A woven fabric has many threads, each crossing many others. Each crossing is a constraint. Each thread has some length of slack or tension between crossings. In principle, the configuration space is enormous — thousands of threads, each with continuous freedom to slide, twist, and buckle.

Tighten the weave enough, and almost all that freedom vanishes (arXiv:2601.15746). In a sufficiently tight plain weave, the crossings lock threads against each other so completely that the entire fabric’s shape is controlled by a single scalar parameter: the relative tension between warp and weft. Change that one number, and the fabric curves into a predictable three-dimensional shape. Leave it unchanged, and the fabric is flat.

This is not an approximation. For geometrically tight weaves — where the threads are thick enough relative to their spacing that neighboring threads physically prevent each other from moving — the mechanical analysis reduces exactly to one degree of freedom. The constraint network that seems to add complexity actually subtracts it. Each additional crossing removes one degree of freedom from the system, and in a tight weave, there are enough crossings to eliminate all but one.

The practical consequence: programming complex 3D shapes into fabric requires only thread-level actuation. Embed a single actuating thread — one that changes length in response to temperature, moisture, or electrical stimulus — and the entire fabric deforms into the target shape. No distributed sensors, no control system, no array of actuators. One thread pulls, and the constraint network translates that pull into a global shape change.

The tighter the weave, the simpler the control. Constraint is not the enemy of programmability — it is the mechanism.