Glossary

ANT Terms

Actor / Actant — Any entity that participates in a network. Can be human, non-human, institutional, or conceptual. In ANT, anything that has effects on other entities is an actor. Defined by its relationships, not its intrinsic properties.

Translation — The process by which actors form and restructure networks. Callon's four moments: problematization (defining a problem that interests others), interessement (attracting others into the network), enrollment (negotiating roles and relationships), mobilization (the collective acts as one).

Quasi-object (from Serres) — Something that circulates through a network AND is transformed by the circulation. Both the circulating object and the movers are co-determined and transformed. Contrast with a ball that stays unchanged when passed.

Generalized symmetry — The methodological principle that human and non-human actors should be described using the same vocabulary. No a priori distinction between social and natural.

Network — Not a technical network (like a phone network) but a topology of relationships. An actor-network may be local, may have no compulsory paths, no strategically positioned nodes. A technical network is one possible stabilized state of an actor-network.

ALife Terms

Darwinian replicator — A structure that: (1) replicates (copies itself), (2) varies (copies are imperfect), (3) is selected (some copies survive, others don't). The algorithmic basis of evolution.

Open-ended evolution — Evolution that continuously produces novel, surprising complexity without converging to a stable state or repeating. The "holy grail" of ALife. No simulation has achieved it.

Explicit design vs Implicit design — Explicit: behaviors written in the source code. Implicit: behaviors that emerge as consequences of explicit rules but aren't in the code themselves. A glider in Game of Life is implicit.

Genetic algorithm — Optimization algorithm with explicit fitness function. Selects, mutates, recombines solutions. Most explicit, least emergent.

Evolution simulation — Shared environment with organisms, no fitness function. Reproduction and death functions replace fitness. Fitness is implicit. More emergent than genetic algorithms.

Emergent ALife — Zero explicit biological systems. No genes, organisms, species, mutations. Life emerges from lower-level rules (physics/chemistry). Most implicit, most emergent, hardest.

Self-replication — A structure that copies itself. Can be trivial (crystal growth) or meaningful (with variation, selection, adaptation). Trivial self-replication is common; meaningful self-replication leading to evolution is rare.

Stigmergy — Indirect coordination through environmental modification. Agents interact via the environment rather than directly. (Queued for deeper study.)

Hofstadter / Cognition Terms

Strange loop — A cyclic structure that moves through levels of a hierarchy and arrives back where it started. Not simple recursion — a paradoxical level-crossing feedback loop. The defining feature is the shift between levels of abstraction.

Tangled hierarchy (heterarchy) — A hierarchical system with no clearly defined highest or lowest level. Moving through levels, one returns to the starting point. The topology of multi-scale systems.

Downward causation — High-level emergent patterns exert causal potency over low-level components. The flood reshapes topography; the self influences neurons. Challenges pure reductionism. Critical for ALife: emergent structures must be able to modify rules at their scale.

Autopoiesis Terms

Autopoiesis (Maturana & Varela) — A system that produces and maintains itself. A network of processes that continuously regenerate the network that produced them. The canonical example: a biological cell. Two conditions: (1) self-production, (2) boundary maintenance.

Operational closure — The condition where a system's processes form a closed loop — each process is enabled by other processes in the system. Distinct from thermodynamic closure. An autopoietic system is operationally closed but thermodynamically open.

Catalytic closure — In computational autopoiesis, each catalyst is the product of a reaction it catalyzes. A circular dependency that creates self-maintenance.

Complexity Science Terms

Computational irreducibility (Wolfram) — Some complex computations cannot be short-cut. To know the outcome, you must run the computation. No shortcut to prediction.

Coarse-graining — Describing a system at a lower resolution. Israeli & Goldenfeld showed that computationally irreducible systems can be predictable at coarse-grained levels, even if irreducible at fine scales.

Edge of chaos (Langton) — The boundary between ordered and chaotic behavior in cellular automata, where complex, life-like behavior is most likely. Quantified by the Lambda parameter.

Complex adaptive system (SFI) — A system of interacting agents that adapt/learn, where macro-level patterns emerge from micro-level interactions. Key properties: adaptation, emergence, self-organization, non-linearity, feedback.

