The Work of AI Writing on Substack
"Don't you even like typing on the mechanical keyboard, OP?"
“He wishes he had never entered the funhouse. But he has. Then he wishes he were dead. But he's not. Therefore he will construct funhouses for others and be their secret operator — though he would rather be among the lovers for whom funhouses are designed.”
― John Barth, Lost in the Funhouse
The AI-assisted Substacker wakes up at 11 a.m. He checks his subscriber count before he brushes his teeth. Three new paid subscriptions overnight. That makes 550 total at five dollars a month. He does the math while the coffee brews. $2,750 monthly before Stripe takes its cut.
He opens his laptop at the kitchen table. The apartment is a one-bedroom in a building where the elevator breaks twice a month. His last online girlfriend hit the virtual road six months ago. She said he spent more time with ChatGPT than with her. She was probably right.
The Substacker has three newsletters running. One about grindset productivity, one about crypto investing, one about becoming an alpha patriarch. He knows nothing about any of these subjects. He worked at a marketing agency for two years before they laid him off. Before that he sold insurance over the phone. His LinkedIn still says "Thought Leader."
He feeds prompts into the AI. "Write 1500 words about why cold plunges at 3 a.m. increase testosterone and Bitcoin gains." The machine spits out paragraphs about ice baths and market cycles. He changes a few words, adds some rocket emojis. The whole process takes twenty minutes.1
His readers don't know. Or maybe they do and don't care. A small percentage of them pay five dollars for the feeling that they're one hack away from greatness. The Substacker gives them that feeling three times a week.2
He has a system. Monday is grindset: wake up at 4 a.m., huff rancid liver inside a paper sack, work eighteen hours high on an intermittent fast. Wednesday is crypto: why $Trump will hit $100,000, how NFTs are the realest of real estate. Friday is relationships: why modern girlbosses swallow SSRIs the way BadlandsChugs downs gravy, how to have eight kids and a submissive tradwife, the evolutionary psychology of sperm wars-powered cuckoldry.3 The topics never change but many of the words do.
Every month he writes a special post about authenticity. He's not like those other gurus, he claims. He does the real work, lives the lifestyle. His 2 a.m. ice baths, his million-dollar crypto portfolio, his pregnant wife who doesn't exist. The subscribers eat it up. Twenty new signups every time. The AI helped him write that post too.
The worst part is the comments. Young men thank him for changing their lives. They share how they're sleeping on the floor now, eating only meat, investing their last thousand in shitcoins. A kid in Arkansas says he dropped out of college to day-trade crypto full time. These messages occasionally make him uncomfortable. He never responds.
Other Substackers know the game. They recognize each other's work, the telltale rhythms of machine-generated prose. They subscribe to each other's newsletters in a circle of mutual support. You scratch my back, I'll scratch yours. The money goes round and round.
He tried writing one piece himself last month about morning routines. It took four hours. The sentences felt clumsy, the ideas wouldn't flow. When he published it, the open rate dropped fifteen percent. The unsubscribes doubled. He went back to the AI the next day.
The Substacker tells himself he's delivering value. Men need guidance. They need someone to tell them to take cold showers and buy Ethereum. What difference does it make who wrote it? Or what wrote it? The information is the same.
But he knows the difference. Real writing has fingerprints. It has the writer's breathing in the punctuation, their personality in the word choice. His newsletters have nothing. They're corpses dressed in gym shorts and Bitcoin logos.
He sees the real writers sometimes. They publish long essays with footnotes, links to obscure books, original research. They interview people, dig through archives, spend weeks on a single piece. Their subscriber counts are lower but their readers stick around. They build communities instead of customer lists. Perhaps they use AI to organize and edit, but their esoteric value propositions (for example, sumo wrestling) are too strange and too marginal for it to be of much help from a composition standpoint. The Substacker envies them but not enough to change.
The Work of the AI Author
They say a new contraption can change the world overnight. Once upon a time, folks talked about the radio in that excited tone, and then they gave the television and the Atari 2600 that same breathless, confused tribute. Now the talk is about AI, more specifically the content it…
The truth is it's not about real or fake. A good AI with good inputs will beat a good AI with bad inputs every time. The real writers feed their work with substance: court documents, interview transcripts, books nobody else has read. The Substacker feeds his AI with other AI articles about mewing and monkey JPEGs. Garbage in, garbage out. The machine can only recycle what it's given. It’s not a miracle worker. It’s not even Patty Duke as Helen Keller in 1962’s The Miracle Worker.
Do the math. First it was 200 subscribers and $1000 a month. Then he hit 500 subscribers and $2500. Now he’s chasing 10,000 "bratty subs" so he can quit whatever day job he still pretends to look for. The math is simple. The execution is mechanical.
