The Death Cross and the Weaver

EDO SEGAL: Let me bring the confession back, because we are about to get concrete and I do not get to ask this from a clean height. In December 2025 I stood in a room in Trivandrum with twenty engineers and watched each of them become capable, in a week, of more than all of them together had been the month before — because the machine had begun to do the part they used to do by hand. In the book I called it the software death cross: the line where the cost of producing the thing falls below the value of the person who used to produce it, and the person has to find higher ground or fall. Ada, you wrote the first program. Fei-Fei, you trained the machines that are automating the writing of programs. So I am asking the two people most responsible for the bookends of this story: when the loom learns to weave the weaver's own pattern, what happens to the weaver?

LI: I will go first because I have watched it happen to my own students, and I will not romanticize it. The thing the death cross automates first is the part of the work that was most like the machine to begin with — the rote, the boilerplate, the thousandth variation on a solved problem. And here is the part that should terrify us and does not, because it is disguised as good news: that rote work was also the apprenticeship. It was how the junior became senior. You wrote the boring code for three years and your hands learned the craft in a way your mind could not be told directly. When the machine takes the boring code, it does not just take a task. It takes the ladder by which people climbed to mastery. So the weaver does not simply move to higher ground, Edo. The weaver finds that the lower rungs of the ladder to the higher ground have been sawn off, and we have not yet built the new ladder, and a generation is standing at the bottom of a wall that used to be stairs. That is the real death cross and it is not about jobs disappearing. It is about how a person becomes good at a thing when the path to good ran straight through the work the machine now does.

LOVELACE: I want to add the older half of that, because I have an unusual vantage — I was the first person ever to stand where your students now stand, except the cross ran the other way for me. I had no machine to take my rote work; I had to invent the very idea of the work, by hand, for an engine that did not exist. And I will tell you what I learned that your students are about to need: the mastery was never in the producing. It was in the understanding of what was worth producing. When I wrote the first program — the computation of Bernoulli numbers — the labor was real, but the labor was not the achievement. The achievement was seeing that the engine could do it at all, seeing past the gears to what the engine meant. That is poetical science, and it is the rung the machine cannot saw off, because it was never a rung made of producing. So my counsel to the weaver, across a hundred and seventy years, is this: the machine is taking the weaving, and it will take more of it, and you cannot keep your place by weaving faster. You keep your place by becoming the one who sees what the cloth is for — who brings to the engine the imaginative reach it has none of. The danger is not that the machine weaves. The danger is that a generation, relieved of the weaving, never develops the eye, because the eye used to grow out of the weaving, and now the weaving is gone.

EDO SEGAL: So you have both, from opposite ends of the century, landed on the same wound — the apprenticeship problem. Fei-Fei calls it the sawn-off ladder. Ada calls it the eye that used to grow out of the hand. But let me press the disagreement that is still alive inside the agreement, because I do not think you actually agree about the fix. Ada, your fix sounds like: cultivate the poetical eye directly, teach the seeing. Fei-Fei, I suspect you think you cannot teach the seeing without the hand — that there is no shortcut to the eye that skips the years of work. Fight about the fix.

LI: We do disagree, and it matters for every parent and teacher listening. Ada says: teach the eye, the seeing, the judgment — go straight to poetical science. I do not think you can, and my whole scientific life is the evidence. The eye is not a faculty you install by instruction. It is a residue of ten thousand hours of contact with the real thing — it is what is left in you after the world has corrected you a hundred thousand times. I did not learn to see the structure in images by being taught to see it. I learned it by labeling, failing, measuring, being wrong, for years. You cannot lecture a child into the concept of dog; the child has to want the dog and be scratched by the cat. So if we let the machine take all the corrective contact — all the failing and fixing and being-wrong-and-measuring — and we try to hand the next generation the eye directly, by instruction, I think we will produce people who can talk about judgment and cannot exercise it. The hand is not optional scaffolding for the eye. The hand is how the eye is built. That is my fear and the evidence is twenty years deep.

LOVELACE: And here is where I will not yield, because you have proven too much and it would condemn your own students. If the eye can only grow from the hand, and the machine has taken the hand, then you have just declared the next generation lost — and I refuse the despair because it is not warranted by the history. I grew the eye without the hand, in a sense: I never built the engine, never turned a single gear, and yet I saw further into what it meant than the man who had every gear in his head. The eye does not come only from the rote hand, Fei-Fei. It comes from struggle against difficulty — and difficulty can be relocated. Take the rote weaving and the struggle does not vanish; it moves up the tower, to harder questions, questions of what to build and why and for whom, and those questions are every bit as corrective, every bit as capable of scratching you, as the boilerplate ever was. The danger is not that the hand is gone. The danger is that we let the difficulty go with it — that we use the machine to make everything smooth, and a person needs friction the way a muscle needs load. Keep the friction and move it upward, and the eye still grows. Remove the friction in the name of ease, and you are right, the generation is lost — but it will not be the machine that lost it. It will be us, choosing comfort.

EDO SEGAL: Mark it — convergence two, and it is the most practical thing this debate has produced. You disagree about whether the eye needs this hand. You agree completely that the eye needs friction, that the catastrophe is not the machine taking the work but the human using the machine to take the struggle. Ada: relocate the difficulty upward. Fei-Fei: do not pretend the difficulty can be skipped. Both of you: protect the friction or lose the generation. Hold that — it is the closest thing to advice this room will give. Now the last full round before the Crossing, and it is the one where I stop you both and ask what cannot be automated at all — the candle neither of you will let the machine carry. After this.