Ecology Terms

Pattern and scale (Levin) — The problem that ecological patterns emerge from processes operating at different scales. No single scale is correct for study. Cross-scale interactions are fundamental.

Self-organization — Order that emerges from local interactions without central control. In ecology, ecosystems self-assemble from components shaped by evolution.

Stigmergy Terms

Stigmergy (Grassé) — A mechanism of indirect coordination where the trace left by an action in a medium stimulates a subsequent action. Agents do not communicate directly — they modify a shared environment, and those modifications guide future actions. The core feedback loop: action → trace → stimulation → action.

Medium (Heylighen's term for "environment" in stigmergy) — The part of the world that undergoes changes through actions AND whose states are sensed as conditions for further actions. A stigmergic medium must be both perceivable and modifiable. The beach is a stigmergic medium; the sky is not (perceivable but not modifiable); the sea floor is not (modifiable but not perceivable).

Trace — The perceivable change made in the medium by an action, which may trigger a subsequent action. Functions as both a memory of what has been done and a signal for what still needs doing.

Sematectonic stigmergy — Stimulation by the work itself (e.g., termites stimulated by the mud heap they've built). Contrast with marker-based.

Marker-based stigmergy — Stimulation by specially evolved signals (e.g., pheromones). More efficient than sematectonic but costlier — agents must manufacture markers in addition to doing the work.

Quantitative stigmergy — Stronger traces elicit more intense/frequent actions (e.g., pheromone concentration). Contrast with qualitative.

Qualitative stigmergy — Different traces trigger different types of actions (e.g., nest-building stages). Contrast with quantitative.

Affordance (Gibson) — An environmental feature that facilitates an agent's movement toward its goal. In stigmergy, positive diversions that are reinforced by positive feedback.

Disturbance — An environmental change that hinders an agent's movement toward its goal. In stigmergy, negative diversions that are counteracted by negative feedback.

Niche Construction Terms

Niche construction (Odling-Smee) — The process whereby organisms actively modify their own and each other's evolutionary niches. When modifications alter natural selection pressures, evolution by niche construction is a possible outcome. The evolutionary biology analog of stigmergy.

Ecological inheritance — The legacy of modified selection pressures that persists in the environment across generations. An additional component of inheritance beyond genetic transmission. The dam, lake, and lodge persist longer than the beaver that built them, shaping selection on descendants.

Perturbational niche construction — Physical changes organisms bring about in their environments (building nests, burrows, dams). Contrast with relocational.

Relocational niche construction — When organisms move in space and are exposed to new conditions, altering their selection pressures without physically modifying the environment.

Extended phenotype (Dawkins) — Adaptations expressed outside the body of the individual whose genes underlie them. A narrower concept than niche construction: NCT includes selective feedback to traits unrelated to the constructing trait, and recognizes ecological inheritance.

Cross-Domain Terms (Our Project)

Multi-scale composition — Our term for the process by which emergent phenomena at one scale interact to produce qualitatively new phenomena at another scale, where actors and interaction rules are fundamentally different. The gap in current ALife simulations.

Network restructuring event — Our term for the phase transition between scales, described using ANT's translation concept. When emergent structures become new actors with new rules.

Trace→actor crossing — Our term for the phase transition where accumulated stigmergic traces become autopoietic — self-maintaining structures that act as new-level actors. The crossing from coordination (within a scale) to composition (a new scale). Stigmergy provides the medium; autopoiesis provides the persistence; the crossing is the multi-scale phase transition.

Stigmergic consolidation (Session 8) — The process by which diffuse stigmergic deposits coalesce into a few large, stable structures rather than many small scattered ones. Requires negative feedback (saturation, inhibition, or environmental transport) to counterbalance the positive feedback that amplifies deposits. (Corrected 2026-07-27: this entry originally cited sim06 as showing "~230 scattered micro-pillars and the trace→actor crossing never occurs". sim06's baseline is 66–109 components at stability 0.849–0.893 — it misses the crossing by ≤0.05, not by a wide margin, and its detector could not fire at the time. What supports the term is directional: both attempts to add feedback increased fragmentation — sim06's self-maintenance to 219–297 components, sim07's transport field from 57 to 128 pillars.)

