Random dots clump; grid dots look artificial. Poisson-disk sampling threads the needle: points that are RANDOM yet never closer than a set distance — ‘blue noise’. It’s how your retina’s cells are arranged, how good renderers place samples, and how stipple art breathes. Even spacing without a grid. Slide the spacing and watch the scatter settle.
Poisson-disk sampling produces points that are uniformly random SUBJECT TO a minimum-distance constraint r — no two samples are closer than r. The result is ‘blue noise’: it has the even coverage of a grid without the aliasing artefacts of one, which is why it’s used for anti-aliased render sampling, dithering, object scattering, and stippling (and matches the arrangement of retinal cone cells). Bridson’s algorithm generates it in O(n). A fail-loud self-check throws unless every pair of accepted points is at least r apart. ◆ real sampling theory, node-verified.
A dart-throwing realisation (Bridson’s grid method is faster); the MINIMUM-DISTANCE guarantee — every pair ≥ r — is exact, which is the definition of the blue-noise property. Point count depends on r and the domain.