The bridge, and three predictions that can fail

Where the theory becomes falsifiable — and where it currently stops

This page is the bridge, and the bridge has not been walked. Everything else on this site is a formal layer: definitions, axioms, machine-checked theorems, and an explicit statement of what they do not cover. None of it says anything about a real system. This page is the only place where the theory touches the world — and it touches it with predictions that are built to be able to fail.

That distinction matters more than any single result here. A theory that cannot be falsified is not a stronger theory. The whole point of writing P1–P3 down in this form is that a laboratory could contradict them.

Where the predictions come from

The programme's downstream assertion is that the core algorithm of mammalian balance under gravity is the infrastructure of the mind's boundary algorithm. That assertion is not derived here. What is derived is a control operator: from least action plus the equation of motion one obtains

u = - k_{p} (κ - κ^{*}) - k_{d} \overset{κ}{˙}

a PD regulator whose proportional gain is the curvature of the potential and whose frequency is the boundary oscillation. This is structurally in agreement with mature balance controllers — PID, and PID with variable stiffness. That agreement is what makes the next three predictions testable rather than merely suggestive.

P1 — the criterion clusters Falsifiable prediction

System	A legged balancing controller, simulated or physical, with an identifiable stiffness gain and an estimator.
Quantity	The operating tension `σ = (c₀/C₀)·κ²`, computed from the measured gain and the measured recoverable-set width.
Criterion	`σ` clusters near `1` rather than spreading over orders of magnitude.
What fails it	A controller population whose `σ` is uniformly far from `1`.

P2 — the frequency is fixed Falsifiable prediction

Statement	The balance-loop frequency is `ω = √2`, with sampling `Δt < T/4 = π√2/4`.
Why it is checkable	This corresponds to a `4×` multiple of the human balance sway (~1–2 Hz) and of control bandwidth (~5–10 Hz).
What fails it	A measured loop frequency inconsistent with that relation to the sway and bandwidth scales.

P3 — the two perturbations couple Falsifiable prediction

System	The same controller as P1, with both an estimator and an actuator that can be perturbed independently.
Quantities	`H`, the entropy of the state estimate, and the actuator stiffness.
Criterion	Across perturbations the two covary as `σ ∝ H⁻²` — equivalently, stiffness `∝ H⁻²`.
What fails it	A measured exponent inconsistent with `−2`, or an absence of coupling.

This is the part that must not be skipped. The mapping holds only after the variable identification — κ with stiffness, and its companions. That identification is precisely the hypothesis under test. It is not derived by the theory. Mapping hypothesis

The theory supplies the form. The assertion supplies the semantics. The verdict belongs to P1–P3 on real systems — not to this page.

Status

Item	Status
The formal layer (Layer I, Layer II, structural theorems)	Machine-Checked
The variable identification κ ↔ stiffness	Assumed, not derived
P1 · P2 · P3	Not run
Any measurement on any real system	None exists

This is what a bridge failure looks like when stated precisely. The mathematics is not in question — it is proved. What is missing is a single measured number on a real controller. That is the whole gap, and it is the reason an earlier submission was rejected for reasons that were about the bridge, not about the proofs.