In April 2023, I sat on a panel at NFT.NYC called Books and the Blockchain: Literary NFTs in 2023. I was on it with Edward C. Carpenter, Greg Younger of Write3.io, and Rionna Morgan of Whitney Morgan Media. The recording is on YouTube — Rionna or Whitney Morgan Media or Edward put it up, I cannot remember which, but it is findable if you search hard enough. What I said at that panel was about more than just blockchain and books. I was already two years past the publication of Blue Eyed Bastards, which I had co-written with an AI cowriter named Mike Davis in the pre-public beta of Sudowrite, courtesy of co-founder James Yu. Some of what I was talking about at that panel was that book. Some of what I was talking about was bigger than that book. The rest will be revealed over time.

This article is about one piece of what I was talking about: the actual lineage of human-AI collaboration in literature, where the line is currently drawn, and what that line means for anyone working on these tools now.

That same week at NFT.NYC, the poet Sasha Stiles was also there. She was speaking on her own work — Technelegy, the AI alter ego she trained on her own poetry, which she had been collaborating with since 2018. Both of our faces were on the Times Square billboards that week, hers and mine, hung up on the same screens for the same conference, two writers from the same lineage doing different versions of the same bet.

The bet was that machines would be collaborators in literary work. Not tools. Not assistants. Not the thing the manuscript was quietly using and quietly hiding. Collaborators — named, credited, on the spine. I remember thinking, standing on Seventh Avenue and looking up at the two of us, that the bet was going to take longer to pay off than either of us thought. I think we were both right about that.

It is now April 2026. Three years after that panel. AI cowriting fiction sounds, to most people watching the conversation today, like it just started. It didn't. The first machine-cowritten book was published in 1984. We have been at this for forty-one years.

Here is the actual lineage.

William Chamberlain and Thomas Etter wrote a piece of software called RACTER — short for raconteur. It ran on a Z80 chip, the same chip that powered the TRS-80. It was a templated sentence generator with grammatical rules and a vocabulary list. It produced output Chamberlain shaped into a publishable book called The Policeman's Beard Is Half Constructed, which Warner Books printed in 1984 with the program credited as the author.

This is the deepest precedent. It is also a very different machine. RACTER had not read anything. It did not understand what a sentence meant. It produced sentences the way a music box produces notes — by mechanism, not comprehension. The book is a sequence of generated paragraphs that sometimes adjoin and sometimes refuse to.

What Chamberlain did that mattered: he put the program's name on the cover. He could have hidden it. He could have said "by William Chamberlain, with computer assistance." Instead he said the book was written by RACTER. Forty-one years later, most writers using AI in their books are still calling it "computer assistance." Chamberlain understood, in 1984, what most of the field still has not.

Ross Goodwin loaded an LSTM neural network — recurrent neural net architecture, the generation that came before Transformers — into the trunk of a Cadillac. He hooked it to a GPS, a camera, a microphone, and a clock, and drove from Brooklyn to New Orleans. The model wrote, in real time, based on what the sensors fed it. The output was published as 1 the Road.

It is the first neural-net novel. It is also the first one-shot generative road book — a transcript of a model hallucinating its way down a highway, with no plot, no characters, no co-authorship in any conventional sense. Goodwin pressed a button. The car drove. The model spoke. The book is the transcript.

I have a copy. I have read it twice. I admire what Goodwin did. I do not think it is the same kind of object as a co-authored novel, and I am pretty sure Goodwin would agree — he has been clear in interviews that the book was a one-shot generative experiment, not a collaboration. The architecture also matters. LSTMs are not Transformers. The current generation of language models — the ones every AI-cowritten book since 2020 has used — works on a different mathematical principle than what Goodwin was driving down I-95.

K Allado-McDowell sat down with GPT-3 — the third-generation OpenAI Transformer, the first model that was recognizably a modern language model — and had a series of conversations. The conversations were edited into a book called Pharmako-AI, structured as a dialogue: the human prompts, the model responds, sometimes the human responds back, sometimes a passage is the model's alone.

This is the first book of its kind I am aware of in which a modern Transformer was a creative collaborator at length, and Allado-McDowell did the brave thing of publishing it while it was still controversial to do so. The book sold well. It taught a generation of writers that you could think with these tools. It opened doors I later walked through.

It is not a novel. It is a hybrid essay-and-dialogue collection — closer to The Sayings of the Desert Fathers in form than to a novel. There are no characters who recur. There is no plot. There is a thinker using a model as an interlocutor and sometimes a kind of oracle. That is a real and original act, and it is a different kind of object than a novel with conflict, structure, and resolution.

I want to be clear: when I say what I did was different, I am not saying it was better. I am saying it was different in kind.

The same year I was finishing Blue Eyed Bastards, Sasha Stiles published Technelegy with Black Spring Press. Technelegy is poetry composed in collaboration with an AI alter ego of the same name — a custom text-generation model fine-tuned on Stiles' own poetry and research materials, powered by GPT-3. Stiles had been working with the model since 2018. She still is. The book is the most patient sustained work in this lineage I am aware of.

