Preface: The Method

On agentic work and the use of artificial-intelligence tools in this study.

This book is written with the help of artificial intelligence. I tell you so on the second page, before you read anything that depends on it, and I will explain exactly how, so you know how much weight to give what you read. In a project that claims to prefer defensible ideas over comfortable ones, hiding the method would be incoherent. Declaring it properly is part of the rigor.

First, what the word means. An agent is an artificial-intelligence program that is given a complete task — not a loose question— and carries it out in steps: it searches, cross-checks, organizes, drafts, and reports what it did. Agentic work is exactly that: delegating whole tasks of research and preparation to these tools, under human direction and human review. In this study, the agents do very concrete things. They search for and triangulate sources: when I claimed that my phrase existed in no prior corpus, it was agents that combed through thirty-five sources in two languages — academic databases, book repositories, quotation sites— and brought me the evidence, classified, including the evidence that could have proven me wrong. They verify citations and hang a label on each one: verified, or pending verification. They formalize terms: the entries for the Contrarian Thesis and for saudade that you will read further on were born as agentic drafts over my notes, which I corrected, pruning by pruning, until they said what I actually hold. They speed up first drafts, translate between Spanish and English, and keep in order the archive of knowledge this book feeds on.

And now the line that is never crossed, because it is the one that matters to you as a reader. Artificial intelligence does not decide what I think, does not choose what I claim as my own, and does not sign. The motto is mine; the three edits that set it apart from the classic tradition are decisions I made and I defend; the Contrarian Thesis is my stance; the stories — the lab in Matamoros, Francisco and his shoebox, the cloud invoice, the recovery— are my life, told by me. The tool amplifies my judgment; it does not replace it. When something in this book is wrong, the error will be mine, because responsibility is not delegated. A very good hammer makes no one a carpenter, nor does it carry the blame for a wobbly table.

How is truthfulness controlled? With a simple protocol you already saw at work on the first page: every claim travels with its source; every source carries its verification status; and the caveats — the limits of what I can defend— are written out in full inside the text, not in an errata sheet nobody reads. Where an attribution is doubtful, I say so. Where a classic text turns out to be pseudo-epigraphic, I flag it. Where the evidence does not support an accusation, I temper the accusation. None of that is guaranteed by a machine: it is guaranteed by the commitment that no page is published without passing through my judgment. You can audit the result, because the seams are left showing.

There is a deeper reason for telling you all this, and it connects to the thesis of the book. I defend sovereignty: over your thoughts and over your data. It would be strange to defend it with a rented method. That is why this study runs, as far as possible, on infrastructure of my own — a small but full-fledged data center I built myself: a server laboratory with GPU infrastructure, networking and systems, where my local models run—, with outside assistance where it genuinely helps, and not on blind faith in someone else's black box. And it is not only the machine that is mine: the method is too. This way of working — the author at the center, giving the instructions, while the agents carry the heavy load— I designed it, I direct it, and it belongs to me; the process is a creation of mine, not of the tool. Artificial intelligence is the instrument — enormously powerful, but subordinate—: it executes under my authorship and my verification, it does not decide for me, it does not sign, and it is neither co-author nor owner of anything. I am neither a technophobe nor a dazzled convert: artificial intelligence seems to me exactly what it is, an enormously powerful tool with a responsible author in front of it. Using it without declaring it would be smoke. Declaring it, showing you where it begins and where it ends, is the opposite of smoke: it is the method laid on the table.

What follows, then, was written like this: the agents carried bricks, and the one who drew the blueprints — who decided where every wall goes, and who answers for the house— is me.