Writing after the homework apocalypse

AI Log reviews "More Than Words" by John Warner

I’m selling my pork chops, but I’m giving my gravy away.

—Memphis Minnie

Reading during the homework apocalypse

In the weeks following the launch of ChatGPT, I devoured every bit of news and analysis I could find about generative pre-trained transformers. I had not been paying much attention to AI, or what I still thought of as machine learning. As I worked my way through blogs and journals, I discovered a few people who had seen it coming.

Timothy Burke played around with GPT-3 for an hour or two in September 2021 and reported that “faculty are going to have to stop assigning writing that is used as proof of student learning. Soon.” A year later, Arvind Narayanan was prompted by an article in Vice titled “Students Are Using AI to Write Their Papers, Because Of Course They Are” to observe that “students around the world are using AI models such as GPT-3 to write essays, and getting good grades on them.” He concluded with,

These are deep and long-standing problems with our educational system. AI tools didn’t create them and banning their use by students won’t solve them.

Both Burke and Narayan pointed to Why They Can't Write: Killing the Five-Paragraph Essay and Other Necessities by John Warner as a critical context for understanding why ChatGPT was not the end of education but rather a chance to fix what our educational system has been getting wrong. On December 11, 2022, Warner himself published what became a touchstone for understanding the Homework Apocalypse.¹

In ChatGPT Can't Kill Anything Worth Preserving, Warner offered the best sentence I have read explaining why ChatGPT has had such an impact on education:

The reason the appearance of this tech is so shocking is because it forces us to confront what we value, rather than letting the status quo churn along unexamined.

ChatGPT shone a bright light on the ways writing instruction has become overly focused on the final product to be graded with little consideration of the process of learning to write. This problem has been getting worse without many people noticing. The panic over ChatGPT helps us see the difference between what we say the purpose of school is and what actually happens in our classrooms. It hasn’t helped that enthusiasts make a lot of gleeful noise about how this new technology, in the form of game-changing chatbot tutors, will save education, not destroy it through the magic of scaling.

Warner has been among those leading the critical response to this techno-daydream with its long history of failure. And we have desperately needed a response. The enthusiasm for AI disrupting education has drowned out genuine exploration of how generative AI might or might not help change instructional practices so that we can achieve what we say we value. Worse, it has distracted us from reform work much needed to bring our practices in line with what we value.

A worthy sequel

More Than Words: How to Think About Writing in the Age of AI is a worthy sequel to that 2022 essay, updating and extending Warner’s views on the limits of LLMs and their potential to prompt reform. I found myself embarrassed while reading it by how often I found a reference or an argument in the book that I had made recently. As he says, “Given the fact that it takes many months for a book like this to go from manuscript to on sale, the risk of the contents going stale before the book meets its first reader are real.”

There is nothing stale in More Than Words. Despite the fact that I started the book already convinced of many of its arguments, reading it helped me think clearly about what follows from seeing generative AI not “as a threat that will destroy things of value” but as “an opportunity to reconsider exactly what we value and why we value those things.” The thing we value, of course, is writing. The question is how ChatGPT changes writing’s value. If you agree with me and Warner, the answer is not at all, at least when it comes to education.

Writing creates value in different ways, but Warner emphasizes its value “as a process, an experience, and activity where the experience itself is meaningful for both the writer and the audience for that writing.” When ChatGPT's artificially generated text is submitted by a student for a grade, nothing about the educational value of writing itself changes. It is just that the student misses out on the value of the process. The value of an 8-year-old writing their first sentence, an 11-year-old writing their first paragraph, or an undergraduate writing an essay does not rest in the quality of the output.

The problem with the status quo is that schools and colleges have developed an obsessive focus on assessing the output and ignore the process. If the final product is all that matters then, yes, ChatGPT is a pale horse. Since it is now nearly impossible to determine whether digital text is the result of hours of human struggle or a few minutes of prompting a large language model, we are in the end times of outcomes assessment.

This focus on the output is what Warner wants to change. He sees the attention on ChatGPT as an opportunity to make the case for the value of writing as a process of turning thoughts into words on a page and sharing the results with other humans.

