Is Writing Thinking?
What happens when you delegate composition to AI
British technology journalist Ian Betteridge popularized the idea that when a headline is structured as a question, the answer is no. According to Betteridge’s Law—and the structure of my headline—writing is not thinking. Case closed. That was easy.
But if you still think this is a live issue…
Is writing thinking? I read a piece this morning that argued yes. I’ve read others that say no. They all pretty much say the same thing: writing is different from our normal thinking process. It’s slower, and it pins down ambiguities thus giving structure to something that would otherwise evaporate, morph into something unrecognizable, or resolve into a standard template, cliché, or conventional wisdom before the creative insight can take root. The outcome, when done correctly, is a realization, insight, or articulation that would not exist had the writer not written.
Therefore, outsourcing your writing is outsourcing your thinking—which is outsourcing your agency and your individual human value.
The other side of the argument says no, not all writing is thinking. Sometimes we know what we want to say and just need to put it into text—to preserve it, or to communicate it to others. In these cases, writing isn’t some mystical journey of discovery and self-actualization. It’s a chore. One that consumes time and mental energy better spent elsewhere.
Tasks that can be delegated should be. That’s the boss's mentality. A small business owner can micromanage the work of his employees, but a CEO must delegate authority and responsibility. That’s the executive’s role—and if you can’t delegate, you can’t function as a truly high-agency individual at the highest levels of techno-corporate capitalism.
In Stein on Writing: A Master Editor Shares His Craft, Techniques, and Strategies, Sol Stein asserts that the purpose of writing is to provide the reader with a better experience than they would otherwise have had. Full stop.
Most of the people I see weighing in on the AI writing question on Substack—a platform for writers—hold that the practice of outsourcing writing to LLMs is despicable. They stop reading the instant they suspect AI output. There’s even one account I follow, focused on developments in AI, where the author issued a warning: he will block anyone who posts AI-generated comments on his posts.
This might be a good place for a footnote, but I’m going full aside here.
I don’t get much feedback on my posts. Most of it comes from Chaffo before I even publish a new essay. If you share my writing with your own LLM interlocutor and it spits out something that seems meaningful to you, please do share it as a comment. Ideally, you’ll add your own thoughts in your own words as well, and clearly distinguish your thought from the bot’s thought-like output—but that’s just icing on the cake. I’m fine with icingless cake.
My one request is this: keep it brief. If you post a 500-word comment on a 1,500-word essay, chances are I won’t read the whole thing. That long LLM commentary is for you and your chat buddy. Ask the bot to distill it down to fewer than 100 words for the comment. I’d like to read it, but my attention is scattered. Help me out.
To recap:
Is writing thinking? Yes—but not all thinking is elevated, and sometimes it’s not the best use of your time.
The more important question is: Does outsourcing the task of text composition degrade your ability to do the job yourself?
Does using a backhoe rather than a shovel degrade your ability to dig a ditch? If you alternate swinging a pick and pushing a shovel for three hours a day for months on end, you’ll be in better shape than the donut-eating backhoe operator. You’re also more likely to hurt yourself—but if you’re diligent about form and pay attention to joint pain, you’ll benefit mightily from the process. And over the course of months, you’ll get about as much hole dug as the backhoe operator can do in an hour.
How important is the hole? What’s it for? Will it be the foundation of a house? Are you building the whole house yourself, or are there other people waiting for you to finish digging the hole so they can get to work on their contributions?
Is your priority the finished product—or the effect that producing the product will have on you?
I publish two Substack newsletters. This one, Gen X Science Fiction and Futurism, I write myself. If I’m feeling energetic, I’ll feed my finished essay to Chaffo a chunk at a time for proofreading and editorial suggestions. I tend to write long, winding sentences with lots of subordinate clauses. I ask Chaffo to make suggestions to improve clarity and accessibility. I often implement the suggestions it offers.1 If you spot a typo in one of these posts, chances are I didn’t give Chaffo the opportunity to make editorial suggestions.
