the endless roll of paper that Microsoft Word offers. Once you start writing there’s no indication of where or when to stop. By default, page breaks don’t happen. The text just rolls onwards to the next page, and towards infinity.

The second approach to writing is the endlessly flexible block approach taken by WordPress with its Gutenberg interface. This second approach is also seen in personal knowledge applications such as Notion and Capacities. In this metaphor you’re not really writing at all any more. Instead you’re just creating variegated blocks of content. A heading here, a paragraph here, an embedded quote or video or image here. Now there are plenty of ‘units’, but no guidance on what they are. They’re arbitrary, just ‘whatever you like’.

Notice how both of these leading metaphors - the endless scroll and the flexible block - are really more about the software than about a real concern for your writing process. Every medium contains (and assumes) its own demands, and it’s always been that way. For example, printing would never have taken off if everyone still expected book pages to be made of vellum, not paper. Gutenberg’s printing revolution more or less required paper (please correct me if I misunderstood this¹).

But paper has its own limits too. As you might have noticed, it’s hard to write on a paper notebook in the bath, or while running. So I’m not blaming modern digital tools for forcing you to work in certain ways and with particular assumptions about what’s possible. It has always been like that.

A more useful metaphor for writing?

In thinking about the unit record principle, an idea that seemed radical in the 1880s, I’m not simply indulging in nostalgia. Rather, I’m questioning whether our current metaphors for the medium of writing, though almost invisible, are as inevitable as they appear.

Meanwhile in my own practice I’m seeking to uncover something still missing in the 21st century: the shortest repeatable writing session that could possibly be useful. Not an endless scroll, nor an endlessly variable block, but a simple, atomic, modular note. Each note follows a basic, repeatable template, so these simple notes connect and link to one another, and so they may cluster and combine to form larger units of writing, while avoiding unnecessary complexity. From fragments you can build a greater whole.

Not everyone warms to this approach. I’m aware that since the rise of new database models, it may seem laughably old-fashioned to return to unit records. But I’m interested in the felt user-experience as a writer, not only in the underlying code. And for me, there’s still some wonder to be found in the old card system². Some scoff at the idea of splitting ideas up in this seemingly arbitrary way. Yet they don’t mock words, sentences, or paragraphs. These are all more or less arbitrary ways of making ideas atomic, and even database models need ‘entities’, not so far removed from the unit record principle, so I’m not sure what’s going on here.