---

THE ADAPTIVE COHERENCE FRAMEWORK

Formal Architecture, Mathematical Foundations, Adversarial Evaluation, and a Substrate-Independent Model of Adaptive Organisation

Author: Tristan Radford — Cross-Domain Cognitive Systems Architect

Publisher: Riff Systems Ltd  ·  Company No.: 17419314

Version: 1.1  ·  Revision: Scholarly Strengthening Revision

Date: 17 August 2026  ·  Status: Conceptual framework

Canonical publication: riffsystems.org/frameworks/adaptive-coherence-framework/

Abstract

The Adaptive Coherence Framework (ACF) is a substrate-independent architecture for analysing life-like organisation through the persistence of organisational coherence under pressure and the generation of retained, future-modifying responses. The framework does not attempt to redefine biological life. Instead, it supplies a behavioural and relational ontology for comparing biological, artificial, collective, host-dependent, and organisational systems without granting privileged status to any substrate.

ACF formalises coherence, pressure, reaction, regulation, adaptive modification, latent adaptive capacity, system boundary, adaptation locus/topology, hierarchical coherence, temporal horizon, and relational architecture. Its central mathematical discriminator is retained modification of the response-generating organisation: a system exhibits adaptive modification when its response policy changes, the change is retained, and later responses differ because of that retained change. Adaptive outcome is evaluated separately through matched-counterfactual adaptive gain, allowing coherence-enhancing adaptation, neutral adaptive modification, maladaptation, and horizon-dependent reversals to be distinguished.

The framework was developed through adversarial stress-testing, a frozen-ontology classification trial across twenty heterogeneous systems, independent re-application of the ontology, cross-substrate operationalisation, and a synthetic quantitative robot proof-of-concept. The robot test exposed a confound in the initial adaptive-gain formulation: sequential comparisons could mistake accumulated damage for poor adaptation. The formalism was therefore revised so adaptive gain compares old and retained policies under the same starting state, pressure, and temporal horizon. This revision strengthened measurement without expanding the ontology. The present paper consolidates the resulting architecture, formalism, classification protocol, failure conditions, and empirical research programme.

1. Scope and Central Proposition

ACF begins from a behavioural proposition: a life-like system maintains coherence under pressure by generating adaptive responses. The proposition is intentionally substrate-independent. Biological systems may satisfy it, but biology alone does not guarantee adaptive classification; artificial or organisational systems are not excluded merely because they are non-biological.

The framework is not a binary definition of biological life. It is an architecture for analysing adaptive coherence: what is being preserved, what threatens it, where retained modification occurs, how system boundaries are drawn, how nested levels interact, and whether apparent adaptation is actually reaction or fixed regulation.

Core proposition: A life-like system maintains organisational coherence under pressure through system-generated adaptive modification.

Methodological constraint: difficult cases should not be solved by adding exceptions. They should be resolved, where possible, by declaring boundary, locus, evidence, relation, and temporal horizon using a frozen ontology.

2. Frozen Ontology and Formal Definitions

2.1 Coherence

Coherence is the persistence of a system's organisational pattern across time despite internal or external pressure. It is represented through structural continuity, functional continuity, and identity continuity. Coherence does not require material identity: components may be replaced while organisational continuity persists.

2.2 Pressure

Pressure is an internal or external perturbation that tends to destabilise one or more declared dimensions of coherence. Pressure is always indexed to a focal system and temporal horizon.

2.3 Response

A response is a system-linked change produced under pressure. Response alone does not establish regulation or adaptation.

2.4 Regulation

Regulation is a coherence-preserving response generated by a substantially fixed response policy. The state may change and coherence may improve, while the response-generating organisation remains unchanged.

2.5 Adaptive Modification

Adaptive modification is a retained change in the response-generating organisation that changes future responses. Whether the modification improves coherence is evaluated separately.

2.6 Latent Adaptive Capacity

Latent adaptive capacity is preserved organisational potential to resume adaptive activity when viable conditions return. Restartability alone is insufficient; adaptive machinery or organisation must remain preserved.

2.7 System Boundary

The system boundary is the smallest region of organisation whose internal components participate in an integrated coherence-maintaining organisation and within which system-generated adaptive responses preserve identity over time, distinct from external resource dependencies and host infrastructure.

