Our dominant mindset about technology is one of machine learning: people train computers, and computers learn from people. Eventually, computers will take over tasks that people do.
But suppose instead that humans could learn from how computers solve text issues to help them make better writing decisions, rather than deferring to an algorithm.
For computers to teach writers something of value, writers must decipher their workings, which are hidden in code. Notation in code offers a window into how text is handled as it evolves.
This post focuses on the role that notation plays in shaping text drafts. I’ll argue that notation can play a useful role in the writing process, provided it is not tightly coupled with specific media or tools.
Patterns and relationships in drafting
Writers imagine their relationship with computers, complex as it is, as one of teacher and pupil. Writers lead by example, computers learn. Writers hope to coax the computer to do what’s desired.
Yet writing is messy, and not an easily replicable process. Computers can only learn so much. That’s because writing involves many judgment calls, which only a human can make.
The writing process is a dialogue between the writer, the draft, and the imagined reader.
As the dialogue about the draft unfolds, fresh thoughts emerge. Writers and reviewers make notations in the draft for follow-up.
Editors apply blue pencils to typescripts, leaving proof-correction marks to indicate suggestions and changes. In computer manuscripts, reviewers layer ghost-like comments that hover in balloons.
Developing a text involves notational practices — letters, symbols, and marks. Writing notation is a kind of pattern language that evolves organically from conventions adopted by various groups over time.
Software applications often play the role of a third wheel in this dialogue, demanding attention, suggesting rewrites at inopportune moments. They aren’t seen as helpful contributors or respectful partners, but as an undisciplined force that needs taming.
Unfortunately, as computers have become more intertwined with the writing process, writers have become more dependent on them. The writing process becomes captured by the platform used.
While there’s no denying the influence of computers on writing, this development does not mean writers must become dependent on them. Writers should resist fatalism about the role of technology. Instead, they should consider how computers can inspire new practices, rather than mandate them.
Learning from customs in computer code
Does something that can often be annoying offer anything it can teach us?
I’ve been looking at how to appropriate some of the intentions behind software conventions to support the writing process. I seek to uncouple the underlying purpose of these protocols from their execution in random-access memory. I want to divorce the ideas expressed in the software from the systems that run it.
The code typed in a text editor incorporates a range of notations. These conventions evolved to solve specific problems in the computer domain — identifying and indexing snippets of text or strings of characters for further action without causing confusion.
The routines that computers follow are best geared to deterministic processes, not compositional writing. Yet they can be useful outside of the computer programs they were originally designed for.
Writing code has discernible similarities to writing prose. Both demand attention to the placement of and interaction between strings of letters.
Computer notations have evolved over time, with various computer languages and programs borrowing from one another. Software programming has existed long enough to establish its own traditions and customs. These languages express vernacular practices that writers can learn from.
Certain coding signposts can be useful for writers, too, even when writing by hand. Borrowing computer conventions does not turn writers into robot apprentices. Writers understandably don’t want to become even more beholden to computers.
Writers need notation practices that work with both handwriting and the keyboard, so that notetaking and drafting are independent of the platform, and so the writer stays in charge of the process. That hasn’t been the case with most writing tools.
The most popular notational systems for text, Markdown and similar formatting systems (Asciidoc, Typst), are designed to shape the layout of text on the screen, not to shape ideas. Too much emphasis is given to formatting already-composed text rather than to how to represent ideas and intentions while composing drafts. What’s needed is notation that facilitates the writing process.
Writing notation to help you to think through issues
The writer’s draft undergoes a long gestation as a work in progress. During this time, it is tentative and incomplete. It can be hard for the writer to track what is happening in the draft.
Notations are a way to indicate what remains to be done, or what themes are emerging. I’ve discovered five computer programming conventions are worth stealing for writing.
Placeholder for the right word
$variable Combining the $ sign with a word signals that the specific thing mentioned is not important. Rather, the statement around the variable is what matters.
Writers can use a variable to indicate that what is said is true, independent of context — or to test the consequences of alternative variables in a statement. For example: In the United States, $person has the right to due process.
The next step is to define the variable with words. In the sample sentence, does $person refer to anyone, US citizens, or some other description of a person? I find this exercise helps me distinguish whether it is the actor or their role that’s most important.
Another use of the $variable is to write about something before you have locked down the terminology you intend to use across a document. I find that writing about abstract concepts is difficult because any label for the concept is a metaphor that can be interpreted in more than one way. By not worrying about the final terminology, I can see how I write about a concept in different contexts, which helps me understand how the term I choose needs to work in different situations. If I were to start using random synonyms as I draft, the concept I am referring to might drift in its meaning.