Writing for money

As a learning experience, Warner argues the value of writing “is intrinsic and external and has nothing to do with markets.” Like learning to apply numbers to problems as a way to understand and solve them, learning to craft words as a form of thinking has value far beyond simply being able to produce specific outputs. However, whether you’re a business analyst or a poet, markets matter when it comes to getting paid for your outputs.

In the chapter “On the Future of Writing for Money,” Warner reminds us that no matter how socially valuable, writing for a living has never been easy. “It is easy to romanticize a past that seems like it must have been better, but the struggle of writers to find a way to support their work has been a constant in American life.” The same is true of making it as an artist or musician. If you’re in it to make a buck, cultural work is a tough road.

Nothing about the self-publishing boom and digital social media has changed how unlikely and uncertain success is. Since the days of Charlotte Temple (1791), a few success stories have raised the hopes of aspiring writers. In a data point that should be placed on every single MFA application and summer writing workshop webpage, Warner reports that based on Author’s Guild data, “there is a greater percentage of writers who are earning $250,000 a year than there are who make a sustainable middle-class living from their writing.” And that percentage is lower than any writer wants to think about.

That doesn’t stop writers from aiming at bestsellerdom or prevent a small industry of semi-successful writers and small-time grifters from selling writing advice about how to make it big. It’s tough out there. Warner details his experiences with markets in writing, including when he was starting out. When no one but your teachers and friends has read your work, your first goal is to find an audience, even if that audience is editors and agents. That means giving away your words for free.

Warner says, “I wrote and published for free, or what we called exposure,” which paved “the way to writing for money.” As Memphis Minnie sang, I’m selling my pork chops, but I’m giving my gravy away.

The economic value of gravy

Much of what I write on AI Log is in the grain of Warner’s argument that ChatGPT is a opportunity to correct long-standing educational problems rather than something to panic about. You can read about my alternative approaches to traditional teaching while experimenting with AI in What is an LLM doing in my classroom? I will post the final essay of that series next week.

So, rather than expand on why I find his argument convincing, let me focus on one area where we may not agree: the economics of writing in the age of AI and disaggregated social media.

Generative AI is delivering a lot of text to audiences in the form of outputs, but so far, it doesn’t seem to be taking much work away from writers. The same goes for musicians, though Spotify seems to be producing cheap content for people who barely pay attention to what’s playing on their devices. Markets in cultural products are changing, and AI may be accelerating that change. If that’s true, then the homework apocalypse points the way.

The clearest example of the economic impact LLMs so far is in the contract-cheating market, especially the implosion of Chegg, which is an enterprise based on contract-cheating without contracts and at a much greater scale. Chegg is not a particularly sympathetic victim of disruption. The story behind contract cheating is more complicated. The new documentary Shadow Scholars tells the story of Kenyan cultural workers whose writing was the product that essay mills sold to American students in the days before ChatGPT. Those same Kenyan markets in cultural labor were key in developing generative AI models, as revealed by Billy Perrigo's reporting about the laborers who did the emotionally difficult work of classifying text describing “child sexual abuse, bestiality, murder, suicide, torture, self harm, and incest.”

The hype over what AI can do continues to mask just how much human labor goes into producing these models. As markets slowly discover the price consumers and corporations are willing to pay for these outputs, questions over who gets paid and how much are far from decided. The writers and artists waiting for the tangle of legal questions about intellectual property to be unsnarled by US courts are part of a complex global network of culture workers whose income is impacted by decisions made in Silicon Valley.

It seems unlikely that generative AI will replace most forms of cultural labor. A better bet is that, like cultural technologies that came before, new forms of paid labor emerge out of the efforts to produce culture using new technology. How much of that labor is done in the Global South and how much those workers get paid matters, even if such questions are rarely asked.

