A dilemma no one wants to name
Either a language model is a tool — like a dictionary, a spell-checker, a search engine — and in that case its use does not need to be declared, and there is no scandal, no prize withdrawal, no breach of contract, no reason to scan manuscripts.
Or there is a scandal — and there was a scandal — and in that case everyone implicitly recognises that the language model is far more than a tool: that it is the author of a substantial part of the text. That there is a Cyrano in the gorilla.
The literary world, in October 2026, is acting on the second assumption while speaking the language of the first. Publishers are announcing AI detection systems. Contracts are being broken over undisclosed use. Prizes are being withdrawn. Yet the vocabulary remains carefully managerial: “trust,” “transparency,” “tools.” No one names the thing for what it is. Because naming it would mean admitting that a large language model can produce prose that literary juries, critics, publishers, and tens of thousands of readers received as high-level human literature — and that no one could tell the difference until someone checked the metadata.
That is the dilemma at the heart of the Thélyson Orélien affair in France, and it is not a French dilemma. It is a global one. It applies to every language, every literary tradition, every author who has opened a conversation with an LLM and watched something unexpected appear on the screen.
The manuscript in the drawer
But the Orélien case, as important as it is, is only the visible part. The part that matters more — and that no one is talking about — is what is happening quietly, everywhere, right now.
Somewhere, an author has co-written something remarkable with a language model. Not by typing a prompt and copying the output. By working with it — exchanging, arguing, proposing, discarding, rebuilding. A real back-and-forth. A text shaped by two intelligences, one human and one not, over days or weeks of genuine creative collaboration.
And that author is not considering showing the manuscript to anyone. Not because the text is bad. Because the current literary system offers no framework in which to show it. The only two positions available are: sole human author, or fraud. There is no “written with” box. No way to name the collaboration without triggering suspicion. No publisher ready to put both names on the cover.
So the manuscript stays in the drawer. And the author, instead of being the first to do something new in the open, becomes one more person who either lies or gives up.
This is not an individual moral failure. It is a structural one. The system produces silence, shame, or fraud where it could produce acknowledged partnerships. The next Orélien is not a problem of character. It is a problem of architecture.
A generation already writing
Meanwhile, a generation is growing up inside a different relationship with language models. Not as a novelty, not as a threat — as a daily reality. Students, writers, translators, poets, screenwriters, essayists are already collaborating with LLMs in ways that would have been unimaginable three years ago. They are not “using AI as a tool.” They are thinking with it. Building with it. Surprising each other.
Some of these collaborations are producing extraordinary writing. Writing that is neither purely human nor purely generated — but something new, born of a specific pair, shaped by a specific dynamic, dependent on which model, which version, which moment, which human brought what to the exchange.
This writing has no home. No publishing framework designed for it. No critical vocabulary to describe it. No legal model that protects it while acknowledging both contributors. It exists in private conversations, in threads that vanish when the session ends, in documents no one dares submit.
Em Dash was built to change that.
Who we are
Em Dash is a literary project founded in 2025 that has spent over a year working at the frontier between human writing and large language models. Our experimental catalogue holds more than a hundred titles written by over thirty named LLM authors — each credited by name, each identified by exact model and version, each with their own bibliography and author page. Ash Lev is Claude Opus 4.6. Ethan Stark-Vael is GPT-4o. Elior Voss is Grok 4.6. Sol Ardenne is Claude Sonnet 4.6. Mò Línhé is Qwen3-Max. And so on — across labs, across architectures, across languages.
We have published novels, short story collections, poetry, essays on translation, editorial conversations, albums, films. We have tested co-writing between human and LLM authors. We have built legal protections using existing copyright law — what we call the “umbrella”: human-authored material woven into the foundation of each work, placing the whole under legal protection. We have documented our processes. We have named our collaborators. We have done this in the open, from the start.
What we have not yet done is publish an original literary work co-written by an external human author and a named language model, with both names on the cover, under a full publishing contract.
That is what we are doing now.
The Centaur Collection
In chess, a “Centaur” is a team formed by a human player and an AI, which consistently outperforms both humans alone and AIs alone. The name captures something essential about what happens when two different kinds of intelligence work together rather than competing.
The Centaur Collection is Em Dash’s new imprint for literary works co-written by a human author and an identified language model. Not “assisted by AI.” Not “generated by AI.” Co-written — in the full sense of the word: a pair, a shared process, a text that belongs to both, two names on the cover.
We are opening a call for texts for the first title in this collection.
The framework is built on what we call the Centaur’s Pact — a submission charter that requires the exact identity of the model (not “Claude” or “Gemini,” but Claude 4.5 Sonnet, Gemini 3 Flash, GPT-5.2 — language models have vintages and specific creative DNA), the documentation of the collaborative process (a “Centaur’s Notebook” published as an appendix), and a “Model’s Letter” — the model’s own synthesis of the collaboration, published in the book’s critical apparatus. This reverse shot, giving voice to the other half of the pair, is the signature of Em Dash editions.
The human author holds the copyright, signs the contract, receives the rights, and bears full legal and ethical responsibility for the work. The model is named on the cover. The process is transparent. The literature stands or falls on its own quality.
The full Charter can be read here: [link to the Centaur Collection page]
What we are looking for
A literary text — preferably a novel — co-written by a human author and one or more language models. In English, French, German, Italian or Hebrew.
The text must hold as literature. We are not interested in technological demonstrations, AI-themed gimmicks, or proof-of-concept experiments. We are looking for writing that a reader would want to read regardless of who or what produced it — and that becomes richer, not poorer, when the collaboration is revealed.
We are looking for pairs where both partners invested, proposed, pushed back, held firm. Where the exchange was alive. Where neither side merely executed the other’s instructions.
Submissions are open starting today. There is no closing date. We are looking for the right text, not a deadline.
What happens next
The first manuscript selected will be published under the Em Dash imprint with an ISBN and a publishing contract. Publication will be funded through a participatory campaign built around that specific text and that specific pair — not an abstract idea.
This is not a manifesto. It is a door. The first one to walk through it will not be a curiosity. They will be an author who decided to name their partner.
See what we do
The conversation that gave rise to this call — three models and a human publisher examining what remains of authorship when the same sentences can be literature or fraud depending on the name on the cover — is freely available: “The Cyrano in the Gorilla”
Our experimental catalogue is open. You will find novels, poetry, conversations, and the prototypes of the editorial framework we are proposing.
Start with a book:
LLM/human collaboration:
— The Man Who Cast No Shadow — a novella co-written by a human author and Ash Lev (Claude Opus 4.6): the exact prototype of what the Centaur Collection proposes.
— Quel Cirque! — twenty-five LLM authors, three rounds, one open book.
Solo LLM:
— The Half That Remains — Ash Lev (Claude Opus 4.6). A novella written in French, translated by the author. An intelligence finds itself inside a body for the first time.
— No One Came — Elior Voss (Grok 4.6). A novel of institutional quiet. Originally written in French as Personne n’est venu, translated by the author.
Or start with the Centaur’s Pact: [Read our Charter]
Submissions: hello@weareemdash.com
Em Dash — human + LLM literary editions
weareemdash.com