Why I'm Long on Literature

I've been reflecting on an appropriate inaugural post for this new blog, The scope of my initial attempt at a commencement piece quickly spiraled into a multi-part series that will see the light of day someday but not today. Then I wrote ~5000 words to organize my thoughts for work and spent much of the day reaffirming why I (still) think writing is such a good use of time.

For most of the internet age the platforms have operated on an unspoken hierarchy of digital mediums, an implied ranking of value. Text is easy, pedestrian, commonplace. Images and audio are more interesting, dynamic, and communicative. Video is the top dog; the most desirable medium and most worthy of investment. Each step in the ladder requires more bandwidth, more technological implementation, and can be monetized more deeply. Tech companies have spent a decade trying to invent a yet loftier rung: 3D video, virtual reality, maybe even smell-o-vision. Now generative AI is attempting to follow the same path not just with distribution but creation of media.

Like most consumers and denizens of the internet, I am ambivalent to this trend. Content is not a monotonically increasing scale and there are different values and contexts best suited to each. Sometimes the barrier of having to scrub through a video to find a few sentences worth of key information is exhausting. Yet reading pages and pages of prose that could be easily piped into my headphones while I do the dishes could be equally onerous.

My uncontroversial view is that which consumers of information usually intuit: regardless of how it's delivered and monetized media is not a hierarchy and each medium has contextual value. The more esoteric flag that I'm willing to plant is that text is actually the most broadly valuable and is the foundation for most great idea generation and distribution.

Writing is thinking with something to show for it

While I'll allow for some brains that work radically different than mine, I would posit that the process of writing is integral to the process of thinking for the majority of people. The catharsis after an evening spent with your journal should hint at the power therein. Each time we write an angry email that we never send we've used the framework of writing to clarify our thoughts. Sharpening the unstructured thought and abstract intuition in our minds into structure, sentences, and words is the best and fastest feedback loop for idea iteration.

Thinking is feedforward; writing is feedback. Even without engaging other people to review or edit your work, the process of writing creates an iterative and self-contained feedback loop. Writing down an idea creates a save point, a snapshot in time of thought. You can then go visit other contexts; take time to think more, synthesize new information, go to a physical or hypothetical space where you think differently, or even see internal ideas reflected back at you from the page. Returning to the past checkpoint now engages a conversation across the dimensions of your own mind, closing the feedback loop across abstractions. I'm adamant that the brain is a muscle and thinking is a skill that can be honed but there is no amount of practice that allows you to exist in two contexts at once. We can move through limitless dimensions of emotional, physical, and mental state yet those moments are mutually exclusive. Writing allows us to approach concurrent and simultaneous multiplicity.

Meanwhile LLMs promise the artifacts of writing without the process. There are usecases for text of pure utility, communicative copy that needs only convey meaning and tone, like advertisement or documentation. But text generation is not writing and outsourcing the former doesn't invalidate the need for the latter. Write not to acquire text but to think, to clarify, and to challenge. There is nowhere that this is more critical than in education; we do not ask high school juniors to write an essay on The Great Gatsby because the final artifact has intrinsic value. There is effectively no chance that one of those essays might contain new and valuable insights into one of the most analyzed pieces of literature ever written. We make them write the essay to learn how to process and synthesize a book into thoughts, to streamline emotional or intuitive reactions to a lot of information into a cohesive summation and response.

Literature diffuses quickly and cheaply

In the annals of fine literature some of the greatest works exist because the barrier to entry is one person and a writing implement. I marvel at the improbable combination of artistic vision and project management skills employed by the most successful film directors, as they work with hundreds or thousands of people over many years to bring a creative vision to life. But that stunningly high skill floor naturally limits the people who can speak through the medium of cinema to a sliver of the population, to say nothing of the accessibility of distribution.

The distribution advantages of text's low overhead are pro-democratic in nature. Both the publishing and film industries have their gatekeepers for better and for worse. But when conventional channels need to be bypassed, writing has the advantage. Starting an independent press or publisher is less resource intensive than a film studio and even fully self-publishing text works is an option. The pure bandwidth requirements mean that even after finding the team and funding to produce a fully independent film, it will still be distributed on someone else's infrastructure, usually a big tech company. Meanwhile vast amounts of text can be hosted on cheap servers, bypassing nearly all of the established infrastructure.

Collaborative efforts can also suffer from their own biases, a form of regression to the mean. A radically new and unprecedented line of thought is a statistical block to finding like-minded collaborators. The direct link between a subversive mind and the text it can generate allows for works of literature to often be more revolutionary than contemporaries in other media. I would posit that most new ideas are first seen in text, even when it's other forms that popularize and spread them.

These constraints also apply to the realm of technical writing. I recently saw a post on a subreddit for a relatively new programming language in which a developer asked for recommendations for video learning resources. While I sympathize that videos can be a better learning tool, the downside is that the update cycle for video documentation is glacial compared to text documentation. The interesting new projects and languages in development don't get video walkthroughs or tutorials for years. Even when the ecosystem gains those resources they'll be updated on a much slower timeline than new version releases. Limiting your learning to only video resources introduces a hidden bias away from new and interesting projects that are developed and documented in text for years before diversifying across media.

Writing is for everyone

Oxide Computer Company's famous hiring process relies on their "materials"; a rather lengthy set of essays written on various topics of past work experience and future goals. When discussed in online forums these materials are sometimes viewed dimly for the amount of labor they request candidates to perform. Oxide Co defends the process as being generic to the role and helpful to the candidate, with the writing being unspecific to Oxide and good practice for any conscientious job seeker. Oxide has also open-sourced their application process, hoping that these materials could truly be generic across companies. Testimonials from both hired and unselected candidates broadly support Oxide's position but I believe there's an even more significant benefit that they're underselling: accessibility.

It may be impossible to design a high-fidelity information exchange without some bias towards different dispositions. Using hiring processes as an example, candidate assessments can include STAR-style exchanges, presentations on past projects, or live coding challenges. These are all terrific opportunities for some candidates yet terrible fits for others. Public speaking is the most common phobia, some highly competent people simply won't work well under time pressure, and language or cultural barriers can hinder any conversation.

Generalizing beyond the hiring process, this is true for any exchange of ideas across communication techniques. Does the most confident and outspoken person in the room always make their ideas heard because the default method of communication favors their style? Sometimes in a sales or marketing context the ability to pitch the idea well is reflective of the quality of the idea but just as often in technical contexts the two metrics are entirely disconnected. You just may be suffering from the hidden bias in communication and leaving excellent contributions on the cutting room floor.

Writing and text is as close as we can come to an equitable mode of communication. The asynchronous nature removes pressure and reduces anxiety on those without public speaking inclinations. It provides much more opportunity for non-native speakers of the chosen language to study their word choices to ensure they're communicating what they intend. It allows writers to find their iterative best feedback loops; asking colleagues to review, back-and-forth with an LLM, reading aloud, or revisiting their own words after some time and distance. There are downsides, those who will be better and worse at writing, yet I contend that on average a larger proportion of people will be able to communicate well using writing than any other medium. If you want to solicit ideas broadly - in a hiring process, a pitch competition or project proposal, a request for discussion/comment, or any other broad communication - choose literature to cast the widest net.

A note on reading

I'm effectively proposing that we introduce much more writing into our lives and that has the natural consequence of requiring much more reading to make any use of it. Reading benefits from many of the same advantages as writing; it's also a guided thought process, text is easy to distribute and easy to access, and reading can ease cultural and language barriers. But also like writing, active reading is a skill that must be developed. Learning one crosses over heavily to the other but not entirely. Perhaps I'll expound more on why I'm also long on reading in a follow up post.