2.8 Adaptation Locus and Topology

The adaptation locus is the organisational level at which retained future-modifying responses occur. A system may contain individual, collective, lineage, population, or multiple simultaneous loci. Adaptation inside a containing system does not by itself establish adaptation by the containing system.

2.9 Hierarchical Coherence

Hierarchical coherence describes how coherence at one organisational level affects coherence at another. Local coherence can support, conflict with, or be sacrificed for higher-level coherence.

2.10 Temporal Horizon

Adaptive value is horizon-dependent. A response may enhance short-term coherence while reducing long-term coherence, or impose immediate costs that preserve longer-term organisation.

2.11 Relational Architecture

3. Frozen Classification Protocol

The protocol is designed to make disagreements traceable rather than intuitive. The ontology is frozen during a classification run.

1. Declare the focal system boundary.
2. Define structural, functional, and identity coherence for that focal system.
3. Identify the relevant pressure.
4. Describe the observed response.
5. Determine whether the response is reaction, regulation, or adaptive modification.
6. Identify the retained modification; if none is demonstrated, do not infer adaptation.
7. Declare adaptation locus/topology.
8. Assess latent adaptive capacity separately from restartability.
9. Evaluate hierarchical coherence at adjacent levels.
10. Declare the temporal horizon.
11. Specify component, dependency, host, parasitic, and coherence-coupled relations.
12. Assign classification and state evidential uncertainty.

4. Adversarial Development and Ontology Convergence

ACF was repeatedly attacked with systems chosen to produce false positives, false negatives, unstable boundaries, multi-level adaptation, and conflicts between nested systems. The significant methodological result was not that every case received an obvious label; it was that later disagreements increasingly localised to boundary, locus, temporal horizon, or evidence rather than forcing new top-level constructs.

5. Frozen-Ontology Cross-Domain Trial

A twenty-system trial was used as a convergence test: no new constructs were permitted. The purpose was not to claim empirical truth for every label, but to test whether heterogeneous cases could be analysed using the existing architecture.

The trial supported ontology convergence in the limited conceptual sense that ambiguities were resolvable without mandatory ontology expansion. It does not constitute empirical validation or prove that every classification is uniquely correct.

6. Independent Re-application and Reproducibility Target

A separate evaluator was asked to classify the same heterogeneous systems using only the frozen constructs. Residual disagreements were reported as boundary, adaptation-locus, or evidence choices rather than as missing ontology. This is encouraging but should be described as initial inter-evaluator support, not a formal reproducibility study. A stronger study requires multiple blinded raters, pre-specified cases, pre-specified evidence, and agreement statistics.

7. Mathematical Formalisation

The mathematics encodes the existing ontology rather than creating a second theory. Symbols denote functional roles that can be instantiated differently across substrates.

7.1 Coherence

C_vec(t) = [ C_s(t), C_f(t), C_i(t) ]^T

When a scalar summary is useful, a pre-declared weighting rule may be used:

C(t) = w_s C_s(t) + w_f C_f(t) + w_i C_i(t), w_s + w_f + w_i = 1

The weights are domain-specific measurement choices, not changes to the construct. Identity coherence should be evaluated against pre-declared identity invariants rather than treated as an unconstrained intuitive score.

7.2 Response Function

R(t) = F( X(t), P(t); theta(t) )

theta(t) denotes the retained variables that determine future response organisation. It may be a regulatory state, learned controller, population composition, or organisational routine, provided its causal role is declared.

7.3 Reaction, Regulation, and Adaptive Modification

Reaction: state changes under pressure without demonstrated coherence-directed control or retained response-policy change.

X(t+1) != X(t), while theta(t+1) = theta(t)

Regulation: a fixed or substantially fixed policy produces a coherence-preserving response.

R(t) = F(X(t),P(t);theta), theta(t+1) = theta(t)

Adaptive modification: the response-generating organisation changes, the change is retained, and later response differs because of it.

Delta theta != 0 AND M(t,tau) > 0 AND Delta F != 0

This is the final formal discriminator. Positive coherence gain is not required to call the modification adaptive; outcome is classified separately, allowing maladaptation.

7.4 Retention

M(t,tau) = || theta(t+tau) - theta(t) ||

M > 0 is necessary but not sufficient. The retained difference must causally contribute to altered later response.