He has a template for growth. Share excerpts on X and elsewhere with matrix references. Comment on bigger newsletters about escaping the plantation. Use keywords that trend: pepe, based, WAGMI, blowing clouds, live the lifestyle, LFG, OMG, OGC, GGG.4 The AI helps with all of it. It writes his tweets, his comments, his subject lines. He's automated himself out of his own existence.
His about page is a masterpiece of deception. "I'm not here to give you recycled advice," it says. "Some people will do that, but I lived this. Lost everything in crypto. Built it back. Now have three kids and a trad wife on our earthship homestead." The prose sounds tinny, like a robot trying to convince you it's human. Which is exactly what it is. But he adds personal anecdotes the AI invented, references to his journey from beta to sigma.
His most popular post is about dopamine fasting while parlay betting cricket and snooker. Eight thousand views, sixty new subscribers. The AI wrote it in twelve minutes. It included quotes from LindyMan (itself plagiarized because that’s Lindy)5, Naval Ravikant, Andrew Huberman, Lex Fridman, anonymous Wolf of Wall Street types. All of it fabricated or stolen from Reddit. The readers loved it.
He meets another Substacker downtown at the raw butter stand. They talk about conversion rates and which herbal testosterone supplements to pretend to take, which steroids to actually use. Neither mentions the AI but both know. It's like two counterfeiters discussing paper quality. The conversation is professional, technical, empty.
The real tell is the consistency. Human writers have off days, weak pieces, experiments that fail.6 The Substacker publishes the same quality every time. Not good, not bad, just adequate. The McDonald's of newsletters, but for guys who think they're too smart for McDonald's. You know what you're getting.
He justifies it different ways. Everyone uses tools. Spell check, grammar check, now AI. Where's the line? He's still choosing the topics, editing the output, hitting publish. He's the curator if not the creator. The museum guide, not the artist.
But late at night, after three beers, he admits the truth. He's running a con. Not illegal, barely unethical, but still a con. He sells words he didn't write to readers who barely read. In truth, he’s not even picking the content. Some SEO tools do that, and then another AI writes the prompts he uses in his base system.
The girlfriend was right to leave. He spends ten hours a day managing fake newsletters about being an alpha male, a sigma male, a veteran of the sperm wars.7 He preaches raw meat diets while ordering $50 pad Thai food from UberEats. He writes about optimal breeding strategies while busting nuts to a steady stream of AI-generated porn.
Other hustles at least require skill. The three-card monte dealer needs fast hands. The pool shark needs to play. Even the circa-2003 pick-up artist actually has to talk to women. The Substacker needs only a laptop and a subscription to OpenAI.
He thinks about quitting sometimes. Getting a real job, writing real things. But the subscribers keep coming, slowly but surely. The payments keep processing. The AI keeps improving. Why fight it?
The future is already here. Newsletters written by machines for readers who might be machines too. Bots subscribing to bots in an endless circle of artificial engagement. The Substacker in the middle, taking his cut.
He knows real writers. They post screenshots of their messy drafts, their crossed-out sentences, their struggles with structure. They share their research notes, their interview transcripts, their detailed prompts for coaxing specific, not-at-all-AI-seeming actions out of the AI. The process is part of the product. The Substacker has no process to share except prompt engineering.
Every few weeks he publishes a manifesto. "Why I Write," "The Death of Real Masculinity," "Authenticity in the Age of Soy." He rails against fake gurus while being one. He calls out other newsletters for being psyops while his own is pure manipulation. Some readers applaud the honesty. The AI wrote those pieces too, prompted to sound indignant about its own existence. The AI even wrote some of the earliest positive comments on the articles, helping to seed all the others.
At the end of the day, the money isn't all that good. After taxes and expenses, he makes less than he did at the marketing agency. Less than he did at the cell phone kiosk in the Ross Park Mall. But it feels different. Entrepreneurial. Independent. He's his own boss even if the AI does all the work. He tells himself he's building something.
He subscribes to one real newsletter. A semi-retired journalist who writes about wrestling, posts old fight footage, interviews forgotten champions. 1,500 words every Sunday, probably takes him all week. The writing has texture, personality, insider knowledge. It costs seven dollars a month. The Substacker pays it gladly.
That's the joke. Deep down, like so many others, he probably knows good writing when he sees it. He can still tell the difference between real analysis and reheated takes. He just can't produce it himself. Or won't. The distinction blurs after a while.
The wrestling writer includes PDFs of old programs, links to rare books, footnotes that lead to footnotes. He visits archives, tracks down retired wrestlers, watches grainy footage frame by frame. The work shows in every sentence. The Substacker's work shows nothing because there is no work, just LLM-regurgitated content about nofap and market manipulation.
His analytics dashboard is his real homepage. Green arrows pointing up. Subscriber growth, revenue growth, engagement growth. The numbers tell him he's succeeding. The empty apartment with the gaming chair and protein powder tells a different story.
He tried dating apps to help fill that void but eventually gave up. How do you explain what you do? Newsletter publisher sounds halfway impressive until they ask about it. He can't say he writes about jawzercising and $megmacoin. He usually mumbles something about marketing and changes the subject.
