The computation
Three nodes, x₁ and x₃ at the boundary and x₂ inside (A22 §4):
E(x₁, x₂, x₃) = ½k₁(x₁ − x₂)² + ½k₂(x₂ − x₃)².
Holding the boundary fixed, ∂E/∂x₂ = 0 gives x₂* = (k₁x₁ + k₂x₃)/(k₁ + k₂). Then x₁ − x₂* = k₂(x₁ − x₃)/(k₁ + k₂) and x₂* − x₃ = k₁(x₁ − x₃)/(k₁ + k₂), so
E_eff(x₁, x₃) = ½ · k₁k₂/(k₁ + k₂) · (x₁ − x₃)².
The composite acts on its boundary as one spring of stiffness k₁k₂/(k₁ + k₂) joining x₁ and x₃, which are not joined directly in the network. The elimination is a Schur complement, or Kron reduction, the operation behind effective resistance.
With memory. If the interior is not fast, it cannot be replaced by its equilibrium. For a linear system ẋ = Ax + By, ẏ = Cx + Dy, solving for y and substituting gives
ẋ(t) = Ax(t) + Be^{Dt}y(0) + ∫₀ᵗ Be^{D(t−s)}Cx(s) ds.
The boundary now depends on its own past through a kernel set by the interior’s dynamics D (A23 §IV). When D is fast and stable the kernel concentrates near s = t, and the integral tends to −BD⁻¹Cx(t), which is the static elimination. So the composite presents its boundary with one of three things (A23 §IV):
- an approximately instantaneous constraint, when the interior relaxes fast;
- a history-dependent constraint, when its memory is significant;
- further active modes, when the interior cannot be eliminated at all.
All three are constraints on the boundary’s admissible trajectories. A memory kernel restricts which boundary histories can occur just as an instantaneous coupling restricts which boundary states can. What separation of time scales provides is the condition under which the instantaneous approximation is adequate:
| what | gives |
|---|---|
| interconnection | an effective behavioural constraint at the boundary |
| sufficient separation of interior and boundary time scales | an approximately memoryless form of that constraint |
So the claim does not privilege memoryless structure over structure sustained through history. Baez and Fong’s black box shows the same thing exactly. With inductors and capacitors it is a frequency-dependent relation at the terminals, exact at every frequency, and needs no separation (NOTE-1223).
The record holds this family of reductions as Mori–Zwanzig (LIT-1245, read, and its neighbours).
The reading
A22: “To a system interacting only with the boundary, the entire three-node organization behaves like a direct coupling between nodes 1 and 3. … The interior has effectively become apparatus. … An internal organization of interactions becomes a condition governing what external interactions are possible.” A23: “This is not simply a useful linguistic redescription. The composite has an externally measurable boundary relationship derived from its internal organization.”
Baez and Fong (LIT-1644, read in NOTE-1223) give the general result for passive linear networks. Black-boxing a circuit to the relation it imposes at its terminals is a functor, so the external behavior of a composite is the composite of external behaviors. Willems’s behavioural form (LIT-1660) states the same projection for any interconnection. The external behaviour is the projection of the joint behaviour onto the variables an observer can reach.
When the reduced constraint is of a new kind
The chain above reduces to another pairwise coupling, and for scalar networks that is always so: the Schur complement of a graph Laplacian is a graph Laplacian on the boundary (Hansen and Ghrist, Thm 4.2). For cellular sheaves it is not. Hansen and Ghrist’s example (THEORY-1090; Fig. 4.1, p. 20) is a star with stalk ℝ² at the centre and ℝ at three leaves. The centre’s maps are the first coordinate, the second, and their sum. Every consistent assignment is fixed by the centre’s value (p, q), so the boundary values are (p + q, q, p), and the reduced constraint is
v₁ = v₂ + v₃,
a relation among three boundary values. No sheaf on the boundary vertices, whose conditions are all pairwise, has exactly that space of consistent assignments; any that admits it admits more. “The problem is that the internal node allows for constraints between boundary nodes that cannot be expressed by purely pairwise interactions” (p. 20).