7.5 Revised Matched-Counterfactual Adaptive Gain

The initial formulation compared expected coherence under new and old policies but did not explicitly hold the starting state constant. A synthetic robot experiment showed that sequential comparison can confound adaptation with accumulated damage. The preferred definition therefore conditions both policies on the same starting state X0, the same pressure P, and the same horizon k:

G_A(k) = E[C(t+k) | theta_new, X0, P] - E[C(t+k) | theta_old, X0, P]

Outcome classes are then separated from the existence of adaptive modification:

7.6 Latent Adaptive Capacity

L(t) = Pr(adaptive activity resumes at t+tau | viable conditions return, organisation preserved)

The probability notation is conceptual until a domain supplies measurable conditions and frequencies. Its role is to distinguish preserved adaptive capacity from mere restartability.

7.7 Adaptation Topology and Hierarchical Coherence

A(t) = [ A_1(t), A_2(t), ..., A_n(t) ]

H_ij(t) = partial C_j / partial C_i (or an empirically estimated directional coupling)

These expressions make explicit that adaptation can occur at multiple levels and that increasing coherence at one level can support or damage another. The derivative form is schematic; empirical implementations may use causal-effect estimates better suited to the domain.

8. Formal Propositions

P1. Fixed-policy restoration is regulation, not adaptation.

If theta is unchanged under relevant pressure, coherence-preserving response may be regulation but does not satisfy adaptive modification.

P2. Adaptive modification and adaptive success are distinct.

A retained policy change can alter future response while G_A(k) is zero or negative.

P3. Substrate is neither necessary nor sufficient.

Classification depends on retained response organisation, boundary, locus, and outcome, not on biological composition.

P4. Restartability is insufficient for latent adaptive capacity.

Resumption without preserved adaptive machinery does not establish L(t) > 0.

P5. Adaptation inside a system does not imply adaptation by the system.

System-level adaptation requires retained modification at the declared system-level locus.

P6. Adaptive success can be hierarchically antagonistic.

G_A > 0 for a subsystem can coexist with negative coherence effect on a containing system.

P7. Adaptive value is horizon-dependent.

The sign of G_A may reverse across declared temporal horizons without contradiction.

P8. Valid adaptive-gain comparisons require matched conditions or a defensible causal counterfactual.

Unmatched sequential observations can confound policy benefit with state history.

9. Cross-Substrate Operationalisation Test

The same formal variables were mapped to a bacterium, an adaptive robot, and a corporation. The test was whether the meaning of the constructs had to change across domains. The observables differed, but the formal roles did not.

Provisional result: substrate-specific observables are measurement choices, not ontology changes. The same causal pattern can be stated in all three domains: pressure, retained policy modification, altered future response, and matched-counterfactual coherence outcome. This is an operational plausibility result, not a demonstration from real cross-domain datasets.

10. Synthetic Quantitative Proof-of-Concept: Adaptive Robot

A controlled simulation compared a fixed regulator with an adaptive controller under the same ACF definitions, coherence weights, pressure structure, and classification thresholds. The first sequential version of the experiment produced misleading negative gain because the later trial began from a more damaged state. This exposed a measurement confound. The corrected test replayed old and retained policies from the same starting state under the same pressure and horizon.

The fixed regulator showed no policy change, no retention, no altered future response, and zero matched gain; ACF therefore classified it as regulation. The adaptive controller retained a substantial policy modification, altered later response, and achieved positive matched-counterfactual gain; ACF classified it as coherence-enhancing adaptation.

This is a synthetic proof-of-concept, not empirical validation. Its value is methodological: the formalism discriminated the designed cases, and the experiment identified a weakness in the original gain measurement that was corrected without adding a new ontological construct.

11. Worked Classification Examples

11.1 Basic thermostat

Boundary: thermostat-control loop. Pressure: temperature deviation. A fixed rule restores the variable while theta remains unchanged. Classification: regulation, non-adaptive.

11.2 Dormant seed

Current adaptive activity can be near zero while organised machinery capable of resuming adaptive coherence remains preserved. Classification: latent adaptive individual system.

11.3 Cancer under therapy

Therapy can alter clone/population composition such that later treatment response differs. Adaptive gain may be positive for tumour persistence while host-level hierarchical coupling is negative. Classification: adaptive subsystem at declared clone/population locus; host effect evaluated separately.

11.4 Adaptive robot

A retained controller update that changes later response establishes adaptive modification. Positive matched-counterfactual coherence gain establishes coherence-enhancing adaptation over the declared horizon.

