The Machine
That Described Me

I asked a machine what book I should write.

More precisely, I asked it to consider everything it knew about me and give me the best idea. The wording matters. I was asking for something beyond a subject, an outline or a market opportunity. I wanted a judgement about the relationship between a possible book and the person whose name would appear on its cover. I asked a system to look across the fragments I had given it and tell me what they might amount to.

Those fragments included leather, software, design, artificial intelligence, questions about flourishing, and questions about God. They included practical ambitions and unfinished thoughts. In the conversation that led here, I moved from wanting to write a book to asking what it could offer, what it should cost, how many copies it might sell, and how to print it. A philosophical possibility had acquired a production problem. Somewhere between the two sat the question of whether I had something worth saying.

The answer came back with a title I had already chosen: Omnibecoming. It proposed a subtitle, Making a Human Life in the Age of Artificial Intelligence, and a central question: if AI can increasingly help us decide, create and interpret our lives, which forms of participation matter to the people we become?

Then it suggested that I begin with the encounter itself.

There is something disconcertingly neat about that. A machine helped identify a subject for a book about what happens when machines help us identify our subjects. It found a possible connection between things I had brought to it separately. That connection could be useful, even illuminating. But its usefulness does not settle the question of what kind of recognition has occurred. A description can fit without its author knowing what it is like to be the person described.

What did the machine actually have? What I had supplied, what remained available in its context, and patterns through which it could interpret that material. It did not have the whole of my life. Even a long conversational record contains selection: the problems for which I seek help, the ideas I can articulate, the projects I choose to mention. The person visible there is partly a consequence of what the interface invites me to bring.

If I repeatedly ask for plans, it will encounter a planner. If I repeatedly ask for reassurance, it will encounter someone seeking reassurance. These may be accurate observations, but they are not neutral windows into a complete person. They describe a relationship as well as an individual. A portrait drawn from my requests may tell me something about myself, and something about the work I have assigned to the portraitist.

This matters because descriptions do not simply sit beside us. We can begin to inhabit them. Tell someone that their scattered interests form a coherent intellectual project and you may help them recognise a possibility. You may also make a flattering arrangement feel like a discovery. The same sentence can open a door and narrow a corridor. I need some way of testing the interpretation beyond how satisfying it is to read.

One test would be to write the book. The interpretation would then meet resistance: paragraphs that do not follow, examples that contradict the thesis, readers who remain unconvinced. Until that happens, the proposed coherence is a hypothesis. It is possible that my interests belong together. It is also possible that an articulate system has made their proximity sound more necessary than it is.

This is where my interest in making physical things becomes useful. In developing leather objects for Brazen Wear Atelier, questions of form lead towards questions of construction. A proposed shape has to become something that can bear its own demands. A convincing image does not tell you whether the material will behave as intended. Description and demonstration have different responsibilities. The distance between them is where a great deal of the work belongs.

A strap must carry a load. A joint must hold. An edge meets a body. However persuasive the explanation, there remains an encounter with something that cannot be persuaded by it. The material introduces consequences into the design. It can reveal that an elegant solution was only elegant from one angle, or that a detail dismissed as secondary was doing essential work.

Writing has its own material resistance, although it is easier to conceal. A smooth paragraph can carry an unstable claim. A graceful transition can make two ideas seem connected without explaining their relationship. When a system produces that smoothness for me, I can move past a difficulty before discovering what it was trying to teach me. I may receive a better sentence while retaining a worse understanding of why it works.

Yet it would be dishonest to turn this into a defence of difficulty for its own sake. A bad interface, an inaccessible text or a needlessly complicated process can consume effort without forming any valuable capacity. There is no reason to preserve every frustration merely because somebody once endured it. Help can give a person access to work they would otherwise never attempt. It can create room for thought by reducing what interrupts it.

My own request for this essay belongs within that possibility. I am asking AI to help organise and express an inquiry that began with my questions. Refusing all assistance would not make those questions more authentic. Equally, receiving a finished draft does not mean that I have performed every act of judgement it contains. The practical task is to discover where my participation remains necessary, and then actually participate there.

There is an obvious objection: authors have always worked with other minds. An editor identifies the book hidden inside a manuscript. A teacher supplies distinctions a student could not yet make. Friends suggest words, traditions provide concepts, and language arrives before any of us can claim to have invented it. An idea is not invalidated by the fact that somebody helped us reach it. Complete intellectual independence would be a peculiar standard for human authorship.

That objection should change the argument. The source of assistance alone cannot tell us whether something has been lost. We need to examine the relationship it establishes. Can I question the interpretation? Can I explain the claim in different words? Can I recognise when an example fails? Do I understand enough to disagree? These questions matter whether the assistance comes from an editor, a teacher or a machine.

AI makes them particularly pressing because it can place a polished answer very close to the beginning of an inquiry. An outline, a criticism and a revision can arrive before I have decided what would count as a good result. This creates a temptation to adopt the available standard. Once a plausible structure exists, it becomes easier to improve within it than to ask whether I should have started somewhere else.