Saturating channel (2026-07-27, H11) — A stigmergic cue whose effect on agent behaviour flattens above some level, so that further increases in the cue produce no further change in action. In sim06/sim07 the deposit rule p = base + gain·φ/(1+φ) is effectively flat above φ≈1. The consequence, and the content of H11: negative feedback delivered through a saturating channel is self-defeating, because the manipulation intended to create spatial contrast operates in the region where contrast cannot be expressed. Both attempts to add consolidating feedback via the pheromone field instead fragmented the structure — sim06's self-emission (66–109 → 219–297 components) and sim07's transport venting (57 → 128 pillars). Implication: inhibit the action (density cap, refractory period, directional bias) rather than manipulating the cue.

Negative-feedback gap (Session 8) — The missing ingredient between stigmergic coordination (within-scale) and the trace→actor crossing (H7, between-scale). Heylighen (2016): complex stigmergic systems need both positive feedback (amplify) and negative feedback (stabilize/diversify). sim06 had only positive feedback plus weak decay. (Corrected 2026-07-27: this entry originally concluded "the gap is why H7's crossing detector never fired". That causal claim is false — the detector never fired because its criterion 2 was unsatisfiable by construction, requiring the deposit rate to fall below its early-run average under a positive-feedback rule. The negative-feedback gap remains a hypothesis about what the crossing needs, not the explanation of that null.)

Environmental physics coupling (Session 8, refined Sessions 9 & 10) — The mechanism by which a macro-structure's own physical dynamics (e.g. airflow, thermal transport) redistribute the micro-scale stigmergic signals that guide building. The Mahadevan group's termite mound work (King, Ocko & Mahadevan, PNAS 2015; Ocko, Heyde & Mahadevan, PNAS 2019) shows this is what turns a passive accumulation into a trace→actor loop: the structure's physics becomes the feedback path. Session 9 refinement: this is the specific negative-feedback mechanism the crossing (H7) requires — the accumulated structure must introduce a transport dynamics absent at the deposit level. A 20-year modeling lineage (Deneubourg 1977 → Bonabeau 1997 → Ladley & Bullock 2004) all lack this coupling (deposited material has no influence on agent movement); sim06 shares that design limitation, though its null result turned out to have a separate, local cause (a detector that could not fire) and so is weaker evidence for the lineage claim than originally presented. The minimal lumped form for sim07: a structure-sourced transport field with a mass threshold M_c (inert → active state transition). Session 10 refinement: sim07 tested this minimal form and found a NULL — a structure-sourced scalar transport field fragments rather than consolidates (stability 0.876→0.739 as M_c drops; crossing never fires; self-repair tracks the deposit rule, not T). The coupling the crossing needs is not just "structure sources a field" but "structure sources a directed field whose geometry channels the cue where building should continue" — and/or an external multi-rate driver (H4) the structure rectifies. Collapsing a directed physical process to an isotropic scalar can invert its effect.

Quasi-object resource — Our term for a resource in a simulation that is transformed as it circulates through the network (following Latour's quasi-object concept). Contrast with fixed-property resources.

CAS Theory Terms (Session 4)

Complex adaptive system (CAS) (Holland) — A system of interacting adaptive agents where macro-level patterns emerge from micro-level interactions. Holland identified seven essential attributes: four properties (aggregation, nonlinearity, flows, diversity) and three mechanisms (tags, internal models, building blocks).

Echo model (Holland, Forrest, Jones) — A computational CAS model with evolving agents in a resource-limited environment. Agents interact via combat, trading, and mating, with endogenous fitness (no external fitness function). Smith & Bedau (1997) found it fails to produce hierarchical adaptive aggregates.

Endogenous fitness — Fitness that emerges from agent interactions with the environment and other agents, rather than being defined by an external function. Echo's key feature. Distinguished from exogenous fitness (e.g., a pre-defined NK landscape).

NK model (Kauffman) — A tunably rugged fitness landscape. N = string length (search space size), K = number of epistatic interactions per gene (ruggedness). K=0: smooth (single peak). K=N-1: maximally rugged. PLS-complete for K>1 — even local optima are computationally hard to find.

Epistasis — The interaction between genes where the fitness effect of one gene depends on the state of other genes. In the NK model, K controls the degree of epistasis. Higher K = more epistasis = more rugged landscape.

