Loop Network Theory
The Dual Nature of Internal Cycle-Space Structure
Feedback loops do two things at once: they detect problems (sensing) and move information (transport). These are not two separate functions sharing infrastructure. They are the same flow doing both things simultaneously.
Break any edge in a loop and you lose both capabilities at once. Cancel the sprint retrospective and you lose both the problem detection AND the knowledge distribution that the meeting provided. There is no way to preserve one without the other.
This duality is not a metaphor. It is a direct consequence of the Hodge decomposition: cycle-space flow is divergence-free, meaning it has no sources or sinks. It circulates. Circulating flow simultaneously carries comparison information (sensing) and distributes state (transport). The math forces the duality.
Loops in the Hodge Decomposition
Every flow on the contact graph decomposes into three orthogonal components. The cycle-space component is the one that creates loops.
Definitions
The number of independent loops in an organism's contact graph. Measures the dimension of the cycle space — how many independent sensing/transport channels the organism maintains.
The rate at which a loop can detect perturbations in its covered region. Depends on loop length (detection period) and spatial extent (detection range).
The number of independent cycles containing a given edge. Measures how much sensing bandwidth is lost if that edge is removed.
Each independent loop contributes to Bcx. The total complexity budget from loops is:
The Sensor-Transport Duality Theorem
Theorem 7 — Loop Duality
Loop Duality
Let L be a simple cycle in the contact graph G of an organism O, with cycle length |L| and edge-weight function w. Let fL be the cycle-space flow on L. Then:
L detects any perturbation p with support intersecting L within one full circulation period .
Show derivation from priors ▸
Step 1. Let p be a perturbation at node v on L. The circulating flow fL passes through v once per period τL (the sum of edge traversal times around L).
Step 2. After the perturbation, the flow value at v shifts by δf(v). This perturbation propagates around L with the circulating flow.
Step 3. The flow returns to v after period τL carrying the accumulated signal from the entire cycle. By comparing fL(v) before and after one full circulation, the perturbation is detected.
Step 4. If the perturbation has support intersecting ANY edge of L, the circulating flow encounters it within one period. Detection is guaranteed for all perturbations with non-zero overlap with L. QED.
L distributes information from any node on L to every other node on L at zero marginal Bth cost.
Show derivation ▸
Step 1. By the divergence-free property of fcyc, flow entering any node of L equals flow leaving. The flow carries information encoded in its magnitude and modulation.
Step 2. Since fcyc is orthogonal to fgrad (the Bth-counted gradient flow), cycle-space transport incurs zero marginal Bth. It is not source-to-sink transport but closed-path circulation.
Step 3. Every node on L receives the flow once per period. Information injected at any node reaches all other nodes within one circulation. QED.
Sensing and transport are the SAME flow. Destroying one destroys the other. There is no configuration that preserves sensing without transport, or transport without sensing.
Show derivation ▸
Step 1. Suppose edge e in L is removed. L is no longer a cycle — it becomes a path with two endpoints.
Step 2. The divergence-free condition fails at the endpoints: they become source and sink. The former cycle-space flow becomes gradient flow (contributes to Bth, not Bcx).
Step 3. Without circulation, there is no return to origin for comparison — sensing is destroyed. Without the cycle, there is no complete distribution — transport is destroyed.
Step 4. Both capabilities depended on the same structural property (the closed cycle supporting divergence-free flow). Destroying the cycle destroys both. QED.
Optimal Loop Topology
How many loops, and what shape?
Optimal Number of Loops
The optimal cycle rank balances sensing gains against coordination costs. It is found at the SEP where marginal coherence per loop equals marginal Bcx cost per loop.
Organism is “blind”
Detectable perturbations go undetected
Fragile to surprises
H_min = sensing
Sensing bandwidth matches perturbation rate
Marginal loop CL = marginal loop cost
Exchange equalization satisfied
SEP
Organism is “over-coordinated”
Detects everything, cannot respond
Budget consumed by circulation
Paralysis
Optimal Loop Length Distribution
Not all loops are equal. The optimal distribution follows a scale-free power law: many short loops (local sensing), progressively fewer long loops (global sensing).
