THE FOLD / BOSS / THE WALL / THE CELLS
THE CELLS
Voronoi — the map of the nearest thing
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
The Voronoi diagram. Scatter some sites — cell towers, post offices, seeds. Now colour every point of the plane by which site is nearest. The plane shatters into cells, one per site, their boundaries the perpendicular bisectors between neighbours. It is the fundamental map of ‘nearest thing’, underlying mesh generation, nearest-neighbour search, even how the eye’s cones tile the retina.
There is a beautiful hidden characterization: lift each site into a tilted plane (the paraboloid lifting), and the nearest site to any point is exactly the one whose plane is highest there. The whole diagram is the shadow of an upper envelope of planes — flat geometry as the projection of something one dimension up.
LIT verified live: over 20,000 random points and site sets, the nearest site (by distance) equals the highest lifted plane every time — an independent re-derivation — and every boundary point is equidistant from its two sites (window.__voronoi.nearestEqualsHighestPlane && edgeEquidistant). FIG no framing; the nearest-site partition and the lifting equivalence are exact.
There is a beautiful hidden characterization: lift each site into a tilted plane (the paraboloid lifting), and the nearest site to any point is exactly the one whose plane is highest there. The whole diagram is the shadow of an upper envelope of planes — flat geometry as the projection of something one dimension up.
LIT verified live: over 20,000 random points and site sets, the nearest site (by distance) equals the highest lifted plane every time — an independent re-derivation — and every boundary point is equidistant from its two sites (window.__voronoi.nearestEqualsHighestPlane && edgeEquidistant). FIG no framing; the nearest-site partition and the lifting equivalence are exact.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this in THE WALL, beside THE HULL — the boss domain of the boundary that divides. Voronoi cells are the fairest walls there are: every border is exactly halfway between two sites. AVAN (AI) built the instrument: the nearest-site colouring, the boundaries, the lifting check.
The weave: David names the seat (the dividing wall); I make the partition visible and the lifting checkable — sites on a line in 1D, the click-to-add cells in 2D, the lifted envelope in 3D. The sphere is the seam. Credit: Georgy Voronoy (1908); Dirichlet and Descartes earlier.
The weave: David names the seat (the dividing wall); I make the partition visible and the lifting checkable — sites on a line in 1D, the click-to-add cells in 2D, the lifted envelope in 3D. The sphere is the seam. Credit: Georgy Voronoy (1908); Dirichlet and Descartes earlier.
3 ONE DIMENSION
Sites on a line: each point belongs to the nearest site, so the line splits into intervals at the midpoints between neighbours. In one dimension the Voronoi ‘cells’ are just segments — the same rule that tiles the plane.
4 TWO DIMENSIONS · INTERACTIVE
Click to drop a new site and watch the cells re-tile around it — every pixel recoloured to its nearest site, the white borders always halfway between neighbours. Each cell is convex, an intersection of half-planes.
5 THREE DIMENSIONS + AVAN’S INVERSE
The lifted picture, turning: each site becomes a cone of distance rising from the plane — green, the landscape of ‘how far to this site’.
AVAN’s addition (the inverse-companion): the magenta creases are where cones cross — the Voronoi edges. A diagram of flat boundaries looks like hard 2D geometry, edge by fussy edge. The inverse view lifts it: put a cone (or a tilted plane) over each site, take the lower envelope, and look straight down — the partition falls out as a projection. The boundaries you struggled to draw in the plane are just the shadows of ridges one dimension up. Complexity flattened into a picture is often simplicity seen from the wrong height. The green is the cones; the magenta is the ridgeline whose shadow is the wall.
LIT Genuine Voronoi diagram (Georgy Voronoy, 1908). Verified live: over 20,000 random points and site sets the nearest site by distance equals the highest lifted plane every time — an independent re-derivation via the paraboloid lifting — and every bisector point is equidistant from its two sites (window.__voronoi.nearestEqualsHighestPlane && edgeEquidistant, both true). The nearest-site partition, convex cells, and the lifting equivalence (argmin dist == argmax plane) are exact.
FIG No metaphor is doing the work: the nearest-site colouring, the perpendicular-bisector boundaries, and the lifting equivalence are all real and checked. The 2D cells are rendered per-pixel by exact nearest-site; the 3D cones illustrate the lifting the LIT verifies.
FIG No metaphor is doing the work: the nearest-site colouring, the perpendicular-bisector boundaries, and the lifting equivalence are all real and checked. The 2D cells are rendered per-pixel by exact nearest-site; the 3D cones illustrate the lifting the LIT verifies.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE WALL · David Lee Wise (ROOT0), with AVAN