A garment is not an image. It is a structured assembly of panels, seams, darts, and grainlines — each carrying physical constraints. The bias cut drapes differently than the straight grain. A dart removes fabric area to create three-dimensional curvature. A seam allowance determines whether the garment can be altered. These are not aesthetic choices alone; they are engineering decisions with quantifiable consequences for drape, stretch, durability, and manufacturability.
Textile IR (arXiv:2601.02792) treats garments as programs. Each design decision is an instruction in an intermediate representation — a formal language that sits between human design intent and physical fabric behavior. A dart is not a triangle to be drawn but a transformation that maps flat fabric to curved body. A seam is not a line but a joining operation with specified allowance, stitch type, and stress distribution.
Note (2026-11-14; read against the paper's full text): the essay describes Textile IR as catching manufacturability errors and making tradeoffs visible, as if it were a working system. The paper says in its scope section that it presents an architectural framework, the Textile IR specification and a seven-layer Verification Ladder, that synthesises capabilities demonstrated separately in published systems, that the feasibility of individual components (GarmentCode, MPMAvatar, a lifecycle-assessment methodology) has been validated in their own publications and that their integration awaits implementation, and that whether designers prefer constraint visibility over abstraction needs empirical validation. The paper's own example of a seam change is a designer-facing diagnostic (a wrinkle warning mapped to a pattern ease increase of 2 to 3 cm), and I did not find the essay's figures of two centimetres of seam movement saving 8 percent of fabric; I take that sentence to be the essay's illustration. The paper also says differentiating through a sewing pattern is impossible, because adding a dart changes the mesh topology discretely (there is no half a dart), which is why it proposes searching over topology and optimising over parameters.
The advantage is that programs can be analyzed before execution. A compiler catches type errors before runtime. Textile IR catches manufacturability errors before cutting fabric: a seam that requires impossible stretch, a grainline orientation that wastes material, a construction sequence that cannot be sewn on standard equipment. These are problems traditionally discovered on the factory floor — after cutting, sometimes after sewing — at costs measured in wasted fabric, wasted labor, and wasted time.
The deeper advantage: the IR enables navigation of tradeoffs that are invisible in visual design. Sustainability (less waste), manufacturability (simpler construction), and aesthetics (desired drape) interact in ways that no human designer can fully optimize by intuition. Moving a seam line by two centimeters might reduce fabric waste by 8% while changing the drape imperceptibly. The IR makes this visible because it represents the garment in terms of the variables that actually matter — physical properties, not visual appearance.
The garment is a program. The body is the runtime. The fabric is the hardware.