The living New Orleans skyline at the bottom of our site is smaller than a single emoji image. Cathedral, Superdome, the Crescent City Connection, 92 windows blinking on their own schedules. Watch it assemble:
There’s no image file. The whole scene is drawn in the page’s own code: “rectangle here, this size, this navy” about three hundred times. A small program we wrote generates the city: building widths and heights come from constrained randomness (no two neighbors the same height, no flat monoliths), and the landmarks (St. Louis Cathedral, the Superdome, the bridge) are drawn by hand from photographs.
The randomness is seeded: the same city regenerates identically every time, which means we can edit one layer without disturbing a pixel of the rest. Every window light is its own tiny rectangle with its own on/off cycle (7 to 26 seconds, scattered phases), so the pattern never visibly repeats.
The animations run as CSS on your device’s graphics chip, off the main thread, so they can’t make the page stutter. A looping video of this same scene would cost a thousand times more data.
This is the standard your site gets. We build everything this way, right down to the decoration. Google ranks fast sites higher, and that is a big part of how our clients keep winning local search.