So composition followed by reduction can force a change in the type of relational structure an adequate description needs, not only in its parameters. The parts’ relations were all pairwise. The composite’s effective relation is not, and a vocabulary of pairwise relations cannot state it. This is the strongest result the claim holds. It has a direct counterpart in luria, where a record’s relation scheme meets what a composed organization requires and cannot express it (CLAIM-419, CLAIM-479).
Its limit: it concerns what a class of mathematical descriptions can express. It does not show that a new fundamental interaction has arisen, or that the three-way relation is anything but the effective trace of pairwise relations through a hidden interior. Whether composition generally creates effective structure that needs a richer relational vocabulary than its parts is QUESTION-283.
Why it bears on the record
- CLAIM-477 holds that the constraints are realized too, and that relata are realizations of them. This is a case with nothing interpretive in it. The constraint between x₁ and x₃ is realized by the interior and changes when the interior’s springs do. It is not detached from the world, and it is not fixed.
- CLAIM-478 says separation makes constraints. The memory case shows what separation does: it makes the composite’s constraint approximately instantaneous. Without it the composite still constrains its boundary, by a relation over histories.
- CLAIM-471 is about gluing local descriptions. This is the other half A22 found missing: “Sheaf theory tells us how locally defined information can be made compatible. But what happens when we physically connect two systems and then observe them as one system?”
What it does not claim
That the composite is an individual. Effective boundary behaviour is one of the properties A23 §VI listed. The others (persistence, autonomy, boundary robustness, constraint maintenance and self-modification) need their own evidence. Nor does the effective coupling add a new fundamental law. It is an effective restriction, and is no more fundamental than the parts it is realized by (A23 §XII).
Note of 2026-10-27 (evening): the readings of the U15–U23 works
- The computation is in the literature. Baez and Fong’s Example 2.9 (NOTE-1223, pp. 14–15) is two resistors in series. With k = 1/r, its interior potential and boundary power are this claim’s x₂* and E_eff. General elimination is their Thm 3.3(vi) with Lemma 3.4: eliminating a node s adds c_is·c_js/Σ_k c_ks to each coupling (THEORY-1089). The paper’s form is dissipated power, not spring energy. The algebra is the same.
- The memory case needs a qualification. With inductors and capacitors the black box is exact at every frequency, as an impedance over ℝ(s), under zero initial conditions. A composite therefore always presents its boundary with an exact relation, and no separation of time scales is needed for that. What separation decides is whether the relation is approximately instantaneous, which is how the title puts it. The term in y(0) is outside Baez and Fong’s assumptions.
- Black-boxing is a functor (THEORY-1063, Thm 1.1): circuits with the same behaviour are interchangeable in every composite. That supports the defeater’s “there is none” for passive linear networks.
- The constraint can be of higher order than any interaction that realizes it. Hansen and Ghrist (NOTE-1233, §4.2, Fig. 4.1, Thm 4.3; THEORY-1090) show that eliminating an interior vertex of a cellular sheaf can leave the boundary a constraint that no pairwise sheaf expresses, where for scalar networks the result stays pairwise (Thm 4.2). So composition can make a constraint not only new but of a new kind.
- Spivak (NOTE-1225) composes by summing pulled-back potentials and never eliminates an interior. Rigid constraints appear in his paper only as limits of stiff potentials (p. 9).
Note of 2026-11-01: the relaxation time decides whether a dependence counts as constitution or causation
The constitution debate read today draws the line between constitution and causation at simultaneity. “Simultaneous changes cannot be causally related” (Baumgartner, Casini and Krickel, NOTE-1494, p. 6), and a revealing intervention must change part and whole “simultaneously” (their H(ii)). This claim’s chain makes that line depend on the description. If x₂ equilibrates fast, the boundary’s constraint is the memoryless E_eff, x₂ follows x₁ and x₃ at once, and the interior’s dependence is synchronic, on the constitution side. If x₂ relaxes slowly, eliminating it leaves a memory kernel: x₂’s response follows the boundary’s change, on the causation side. Whether the dependence is constitution or causation, in the debate’s sense, is then fixed by the time resolution at which the description treats x₂’s relaxation as instantaneous. CLAIM-528 part 3 rests on this. Krickel 2018 (LIT-1925, unread) reportedly reads mutual manipulability causally, between a component and temporal phases of the phenomenon. If so, it is the same move made from the other side.
No status, title or warrant changes.