The value of already having value

It seems clear that the value of culture created by humans in the age of machine cultural production will persist. Everyone’s favorite example is Magnus Carlsen, a human superstar in a market where the best players have been artificial for decades. Chess has become a billion-dollar industry and is now among the most popular e-sports. Watching humans compete with each other in digital games remains far more marketable than the novelty of watching AI models beat humans or play each other. Even BattleBots is as much about the human builders as it is about fire-breathing, wreaking machines.

In the unlikely future world where LLMs can match the outputs of the best human writers, would anyone pay to read a novel authored by GPT-9? How many want to pay to watch a chess game played by AlphaZero? Whether it is esports or writing, you need to have a following to get paid. You have to get good and get noticed before you get paid. For writers, this means giving your gravy away to learn the craft and gather an audience will remain necessary.

One worry is that AI slop will spill over into the gravy boats, making it that much hard to get the attention of an audience. Warner tells several stories about the way AI-enabled misinformation, scams, and spam create problems for humans. My hope is that human networks of judgment and recommendations gain value over black-box algorithms that simply feed you the next thing using a predictive model. If your sources are other humans whose judgment you trust, avoiding AI slop doesn’t seem that hard, which brings me to a point of disagreement.

After a fascinating (to this aspiring writer) account of the economics of writing a regular column and a newsletter on Substack, Warner says that the disaggregated model “doesn’t appear any more sustainable” for making a living writing than what came before. That sounds right. But then Warner argues that from “an audience’s point of view, a disaggregated universe of writers,” like the one created by Substack and Wattpad, “makes it harder to find the next new voice you’re interested in hearing more from.” That, he says, is because most of the writers making it big on Substack were already big names. People like Margaret Atwood or Paul Krugman convert the audiences they created by selling their words in more traditional markets into subscriptions on Substack. Once this wave of legacy big names fades, book publishing will decline. The incentives will lead fewer writers to write books.

Maybe, but it is also possible that book publishing will change its model to compete. Or, the backlash against digital culture grows into a counterculture movement based on book culture and other longform writing. A boy can’t dream?

In any case, my own experience as a reader is that I spent money more last year on Substack, the same on books, and less on newspaper and magazine subscriptions. An “n of one,” I know, but it seems to me that Substack is a greater threat to legacy magazines than they are to legacy book publishing.

Paying vs. finding writers

Newspapers and magazines did little to satisfy my sudden desire in November of 2022 to learn as much as I could about what thoughtful people were saying about AI in education. As newspaper reporters interviewed ChatGPT and The Atlantic chased the moral panic with headlines about the end of high-school English and the death of the college essay, I served as my own aggregator using online platforms instead of relying on magazine and newspaper editors. I found Arvind Narayanan and Sayash Kapoor, and Warner and Mollick on Substack long before I was able to read and review their books.

My first paid subscription to a newsletter on Substack came from clicking on a link in one of Mollick’s LinkedIn posts, which described a historian named Ben Breen writing about LLMs. Recommendations from fellow humans led me to discover Timothy Burke and Max Read, who I buy a beer for each month. It was through Substack that I discovered Audrey Watters had returned to writing about edtech, and I subscribed even though her newsletter is on a different platform. Prior to finding Warner’s landmark essay from 2022 on Substack, I hadn’t realized the op-ed writer whose pieces I liked in Inside Higher Ed was the author of that book about killing the five-paragraph essay.

All that is to say that my experience runs counter to the idea that disaggregation makes it harder to find new voices. When it comes to finding good writers, there is nothing better than following the recommendations of humans whose writing you think is good. One reason I like Substack is that more of that happens around here than on other platforms.²

Serving as the aggregator of my own cultural consumption is not all that different from earlier forms of cultural sharing and community building. Reading about a band from Georgia called R.E.M. in The Rolling Stone, as I did long ago, is not so different from reading about QuestLove’s documentary about Sly Stone on The Bibilioracle Recommends, as I did last month. Checking the back cover of a novel I like for who blurbed it is not so different from checking a Substack homepage to see who a writer recommends.

