An Editor's Note: What This Publication Is For

This is the first piece, so it had better say what I think the project is. A short answer would be: a place to write slowly about the ethics of intelligent machines, for an audience that is prepared to read slowly too. A longer answer takes the rest of this essay.

I came to the subject from physics, not from computer science, and that origin shapes the things I find interesting about modern artificial intelligence. The questions that animate me are not, on the whole, the ones that animate the trade press. I am less interested in benchmarks than in what counts as a measurement, less interested in the next architecture than in the moral standing of the institutions that deploy it, and less interested in arguments about whether a particular system is "really" thinking than in the smaller, surer claim that systems whose internals we cannot inspect are reshaping decisions that we used to make face-to-face.

The publication is therefore narrow. Quantum & Quill runs four columns. Essays, the long argument form, will appear weekly. Conversations is reserved for serious exchange — sometimes with collaborators, sometimes as a reply to a published piece by someone whose thinking I admire and disagree with. Reviews treat one book, paper, or proposal at a time, and try to do so at a length proportionate to what the work is owed. Syllabus, the most personal of the four, is the reading list I keep for my own work, opened to anyone curious enough to follow it.

The aim is not to demystify the machines. It is to refuse the suggestion that they are mysterious in a way that absolves us of moral responsibility for what they do.

A word, then, about what I am trying to avoid. I am trying to avoid the cadence in which most technical writing about AI is currently composed: the breezy futurism, the speculative civilisational stakes, the assumption that anyone who declines to be impressed must not have been paying attention. I want to read — and where possible to write — in a register closer to that which philosophers like Shannon Vallor, Luciano Floridi, and Onora O'Neill have managed to sustain: careful, citation-dense, willing to use technical detail when the argument requires it but unwilling to mistake technical detail for the argument.

There is, I think, a hunger for this. The collapse of long-form technical journalism into newsletters and short threads has produced an enormous quantity of commentary on AI and almost none of it survives a second reading. Some of the best essays of the last decade on the social implications of computing — Cathy O'Neil on opaque scoring systems, Frank Pasquale on the black-box society, Ruha Benjamin on the politics of design — were written in the form, and at the length, the arguments required. They are not popular reading in the contemporary sense. They are durable reading.

I have no quarrel with newsletters, or with threads, or with the trade press. But I have observed in myself a slow erosion of patience for ideas that take more than three paragraphs to develop, and the corrective — if there is one — is the discipline of reading one essay at a time, properly. This publication is partly an attempt to extend that discipline outward, on the suspicion that I am not the only person who needs it.

Some Practical Notes

A few mechanical things. New essays appear on Sunday mornings, GMT. The Weekly Essay newsletter contains the full piece in plain HTML and nothing else — no preamble, no list of links, no sponsor. There is no comments section here, by design. Reader correspondence is welcome by email; the most useful form is a letter, which is to say something that has been thought about between drafting and sending.

Citations appear at the foot of each essay, numbered. Where I have relied on a particular passage I have tried to say so. Where I am gesturing at a tradition rather than at a single source, I have tried to mark that distinction too. The bibliography for the publication as a whole will eventually be visible from the Syllabus department.

The publication takes no advertising, no sponsorship, and no platform money. I expect this to remain the case. The cost of running it is small enough to be absorbed by my consulting work, which is on a different subject entirely and conducted under a separate name. If you would like to support the writing without subscribing to anything, the most useful thing you can do is to forward a single essay to someone you think will read it carefully.

What Will Be Discussed

The recurring subjects of the next year will probably include: the epistemic responsibilities we acquire when we cannot interrogate the systems we deploy; the ethics of probabilistic harm, where no single decision is clearly wrong but the aggregate is unacceptable; quantum cryptography, both as a technical achievement and as a redistribution of privacy; the strange moral status of synthetic minds that may or may not be persons; and the role of professional virtue in fields whose professional standards have not yet been formed.

Some of these will be approached through reviews of existing work — Bostrom's Superintelligence, now ten years old and ageing in instructive ways; Vallor's Technology and the Virtues; Russell's Human Compatible. Some will be approached through original argument. Some will be approached through conversation with people whose disagreement with me is more useful than my own thinking.

I hope you will read for a while. I am grateful for the attention.

— Zara Nova
Zurich, July 2025