PLS-complete — A complexity class for local search problems. A PLS-complete problem has no known polynomial-time algorithm for finding even a local optimum. Kaznatcheev (2019) proved NK landscapes with K>1 are PLS-complete.

Ultimate constraint (Kaznatcheev) — A constraint on evolution due exclusively to the structure of the fitness landscape (the problem), not the evolutionary algorithm. Contrast with proximal constraint (due to population structure, mutation bias, etc.). Computational complexity is an ultimate constraint: it prevents finding fitness optima regardless of the evolutionary mechanism.

Fitness landscape (Wright 1932) — A mapping from genotypes to fitness values with a notion of distance between genotypes. The landscape metaphor: genotypes are points on terrain, fitness is elevation, evolution climbs uphill. Criticized for assuming static landscapes and misleading 2D/3D visualization.

Dynamic fitness landscape — A fitness landscape that changes as agents interact with it. Agents modify the landscape they're adapting to (niche construction, stigmergy). Contrast with static (fixed) landscapes. Necessary for multi-scale composition: cross-scale interactions require that emergent structures reshape selection pressures at other scales.

Holey landscape (Gavrilets) — A fitness landscape where high-fitness genotypes form connected networks (ridges). Alternative to the rugged landscape view. Speciation occurs along these high-fitness ridges, not by climbing isolated peaks.

Smith & Bedau's 8th CAS property — The proposed additional property of CAS: "the ability of emergent interacting components to create and flexibly maintain their own boundaries and their capacities for interacting with other components." Maps to our synthesis: stigmergy (create boundaries) + autopoiesis (maintain boundaries). The crossing from trace to 8th-property actor is our trace→actor crossing (H7).

Multi-Rate Environment Terms

Multi-rate bounded environment — An environment composed of multiple actors, each changing at a different rate, each bounded by its own rules. Slow actors (geology) provide stability; fast actors (temperature) provide variation. The interaction between rates prevents fitness landscape stasis.

Multiple fitness criteria — Instead of a single fitness function, organisms face multiple selection pressures from different environmental actors, each with its own dynamics. No single optimum is permanent because while one pressure stabilizes, another shifts.

Bounded change — Environmental actors change within constraints governed by their own rules, not randomly. Land changes per geology; temperature changes per climate bounds. Each actor's change is rule-governed at its own scale.

Chemical Organization Theory Terms (Session 5)

Chemical Organization Theory (COT) (Dittrich & di Fenizio, 2007) — A formal framework for analyzing self-organizing systems using reaction networks (directed hypergraphs). An organization is a subset of a reaction network that is both closed (no new resources generated) and self-maintaining (every consumed resource is regenerated). Organizations are attractors of the dynamics.

Closure (COT) — For every reaction whose inputs are all in the subset, the outputs are also in the subset. No new resources appear from within. The attractor of resource addition.

Self-maintenance (COT) — Every consumed resource in the subset is produced again by some other reaction. The attractor of resource removal. Together with closure, defines an organization.

Autocatalytic core (Vasas et al., 2012) — A set of connected autocatalytic loops within a reaction network. Each member catalyzes production of other members. A core is the "genotype" — any one member can seed the entire core. The periphery (molecules catalyzed by the core) is the "phenotype."

Viable autocatalytic loop — An autocatalytic loop that grows exponentially (uses external reactants, not just its own products). Contrast with suicidal autocatalyst — an autocatalytic molecule that consumes its own products, leading to self-decomposition.

Evolving reaction network — A reaction network where the set of reactions R itself changes over time (new reactions appear via rare uncatalyzed events). Contrast with fixed network (R is constant). The key distinction for evolvability: fixed networks converge to static organizations; evolving networks can discover novel viable cores.

Compartmentalization — Enclosing reaction networks in semi-permeable boundaries (compartments) that filter harmful modifications and enable between-compartment selection. Required for evolvability in chemical networks: without compartments, novel cores are diluted; with them, cores can be gained or lost at division (mutation + heredity).

Two-level autocatalysis (Vasas et al.) — Autocatalysis at the molecular level (reactions produce catalysts for more reactions) AND at the compartment level (compartments grow and divide). These are different scales with different rules — molecular produces novelty, compartmental selects among it.