Show derivation ▸
Step 1. Poke arrival rate per unit area is approximately uniform (A4: pokes are local, no preferred location). Total pokes per tick scales with organism area ~ Rcascade2.
Step 2. A loop of length l covers area ~ l2 and detects pokes in that area within period τ(l) ~ l.
Step 3. Uniform detection coverage requires n(l) · l2 ≥ Rcascade2 at each scale, giving n(l) ≥ l-2.
Step 4. Per-loop cost Bcx(l) ~ l. Total Bcx at scale l is n(l) · l ~ l-1.
Step 5. SEP adjustment (balancing Bth and Bleak marginal returns) shifts the pure coverage exponent α = 2 to the range [1.5, 2.5]. QED.
Supporting Theorems
Minimum Viable Loop Network
An organism achieves Sel > 0 iff it has at least one loop (β1 ≥ 1) with detection period less than the fastest perturbation arrival rate.
Show derivation ▸
Step 1. By A9, perturbations exist. By A10, the organism must adapt. Adaptation requires detection (sensing).
Step 2. A tree graph (β1 = 0) has no cycle-space flow and therefore no loop-based sensing. All sensing must come from gradient flow (Bthtransport from boundary to interior) — slow and expensive.
Step 3. A single loop provides a self-contained sensor at zero marginal Bth. The organism fails selection if perturbations arrive faster than its detection response time. Therefore β1 ≥ 1 is necessary. QED.
Loop Failure = Blind Spot
If loop L is broken (an edge removed), the organism loses sensing capability in the region covered by L. If no other loop covers that region, the organism acquires a permanent blind spot there.
Show derivation ▸
Step 1. By Part C of Theorem 7, sensing and transport on L are inseparable. Removing an edge destroys the cycle (β1 decreases by 1).
Step 2. The formerly cycle-space flow on L becomes gradient flow. The organism can still transport information through the broken path via Bth, but loses zero-cost sensing.
Step 3. By A9, perturbations in the uncovered region continue to arrive. By T1 (editor opacity), the organism cannot detect what it cannot sense. Perturbations accumulate undetected until they cascade. QED.
Loop Emergence from Traffic
In a growing organism (T2), new loops form preferentially along the highest-traffic edges. The busiest transport paths become the sensory network.
Show derivation ▸
Step 1. Growth adds edges. A new edge between two already-connected nodes creates a cycle (β1 increases by 1).
Step 2. By A3, the probability of a new edge forming between nodes u and v is proportional to their interaction strength. High-traffic edges have highest interaction strength.
Step 3. When a cycle forms along high-traffic edges, the cycle-space component absorbs part of the former gradient flow, converting expensive Bth into efficient Bcx.
Step 4. This conversion is energetically favorable when sensing gains exceed Bcx cost. The edges carrying the most information benefit most from becoming cycle-based sensors. QED.
Cascade Vulnerability
An organism's loop network is vulnerable to cascade failure iff there exists an edge e with — a single edge whose removal destroys more than half the sensing bandwidth.
Show derivation ▸
Step 1. By A7, the organism can sustain bounded loss of sensing before Sel drops below zero.
Step 2. The survival threshold is approximately β1/2: the organism can lose half its loops and still detect enough perturbations via the remaining loops.
Step 3. An edge with K(e) > β1/2 creates a single point of failure that exceeds this threshold. Its removal simultaneously breaks more than half the independent cycles.
Corollary: Robust organisms distribute their loops across independent edges. The optimal topology has K(e) approximately uniform across all cycle edges. QED.
Loop Failure Taxonomy
Three types, mapped to the three budgets
An edge in the loop is destroyed or becomes too expensive. The cycle breaks. β1 decreases by 1.
Cause: Scaffold instability, node removal, deliberate pruning
Effect: Blind spot (Theorem 7B). Transport reroutes to gradient flow, increasing Bth.
