Integrated Writing Environments - A Digital Writing Studio

Why is Writing so Hard?

I sit down with an amazing idea in my head. I want to explain it to the world. I open up a new document to write, but I have nothing but a blank page starting back. My notes are scattered in twelve different half-finished files and drafts, my reference material is in a neglected pile elsewhere, and my outline is a decrepit structure with no meat on its bones. I can put the idea into words talking about it to a friend, why I can’t put it into words on a blank page?

Writing is something that doesn't feel like it should be as difficult as it is. Computers and digital tools have made creating almost anything easier. In many creative fields, there is a new problem: it feels hard to get going because you have too many options. But with writing, we get a blank typewriter page and the most barebones set of editing tools staring back at us. Writing in Google Docs doesn’t feel like flying the spaceships that are Photoshop, Ableton Live, or Vegas Pro.

But writing feels well-positioned to gain from digital tools. We can use visual mind mapping tools and text-based machine learning with language models like GPT-3 to make the writing process easier, more personal, and more fun. Digital writing stands to gain in three major areas - representing the writing’s structure, searching our reference material for insights and support, and editing and stylizing our work-in-progress.

Structure - The Physical and the Spatial

Why is the text editor still like the typewriter if modern writing is nothing like typewriting?

The end product of writing is a linear experience. It is meant to be read in sequence from start to finish to present an argument or story and support it. But for the author, constructing that finished work is anything but linear. As writers, we dump ideas, pull notes, and shuffle paragraphs and sentences like 52 cards in a magician's hands. We may even make a few of them disappear too.

In Google Docs, the best we can do to accommodate this is move our cursor up and down the page. Cut this here, and paste that there. We have a single screen to work with, and ideas that aren't on screen right now may as well not exist. We struggle to keep structure and form in mind as we write.

You often hear of writers in the real world putting ideas and arguments on notecards. They move these cards around on a desk to visualize structure. What would this concept look like with these two other arguments? Writing and editing is a matter of hanging your ideas on the wall like paintings. You have to find the right place for them, and make sure they look right in the presence of others.

What if our digital writing tools accommodated this kind of dynamic arrangement? What if we could pull concepts from our notes, and spread them out on screen like a stack of note cards? What if we could add and remove these idea primitives like a Photoshop wizard can add, remove, and rearrange layers? In a digital writing studio, it could be the press of a button to switch between a traditional word processing view, and a graph view of those ideas represented with colors, shapes, and connections.

Less Monologue, More Conversation

What if we could make writing essays and articles less solitary as well? When you open up a new word document and get the blank white screen and ticking cursor, you may as well be staring into the void. Even when you manage to brain dump an outline or a draft, it still feels like the longest road to go from draft to finished product.

When you have a good conversation with a buddy or colleague, you don't feel that same dread. Ideas lead to ideas. You make new connections you had not made before. There is no feeling of needing to get to the end. If writing is a conversation with the self, why not have a digital conversation partner?

Here's where AI language models like GPT-3 come in. Language models may not truly understand English, but they understand context. You can prompt a language model with context, and it will give you new text or answers that often make sense. Contextual searching is much more powerful than text searching or filtering, our traditional tools for talking to a knowledge base.

Imagine you have dumped some ideas onto the page and now you are stuck. Now imagine a digital conversation partner that you could bounce those ideas off. It would search your notes, your reading highlights, and the conversations your friends and the world are having on Twitter. It would come back with what it thinks are the most relevant supporting points, case studies, and counter-arguments. You are instantly unstuck and writing from abundance once again, with no effort on your part beyond asking the right questions.

How often do we read something and think 'this would be a great point for this article I'm going to write'? Then when we get around to writing that article days or weeks later, we've lost that point and don't remember that amazing example we saved.

We have two options right now for bringing in support for what we're writing. One is to find it in the moment while we're writing, which requires us to break flow to go looking for it. The other is to search for all relevant supporting material ahead of time, which is a clunky and time-consuming process.

What if there were a third path, where our writing tool examines what we're saying, and is able to offer supporting points from our notes for anything with the press of a hotkey? This is all possible with language models and context-based searching, combined with a mindful practice of regularly saving what you read and think about.

Dressing Up Your Words

Now that we have something readable, how can we make it worth reading?

In writing, we walk the hero’s journey from brain dump to publication. But just because something can be read doesn't mean it's worth reading. Writing worth reading has humor, personality, specific examples, and it lives and breathes.

While style and personality are dependent on the piece and author, we have some tools already to help us make our writing clear and concise. Tools like Grammarly or Hemmingway Editor will take what you've written and help you find mechanical flaws. This is useful, but it breaks flow to use these tools. You have to take what you've written, copy it over, examine the changes, decide what to keep, make changes, then copy it back over to keep working.

Imagine a writing tool that establishes an API-like standard for editing plugins. You provide text, and the plugin returns annotations on that text. You could get Hemmingway's suggestions directly in your editor. You could look at Hemmingway and Grammarly's suggestions at the same time, or any other digital writing editor you love. And you could do this all without breaking flow to copy and paste your work to an external website.

When we write, we also publish to many channels. The article may go to our blog, but we also write a Tweetstorm for the article. This is a new format that requires understanding what we've already written, then repackaging it from scratch. What if we had an AI model that understood how to translate articles into effective Tweetstorms instead? We could feed it our finished article, and get a skeleton Tweetstorm to start with and avoid the blank canvas problem. Wordpress has already experimented with this idea. If you publish to multiple platforms, tuning your work to each platform should be effortless.

An Integrated Writing Environment

In software development, there is the concept of the Integrated Development Environment, or IDE. Software is written in code files in a text editor, like writing is paragraphs and sentences in a word processor. An Integrated Development Environment combines the editor itself with two things: intellisense and plugins. Intellisense is the IDE understanding the language you’re writing, and offering  suggestions when the syntax or formatting is off. The IDE also offers a suite of plugins for tasks like debugging, searching, and source control.

I envision digital writing moving towards a similar model, towards an Integrated Writing Environment. I want my word processor, but I also want writing intellisense and plugins. I want to write somewhere that makes it as easy to create good writing as Photoshop makes creating digital art. Like Photoshop, it starts as a blank canvas, but becomes so much more once the creative process takes flight. I want tools that make digital writing more efficient, more fluid, and ultimately more joyous.