How a world goes from a pencil doodle to 1,536 pixels.
One canvas size
Everything is 48 × 32. At eight times magnification that's 384 × 256, which fits a phone screen, and the canvas is small enough to finish in an evening or two. The limit is the point: when a lighthouse gets three pixels, you find out which three matter.
Palette before pixels
I pick colours first, usually somewhere between 18 and 28 per mood. Sky gets three bands, the ground two or three, and the remaining colours go to whatever the eye should land on. Each world is drawn once with named colours (“sky top”, “window”, “lamp”), and each mood is just a different list of values for those names.
Sleeping colours
The trick behind every day/night pair: things that light up at night – windows, lamps, stars, reflections – get their own names, but in the day palette they're set to the same colour as their surroundings. They're invisible until the night palette wakes them up. No redrawing, no layers.
Dithering, sparingly
Where two sky bands meet there's one checkerboard row to soften the step. I use the same pattern for light that should feel thin – the lighthouse beam, a lamp's glow – and almost nowhere else. Dither everywhere and the picture turns to gravel.
Two or three frames
Animation is kept to two or three frames: a beam swinging, rain falling, flames, snow, smoke. The frames are stacked inside the picture and simply take turns. If your system asks for reduced motion, they hold still on the first frame.
Exporting
Every world is saved as an SVG in which each run of same-coloured pixels is a rectangle. It stays perfectly sharp at any zoom and a whole world is about 5–10 KB, smaller than most thumbnails.
Tools
Paper first: a rough sketch with the 48 × 32 grid ruled in pencil. After that, any grid editor works. The tool matters much less than deciding what each pixel is for.