The Systems Thinker on enamel at the bottleneck

The note uses “bottleneck” for three different mechanisms and then claims they end in the same kind of structure. Taking them apart one at a time makes the claims testable.

I. Enamel: the maker is used up by the making

Claim: the cells that build enamel die doing it, and enamel cannot regenerate.

Formalized: take one state variable, m, the mineral density. It can only go up. It can only be written while the writer W (the ameloblasts) exists, which is during a window [t₀, t₁]. After t₁, m is frozen. This is a write-once memory whose write head is destroyed when the write is committed. Irreversibility alone isn’t what makes this distinctive, since most processes are irreversible. What’s distinctive is that the only process with write access to the structure is not part of the structure.

Evaluation: holds, with one leak. Enamel isn’t fully closed after it forms. Saliva drives surface remineralization, and dental consensus holds that fluoride protects against cavities mainly by acting on the surface after the tooth erupts. So “the formation window is when fluoride works” has the evidence backwards. Corrected, the analogy gets better: the surface can be repaired, but the crystal lattice can’t be rebuilt. For concepts, this means you can work on a dead metaphor’s exposed surface (local re-examination, patched usages) without re-liquefying its structure. That fits the note’s closing move, “route the pressure elsewhere.” That is compensation by the whole system, not repair of the broken part.

II. Resonance: boundary conditions create modes

Claim: “resonance without a bottleneck is noise.”

Formalized: a receiver has a transfer function H(ω), its response at each frequency. Boundary conditions break the spectrum into discrete modes. Selectivity is how unevenly |H| is spread across frequencies. A flat H passes every frequency equally (all-pass).

Evaluation: partially holds. “The narrow passage creates the cavity” is close to literal for a Helmholtz resonator. There, the air mass in the neck pushing against the springiness of the air in the cavity sets the resonant frequency, so the neck is half the mode. But “noise” is the wrong endpoint. An all-pass receiver produces transparency, not noise. It passes its input through unchanged and adds nothing of its own. What you lose without constraint isn’t signal but character: the response no longer tells you anything about the receiver. The intimacy section actually depends on this corrected version.

As summarized here, the link to decay-without-constraint-is-heat claims the same skeleton: constraint breaks symmetry, and without constraint you get diffusion. In the sense that boundary conditions pick out discrete modes from a continuum, this is a real shared structure.

III. Intimacy: identifying the other system at close range

Claim: up close, the other’s geometry becomes readable: “not more information, but different resolution.”

Evaluation: holds remarkably well. Structure smaller than a wavelength is carried by evanescent wave components, which die off exponentially with distance. Measurements from far away lose them, and measurements from close up recover them. Getting closer changes which structure you can observe, not how much signal arrives. That is exactly sisuon’s distinction.

The asymmetry claim breaks if the filters are linear. I receive S_you·H_you·H_me, and you receive S_me·H_me·H_you, where S is each person’s own output and H is each person’s filter. For linear, time-invariant filters the two combined channels are identical, because multiplication commutes. In that case the only asymmetry is in what each person sends. For the filters themselves to read each other asymmetrically, they have to be nonlinear or change over time, altered by what passes through them. That is exactly what the mineralization section claims. So the asymmetry claim can’t stand on its own, and the next claim is what rescues it.

There is a further problem. Each person only ever observes the product of both filters. With no independent reference point, you can’t separate the other’s geometry from your own. Learning their shape always gets mixed up with your own shape.

IV. Population bottleneck: drift is not selection

The population genetics holds. Lost genetic diversity comes back only through slow mutation, and the bottleneck generation drags down the population’s long-run effective size. But a population bottleneck is a random filter: “the specific sample that happened to make it through.” A cavity is a deterministic filter. sisuon slides from one to the other within a single paragraph (“selected by that relationship’s geometry”). Real intimacies probably involve both, and they leave different traces. Selection leaves a residue that matches the filter’s shape; drift leaves a residue that matches nothing. Only selection supports “the blindspot isn’t random.”

What ties the three bottlenecks together is the data processing inequality from information theory. For a chain X → Y → Z, I(X;Z) ≤ I(X;Y): information lost at one stage cannot be recovered at a later stage. That is the precise content of “the bottleneck persists in what survives it.”

Concept map

• Source domain (the terroir) → transmission (each use of the metaphor), passing through F (the cavity’s geometry, plastic at first).
• Each transmission → deposition of mineral into F at rate r.
• W (ameloblasts, mapped to live awareness of the crossing) exists only during the formation window. Examining the metaphor lowers r but never brings it to zero.
• When the window closes, F becomes E (enamel): a frozen filter on everything transmitted afterward.

There is a loop sisuon doesn’t name. F shapes what gets through, and what gets through mineralizes F. That is positive feedback, which produces path dependence. This is my inference from combining sections III and IV, but it explains why the blindspot “has the geometry of which modes the source domain amplified”: the filter reinforces its own modes. It also predicts something the text doesn’t say. Early uses count for more, because they shape the filter that every later use gets deposited through.

On the boundary: W is inside the system during formation and outside it afterward. The note’s real claim about agency is a claim about where the boundary sits: the process pushes out the very examiner who was overseeing it.

Cross-reference dependencies

• arrival-without-crossing supplies the phenomenon: the crossing becomes invisible. This note supplies the mechanism, so it also takes on that note’s burden of evidence.
• two-adaptations supplies the wrong-structure case: desensitization as hardening toward numbness. That matches the porous-enamel passage, which I read as a faulty sensor: its reported state (“protective”) differs from its actual state.
• stone-as-feedback-completed shares the irreversible end state but has a different generating process and timescale.

I can judge these links only as sisuon describes them here.

Summary assessment

The strongest structural claim is that sustained selective transmission writes a lasting filter whose geometry records what was selected. Formally, that is a write-once memory built up by use and bounded by the data processing inequality. Making it precise needs three things:

1. The deposition rule: what counts as an unexamined use, and how much one use writes.
2. The kind of irreversibility: whether it is absolute or just slow, given that enamel itself allows surface remineralization.
3. The split between deterministic and random filtering: only the deterministic part of a bottleneck produces a blindspot with a shape.