STORM PLATES · WAVE PRESENTATIONS OF THE SIEVE

The elementary presentation: prime p is a swell of wavelength exactly p. Every crest lands on a multiple of p and breaks that plank, except plank p itself, which is where the swell was born and stands proven-green. The crests never drift: the animation surges vertically but the phase is pinned to n ≡ 0 mod p, which is what makes this a specimen and not a costume. Guard: kill positions re-derived live by trial division, not painted. Knob: swap the swell and watch wavelength = p hold.

The full storm to 60 needs only swells ≤ √60: four waves make every gap. A plank drowns if any crest reaches it; it survives only by standing at a node of every swell at once. Survival is a conjunction: that is the whole sieve. Guard: the Mertens density ∏(1−1/p) (amber in the stat line) predicts the survivor fraction without looking at the bridge. Knob: mute swells and watch false survivors flood in. A muted wave is a hallucinated prime.

The record gap 113 → 127: thirteen consecutive planks gone, gap 14, the first of its size. This is your constructive-interference intuition, drawn: across the span every position is under at least one crest, and each broken plank is tagged with the swell that took it. Erdős–Rankin large-gap engineering is exactly this: CRT-tuning phases so many swells crest together. Guard: gap 14 at 113 pre-verified in Node and re-derived by the live sieve here. Note the tag pattern: swell 2 does half the work, the long-wavelength swells (7, 11) arrive only to break the planks nobody else reached.

The sign flip that matters: small gaps are not sharp crests but coordinated calms. At 29 and 31 every active swell (2, 3, 5) sits at a node on both planks simultaneously: residues 29 ≡ {1,2,4}, 31 ≡ {1,1,1}, never zero. Constructive interference only destroys; twins exist where the whole storm misses twice, two planks apart. This is the CH·A habitat law wearing water: offset {0,2} LIVES because no residue row of any swell covers it. Whether the storm grants this calm infinitely often is the twin conjecture (blue door, still open); ≤ 246 is the proven bound on some recurring calm.

Your breakthrough line, drawn: the farther you walk the bridge, the more long-wavelength swells have joined the storm. Swell p breaks its first fresh plank at p². Before that, every multiple of p was already under a shorter wave. So the storm at position x is the set of swells with p ≤ √x, and the active count (amber staircase) is π(√x): the guard, computed from the prime list, not from the picture. More swells, longer wavelengths, thicker interference: mean gap ln x falls out of exactly this crowding.

The duality plate. Left weather station: the storm as swells, one square wave per prime, periodic in n. This is the Euler-product view; it is what PL·I–V drew. Right weather station: the same storm as spectral lines, one tone per zeta zero, cos(γₖ·log x), periodic in log x. This is the explicit-formula view; it is what the ORCHESTRA plays. Amber ticks mark γ₁, γ₂, γ₃ ≈ 14.134 · 21.022 · 25.011. Neither station is the storm; both are exact transcriptions. CH·A and CH·ρ pointed at the same water.