The other Substackers are his only community. They share tips in channels like #BigMoney and #TopSeller, celebrate arbitrary (but not capricious!) milestones, complain about algorithm changes. None of them talk about what they write, only how it performs. The content is beside the point.
He's started believing his own pitch. Maybe he is different. Maybe his AI writing is better than their AI writing. He’s the GPT-9 poster boy while they’re still stuck on GPT-4. The delusion fits like an old shoe.
The game will end eventually. Readers will wise up or better cons will emerge. The Substacker will try to adapt. He has pivoted before. Find the next platform, the next tool, the next angle. The hustle continues.
For now, he schedules tomorrow's post. "Why I Sold Everything for PrepuceCoin and Started Eating Fermented Raw Beef." He's never owned crypto. He had week-old Chipotle for lunch; it might have gone bad, but it wasn’t raw. That's enough for $2,750, minus Stripe's cut. It's not honest work, but it's work.
Tomorrow he'll again wake up at 11 a.m. Check the subscriber count. Feed the machine. Collect the payments. Pretend it matters. The American dream, automated and empty, delivered three times a week to inboxes of lost young boys and girls who delete it, largely unread, after they experience that first quick dopamine hit of skimming someone else's fake success.
Look, these days everyone's using AI to paper over the gaps. You can spot it right away. The sentences all have the same rhythm. The vocabulary's just a little too consistent, at times repetitive. But the real tell is the formatting. Nearly every section has a colon in the heading. Bullets everywhere because ChatGPT in particular loves bullets. Random words in bold like they're trying to emphasize something important. Italics for key concepts.
Here's some “work” I ginned up on Claude:
The Modern Man’s Dilemma: Reclaiming Purpose in the Digital Age
In today's fast-paced world, men face unprecedented challenges:
Loss of traditional masculine rolesDigital distractions destroying focusDecline in testosterone levels (THIS IS CRUCIAL)Modern women don't appreciaterealmasculinity
The Solution: A Return to First Principles
It’s essential to understand that our ancestors had it figured out...
You see that formatting, you know nobody human wrote it. Real writers don't write like they're making a PowerPoint presentation. But these guys feed their Twitter thoughts into ChatGPT, and out comes this corporate workshop garbage. They don't even know enough to strip out the formatting, make it look like actual prose.
Want to read the Real Work? Subscribe to any/all of the Substacks I recommend.
Read that link! Wild stuff, the kind I love to see.
As far as OGC goes, arrange the letters vertically. What does that look like? Someone gooning? Exactly. As for GGG, old heads know…
The LindyMan situation was wonderful. Here's this Twitter personality (we called it Twitter back then) who is getting thousands of likes for his takes on modern life and traditional values. Builds up enough of a following to get profiled in The (Failing) New York Times and launch a Substack. The paid subscriptions start rolling in like tides on the seashore. He's making real money now, the kind you have to report to the IRS after a certain threshold. This was manna from heaven in that benighted age before Elon monetized X for the King Poasters.
Then someone notices his posts sound familiar. Not just the ideas (everybody's recycling ideas) but the actual sentences. Turns out LindyMan was lifting entire paragraphs from old magazine articles, obscure blog posts, wherever he could find them. The discovery came when a reader recognized a passage from a 2015 Atlantic article about social atomization. Once people started looking, they found more. Chunks from First Things, old American Conservative pieces, even some Substack posts from smaller writers who must have been especially galled to see their work stolen by someone with a bigger platform.
When he got caught, he tried to blame it on his "research assistant." He didn't have a research assistant. Then he went quiet for a day before coming back with the most lindy move possible: Never apologize, never explain, just keep posting and block anyone who gives you any lips. Someone documented the whole thing, tracking down every stolen paragraph, but LindyMan was still trucking back when I last checked in 2024.
Pure plagiarism followed by denial or no comment whatsoever? I say again: as lindy as it gets. Ask for forgiveness, or don't, but never, ever ask for permission. The old ways are the best ways, especially when you're stealing other people's words about the old ways.
Substack banned him, but not before he'd collected a few months of their do re mi. Nowadays, AI does plagiarism at scale (it’s built on something, after all — your somethings) and Substack couldn’t possibly police that, but plagiarism, even when I was a professor, has never been anything I cared about because everybody, even Harvard presidents, plagiarizes. The suckers among us, “ya boi here” included, waste time trying to make our borrowings look so fresh and so clean.
Maybe this is one of them!
Are these different things or the same thing? Who the heck knows. Ask Claude Opus, man.
I found it so interesting to be linked here as the antithesis to this. I was reading the piece and thinking about my own Substack experience and how foreign all this sounded.
As always, very interesting. The breadth of your output remains astonishing!
Love it! One can almost see the point where the AI is fed by other AI writing, like a snake eating its own tail. Similar to what’s already underway in AI generated “art” as well