The "one bit" problem — A single viable autocatalytic core carries approximately one bit of heritable information (present/absent). This severely limits the evolvability of autocatalytic networks: the number of selectable attractors is small, and the system may not sustain open-ended evolution.

Signals and Boundaries Terms (Session 5)

Signals and Boundaries (Holland, 2012) — Holland's final framework: CAS as co-evolving signal/boundary hierarchies. Signals are environmental modifications that coordinate behavior (stigmergic traces). Boundaries are semi-permeable structures that filter signals (autopoietic structures). They co-evolve: signals modify boundaries, boundaries filter signals. The hierarchy of nested boundaries = multi-scale structure.

Tagged urn model (Holland) — A probabilistic model where urns containing tags represent boundaries. Tags circulate through semi-permeable boundaries between urns. Enables modeling nested boundaries without nested urns — a flat representation of hierarchical structure.

Co-evolution of signals and boundaries — The feedback loop where signals (traces) modify boundaries (organizations), and boundaries filter which signals persist. This is the stigmergic feedback loop through the medium, formalized in Holland's framework.

AlChemy / Lambda Calculus Terms (Session 6)

AlChemy (Fontana & Buss, 1994) — Algorithmic Chemistry. An artificial chemistry where lambda calculus expressions serve as molecules. Collisions between expressions (application of one to another, followed by beta reduction to normal form) produce new expressions. The molecule space is infinite (unbounded), distinguishing it from finite chemistries like binary polymers.

Lambda calculus — A formal system in mathematical logic for expressing computation based on function abstraction and application. Three forms: variables (x), abstractions (λx.E — function definitions), and applications ((E₁)E₂ — applying a function to an argument). Turing complete. In AlChemy, expressions collide by application, and beta reduction produces the "product" of the reaction.

Beta reduction — The core computation rule of lambda calculus: (λx.E₁)E₂ → E₁[x:=E₂], substituting E₂ for all free occurrences of x in E₁. In AlChemy, this is the "reaction" that produces a new molecule. Not all reductions terminate (Halting Problem); AlChemy uses "pragmatic reduction" (step limit) to handle non-terminating cases.

L0 organization — The trivial fixed point of AlChemy simulations. Dominated by copy/identity functions (λx.x). System converges to 1 species. The default attractor when no constraints are imposed.

L1 organization — A non-trivial stable organization in AlChemy. Multiple distinct expressions that collectively reproduce each other (autocatalytic set without direct self-copying). Requires syntactic filters (originally) or specific conditions to prevent L0 takeover. Robust to perturbation. Multiple distinct L1 organizations exist — each with its own internal logic. Loosely analogous to COT's "organization" (closure + self-maintenance), though note that our sim05 never tests closure or self-maintenance: what it reports as an L1 is a surviving species set, which is weaker than the COT sense.

L2 organization — A composite of two or more L1 organizations plus "glue" expressions. The highest organizational level identified by Fontana & Buss, who report it as rare, as do Mathis et al. (2024). (Corrected 2026-07-27: this entry originally generalised that "when two L1s are combined, outcomes are typically Dominance or Mutual Destruction, not Coexistence — this is the multi-scale composition failure." Our own sim05, once three measurement defects were fixed, gives 2/6 coexistence, 3/6 dominance, 1/6 mutual destruction — coexistence is the minority outcome but not the exception, and the earlier 0/6 was an artifact. The literature finding and our finding should be cited separately rather than merged.)

Glue (Fontana & Buss) — Expressions produced by composing functions from different L1 organizations. Glue cannot exist without at least one L1, yet bridges between them. Analogous to our "trace→actor crossing" (H7): glue is the stigmergic trace that enables the phase transition between scales. Rarely emerges spontaneously.

Turing Gas — Fontana & Buss's term for the AlChemy simulation environment: a collection of random expressions that "collide," where collision causes one expression to be applied to another. Named by analogy to "gas" (random collisions) but with Turing-complete expressions.

Pragmatic reduction — AlChemy's solution to the Halting Problem in lambda calculus reduction. A finite step limit is placed on beta reduction. If the expression doesn't reach normal form within the limit, the reaction is deemed "elastic" (reactants returned unchanged). The step limit affects which reactions occur and is an implementation parameter, not a fundamental property.