Examples: Network cable cut. Employee quits. Synapse pruned.
The loop exists topologically but cycle-space flow has ceased. Edges intact, no information circulates. Structurally present, functionally absent.
Cause: Bcx starvation. Information production rate below circulation threshold.
Effect: The organism BELIEVES it has sensing coverage but does NOT.
Examples: Meeting nobody prepares for. CI pipeline nobody reads. Nerve the brain ignores.
The loop crosses a domain wall. Signal leaks at boundary crossings. The loop functions but with degraded signal-to-noise.
Cause: Domain wall instability. Misalignment between sub-domains.
Effect: Sensing bandwidth reduced. Perturbations detected but information corrupted.
Examples: Report passing through three departments. Feedback loop across microservice boundaries.
Editors Require Loops
Element III x Element V coupling
Hidden editors (Element V) detect misalignment and correct it. But detection requires sensing, and sensing requires loops. An editor without a loop is a blind corrector — applying random corrections that increase Bcxwithout increasing CL.
An editor can only correct misalignment in regions covered by at least one functional loop. Regions with no loop coverage have no editor capability, regardless of how many editors nominally exist.
The tightest loop in an organism is the sensing-correction cycle:
Falsifiable Predictions
Each prediction specifies its falsification condition
P7.1 — Organizations: Killing Loops Kills Both Functions
Eliminating a feedback loop (sprint retros, customer surveys, etc.) causes BOTH sensing degradation (problems undetected longer) AND coordination degradation (knowledge distributed slower). These cannot be separated.
Falsified if: an org eliminates a feedback loop and observes improved sensing or coordination.
P7.2 — Organizations: High-Traffic Paths Become Loops
The busiest informal communication paths (most-active Slack channels, most-forwarded emails) preferentially formalize into recurring feedback loops (become scheduled meetings, dashboards, automated reports).
Falsified if: high-traffic paths systematically remain informal and never formalize.
P7.3 — Organizations: Single-Point Cascade
An organization with a single critical communication node (one person connecting two departments) loses BOTH transport and sensing when that person leaves — not just information flow, but problem visibility.
Falsified if: removing the bottleneck causes no sensing degradation.
P7.4 — Biology: Sensorimotor Loop Duality
Disrupting motor output of a sensorimotor loop (paralyzing a limb) degrades sensory acuity in that limb's domain, not just motor capability. Active sensing requires motor action as part of the sensing cycle.
Falsified if: paralysis has zero effect on sensory acuity in the affected domain.
P7.5 — Biology: Neural Loop Length Distribution
Cortical loop lengths follow an approximate power law with exponent : many short-range loops (local circuits), progressively fewer long-range loops (thalamocortical, interhemispheric).
Falsified if: cortical loop length distribution is uniform, exponential, or non-power-law.
P7.6 — Software: CI/CD as Inseparable Loop
Organizations that separate CI from CD into independent processes (no shared cycle) experience increased Bth (redundant transport) and degraded sensing (test results disconnected from deploy decisions).
Falsified if: separating CI from CD improves both coverage and reliability.
P7.7 — Markets: Price Controls Break Dual Function
A market with price controls loses BOTH the allocation function AND the regulator's ability to sense true supply/demand. The price signal that would reveal the state has been suppressed.
Falsified if: price controls degrade allocation but preserve sensing of true conditions.
Connections to Other Research Seeds
Under B_cx starvation, loops are the first element pruned (they ARE B_cx). Under B_leak dominance, loops are the highest-priority investment (sensing is critical for boundary defense).
The anti-binder signal travels through loops. If the loop connecting binder to anti-binder breaks, the organism loses its most important peripheral sensing. T5 implicitly requires functioning loops.
The bounce threshold CL_bounce depends on internal loop density. Without sufficient loops, an organism cannot achieve scaffold independence because internal transport relies on external routing.
Loops crossing domain walls have degraded signal (Type 3 failure). Wall permeability determines how much sensing bandwidth is preserved across the wall.