The self-guided aggregation that is my own blog helps me connect with fellow travelers on the road to turning the freakout over ChatGPT into an opportunity to change education for the better. It helps me find and connect with writers working on what Daron Acemoglu calls “an anti-AGI, pro-human agenda for AI.” In the penultimate section of the book, Warner calls this “finding your guides.” I found everyone he lists as a guide this past year through their writing on disaggregated platforms.

A book is seldom a pork chop

I recently added “writer” to the employment section of my LinkedIn profile, giving it first place in the list of things I do these days. I did so without the expectation, or even hope, that writing would ever pay for one of those beers or cups of coffee. That's because the most likely outcome of my turning on a paywall is I get discouraged by how few people want to buy me a drink. Even the best-case scenario is not a good one. Warner describes it more realistically than most Substacks about Substacking do. He says authors who attract a paying audience “put themselves on a treadmill of constant production and audience service in order to realize income from their work.”

Substack and LinkedIn, the two places I publish, already do everything they can to put me on that treadmill for the meager reward of showing me charts where numbers go up. It is a constant battle to ignore all that “encouragement to grow your audience” to focus on writing.

My plan is to use Substack as a gravy pump station. I want to find as many readers as I can in the hopes that they will buy and help promote my book, should I have the good fortune to finish and publish one. Even then, I don’t think of my book as a pork chop. It’s a lottery ticket with most of the payouts consisting of getting some of your money back. So, I don’t spend much time thinking about the jackpot. Instead, I think of writing a book as generating gravy, just another way to advertise my pork chops.

I suppose, like vinyl records and T-shirts, books might become merch I can sell on the side of my public speaking and consulting engagements. I like this analogy because it hearkens back to my days of hanging out near some of the musicians who lived in Athens as part of what was called the college rock scene. There are plenty of lessons to be learned from the history of the music industry of that era—for that matter, every other era since the nineteenth century—regarding how artists make a living in and around markets structured by corporate capitalism.

There were impassioned fights in Athens in the late 1980s over the question of R.E.M. selling out and plenty of examples of corporations exploiting artists hoping to do the same. Among the lessons I took from that place and time was that the value of communities organized around producing culture is not measured by how much artists make by selling commodities. The smart move as an aspiring college rock musician was to have a day job flexible enough to play live gigs when the opportunity came along. A record release for your band, should you be so lucky, was your chance to go on tour where your cut at the gate and merchandise sales could put enough money in your pocket to pay for groceries next year, or maybe just next month.

Spotify and Bandcamp haven’t changed the dynamic all that much, but they have continued a structure where musicians mostly give their gravy away, and corporations keep the vast majority of the proceeds from any pork chops sold. The most important question in this relationship is, “What’s your cut?” It’s the same as it is for writers on Substack or photographers and illustrators on Unsplash. Again, this describes not some new unfortunate development but the way it has been since mass markets in music, art, and words developed two hundred years ago.

Gravy as a social good

The economic value of exposure to an artist is tied to its potential to land some paying work. Its social value depends a great deal on what sort of community the artist is looking to build. Few musicians in Athens during the 80s and 90s were reading Frankfurt School social theory like I was, but they lived the problems of making culture in an age of mass production. An anti-corporate or anti-establishment ethos can create compelling art, but artists with that frame have to assume its value won’t be rewarded by markets, even if it finds an audience. When the markets notice the crowds and come calling with record deals, it gets confusing for everyone involved. Modern cultural labor is a process of responding to and pushing against the values represented in existing culture, but it is also a process of adapting to structural and technological change.

Changes in the technology of cultural distribution, just like the changes in the technology of cultural production, are “an opportunity to reconsider exactly what we value and why we value those things.” In the 1980s, new distribution through college radio, small magazines, and small clubs in college towns were the means artists created audiences for music that didn’t fit the model corporations had set up to sell an increasingly banal version of rock and roll. New ideas about music and new channels of distribution led to large amounts of cultural labor that paid little to most, was experienced by many people, and was wildly successful for a handful of artists.

