The SEGA AI Judge System · Patent Pending PI2026005219 · UYSF & SEGA Intercontinental Tournament — Intercontinental AI (Live) Register →
Teaching instrument · interactive

Anatomy of the asana

One figure moves through a flowing sequence, and under it four systems are revealed in turn — skin, muscle, nerve, bone — and then stacked together. Below the sequence there is a library of twenty-four asanas: pick any one and the figure holds it while the layers keep running, so you can stop on the nerve layer of a headstand rather than of a standing pose. Nothing here is a photograph or a scan: it is a working diagram of where load travels, drawn from joint positions and segment widths, which is a more honest thing to put in front of a student than an illustration that implies more precision than anyone has.

The stack, in order of dependency
01 Integumentwhat the room sees
02 Musculaturewhat holds the shape
03 Nervouswhat corrects it
04 Skeletalwhat lets muscle stop
Read upward: each layer only rests when the one below it is doing its job.
Stand still. Everything under the skin is already working.
Why it is built this way

A diagram you can stop, not a video you can only watch

A video of anatomy is watched once and forgotten. What a teacher actually needs is to stop on a layer — to hold the figure at the moment the nerve pathways light and say this is why the correction arrives before you have decided anything. So this is scrubbable and pausable, the four layers are addressable by name, and the whole thing runs from about eleven kilobytes of code with no video file, no framework and no network request. It will load on a phone at a tournament venue.

It also stops when it scrolls out of view, and it never starts at all for a reader whose system asks for reduced motion — they get a still frame and the same controls. An instrument that runs regardless of what the person in front of it needs is not an instrument.

Layer 01 · Integument

Skin and outline

The shape a room actually sees, and the only layer most correction ever addresses. It is drawn faintly on purpose — it is the least informative of the four, and it is the one everybody looks at.

Layer 02 · Musculature

Fibres, building in order

Bundles thicken along each segment as the figure moves into Virabhadrasana II. What holds a posture is not one muscle but a sequence of them shortening and lengthening against each other, which is why a pose can fail in a place that does not hurt.

Layer 03 · Nervous

Spine to limb, in milliseconds

Pulses run the chains from the pelvis outward. Postural correction happens far below decision: by the time a student notices they are falling, the correction has already been sent and either arrived or not.

Layer 04 · Skeletal

Stacking, so muscle can stop

Joints resolve as the figure folds into Vrksasana. The whole economy of an asana is here: when the skeleton stacks, muscle stops gripping, and what looked like strength turns out to have been alignment.

What this diagram is not.

It is a schematic, not an anatomical atlas. The muscle layer shows bundles along segments rather than named muscles with real origins and insertions; the nerve layer shows conduction pathways schematically rather than actual innervation. It is built to teach the principle — that four systems are working under one shape, in a fixed order of dependency — and it is accurate about that. For named structures, use a proper atlas, and for anything about your own body, use a clinician who can examine it.

Where it came from

Designed in one tool, rebuilt to ship

This composition was authored as a design file and exported as a self-contained page. Unpacked, that export was a little over three megabytes: React, an in-browser compiler, a general-purpose animation engine, a control-panel framework and thirteen font files — wrapped around sixteen kilobytes of actual scene.

None of that is wrong for a design tool, whose job is to let a composition be edited. It is wrong for a public page. So the scene was ported rather than embedded: the same joint coordinates, the same three poses, the same layer choreography and camera moves, rewritten as plain SVG and plain JavaScript against this site's own palette and type. It is the same piece; it just no longer carries the workshop it was made in.