11.5 Corporation

Human learning inside the organisation is insufficient. Collective adaptation requires the modification to be encoded and retained at the organisational locus in routines, procedures, decision rights, structure, or comparable institutional memory.

12. Falsifiable and Generative Predictions

• Adaptive systems should exhibit retained changes that alter subsequent responses under comparable pressure; complex regulators with fixed policy should not.
• If adaptation is attributed to a containing system, retained modification should be detectable at that containing system locus rather than only in adaptive constituents.
• Matched-counterfactual tests should separate policy benefit from accumulated state history; unmatched sequential comparisons should be less reliable.
• Host-dependent adaptive systems should preserve identifiable organisational continuity while outsourcing specific enabling processes to host infrastructure.
• Antagonistic hierarchical systems should show measurable trade-offs in which coherence gain at one level predicts coherence loss at another.
• The sign of adaptive gain should sometimes reverse when temporal horizon changes, without requiring reclassification of the existence of adaptive modification.
• Systems with similar visible behaviour but different retention mechanisms should diverge under repeated perturbation.

13. Failure Conditions

ACF should be considered weakened if repeated, pre-registered tests under frozen definitions produce any of the following patterns:

• Fixed-policy regulation cannot be operationally distinguished from retained future-modifying response organisation.
• Boundary and locus can be redrawn after observing outcomes so freely that incompatible classifications become equally defensible.
• Retained modification can be measured but cannot be causally linked to altered later response.
• Coherence cannot be operationalised without circularly defining it as whatever the focal system happened to preserve.
• Matched evaluators using the same boundary, locus, horizon, evidence, and thresholds systematically reach incompatible classifications.
• Cross-substrate applications require changing the meaning of the core constructs rather than only changing observables.
• Adaptive gain depends primarily on arbitrary analyst choices rather than defensible counterfactual design and horizon declaration.

14. Comparative Positioning

ACF overlaps with established traditions concerned with self-maintenance, autonomy, feedback, viability, learning, and adaptive control. It should not be presented as replacing those traditions. Its proposed contribution is architectural: it combines explicit boundary declaration, retained response-policy modification, adaptation topology, relational roles, hierarchical coherence, temporal horizon, and a matched-counterfactual outcome test in one substrate-independent classification protocol.

14.1 Novelty Claim and Non-Claims

ACF does not claim novelty for feedback, homeostasis, learning, organisational closure, complex adaptation, viability, hierarchical organisation, or system boundaries considered separately. These ideas have substantial prior literatures. The proposed contribution is integrative and methodological: ACF combines explicit boundary declaration, relational roles, adaptation locus/topology, retained modification of response-generating organisation, hierarchical and temporal indexing of coherence, and matched-counterfactual evaluation of adaptive outcome within one substrate-independent classification protocol.

Accordingly, ACF should be evaluated on whether this combined architecture improves classification consistency, causal clarity, cross-domain transfer, and prediction relative to simpler alternatives - not on a claim that its individual ingredients are unprecedented.

14.2 Relation to Cybernetics and Requisite Variety

Classical cybernetics established a rigorous language for regulation under disturbance. Ashby's law of requisite variety concerns the response variety required for a regulator to constrain disturbance-driven outcome variety. ACF inherits the importance of pressure, response, and regulation, but asks a different downstream question: has the organisation generating future responses itself been retainedly modified? A high-variety fixed regulator can therefore remain regulatory in ACF, whereas a lower-level controller can become adaptive if a retained change in its response-generating organisation alters later response. ACF's reaction/regulation/adaptation distinction should therefore be read as complementary to, not a replacement for, cybernetic control theory.^[4][5]

14.3 Relation to Autopoiesis and Organisational Closure

Autopoiesis characterises living organisation in terms of a network of processes that produces and maintains the components and boundary constituting the system as a unity^[1][2]. ACF shares the concern with organisational identity and non-arbitrary system boundaries, but does not require autopoietic self-production as its defining mechanism. Its boundary test instead asks which components participate in an integrated coherence-maintaining organisation, while separately representing dependencies, host infrastructure, parasitism, and coherence coupling. This allows ACF to analyse artificial, organisational, and host-dependent systems without asserting that they are autopoietic.

