There is an image that explains, better than any argument, what separates the meticulous construction of a series from the mere accumulation of text: a path opening through undergrowth, not because it already existed waiting to be found, but because someone advances with a sword and decides, cut by cut, which branch falls and which remains standing, and it is that constant deciding, nothing else, that turns undifferentiated vegetation into a route with direction and sense. A language model, faced with the same undergrowth, opens no path at all: it flattens the whole thing in a single stroke, producing a clearing that looks perfectly coherent, frictionless, with no trace of the thousand forks it discarded, because it never had to choose between them in the first place — its task is not to go anywhere in particular but simply to fill the blank with whatever continuation is statistically most plausible.
That instant clearing may resemble a book, but it is not a path, because a path implies a point of departure and a point one wanted to reach, and between the two, millions of possible trajectories — tones, eras, turns, references, figures, ways of naming the same thing differently — of which only one was actually walked, at cost, and that cost is precisely what the automatic shortcut cannot simulate no matter how fluent its final output. The sword, here, is not violence against the material but precision: cutting well requires knowing exactly what one wants to leave out, which is far harder than knowing what to include, since inclusion is the default that requires no judgment at all, while excluding fairly — keeping forty-eight references out of a hundred and sixty, choosing three sources out of twelve that actually bear the weight of the argument — demands having first understood the whole field before deciding, with a steady hand, which part of it can be let go. This is the exact sense in which swordsmanship is a method: not brute force against the sea of existing text, but aim — the ability to find the precise angle at which a minimal gesture opens a passage where before there was only undifferentiated thicket, a precision acquired only by walking, that is, by revising, synthesising, discarding, retracing one's steps when a cut landed at the wrong angle, and trying again, as many times as it takes, until the path begins to hold its own shape. What is interesting is that this path, once opened, resembles no other — not because the material it is made of, the words, the references, the cited figures, is secret or inaccessible to any other walker with the same sword, but because the exact trajectory across that sea of possibility depends on a particular sensibility, on a particular order of decisions made in a particular sequence, and the odds that two people — or two models, or a person and a model working alone — would reproduce the very same sequence of cuts across millions of available forks are, if the work has been fine enough, close to zero. That is the real difference between a generic library and one's own: the generic library is the uncut thicket, everything still possible and undifferentiated, available to anyone in exactly the same form; one's own library is the path already traced, which exists only because someone decided, at a precise point in the forest, to go left rather than right, to let this branch fall and not that other one, and kept doing so thousands of times until the route became recognisable as a path rather than any path at all. This carries an uncomfortable consequence for any fantasy of accelerated production: there is no real shortcut to a path of one's own, because the shortcut, by definition, is what replaces the cut with the bulldozer, and a bulldozer leaves no path — it leaves uniform ground where a forest of possibilities used to be. The collaboration between the hand holding the sword and the force that could flatten everything in one pass does not, then, consist in letting that force do the work while the walker watches; it consists in using that force against its own grain — not to produce the instant clearing, but to briefly illuminate the whole thicket, to show every branch, every fork available at that point in the forest, so that the cut, when it comes, is an informed cut and not a blind one. Seeing more of the undergrowth is not the same as clearing all of it; seeing the thousand conceptual possibilities opening at each sentence, each reference, each shift of tone, is precisely what makes the choice a real choice rather than a default one. And in this the path finally resembles any art that depends on a precise gesture repeated with patience until the hand no longer needs to think the cut through: it is not the material that makes the path unique, but the hand that decided, among a million possible openings, exactly this one.
