The Golden Placer demonstrates that placing the n-th item at the golden angle produces provably even, non-clustering distribution, and applies it to consistent hashing without virtual nodes, key hashing of strided keys, two-dimensional and spherical point placement, and task scheduling — each shown side by side against the standard random or modulo baseline with its clustering.

the golden placer

One primitive: place the n-th thing at the golden angle (137.5°) and it lands in the biggest remaining gap — so it never clusters, for any count, online (the three-gap theorem). Pointed at every problem where clustering is the enemy and even spread is worth money.
Consistent hashing even load, 1 point/server (not 100s of virtual nodes)
Key hashing strided keys spread where % collapses them
Placement slots / sensors / samples, collision-free
Scheduling spread tasks, kill the thundering herd

The primitive — golden vs random

watch the golden spiral stay even at every count; random clumps and leaves holes
points 240
golden angle—largest/smallest gap
random—largest/smallest gap

Consistent hashing — without virtual nodes

each server owns the arc back to the previous one; even arcs = even load. Golden gets it with ONE point per server.
golden ring · 16 servers—load CV (lower = even)
random ring · 16 servers—load CV

Key hashing — Fibonacci vs modulo on strided keys

sequential IDs, aligned pointers, timestamps — keys with a stride. key % k collapses them onto a few bins; golden spreads them flat.
stride · bins
golden (Fibonacci hash)
CV —
modulo key % k
CV —

Spatial placement — collision-free by construction

warehouse slots, sensors, sample points: golden maximizes the nearest-pair distance; random leaves near-collisions
points 256
golden lattice—nearest-pair distance
random—nearest-pair distance