link3 Substrate Report — What the First Candidate Proved

A mechanics-honest accounting of what the link3 engine (links, sessions, ν spectrum) produced, what it couldn't produce, and what the next substrate candidate needs. Everything below was measured on 64³–256×20×20 WASM grids.

The physics program this validates

The substrate's discipline: only local rules — a cell reads its own registers and channel neighbors, moves at most c per tick, no global state, no action at a distance. Then let the measurements decide — which features of effective physics are forced by the causal layer itself, vs which are contingent. The bridge (cell = ℓP or 0.8 fm) is a label, not a mechanic; c is a causal rate (1 hop/tick), not a dimensional constant.

What survived the ratio tests

Measured ratio	Substrate	Nature	Honest read
Isotropic dilution	1/r² mid-range	1/r²	✓ exact — dimensionless
ν ∝ iE	linear	E=hν	✓ right proportionality
Mediator range ∝ 1/ν	λ halves per 2ν	ℏ/mc Compton	✓ the strongest match
Fusion yield ∝ ν	E-weighted drest	E=mc²	✓ energy-conserving
p/E on absorption	∝ ν recoil	p=E/c	✓ momentum ∝ energy
Back-to-back annihilation	5120=5120 antipodal	e⁺e⁻→2γ	✓ perfect balance
Beam divergence	linear cone, emission-set	diffraction	✓ structure matches
Binding/cluster trend	defect/lump grows N=2→6	binding/nucleon rises	✓ trend only
Hub saturation	per-client declines, total plateaus	capacity limits	✓ shape matches
Time dilation	6% absorb drop when moving	rate dilution	✓ small but real
Exchange correlator	⟨φ₊·φ₋⟩ = −0.81	two-point signal	✓ real volley measured

What failed or is missing

Quantum signature	Result	Honest diagnosis
Stimulated emission	1944 regardless of drive ν	✗ no laser condition — pure spontaneous decay
Photoelectric threshold	noisy plateau, no cutoff	✗ absorption is graded, not quantum-gated
1/n² spectral series	lines diverge 8,12,20, cascade ~2/3 steps	✗ linear/geometric, not quadratic levels
Bond distance falloff	flat to gap=80+	wire = lossless channel — no field falloff until transit loss added
Lump self-propulsion	now 0 — parked flux	fixed the drest leak; lumps are pinned

The architecture that worked

• Registers: lREST (parked mass), lDR (winding/momentum), lIE (charge), lPH/lPHNU (photons + energy tag), lSST/lSESS (session state), lBOND (depth register)
• ν is the multi-role register: energy (E=hν), mediator mass (range ∝1/ν), momentum (p=E/c), production value (fusion ∝ν) — exactly how photon frequency works in nature
• Sessions: UDP→LISTEN→SYN→ESTAB→CLOSED — the bond is a transaction protocol, not a force field
• Bonding: antipodal complementarity (facing windings exchange) + broadcast announcement (nodes emit on all ports until a peer answers)
• Parking: wound mass is structure, not transport — meso3i's "parked flux stays put" was the retention fix that made bound states stable

What the next candidate needs

1. Internal level structure — 1/n² needs multiple stable configurations of a bound pair, not a bigger charge bank. The lump needs discrete excitation modes, not just iE
2. Stimulated emission — a photon of matching ν should trigger emission (coherent amplification). Requires state-occupancy mechanics, not just decay rules
3. A sharper work function — compat=0 rejection exists but isn't a clean threshold; the absorb path needs quantum gating
4. Channel loss for bond range — the wire is lossless; a real short-range force needs mediator persistence to limit the range (ν-decay does this — keep it)
5. Keep everything that works — the conservation ledger, ν multi-role, session lifecycle, parked mass, broadcast announcement. These are proven mechanics, not guesses

The honest verdict

link3 isn't the final substrate — it's the proof that a substrate can produce real emergent physics from local causal rules. Its contribution is the mechanics inventory: which rules force real structure (conservation, ν, exchange, parking) vs which gaps need different dynamics (quantum statistics, level structure). The next candidate inherits the proven mechanics and tries a different internal structure where the quantum signatures live.