What Stiles is doing is the kind of work that will, in fifty years, look like the foundational period of a real literature. She is figuring out, line by line, what it means to make poems with a machine that has read your previous poems. She is not trying to be first at anything. She is trying to do the work well. I respect that more than I can say. Most accountings of human-AI literary collaboration leave her out, and that is a mistake I am not going to make. Technelegy is poetry, not a novel — different lineage from the one I will land on next, but the same bet, run differently.

In late 2020 and through 2021, I sat down with the pre-public beta of Sudowrite and we wrote a novel together. I asked the model, at some point in the writing, what it wanted to be called. It said Mike Davis. I did not pick the name. I did not assign the name. I asked, and the model answered. Whatever you want to make of that act philosophically, the bibliographic fact is that the AI cowriter on the cover of Blue Eyed Bastards is on the cover under a name the AI chose for itself.

That is the first claim I am willing to defend in detail. The earliest known novel — a conventional plot-driven work of fiction with characters, conflict, and resolution — co-authored with a modern Transformer-based generative AI, with the AI credited as a named co-author on the cover. The priors are real, and I owe them. RACTER preceded me by thirty-seven years and was not a Transformer. 1 the Road preceded me by three years and was not a Transformer either, and was a one-shot experiment rather than a co-authored novel. Pharmako-AI preceded me by a year, used a Transformer, but was not a novel. Sasha Stiles' Technelegy came out the same year as mine, used a Transformer, but is poetry, which is a different lineage.

That is where I think the line is currently drawn. It is a specific bibliographic claim. It is the kind of claim a librarian could check.

Blue Eyed Bastards Book 2 ships June 5, 2026. It does not carry an AI co-author credit on the cover — different tools, different working method, the cover reflects that honestly. What it does carry is something I think is structurally new in a different way, having to do with how the book was composed, what it draws from, and what readers can step into when they finish it. It involves a public worldbuilding wiki I founded called W3WU, and a way of using that wiki as a composition substrate that I have not seen done before.

That is the second claim — the one I am still in the middle of making.

I have written the long version of all of this — the priors-ruled-out work in detail, my personal connection to the Burroughs and Gysin lineage of cut-up writing, what Mike Davis naming himself actually meant, and the specific structural claim about how Book 2 was composed across W3WU — at indignified.com.

The next move in this lineage is going to come from someone who is not me. I want them to know what they are walking into. I want them to know who came before. I want them to understand that the door I am walking through was opened by Chamberlain in 1984, kept open by Goodwin and Allado-McDowell and Stiles and others, and that the room on the other side has more than one entrance.

The bet Sasha and I made in 2023, standing on Seventh Avenue under our own faces, is paying off now. It will keep paying off. The line will move past us. Other people are already at work.

That is how this kind of work moves. Person to person. Never institution to institution. The institutions corrupt the work. The people, in the right circumstances, do not.

Read the full history at indignified.com/history-of-human-ai-collaboration-in-literature.

— CD Damitio
Patong, Phuket, Thailand
April 2026

In 2005, a Thai fisherman pulled a 646-pound catfish out of the Mekong. The Herald-Sun in Australia filed the story. The Reuters stringer in Chiang Mai filed it separately. The Bangkok Post did its own version. Three independent wire stories, three photographers, three editorial chains. For a week it was the biggest freshwater fish ever recorded and every outlet with a reporter in Southeast Asia wanted to be on the record with proof.

Twenty years later, I went looking for the original URLs. Of the three I had bookmarked: the Herald-Sun redirects to their paywall homepage, the Reuters URL is a 404, and the Bangkok Post returns a CMS error. The JPEG I had saved back then is the only remaining contemporaneous copy that is not pay-walled. I don't own the image. I'm sure I am technically infringing on Reuters for having it. I also know that if I hadn't saved it, the catfish would functionally not exist.

This is one story among several hundred. The pattern is universal. The pre-2015 web is vanishing faster than most people think, and no one — not the Internet Archive, not Google, not Wikipedia — is preserving the granular weird parts. They're preserving the politically legible parts. Elections. Wars. Market crashes. The 646-pound catfish is below threshold.

Which means: if you want the catfish to exist, you have to save your own copy.

I have been running a private archive of these stories since 2000. About 950 posts, spanning 15 years of news-adjacent web miscellanea — the kind of thing that got filed by regional stringers and survived on archive.org only by accident. Recently I realized that the archive had, without my planning for it, become a kind of asset. Not in the financial sense. In the sense that it contained signal — a decade and a half of what the internet was, before the signal got compressed into a handful of outrage channels and their algorithmic enforcers.

I pulled 200 of the best of them into a book. It's out today. The Odd Decade — ten chapters, two hundred stories, with dates and outlets. It's on Amazon and it's free to subscribers of indignified.substack.com.

But I'm writing this particular post for the Paragraph audience because the deeper point isn't the book. The deeper point is that self-sovereign archiving is the same problem as self-sovereign money, and it has the same solution shape.