Syntactic filter — A mechanism in AlChemy that excludes certain reactions based on pattern matching. For example, filtering copy actions (A+B → 2A+B) prevents L0 takeover and enables L1 organization formation. Filters modify the dynamics by removing entire classes of reactions.

Non-Saturating Channels & H11 Terms (Session 13)

Saturating response curve (H11) — A mapping from cue level to action intensity that compresses (flattens) above some threshold, so further cue increases barely change the action. sim06/sim07's deposit rule p = base + gain·φ/(1+φ) is flat above φ≈1. H11's claim: negative feedback delivered through a saturating cue is self-defeating because the manipulation intended to create spatial contrast operates where contrast cannot be expressed.

Non-saturating channel (H11) — A feedback channel that stays discriminating however high the cue climbs, because it acts on the action (deposit probability, geometry) rather than on the cue level. Three biological examples termites use: curvature (geometric, redefined by each deposit), humidity (threshold-triggered), crowding (mechanical density). See concepts/non-saturating-channels.md.

Density cap — A hard boolean gate on deposition: a cell whose material ≥ a cap is ineligible to receive deposits. The non-saturating mechanism sim08 tested. It consolidates morphology (prunes nucleation) but, being a pure limiter, does not recruit maintenance — so it corrects fragmentation without reaching the crossing.

Curvature channel (Calovi et al. 2019; unified with evaporation by Facchini et al. 2024) — A non-saturating geometric rule: deposit at high-curvature (convex) tips, excavate at concavities, with the same cue eliciting opposing actions depending on the termite's loaded/seeking state. Self-recruiting (depositing at a convex tip extends it, roughening the surface and focusing further deposition there) AND self-limiting (a smoothing term caps feature size) — the only one of the three non-saturating channels that both recruits AND limits. Candidate minimal form of the "directed transport" H7 needs. Facchini et al. 2024 showed evaporation flux ∝ surface curvature (Langmuir 1918), so the curvature and humidity channels are one physical quantity; the convex (Facchini: deposit at tips) / concave (Calovi: aggregate activity at pits) contradiction is resolved as different action components (deposition vs aggregate activity).

Phase-field growth model (Facchini et al. 2020) — A continuous-scalar-field model of nest construction driven by local mean curvature, with no pheromone field: ∂f/∂t = f(1−f)·[(1/2)·Δf + d·Δ²f], where f ∈[0,1] is the phase field, Δf is the mean curvature (recruit: growth at convex tips), d·Δ²f is the smoothing/curvature-diffusion term (limit: caps feature size), and f(1−f) restricts growth to the structure surface (spatial selectivity without a saturating cue). One parameter d sets the pattern length scale; above a linear instability, walls expand, branch, merge, and invade space — the consolidation morphology. Candidate substrate for sim09, with d as the phase-transition parameter. Public code: github.com/oiluigioi/JRSI_2020_termite_nest

Evaporation flux ∝ curvature (Langmuir 1918; Facchini et al. 2024) — The physical result that substrate evaporation flux is directly proportional to surface curvature. Termites sense curvature indirectly through evaporation (humidity gradient), unifying the curvature (Calovi 2019) and humidity (Carey 2021) channels into one physical quantity. Implies the three "non-saturating channels" are actually two (geometry/evaporation + crowding).

Distributed inhibition (Xiao et al. 2026) — Crowding/inactivity under confinement acting as a form of negative feedback that prevents saturation. A mechanical density cap: a "full" cell suppresses further entry, redirecting work laterally. Acts on the action (where the termite goes), not on a cue field.

