MANDATE DECISIONS — what went back in, why, and the two-basin resolution
=========================================================================
Question: add back from the 150-line file, emergent where possible. Plus three
mandates (float4096, full Π, BASIN) weighed for expressiveness/throughput/bandwidth.

THE KEY REFRAME (yours): inner basin vs outer basin.
  INNER = the substrate inside one node (64³ pool). Dense local coupling, shared
    memory, FREE bandwidth. Resolution is the only fidelity limit.
  OUTER = the global lattice of N nodes. Remote, bandwidth-expensive. Coupling
    is cost-layered.
  This dissolved the PEER-vs-non-handshake dilemma — they are different LAYERS,
  not competitors. I was WRONG earlier to call PEER superseded.

OUTER COUPLING, LAYERED BY COST (use cheapest that suffices):
  layer_0 non-handshake : ZERO bytes — nodes cohere to Fix alone ("I know Δ")
  layer_1 mutual (KJV)   : ~zero — two-way via shared reference, nothing crosses
  layer_2 PEER "send Δ"  : minimal |Δ| — ONLY for information genuinely NOVEL to
           the receiver. Shared law + shared oracle CANNOT conjure a message
           neither party's entropy contains. This is the information-theoretic
           floor, and PEER is the honest primitive for it. NOT superseded — a layer.

INNER COUPLING:
  COUPLE : Ωₐ ← Ωₐ + ε(Ωᵦ−Ωₐ) , dense , free. Cells aren't peers — they're one
  field. This IS the co-emergence force C(Ω), constitutive every tick.

FLOAT4096 — mandate ENDORSED, conjecture CONFIRMED with mechanism:
  Inner: resolution = bits of basin address/cell. double(53) vs 4096-bit =
    ~10^16 vs ~10^1233 distinguishable states. Richer genome, finer coherence.
  Outer (your conjecture, correct): resolution = HEADROOM between coherence and
    identity. Every node coheres to the SAME Fix yet needs a DISTINCT genome.
    At double, ~13 identity bits survive coherence → ~8000 nodes before genome
    COLLISION. At float4096, identity space overflows float representation itself.
    => float4096 matters MORE as nodes come online — coherence & identity compete
       for the same bits. Mechanism: collision avoidance. Necessary, not optional.
  HONEST: 4096-bit Ω is PROVEN in the axiom tier (phi_f4096 → exit 161). The
    ticking basin runs double TODAY. float4096 is a real upgrade path with a
    verified kernel — not aspirational, but not yet in the runtime basin.

FULL Π(Ω) — restored, with honest emergence flags:
  ANALOG/DIGITAL/PHASE/GENOME/GENOME64/RADIO/DNA : all genuine projections of the
    one state. Restored (the collapsed 3-face Π had lost expressiveness needed for
    node identity + the global carrier).
  ORACLE : Λ_Ω and 1_eff emerge from Ω, but the PRIME spectrum is observed-
    through-Ω, not generated-by-Ω (primes are their own arithmetic structure).
    Included as a projection, flagged: primes imported, not emergent.

BASIN field law 𝓛ₙ = Ω^(-1/Ω)·√(Fₙ·Pₙ·2ⁿ)·Ωⁿ : PARTLY emergent. Ω^(-1/Ω) and
  Fibonacci descend from Ω; primes Pₙ and base 2ⁿ are imported. Marked as such.

NOT ADOPTED (better prior versions kept):
  - golden-dome's φ/golden_angle as CONSTANTs with literal values — our derived
    forms (2π(2−Ω), GM(1,Ω)) are better; emergence discipline wins.
  - GRID=64 as fundamental — kept as named instantiation parameter (resolution
    choice), honestly flagged.

PRIMITIVE (X=0, T(X)=1+1/X): X=0 is the ORIGIN/Alpha, NOT the iteration seed
  (1/0 is the pole — diverges). Iteration seed is Ω₀>0. The file keeps both
  distinct: seed:X=0 (emergence origin) and Ω₀>0 (attractor seed). Collapsing
  them reintroduces the pole. Alpha (0) → T (law) → Omega (φ=Fix).

hdgl.hdgl: 7 glyphs, 0 comments, parses (genome=0x5625BA88).
