Let me tell you how this works.
I have a thought — usually dense, usually half-formed, sometimes barely grammatical. I write it down in what I call a seed: a compressed file of references, connections, structural intuitions, and emotional register. It is often messy. It is always specific. I know what I want to say; I do not always know how to say it in a way that a reader will want to receive.
Then I give it to Claude.
The model reads the seed, finds the structure latent in it, writes a draft, receives my corrections — this line is too soft, that reference needs more weight, this section buries the argument — and produces a revised text. I read everything. I cut what doesn’t land. I restore what got smoothed away. I push it out in three languages, which the model also handles.
This is the process. I am not hiding it. This post is the disclosure.
I. Why the confession is necessary.
The homepage already has a line: Thinker, not prose stylist. The ideas are mine; the articulation is assisted by AI — an honest arrangement, as old as the first scribe. That is the short version. This post is the long version — because the short version invites a question the short version cannot answer: what exactly makes the ideas “mine” if I didn’t write the sentences?
The question deserves a real answer, not a defensive one. And the answer requires some history.
II. The oldest arrangement.
Scribes in Mesopotamia wrote what rulers and merchants dictated. The escribano in colonial Latin America — a licensed notarial officer, a figure with no direct equivalent in the Anglo-Saxon legal tradition — gave legal and literary form to ideas that landowners, merchants, and ordinary people could not express in the formal register the law required. Without the escribano, the idea had no legal power. Without the idea, the escribano had nothing to write. The arrangement was explicit, professional, and entirely respectable. It was also the standard operating condition for the transmission of ideas across most of human history.
Plato wrote Socrates. Socrates wrote nothing. The most influential philosophical voice in Western history left no manuscripts; what we have is one man’s interpretation of another man’s spoken thought, shaped into dialogues that Socrates never read or approved. We call this philosophy, not fraud.
Jorge Luis Borges, after losing his sight in his fifties, dictated everything to his wife and, later, his secretary. His prose did not diminish. His ideas remained entirely his own. The hand that held the pen changed; the mind that generated the sentences did not.
Every American president since Franklin Roosevelt has used speechwriters. FDR had Robert Sherwood. Kennedy had Ted Sorensen — who wrote “Ask not what your country can do for you.” Obama had Jon Favreau, who started at twenty-three and became chief speechwriter at twenty-seven. Churchill dictated; secretaries typed. Every major celebrity memoir has a ghostwriter whose name does not appear on the cover. The industry is enormous, mostly invisible, and has never been seriously called fraudulent — because everyone involved understands that the public figure is the source of the ideas, the experience, and the persona, while the writer is the craftsperson of the form.
The objection that “the writing must be yours” is extremely recent, historically speaking, and was never universally observed even when nominally endorsed. The Romantic notion of solitary authorial genius — the writer alone at the desk, suffering for the craft, producing sentences that are extensions of the self — is a brief cultural episode, roughly two hundred years old, in a history of writing that is five thousand years old. It is not a natural law. It is a fashion, and like most fashions, it is already passing.
III. Where the work actually lives.
The meaningful distinction is not between “writing yourself” and “using a tool.” It is between having a thought worth having and not.
If the prompt is “write something interesting about epistemology,” the output belongs to the machine. There is no human seed — only a request for generation. The machine averages the corpus; the result is average. Nothing of value has been contributed.
If the prompt is a specific argument — a particular claim about string theory and institutional drift, a particular reading of Emerson applied to colonial gratitude, a particular analysis of the duck test as the thing that high-status systems learn to spoof — then the LLM is executing a blueprint drawn by a human mind. The architect does not lay the bricks. The composer does not manufacture the instruments. The question is not who held the pen but who determined what the pen would say and why.
Roland Barthes declared “the death of the author” in 1967, arguing that the author’s intention is irrelevant to a text’s meaning — meaning is produced in the reader, not the writer. The LLM debate makes this theoretical position suddenly very practical. If authorship is a function, as Foucault argued in the companion essay two years later, rather than a biological attribute — a role, a legal category, a set of responsibilities — then the question “who is the author?” is answered not by asking who typed the words but by asking who holds the intellectual responsibility for the claims.
I hold it. The claims in these posts are mine to defend or withdraw. If an argument is wrong, I am wrong, not the model. The model does not have positions; it has patterns. The positions are mine.
The multilingual dimension makes this more concrete. Publishing in English, Spanish, and French without professional translators for each language was, before LLMs, simply not available to a non-institutional author. The ideas do not change language — they are translated. I did not write three posts; I wrote one thought, which three versions of the same tool rendered in three grammars. This is not deception. It is access.
IV. When does it become cheating?
The “cheat” label attaches when there is a gap between the persona presented and the reality of the process.
If I claimed to be a stylist — a craftsperson of sentences, a master of prose rhythm — and used LLMs to fake that mastery, the deception would be real. The persona and the reality would not match, and a reader attracted by the writing as writing would be misled.
But I am not claiming to be a prose stylist. The claim, the only claim, is: I have thoughts worth reading. That claim is either true or false independently of who arranged the syntax. The reader who evaluates the ideas on their merits is not being deceived. The reader who is attached to the idea that I personally agonized over every comma is importing an expectation the text never invited.
Chögyam Trungpa warned against spiritual materialism — using the tools of liberation to reinforce the ego rather than dissolve it. The equivalent here is authorial materialism: being more attached to the identity of being a Writer (capital W, solitary, suffering for the craft) than to the actual work of thinking and communicating something worth the reader’s time. The LLM dissolves the authorial ego. What remains is the idea, naked, available for evaluation. That is not a loss. That is the point.
V. (Coda) This post.
This seed was developed in conversation with Claude. The post you are reading was drafted by Claude, corrected by me, and will be published in three languages by the same process described above. The post about using AI as a ghostwriter is itself ghostwritten by an AI.
I considered whether this undermined the argument and concluded that it is, in fact, the argument. The tool is transparent. The thought is mine. The escribano does not pretend to be the client. The client does not pretend to be the escribano. This is the oldest arrangement in intellectual history.
It is also, now, the cheapest.