I like to think that over the past two years, a similar anti-corporate project has formed around educational technology. It’s ethos is directly opposed to Silicon Valley’s vision of educational technology as personalized chatbots and digital teaching assistants and its belief that automation at scale is the solution to every educational problem. This process is dialectical if you want to talk about it in terms of social theory. I’ll leave that to others. My thing is writing about it as a lived experience.

In his conclusion, Warner argues for the importance of living in a shared ecosystem of human-created communities. He writes, “I find this to be a superior way to spend my days, while my time on the questions that interest me, rather than outsourcing my mind to something that cannot think, feel, reason, or have experiences. I am more than my economic outputs.”

Given the odds of success in the marketplace, this is important advice to anyone trying to make a living selling their cultural outputs. As a writer who gives it away for free, I would add that I am—we internet writers are—more than our social media platform data. The obsession with making our numbers go up, an obsession Substack and every other social media platform cultivate with every payment processed and every new feature added, is designed to get you back on that treadmill.

That’s why I won’t be turning on my paywall. I want to avoid the grind Warner warns about in order to develop connections, some of which may lead to paying gigs. The best thing to come out of all my reading and writing on the internet since 2022 is the chance to meet people and develop shared projects. Too much of it happens on Zoom, just as too much of our shared labor in general happens on Zoom these days. That’s why my rate card as a consultant is lower for in-person engagements than for remote work. That’s why I charge less for an in-person speaking engagement than for doing an online webinar.

I adjust the price of my pork chops this way because the value I get from human created community is much greater when it happens in person. This is not the same as saying communities based in digital experience have less value. I know too many people who tell me that digital communities on the internet saw them through periods of isolation and alienation. Still, there is value in sharing physical space with other humans in an increasingly digital world.

My hope is that as “we reconsider exactly what we value and why we value those things,” colleges and college towns become places where producing things of value takes place in a human-centered community. They are entering hard times as measured by their economic outputs. College towns, like Athens, are the loci of a shared ecosystem that could develop an educational movement to counter what corporations want to do with generative AI and digital social media while also grounding a political movement to counter the dangerous turn the US has taken.

Is that too utopian? Definitely. It is too simple, too backwards looking. Too much a story of my own lost youth. Yet, I take a great deal of satisfaction in noting that despite the long history of writers not getting paid much for their economic outputs, writing retains its value, even and especially in the age of AI. That value, as Warner says, is based in activities “where the experience itself is meaningful for both the writer and the audience for that writing.”

Turning such activities into a social movement feels like work that could happen most readily where markets have located people who care about writing, that is, places where writers and teachers of writing have been employed. You might say the same for artists and musicians wherever they are working together. Where, exactly, this happens is far less important than that it happens.

AI Log, LLC. ©2025 All rights reserved.

1. This is Ethan Mollick’s term for the collective freakout about what a generative pre-trained transformer hooked up to a chat interface means for education. The important point that Mollick and many others make is that cheating is nothing new. ChatGPT just made it much more visible.

2. I try to stay out of Substacking about Substack on Substack game, but I don’t want to be misread as a cheerleader. As far back as Matthew Carey negotiating with Susanna Rowson over her cut for the US publication of Charlotte Temple, the important question has been how much an author receives for turning over their valuable words to be distributed to a paying audience.

Warner describes the economics of Substack accurately as yet another attempt by Silicon Valley to capture a large audience by giving away something of value for free or reduced cost. This lose-money-to-make-money strategy means that at some point the company will need to turn that audience into profit by generating as much revenue as possible through ads and squeezing workers. For the game to pay off its investors, the company goes public. Index funds and less savvy investors are left holding the steadily emptying bag as audiences move on to the next thing.

Writers understand this game, which is why Substack set it up so that writers keep the emails of their subscribers and can use their own domain names. It makes it easier to decamp when Substack comes to the same conclusion printers did during the nineteenth century: they can only make this work if they keep most of the revenue.

What will writers do when Substack increases their take and starts charging the free-riders like me? We’ll see. I’ll go looking for the cheapest, ad-free platform and hope that email and the web survive. If Substack turns out to be a genuinely new model where authors keep more than half the revenue, then maybe I start selling some pork chops around here.