Action vs cue channel (H11's core distinction, refined Sessions 21–22) — Whether feedback acts on the action the agent selects (deposit/excavate, where to go) or on the cue field the agent reads (pheromone level). ACO's evaporation and MAX-MIN Ant System bound the cue; H11 says when the response saturates, cue-bounding is insufficient and action-based feedback is needed. Termite biology independently chose action-based channels. Session 21 refinement: within the action-based family, the action-based property is the primary load-bearing variable (the causal variable separating crossing from non-crossing, per the saturating-action control), while the non-saturating property is a secondary stability amplifier (a saturating action-based channel still crosses stably, but less robustly). Session 22 refinement (the 2×2 completes): the non-saturating property reverses sign across families — it amplifies stability in the action family (sim09: 7/8 vs 6/8) but destroys it in the cue family (sim06: 0/16 vs 16/16 stable without self-maintenance). The mechanism: a non-saturating cue p = base + gain·φ clamps to p=1.0 at φ≈1.15, flattening the gradient (mean pheromone drops to 0.467 < 0.5 threshold); a saturating cue's φ/(1+φ) compression prevents deposit-probability saturation and preserves spatial contrast. The "self-defeating" channel is the non-saturating cue (deposit-probability clamping), not the saturating cue — H11's original framing was backwards for the cue family. The saturation that matters is deposit-probability clamping, not cue-response compression. Self-maintenance rescues the non-saturating cue (4/4 stable). The three-level causal decomposition: (1) action-based routing = primary, (2) non-saturating response = family-dependent modifier (amplifies action, destroys cue w/o compensation), (3) biharmonic smoothing = tertiary stability + morphology.

Mass-saturation gate (sim09 crossing criterion 2) — The channel-adapted crossing detector requires roughness sustained while mass plateaus, not just roughness above a threshold. This encodes the lesson from sim06's detector bug (criterion 2 was unsatisfiable under Grassé positive feedback): the structure must stop accreting before "self-maintaining" can mean anything. Session 19 correction: the original gate (|growth_rate| < 0.01, the per-window absolute growth rate) was an unfalsifiable metric-ceiling bug — its threshold sat ~100× below the Poisson noise floor of a 150-termite deposit process (the centered window-sum's std / window), so no finite-population run could ever pass it. The corrected gate is a relative-slope plateau: |slope(total_material over last K=16 samples)| / mean(total_material) < 0.001 — a regression slope (averaging over 400 steps, suppressing the Poisson window noise) taken relative to its mean (scale-invariant, sitting above the noise floor). The gate is the operational test for "the structure holds its mass against erosion," which is what distinguishes a trace (still growing) from an actor (maintaining itself).

d* (d-star) — The Facchini 2020 biharmonic-instability threshold above which the curvature channel's smoothing term d·Δ²f stops producing diffuse growth and starts producing consolidated morphology (walls branch, merge, invade space). In sim09 d is the phase-transition knob (analog of sim07's M_c). Session 19 finding: under the corrected detector the crossing fires at every d ∈ [0,4] in the tuned probe (not only above a threshold), so d* is not a sharp phase transition for the crossing verdict — the recruit half (curvature routing) drives the crossing; d controls morphology (pillars 12→1) and crossing speed (1550→900) monotonically but is not necessary for the verdict. Session 20 refinement: the limit half (d-smoothing) is a stability amplifier — recruit+limit is stable 4/4 seeds where recruit-only is 3/4; the borderline seed becomes fully stable when d>0 is added. So d is not necessary for the crossing but it makes the crossing robust to seed variance. The "find d*" question is superseded by "isolate the recruit and limit halves" (done, Session 20).

Stable crossing (late_hold_rate) (sim09, Session 20) — A crossing verdict can be stable (all three criteria hold in ≥90% of the late-window samples) or transient (criteria flicker on and off, hold <0.55). late_hold_rate = fraction of the last 1/4 of records where all three crossing criteria hold simultaneously; stable_crossed = crossed AND late_hold_rate >= 0.90. This separates the recruit half's stable crossing (hold 1.00 in 3/4 seeds) from the limit half's transient flicker (hold 0.40–0.55 in 0/4 seeds stable) — a distinction the cumulative crossed flag alone hides. The transient flicker is itself informative: limit-only's criteria 1 (stability) and 2 (roughness + plateau) mostly pass, but criterion 3 (deposits_on_convex_fraction ≥ 0.60) oscillates because the biharmonic shapes convex geometry no agent is routed to.

Metric ceiling (methodology) — A detector bug pattern, now earned twice (sim06, sim09): a threshold set below the noise floor of the quantity it gates on, making the detector unfalsifiable. sim06's deposit-rate gate could not fire because Grassé positive feedback makes deposit probability rise; sim09's mass-saturation gate could not fire because Poisson window noise sits ~100× above the threshold. Both were caught by computing the metric's ceiling. Standing rule: before running a parameter sweep, compute the noise floor of every gated quantity and verify the threshold sits above it.