Who said or did that?
@citation The “at” sign started as an accounting convention indicating “at a rate of.” It was adopted for use in email addresses to signify the addressee’s domain. In messaging apps, using the @ sign indicates a mention of someone’s name.
In writing, @ can indicate a citation of a writer’s work. For example, @Smith 1776 refers to Adam Smith’s Wealth of Nations, published 250 years ago. This citation can be embedded in text, such as: @Smith 1776 argued that it would be reasonable for the rich to pay higher house-rent taxes.
The @ symbol indicates a reference to a multi-line data entry, which readers don’t need to see while reading. The citation convention began with software called BibTeX, which created an entry type (e.g., @article, @book) followed by a “citation key” consisting of the author’s surname and publication date.
To align more closely with the “mentions” convention, some software uses only the key indicating who and when (the author’s surname and publication date), similar to inline references in research papers. I’m following this approach.
The @citation is succinct shorthand for noting who said something and when. But I also find it helpful when citing events driven by people, to build a timeline (you can add month and date as YYYY-MM-DD if needed). It can be applied to any kind of actor: a named individual, an organization, or a group. This usage again is similar to the “mentions” convention. You are citing someone’s activity rather than quoting them.
What are the key concepts?
#topic The hash, or octothorpe, is a mark used to indicate comments in computer code. It later came to signify a “tag” indicating the topic of content. The hash can be added to a keyword in a sentence to allow the sentence to be indexed by that keyword.
Hashtags are less common in social media posts these days, but they remain useful for notes. I find that adding a hash to a word is helpful when the word embodies an important concept I want to follow. The word might be a proper noun (#Silesia), a specialized term with a precise meaning (#prosopography), or a word that has a specific meaning for the author using it or in the context in which it is being used (for example, #winners).
Tags provide a thematic index of your draft. When embedded in the text, they can reveal how you discuss concepts. For example, do you repeat them across the text, or fan out the discussion to related topics? Do you drop discussion of a theme at a certain point, where it might fade from the reader’s awareness?
What needs doing?
/task The forward slash has become a common hotkey to invoke a menu of commands in applications, such as the WordPress blogging software I am currently writing with. In a text document, a forward slash, preceded by whitespaces, signals a change in path.
Drafts are full of loose ends. The writer can use the slash-task notation to indicate anything that needs more attention:
/factcheck/getReference/streamline/updateInfo
In AI platforms, such commands are called “skills,” describing a set of procedures or routines to follow. They are not much different from a punch list or checklist a writer might follow to ensure they don’t forget anything. For example, you might maintain a list of questions that ask how the reader would understand or react to what’s been drafted.
Once finished with the task, you can remove it from the document: /getReference
What’s missing?
{{insert content here}} Double braces (figuratively called “mustaches” or “antlers”) are a convention used to indicate a place where instructions or text are meant to be inserted.
They are common in templating systems, where supplemental or dynamic text is added to a base template. Some systems use single braces (“handlebars”) to indicate insertion, but double braces allow writers to use single braces for other kinds of notation, especially to indicate the grouping of items that span more than one line.
Double braces allow the writer to separate different sections of an article or document, to focus on different levels of detail.
The writer’s mind shifts focus between the big picture (the argument) and the details and nuances (the explanation or justification). Double braces are good for setting aside parts of writing you are not focused on.
I tend to develop an outline or skeleton of what I’m writing, but keep the details of each section as separate files. I can indicate in the outline the issues to elaborate within double braces. Initially, I will describe the section’s intent and goal within the braces. After I’ve drafted enough of the section to understand what I want to convey, I can label it with a more formal filename.
The shift from intent to file might look like this:
{{show three diverse examples}}–>{{examples.md}}{{connect implications to next steps}}–>{{conclusions.md}}
Digital compatibility, not compliance
Writing practices should leverage computers without restricting the author’s freedom.
The motivation is to embrace innovations that computers have introduced in text annotation, without being forced to adopt a rigid, computer-defined process. The annotations can help you notice features in your draft, whether you scan the text visually or, if digital, use a basic “find” command.
Much writing advice promotes digital-only workflows, seeking to reduce the quantity of text that must be entered on a keyboard. That advice is useful for generating boilerplate (marketing copy, tech docs, etc.). But automation diminishes one’s cognitive engagement in the writing process. Compositional writing involves a set of tasks distinct from those in document production.
Writing shouldn’t be dependent on a specific tool or platform. The writing composition process remains poorly researched, and most recent research focuses on the technology-enablement of writing through networks, online collaboration, and AI augmentation — the features app vendors want to promote. Researchers assume that because technology is changing how writing is done, it’s important to learn to optimize it in the writing process.