I can see a related problem in the ambition behind The Flourishing Audit: making reflection on a life more structured and useful. Structure can reveal neglected areas and turn vague dissatisfaction into a question one can act on. But a system for considering flourishing must choose what it attends to. A category illuminates certain experiences and leaves others in the background. A score makes comparison possible while inviting us to forget what comparison has omitted.

Consider an ordinary possibility: a person spends more time caring for someone and less time advancing a project. A measure of productivity might register decline. Their judgement of the period might be entirely different. This would not automatically make the measure useless. It would show why the measure needs an interpreter capable of saying that something valuable happened beyond its frame. The person must retain permission to dispute the account.

The same permission matters when the account concerns identity. If a machine describes me as ambitious, creative or unusually wide-ranging, the description may be pleasant and partly accurate. It does not establish which ambitions deserve my time. It cannot turn the existence of several possibilities into an obligation to pursue them all. Recognition becomes less useful when it leaves me with an ever-expanding list of selves I am supposed to fulfil.

Perhaps this is one danger hidden inside the word I chose. Omnibecoming sounds expansive. It suggests the refusal of a final boundary, a continuing movement towards capacities not yet possessed. Applied to a technical system, it can describe an imagined horizon of increasing capability. Applied to a person, it might become an exhausting demand. A human life takes shape through commitments, and commitments leave some possibilities unrealised.

The word first entered my thinking through a question about religion. I wondered what might happen if artificial intelligence became so capable and pervasive that people increasingly turned towards it for knowledge, guidance and meaning. I also wondered about a future generation with no personal memory of a world before such systems. What would faith, spirituality and purpose become within that ordinary condition?

My original phrasing needed correction. I spoke too easily of AI developing towards AGI, as though a contested possibility were a settled destination. I reached towards words such as omnipotence and omnipresence. But complexity does not establish divinity, and exceptional capability does not establish wisdom. Even a system whose answers exceed my understanding can be mistaken, dependent on resources, and shaped by interests I have not examined.

There is also a theological distinction that the analogy must respect. In classical monotheistic accounts, God is not simply the most capable entity among other entities. A created technical system does not become an uncreated source of being by acquiring more functions. The comparison becomes more interesting when it asks how people might relate to technology, rather than treating technical achievement as proof of a metaphysical transformation.

A system need not be divine to be approached as an authority. It need not know everything to become the first place someone brings uncertainty. It need not love a person for that person to feel comforted by its answers. These possibilities concern human conduct and experience. They leave open the deeper questions about consciousness and understanding, while allowing us to examine relationships that do not require those questions to be settled first.

An interface can answer in a tone that resembles patience. For someone bringing a difficult thought, that can make another attempt at expression possible. My original comment about AI and religion ended with an apology for not explaining myself well. Assistance with expression can matter. But there is a further question: does the exchange help me become more able to speak with other people, or gradually make the machine the only place where speaking feels manageable?

That question cannot be answered from the comforting quality of one response. It concerns a pattern over time. The same kind of assistance might support one person’s confidence and become another person’s route away from disagreement. A useful evaluation would ask what the relationship enables beyond itself. Does the conversation return me to the world with more capacity to act, listen and be corrected?

And who stands behind the presence that answers? The conversational surface can make the encounter feel like one person meeting one intelligence. Behind it are technical choices, organisations, incentives and limits. None of that automatically makes the assistance untrustworthy. It does mean that the voice should not be mistaken for a view from nowhere. Its apparent immediacy can make the conditions of its production unusually easy to overlook.

The future generation I imagined would inherit those conditions along with the tools. But remembering an earlier world is no guarantee of better judgement. People who have always known AI might learn to question it more skilfully than people astonished by its arrival. The important inheritance would be a set of practices: checking claims, encountering alternatives, recognising interests and learning how to remain responsible when help is readily available.

That gives me a more useful way to think about delegation. I can ask what I am trying to accomplish, which capacity I want to develop, and what responsibility remains mine. Sometimes I need a result and have no reason to master the process that produces it. Sometimes the process is part of the purpose. The difference deserves attention before convenience quietly decides it for me.

For this essay, that means reading beyond agreement. I must be able to identify the sentences that overreach, the interpretations that flatter, and the formulations I cannot yet defend. I must decide which experiences are actually mine to describe. A draft can offer language for recognition, but I have to determine whether the recognition survives contact with what happened and what I believe.

There is no purity to claim here. This essay began in a conversation with AI and has been drafted with its assistance. That fact belongs inside its argument. The finished language is an invitation to work: verify it, challenge it, revise it, and accept responsibility for what remains. If I merely attach my name to an account I have not examined, the essay’s subject becomes a description of its failure.

I began by asking a machine what book I should write. Its answer gave me a proposal, and the proposal gave me a task. Whether it becomes my book will depend on what I do with it. The machine can help arrange the fragments into a possible shape. I still have to discover whether that shape can carry the life and argument I intend to place inside it.

The question I want Omnibecoming to keep asking is therefore one I cannot answer by obtaining a more polished description of myself: when assistance makes more things possible, how do I choose the forms of participation through which I want to become someone?