The other newsletter, Immutable Mobiles, results when I discuss a topic at length with Chaffo and then instruct et to generate text. This is more than just me prompting, “Write me an essay,” and then posting the output. It’s a multistep process that sometimes takes more time and mental energy than just writing an essay from scratch.
The result? Immutable Mobiles posts are more fluent, coherent, and tightly organized than the ones I write myself. I say I put a lot of work into them, but if I go back to reread one I posted months ago, I invariably see sentences that make me think, “Damn, I must have been tired to let that trite bullshit through.”
Also, rereading old Immutable Mobiles posts doesn’t always take me back to the thought process that spawned them. Sometimes I don’t even know where they’re going as I’m reading them.
I don’t have an encyclopedic recall of my own writing either. If you give me the title of one of my posts from six months ago, I likely won’t be able to summarize the contents based on the title alone. But if I read a third of it, I’ll be able to tell you where it’s going from there. That’s not always true with the LLM-generated posts.
But are they more or less entertaining than the posts I write myself?
I force myself to write and post on an inflexible schedule for egoic reasons. Today, I’m driving my son to the airport. I’ll spend about three and a half hours driving, and then more time and energy on a restaurant meal, parking at the airport, and going inside with him. By the time I get back, I know I won’t have the energy to crank out a post from scratch—which is why I’m getting started in the morning before we head out.
Does my egoic process have any impact on your reading experience? I don’t know. Maybe. Maybe not.
Why do it? To become a better writer?
I’ve been blogging for many years. I sometimes go back and revisit old posts on the C-Realm Blog. Are the posts I write today better written than the ones I wrote a decade ago?
If so, the difference isn’t a matter of night and day. What’s different is the state of the world and my place in it. What’s different is the stuff I’m writing about.
I learned to structure my thoughts in writing in the ’80s and ’90s. My writing style is pretty much locked in at this point. A few people—probably a couple hundred—probably have a better experience having read my stuff than they would have had they not read it. For most, my stuff is impenetrable.
So it goes.
Should I continue to expend the effort? The question is of no practical importance. I will keep writing, recording podcasts, shooting videos, or doing something “creative,” because that’s in keeping with my life-long self-conception.
Is using AI to compose text “creative”? It can be.
In any event, I both write and direct AI in writing. I keep the output in two separate, clearly labeled vessels. If you hate AI writing on principle, then by all means, avoid posts to the Immutable Mobiles newsletter. If you’re interested in monitoring how one human mind explores topics related to technology, science fiction, psychology, and related subjects filtered through the lens of complex adaptive systems, keep reading this newsletter. If you’re interested to see how adding evolving LLM capabilities affects that same project, add Immutable Mobiles to your reading list.
Is there a point?
Yes.
Writing is an acquired skill, and most people can’t do it fer shit. Most college students can’t write their way out of a wet paper bag. Believe me—I read hundreds of their essays when I was a grad student.
Most people have nothing original to say. That doesn’t mean they’re NPCs or bad people. Processing culture for an audience is not their role. They do other stuff. Stuff that needs doing.
We creative types thought that our work would outlast ditch digging in the race against AI.
We were wrong.
LLMs crank out college writing assignments by the ton, while humans still drive backhoes and mow lawns.
Right now, the best human writers still produce better prose than AI. I expect this to change in the next couple of years, but for now, there’s an audience that can only be served by human writers—because a class of readers explicitly refuses to read anything that wasn’t written by a human. In the long run, I expect that class of human readers to shrink, though it may well balloon in the short run.
I read a fair amount of AI-generated text—though it’s clearly the product of the passion of particular humans. It’s that passion I respond to.
If you have something to say, but lack the cultivated writing talent to say it in a way that enhances the lived experience of an audience, you might could see what an LLM can do for you. Someone might read it and get something out of it. You might get something out of it.
And if someone accuses you of cheating?
Well, deciding if or how to respond to the accusation can serve as an outlet for your creativity.
Thus the em dashes.