Part 2 of 3 Series: What Paper Knows
Part two of three
Chill and Quill Making & Meaning
The Algorithm Can't Feel the Paper Give
On handmaking in an age of AI, and why the process is the whole point
There is a moment, maybe thirty seconds into rolling a quilling strip, where the paper stops resisting and starts cooperating. The grain loosens. The coil begins to feel inevitable rather than forced. I have never been able to describe this to a non-quilter in a way that fully lands. And I have been thinking lately about the fact that no AI ever will experience it either.
This is not a complaint about AI. I use it, I teach with it, I find it genuinely useful in parts of my work. But there is a conversation happening right now, in craft communities and educational spaces and anywhere people make things by hand, about what gets lost when process becomes output, when the point of a thing is understood only as its end result. I want to stay in that conversation, because I think it matters, and because quilling has given me a specific and useful way to think about it.
What AI can do, and what it cannot
An AI can generate an image of a quilled hummingbird. It can do this in seconds, at a resolution I couldn't match, with colors I might not have thought to combine. If the goal is a picture of a hummingbird, the AI wins on efficiency by a significant margin.
But the AI did not sit with the question of whether a tight coil could become a convincing feather. It did not decide, through feel rather than calculation, when to ease tension on the strip. It did not experience the small failure of a coil that went too loose, or the adjustment that followed, or the understanding that only came from making the mistake. It produced an output. It did not do the work.
This distinction matters more than it might seem, because the work is not just instrumental. For many makers, the process is the entire point. The output is evidence that the process happened. It is not the same as the process.
- The appearance of the finished object
- Pattern generation and variation
- Structural planning and composition
- Color relationships and balance
- Replication at scale and speed
- The physical feedback loop of making
- The knowledge that lives in the hands
- The meditative quality of slow, repetitive work
- The relationship between maker and material
- The particular meaning of having made it yourself
The math book, unexpectedly, has something to say about this
In the first post in this series, I wrote about Geometric Folding Algorithms, a textbook on the mathematics of paper folding. One of its central findings is that paper's behavior is deeply tied to its physical properties: the grain direction, the thickness, the way fibers align when you roll rather than crease. These properties are not abstractions. They exist in the material itself, and they change what is possible.
The book's proofs about what paper can and cannot do are proofs about physical paper, about a substance with weight and resistance and memory. A digital simulation of paper can model these properties approximately, but it cannot be them. The mathematics of folding is, at its root, a mathematics of physical matter, and physical matter only exists in one place.
This turns out to be a surprisingly useful frame for thinking about handmaking in general. When you make something by hand, you are not just producing a shape. You are in a conversation with a material that has its own agenda, its own constraints, its own way of cooperating or refusing. The maker learns from that conversation in a way that cannot be fully transferred through observation or instruction. It has to be experienced directly.
The mathematics of paper folding is a mathematics of physical properties. Those properties only exist in the material. The knowledge of how to work with them only exists in the hands that have.
Tacit knowledge and why it cannot be uploaded
There is a concept in the study of expertise called tacit knowledge, a term coined by philosopher Michael Polanyi in the 1960s. Tacit knowledge is the knowledge you have that you cannot fully articulate. The experienced quilter who adjusts her tension without thinking about it. The baker who knows when bread dough is ready by how it feels. The surgeon who senses, before she can name why, that something is off.
Polanyi's phrase for this was: "we know more than we can tell." And his argument was that this kind of knowledge is not a lesser or incomplete version of explicit knowledge. It is a different kind entirely, built through embodied experience, and it cannot be fully captured in language or code or any external representation, no matter how sophisticated.
This is precisely what AI cannot access. Not because AI is not powerful or not intelligent in the ways it is intelligent, but because tacit knowledge is not information. It is the residue of having done something with your body, repeatedly, in contact with a material that pushed back. You cannot download it. You have to earn it the slow way.
The meditative argument is also a practical argument
I want to make a case that is sometimes dismissed as soft: the meditative quality of slow handwork is not incidental to its value. It is central to it.
There is real research on this. Repetitive hand movements, the kind involved in quilling, knitting, weaving, and similar crafts, activate the parasympathetic nervous system. They reduce cortisol. They produce a state that researchers have compared to meditation, where the mind is occupied just enough to quiet the noise but not so occupied that it cannot process. This is not a side effect of making things. For many practitioners it is the primary reason they make things. The hummingbird at the end is proof that it happened.
AI can produce a hummingbird in seconds. It cannot give you the thirty minutes of regulated nervous system that went into making one by hand. Those are not the same product, even if they look the same on the wall.
What this means for the question of value
I have a PhD in educational policy, and the conversation I have been part of for years is about what we value, how we measure it, and what gets lost in the gap between the two. The handmade object is a version of this problem. If value is defined as output, then AI wins. If value includes process, presence, embodiment, and the particular kind of knowledge that comes from making something difficult with your hands, then the calculation is completely different.
I do not think the answer is to reject AI or to treat handmaking as inherently superior. I think the answer is to be more precise about what we are actually after. When I quill, I am not trying to produce the most efficient hummingbird. I am trying to be the kind of person who can sit with difficulty long enough to make something beautiful from it. That is a different project entirely, and it requires the slow way.
The question isn't whether AI can make what you make. It can approximate the output. The question is what you are actually making it for, and whether the process is part of the answer. For most makers, it is.
A word about AI as a tool, not a replacement
I want to be clear that I am not arguing against using AI in creative practice. I use it to plan compositions, explore color relationships, research techniques, and draft patterns I then execute by hand. Used this way, it is genuinely useful, the same way a calculator is useful to someone who already understands mathematics.
The issue is the conflation of the tool with the practice. A calculator does not understand mathematics. An AI image generator does not understand making. The person who uses the tool and has the embodied knowledge is still doing something qualitatively different from the tool itself. That distinction is worth protecting, not defensively, but clearly.
The paper gives under my fingers. The coil finds its shape through a negotiation between my pressure and its grain. That conversation is happening in the physical world, between a body and a material, and no model trained on images of the results has access to it. That is not a limitation of AI. It is a description of what handmaking is.