Deterministic emergence: how the organs are derived from measured DNA, with no tuning

This whitepaper follows one discipline: LOCK, Derive, Gate. The organ identities and their developmental order are read out from measured human-DNA stacking stiffness γ — SantaLucia nearest-neighbour ΔG37 over the real promoter, never fitted — and every page reproduces bit-for-bit (2×sha256 identical). No number is chosen to hit a target.

The program rests on three commitments that turn “reproducibility” into a structure. First, no tuning: every quantity is either a measured input or a value derived from one, never adjusted to match an answer. Second, determinism: a fixed seed makes the engine emit a byte-identical result twice over (2×sha256 identical, 6a68bc48…). Third, explicit grading: [V] verified in-simulation, [L] calibrated or cited, [O] open with a stated obstacle, [H] hypothesis — so the reach of every claim is legible.

A new theory earns trust by how tightly it is constrained, not by how much it explains. The point of this section is to make the constraints explicit, so that the results in the chapters that follow read as a grounded derivation rather than an illustrative simulation.

What γ is, and how it is measured

γ is the DNA stacking stiffness of a gene: the negative mean of the nearest-neighbour stacking free energies ΔG37 along its human proximal promoter (SantaLucia 1998 nearest-neighbour thermodynamics, the standard for DNA base-pair stacking). It is a property of the real, published genomic sequence — not a knob. The five master-gene values used here (PAX6 1.511, RAX 1.4541, EYA1 1.3638, SOX2 1.4573, TAS1R3 1.5555) are read once and cached, so γ reproduces offline and bit-for-bit.

The no-tuning discipline

Every quantity in the program is one of two things: a measured input (a sequence-derived γ, a cited physiological threshold, a classical-optics constant) or a derived value computed from those inputs by a fixed rule. No quantity is ever chosen to make a result come out right. The developmental order, for instance, is simply argsort(γ); the cochlear compression exponent is fixed by a normal form, not selected.

LOCK → Derive → Gate

• LOCK — inputs are frozen: the measured γ values, the cited anchors, and the substrate equations are read-only for the run.
• Derive — the engine computes the consequences deterministically (node order, transducer bistability, the Hopf exponent, the disease law).
• Gate — a stress battery and a determinism check must pass before any result is written; the canonical HTML then renders only what passed.

Bit-for-bit reproducibility

The engine is run twice in separate processes and the two outputs are hashed; the program is accepted only if the hashes are identical. The signed-off result for this volume is 6a68bc48… (2×sha256 identical), and the whole site is generated from that frozen result with a fixed build date, so the canonical HTML itself rebuilds byte-for-byte. Anyone can re-run it: from the package root, python repro/run_all.py.

The grading vocabulary is what makes it falsifiable

Each claim carries a grade, and the grades mean exactly what they say:

Because the boundaries are explicit, the program is refutable: if the taste organ were specified earliest, the order claim would fail; if a transducer were a graded rather than an all-or-none gate, the unification would fail; if the measured compression exponent were far from 0.3333, the Hopf account would fail.