On Writing With Machines

There is a question that keeps surfacing in conversations about AI and creative work, and it is not the one people think they are asking. The surface question is: can a machine write? The real question is: what do we mean when we say someone wrote something?

Every writer uses instruments. The quill, the typewriter, the word processor — each one changed the texture of prose. Scholars have argued that the typewriter made Nietzsche's sentences shorter. The word processor made revision frictionless and, some would say, made writers less decisive. The instrument shapes the work.

AI writing tools are instruments of a different order. They do not merely transcribe or format; they generate. They propose. They complete. The question of authorship becomes genuinely complicated when the instrument can produce a plausible sentence you had not yet thought of.

Here is what I think survives: judgement. The capacity to look at a generated sentence and know whether it is true — not factually, but tonally, intellectually, in terms of what the piece is trying to do. That capacity is not automatable, because it requires knowing what you are trying to say before you can evaluate whether something says it.

This is why the most interesting use of AI in writing is not generation but interrogation. Using the tool to surface what you actually think by showing you what you do not mean.

The studio operates on a simple principle: AI accelerates the distance between the first draft and the final one. It does not replace the thinking that precedes the first draft, or the judgement that shapes the last.

That thinking — the research, the argument, the decision about what matters — remains entirely human. The machine handles the scaffolding. The structure is ours.