Honest scope. This is
open-loop even distribution — provably non-clustering, deterministic, incremental, parameter-free. "Even" means
bounded variance (the three-gap theorem: at most 3 gap sizes at ratio φ),
not zero — golden's load CV floors near 0.3 with one point per node, still ~3× tighter than random. It is not adaptive load feedback; for weighted or adversarial load you put policy on top. The novelty isn't the math (phyllotaxis, Steinhaus, Knuth's Fibonacci hashing are old) — it's using golden-angle placement as
one general-purpose primitive across all of these at once. Gate
23/23. Pure kernel, zero-dep, offline. Code:
github.com/sjgant80-hub/golden-placer.