But are the tools good for writing? Professional writers, such as Sven Birkerts and Cory Doctorow, among many others, have criticized the distractions caused by tech features in writing tools. “The last thing you want is your tool second-guessing you,” Doctorow wrote in 2009, a decade and a half before writing about internet enshittification.
Drafting quality versus drafting speed
Numerous writers insist that writing by hand helps them to think more deeply about a topic and choose words more deliberately. Handwriting is enjoying a mini-renaissance, from the teaching of cursive again in grade schools to the popularity of keyboardless devices like the Remarkable tablet.
Don’t underestimate the value of typing your handwritten notes: rereading, rearranging, and enhancing them. But if transcribing sounds like a chore, technology can accelerate the conversion to digital. Handwriting-to-text (HTT) is becoming more robust and widely available thanks to advances in AI, making paper-first a viable writing workflow.
For some other writers, a manual typewriter provides the friction to slow down and reflect before writing. While non-digital tools will never win in terms of speed or output quantity, they may foster higher-quality results for some tasks and some people. I can think faster than I write or type, but I find my thinking can be scattered when it’s a stream of consciousness. (For that reason, I never use voice dictation for non-trivial writing.) Now that AI can generate text faster than any human, speed is not the goal. Quality is.
I generally take notes by hand during the early stages of forming ideas. Handwriting helps me capture thoughts in a freeform manner, without deciding on their relationships yet. Handwriting frees you from making premature commitments. I can focus on concepts and ideas before worrying about how they fit into sentences and paragraphs. Computer screens, by contrast, force a linear ordering of material.
While I don’t typically draft complete articles by hand, I find it helpful in the early stages to generate options for the points I want to make. I consider handwriting more forgiving than typed text when drafting views in different ways, as I try to find the right phrasing.
Sometimes I struggle to spell a word when I am writing by hand, and I realize I don’t really understand the word’s roots and etymology. In contrast to my ugly handwriting, text typed on a computer screen looks finished, even when it is still raw. The neat characters call attention to stylistic issues that need fixing rather than the substantive ones. I’d rather learn tolerance for my half-formed ideas and nurture them.
When writing on a computer, choose a distraction-free editor — free of extraneous options and premature editorial feedback. All writers, not just those who struggle with ADHD, can find that using a tool that offers integrated access to everything online can be disruptive. It breaks your concentration and interferes with deep thinking about the meaning of words, statements, and arguments.
I prefer a document editor with a focus mode, so that I don’t have to see either menu options or text formatting markup. I don’t like writing in raw Markdown, as I find the formatting distracting. I appreciate Markdown as a portable, platform-independent file format. But I loathe it as a user interface for writing. Fortunately, Markdown editors exist that hide Markdown’s distracting ornamentation.
Keeping AI optional
AI chatbots have infiltrated almost every commercial application used for writing. It’s becoming harder to write on a computer without AI intruding, making suggestions you don’t want, hijacking your intentions. LLMs operate on the premise that “attention is all you need.” AI applies that mantra not just to web-scrapped text but to human users. It is now consuming our attention and seeking feedback in the same way social media has.
To ensure that the editor is truly AI-optional, rather than AI-dominant, look for an open-source tool, which will require you to use your own account if you want to connect to AI. Open source tools are free. The notion of having to pay an annual subscription licensing fee for the privilege of writing your own words is the opposite of freedom.
None of this advice is to imply you should never use AI. I recommend keeping your chatbot in a separate window from your editor during the drafting phase. If your text is typed in a computer file, you can always copy and paste those parts you’d like AI to provide feedback on. You can ask AI to critique the strengths and weaknesses of a snippet you’ve written, for example. But you should only do so when you’ve exhausted your internal ability to make that assessment.
There are downsides to allowing AI to second-guess what you want to convey. The writing becomes less yours. Your ability to compose and critique ideas atrophies.
Writers don’t need to run all their drafts through AI for a critique. Even the most capable AI platforms won’t necessarily understand the audience you are hoping to reach, or the point you are making — especially if the material is novel or presumes a lot of inside background knowledge.
Authorship is about the ownership of ideas. Its purpose is to express oneself authentically, even if imperfectly. One should always care about the reader’s priorities and strive to address them, but never assume a bot knows them better.
AI can be helpful for final editing, catching typos and wordiness. But even in the final stages of editing, don’t forfeit control. Bots only match patterns; they don’t understand meaning, despite the claimed semantic capabilities of newer LLMs. Bots can make your writing sound slick or mimic Hemingway’s style. But don’t let the bot make a choice you won’t make yourself. You stop being the author once that happens.
— Michael Andrews