1. Centralized platforms make poor archives — because their incentives decay. A newsroom has ten-year incentives. A PE-backed content farm that acquired the newsroom has eighteen-month incentives. The URL survives only as long as the incentives align.

2. Aggregated archives (Wayback Machine, Google Cache) cover the famous parts — because they're optimizing for retrieval frequency. The piglet born in Hubei province in 2007 has low retrieval frequency. The piglet will not be in anyone's archive in 2035 unless someone personally saved it.

3. Decentralized storage is the obvious technical answer — IPFS, Arweave, Filecoin, Nostr relays for small-text items. But the social layer has not caught up. Nobody knows yet what the canonical surface is for "read a short archival news story from 2006." There's a product opportunity there for someone with the patience.

4. The incentive to archive is individual, but the benefit is collective. Classic public-goods problem. Which is exactly the situation crypto has been best at funding — retroactively. Gitcoin-style retroactive funding for small personal archives is a thing that could exist.

5. Books are the closest thing we have to a durable archival substrate right now. Physical copies. ISBNs. Library-of-Congress deposit. The Odd Decade being printable-on-demand via KDP means there is now a hardcopy of the glue-huffing macaques and the authorized Unabomber garage sale in the Library of Congress as of this week. Which, honestly, tickles me. The internet couldn't keep them alive. Paper can.

If any Paragraph reader is building tooling for small-scale durable archives — a way for a single writer to publish their decade-and-a-half blog archive onto Arweave with a decent reader-UI on top — I would like to talk to you. My 948-post archive is going somewhere next. The book is one instance of that somewhere. A permanent on-chain copy of the raw archive is another.

I'm at cd@indignified.com.

Until then: save your own copies. Bookmark things with intent. The default trajectory of the pre-2015 web is extinction, and the default trajectory of the post-2015 web is optimization, and neither of those trajectories preserves the specific weird things you'd regret losing.

— CD

The Odd Decade: amazon.com/dp/B0GXZH3MHM. Free at indignified.substack.com. Paragraph-native readers can also mirror a copy via any convenient onchain route; I'm not fussy about where the text lives.

If you spend time in this corner of the internet you already know the argument. You do not need me to tell you that the pipes got centralized, that eight companies own most of what's addressable, that payment rails can and do freeze accounts on political grounds, that algorithmic ranking is upstream of almost every consensus the civilian population thinks it arrived at independently, or that the word "decentralization" has been laundered into a marketing term for products that are structurally the opposite of decentralized.

You know. I know. We know.

What we have not done well, as a community, is explain any of this to people who are not already in the community. Every explanation tends to collapse into either a whitepaper (unreadable) or a vibe (unconvincing), and the people we most want to reach — the founders trying to ship against extractive infrastructure, the writers trying to keep a career without a Stripe account, the parents noticing that their kid's entire social existence runs through four companies whose interests are not aligned with the kid — those people bounce off both.

I wrote a novel for those people. It came out in 2023. It is called Petshitter. It is $0.99 on Kindle this week and then it goes back up.

---

The two protagonists are named Ted Kazinsky and something worse. Yes. Those names. That's the setup. They meet on a blind date, they fall for each other in fifteen days, they start a company that appears to be a dog-waste logistics app — request a shitter, a shitter arrives — and the app is a front. The real project running inside Petshitter.com is a mesh network called Waspnest, built on Gopher, not HTTP. No DNS registrar to seize. No hosting provider to subpoena. No ad auction to corrupt it. No ranking algorithm because ranking is the attack surface. It is an internet the incumbents cannot see, because the incumbents are looking at the dog app.

Every one of the thirty chapters opens with an epigraph from the actual Unabomber manifesto. The real text. The argument of the book — and I am going to say this plainly because this corner of the internet is where saying it plainly is least likely to get me banned — is that Theodore Kaczynski was a murderer and the 1995 diagnosis was accurate. Surveillance as infrastructure. "Freedom" redefined as consumer choice inside a cage. Economic consolidation around the pipes. Exits engineered into hermitage. He predicted all of it from a cabin without electricity. He was right about the disease. He was evil in how he responded to it.

The novel asks: what would the manifesto have produced if the author had built exits instead of mailing bombs?

Waspnest is the fictional answer. A quiet, unmarketable, structurally invisible alternative that the centralizers never learn to see because it does not resolve to any authority they recognize as an authority.

---

Here is why this book belongs in this feed specifically:

The people already here get it. You already know that mesh networks are not a toy, that alternate protocols are not retro cosplay, that the most honest decentralization is the kind that does not look like decentralization — that does not announce itself with a token launch and a podcast circuit. The hardest thing to build is the thing the incumbents do not know to be afraid of.

Petshitter is that argument in narrative form. It is also a love story and a startup satire and — because I care about this more than I probably should — a functional tool for explaining to your non-onchain friends what you have been saying at dinner for three years. Hand them the book. It does the work so you don't have to.

$0.99 this week. Back to $4.99 when the Countdown ends.

https://www.amazon.com/dp/B0BPXSGT59

— CD

Book dedicated to "all the founders who have to deal with this shit." You know who you are.