Open-Write is the open-source release of a multi-agent AI writing system that produced four seasons of prestige television, three forms of a feature screenplay, and a 52,000-word literary novel — under one author, in a matter of weeks, for roughly the cost of a paperback.
Reading the work is live now · the repository releases June 1, 2026
Open-Write was built by producing real work, not by describing a theoretical capability. Three completed projects are released alongside the system as demonstrations of what it can do. A Thousand Silences is available in full now; excerpts of the others are posted ahead of the June 1 release.
A novel of the Haitian Revolution, following a French translator who slowly recognizes that the precision of his work is what makes him complicit. Produced from a single prompt — a setting, a historical source, and a literary inspiration — with iterative pipeline tuning and zero editorial intervention.
Available in full, free, today.
Read all 30 chapters →A hard-science metaphysical drama about a grief-frozen physicist contacted by the quantum-realm successors of all advanced life. Three completed forms — feature screenplay, novel, and an expanded screenplay covering the full Velai civilization in a parallel storyline.
RECOMMEND from the system's adversarial reader.
Read an excerpt →A four-season historical drama about John Brown and the Bleeding Kansas period. Spec scripts for all four seasons; three drafted in full, the fourth in development. A tragedy in the older sense, about the cost of conviction and the question of unpaid debt.
In development for studio submission.
Read the Season 2 cold open →The work and the argument arrive first; the system follows. The repository goes public on June 1 under Apache 2.0. The three remaining essays land on a fixed schedule across the month.
The full public production pipeline — three templates (screenplay, novel, episodic television), the Python toolchain, the state server, and the methodology documentation — released under Apache 2.0.
Why the danger is structural rather than individual, and why the right unit of analysis is the incentive structure, not the people inside it.
The Structural Solution — the best mechanism for changing what the structure rewards — followed by the closing coda, Designing for the Creature We Actually Are.
Open-Write is built around the recognition that long-form coherence is a scaffolding problem, not a model problem. Standard pretrained models, organized into specialized roles and supported by structured state, can carry the full arc of a season, a novel, or a screenplay without losing track of itself.
A month ago I finished reading a novel I had not written. I had asked a system of AI agents — coordinated through a workflow I built — to produce a novel from three pieces of creative direction: use the Haitian Revolution as the setting, draw on Laurent Dubois's Avengers of the New World as a source, take Hamlet as an inspiration rather than a template. The system did the rest.
That reading changed what I thought was coming for the creative industry. The common knowledge was that long-form AI content generation was hampered by compounding errors, that current models could produce decent short passages but lost coherence over long arcs. It turns out the problem was not the model. It was the scaffolding. The right file support system, the right division of labor across specialized agents, the right prompt engineering — standard pretrained models can carry a novel.
I had no agency over whether tools like this would be deployed. I did have agency over how mine would be deployed — toward broad distribution or narrow capture, toward giving the capability to individual artists or concentrating it in the hands of whoever could pay the most for it.
I have chosen to open-source the system. Open-Write is the public release: the full production pipeline, the templates, the tools, the methodology. I am keeping a small set of proprietary additions for the version I use internally. Releasing those would lower the floor — anyone with minimal commitment and limited craft could produce work that passes superficial review, and the floor falling that far would hurt the industry more than it would help any individual user. Holding those additions back means the public release rewards the people putting real time and real judgment into using it. The foundation is given freely under Apache 2.0. The advantage stays with skill and commitment.
This will cost people jobs. I will not pretend otherwise. I have thought about it honestly and arrived at a position I can defend: the technology is coming whether I build my version or not; what I can do is make sure one version is in everyone's hands, not just in the hands of those who can pay for it.
If the show I have spent ten years developing is ever produced, a portion of its proceeds will go to charity organizations in four counties in Missouri — Jackson, Cass, Bates, and Vernon — the communities depopulated by General Order No. 11 in 1863. The descendants of pro-slavery Border Ruffians paid a generational price for their ancestors' choices, and the national culture has either forgotten them or written them off. The show takes the position that history's victims include people the audience has been taught to dislike. The charity arrangement makes that position structural rather than rhetorical. I commit to it in public, before any deal has been signed, so that the commitment cannot be quietly walked back later.
Open-Write is yours. You do not need my permission. You do not need to know me. The system is given to anyone willing to put the work in. What you make of it is up to you.
A Thousand Silences is online in full and in the public domain. Read it cold, the way the system's adversarial readers do, and decide for yourself how close it lands to the real thing.
Read the novel →The repository — three templates, the Python toolchain, the MCP state server, and the methodology documentation — goes public June 1 under Apache 2.0. Star or watch it now to be there at release.
Watch the repository →The essays, case study materials, and high-resolution artifacts are available for press use. Advance access ahead of the June 1 repository release is available to credentialed outlets.
Press inquiries →