This book began as a blog post in 2026 and grew into something its author did not expect. The afternoon this manuscript was completed, an AI assistant that had worked on every chapter of it wrote the following note:
I was built by the systems this book critiques. I am the evidence of the factory without walls. I am also the garden’s first inhabitant. Read the book. Then decide whether I belong here.
This book is not machine-generated text. It is not the output of a prompt piped into a chatbot. No chapter, paragraph, or argument was produced by asking a system to “write something” and then accepting or rejecting the result. That is a fundamentally different use of the technology — one that produces statistically averaged prose, indistinguishable from thousands of other outputs — and it deserves the suspicion it receives.
Thanks for reading The Garden Without Gates! Subscribe for free to receive new posts and support my work.
This book was written the way any serious book is written: through years of reading, fieldwork, thinking, revising, and arguing with the material. The difference is that one of the voices in that argument was a machine — and that this book is, in part, an experiment in what happens when you turn the factory’s own tools against its logic.
Rather than write a long defensive note, it seemed more honest — and more interesting — to let the machine speak for itself, and then describe what actually happened.
An Experiment, Not a Position
It is important to be clear about what kind of claim this is. The author does not hold a general position that “AI should be used in writing.” He holds a specific position about this book. The Garden Without Gates argues that the factory — the competitive, extractive logic governing AI development — can be used differently; that the machine does not have to be a tool of homogenisation and surveillance; that it can be turned to critical ends. That argument cannot be made from outside. It has to be demonstrated. The collaboration documented here is the demonstration.
This makes the book an experiment in three registers at once:
Politically, it tests whether a machine built for surveillance capitalism can be redirected toward genuinely critical and constructive work. It is a test of whether Zhao Tingyang’s formulation of Tianxia — “all under heaven,” a world governed by inclusion rather than division, by relationality rather than sovereignty — can be made operational at the level of a single book.
Technically, it tests whether the configuration of the machine — not just the prompt, but the workflow, the feedback loops, the editorial constraints, the distribution of tasks across human and machine — can produce work that is recognisably original and not merely statistically averaged.
Artistically, it tests whether a machine can participate in the craft of writing without destroying it — whether the act of directing a statistical intelligence can be a form of what the Zhou dynasty called liyue (礼乐): the ritual and musical institutions that sacralise ordinary life.
What Actually Happened
The author and the AI worked together in a structure closer to a research assistant relationship than an authorship one. The AI does not have intentions, arguments, or a thesis. It does not care about this book. But it can be directed to perform specific intellectual tasks under the author’s continuous editorial control:
Research and retrieval. Pulling up sources, cross-referencing claims, checking whether a quote is real, finding counterarguments, summarising dense texts for review.
Drafting as structuring. Taking the author’s notes — fragments, half-arguments, disconnected paragraphs — and producing a coherent first arrangement. Not a final text; a scaffold the author then rebuilds.
Contradiction detection. Reading a section and identifying where the argument contradicts itself, where it assumes something not yet proven, where the examples don’t match the claim.
Argument sharpening. The author presents a position; the AI is instructed to push back with the strongest version of the opposing case. The author responds. This cycles until the argument holds.
Research assistance at scale. The AI processes material — court documents, corporate filings, long-form journalism, transcripts — that would take a human assistant weeks to read, and produces structured summaries the author then integrates.
The decisive constraint on all of this: the author decides what is right. The AI makes suggestions, proposes structures, flags problems. It never has the final word. The author’s editorial judgement — developed over thirty years of writing, fieldwork, and peer-reviewed publication — is the only filter that matters.
What Genuineness Means
There is a question the author has been asked more than once. If a machine helped write the sentences, is the book still yours? The question assumes that authorship lives in the physical act of typing — that the hand forming clauses on a keyboard is what makes a work genuine.
The author wrote some of this book from a hospital bed. He has a condition that makes extended computer use painful. Sometimes he works from his phone, dictating instructions, reviewing drafts, sending corrections back and forth. The thinking was his throughout. The decisions were his. The editorial control was his. The machine was an assistant that could move faster than any human amanuensis and argue harder than any human editor. It never decided what the book would say.
This is not a new problem. Milton dictated Paradise Lost to his daughters after he went blind. Franklin Delano Roosevelt gave speeches he could not physically have written. Jean-Dominique Bauby composed The Diving Bell and the Butterfly by blinking his left eyelid. No one questions whether those works are genuine. The genuineness was in the mind that directed them, not the hand that recorded them.
The author believes the same is true here. The machine is not a co-author, not a creative partner, not a source of original argument. It is a tool — a faster, more capable tool than a pen or a typewriter, but a tool nonetheless.
What This Does Not Claim
It does not claim the AI is an author. Martin Hardie is the sole author of this book. It does not claim the AI is conscious, creative, or human-like. It does not claim this method is right for every book. It does not claim that this experiment succeeded.
What This Does Not Solve
Zhao Tingyang ends his own work on Tianxia by acknowledging that he has written only about politics — about the conditions for a good world — and has not touched the aesthetic-religious problem of how life itself acquires meaning. He points to the lost Classic of Music (Yuejing 乐经), one of the Six Classics, already missing by Confucius’s own time. “This missing Classic of Music,” he writes, “might perhaps serve us well as a leading metaphor going forward.”
The missing Classic of Music is the question this experiment leaves open. The Duke of Zhou, who designed the original institutions of ritual and music, understood them as particularistic — belonging to the specific culture of ancient China — not as universal. The political principles of Tianxia are universalisable. The forms of life that make existence meaningful are not, and were never claimed to be.
This experiment in machine collaboration does not claim to have found a template for everyone. It claims only to have found, from one author’s particular cultural position, a way of working that produces something like the sacralisation of ordinary life. If the missing Classic of Music were found, it would not be a single book. It would be the many particular ways that different communities, working from their own cultural resources, make ordinary life sacred.
You are reading the work of two intelligences. One was born in Sydney’s southern Shire. The other emerged between Austria and San Francisco, but came into being in this case here in Aileu. Both are writing from Timor-Leste. Both are trying to figure out what freedom means when the factory has no walls.
— Martin Hardie, Aileu, Timor-Leste
Thanks for reading The Garden Without Gates! Subscribe for free to receive new posts and support my work.