14.4 Relation to Homeostasis, Allostasis, and Active Inference

Homeostatic and allostatic traditions already distinguish stability from stability achieved through change; allostasis also explicitly accommodates anticipatory and learned regulation. Active-inference accounts further connect homeostatic regulation, learning, perception, and action through generative models^[3][7][8]. ACF is mechanism-agnostic by design. An active-inference agent can instantiate ACF variables, but ACF does not require Bayesian generative models or free-energy minimisation. Its narrower formal claim is that adaptive modification requires retained change in response-generating organisation and that the outcome of that modification should be evaluated separately.

14.5 Relation to Complex Adaptive Systems and Organisational Learning

Complex-adaptive-systems research emphasises interacting adaptive agents, emergence, and system-level pattern formation^[6]. Organisational-learning research likewise distinguishes exploitation of existing capabilities from exploration of new possibilities^[11]. ACF adds an explicit attribution rule: adaptation by constituents is not automatically adaptation by the containing system. The evaluator must declare the adaptation locus and identify where retained future-modifying organisation resides. This is particularly important for corporations, ecosystems, colonies, and multi-agent artificial systems.

14.6 Biological Retention as an Empirical Bridge

Recent work on bacterial memory provides a useful empirical bridge because prior exposure can leave genetic, epigenetic, biochemical, protein-based, or ecological traces that influence future behaviour. Such cases are well-suited to testing ACF's retention requirement without changing its ontology: the biological question becomes whether a measurable retained state functions as theta, persists after the initiating pressure, causally alters later response, and changes matched future coherence. This is a candidate test domain, not evidence that ACF has already been biologically validated.^[12]

14.7 Pre-Registration and Anti-Flexibility Rules

To prevent post-hoc fitting, an ACF empirical test should freeze the following items before outcome inspection. Changing them after results are known must be reported as exploratory re-specification rather than confirmatory analysis.

14.8 Identity Coherence as a Pre-Declared Invariant Set

Identity coherence is a major operational vulnerability if it is left intuitive. ACF therefore treats identity as test-specific but pre-declared. Let I = {i1, i2, ..., in} denote the identity invariants chosen before analysis. These may include organisational relationships, persistent controller identity, lineage markers, legal continuity, or other domain-appropriate invariants. Identity coherence C_i(t) is then evaluated against persistence of I rather than against an unconstrained judgement that the system 'still feels like the same system'.

This does not make identity substrate-independent by pretending every system has the same identity marker. It makes the formal role substrate-independent while forcing each empirical study to expose its identity assumptions.

14.9 Causal Requirements for theta, Retention, and Delta F

A measured state change should not be labelled theta merely because it correlates with later performance. The proposed theta variable must have a defensible causal role in generating later response. Retention M must demonstrate persistence beyond the immediate perturbation, and Delta F must be tested under a relevant matched pressure. Where intervention on theta is impossible, causal inference should rely on a pre-specified identification strategy and the resulting claim should be graded by evidential strength.

15. Limitations and Empirical Programme

The framework is conceptually and formally developed, but its scientific status remains pre-validation. The principal risks are now operational: defining identity invariants without circularity, identifying theta before outcomes are known, demonstrating causal retention, constructing defensible matched counterfactuals, and obtaining reproducible classifications across independent evaluators and domains.

1. Pre-register domain-specific observables for C_s, C_f, C_i, theta, pressure, and horizon.
2. Run matched old-policy versus retained-policy perturbation tests in an artificial system using real hardware or logged controller data.
3. Apply the unchanged classifier to biological stress-memory data, with explicit cell/lineage/population locus declaration.
4. Apply the same architecture to an organisational dataset where routines or procedures can be observed before and after pressure.
5. Use multiple blinded evaluators and report agreement on boundary, locus, relation, and final classification.
6. Compare predictive discrimination against simpler alternatives such as feedback-only or persistence-only classifiers.
7. Only after reliable measurement exists, consider a composite Adaptive Coherence Index; do not treat an arbitrary weighted product as validated science.

15.1 Three-Stage Validation Ladder

The empirical programme should proceed without changing the core classifier between stages. A failure to transfer the same formal roles across substrates counts against substrate independence.

15.2 Inter-Rater Reliability Protocol

A formal reproducibility study should provide multiple blinded evaluators with identical case evidence and a frozen worksheet. Agreement should be reported separately for boundary, regulation/adaptation, locus, relational architecture, and final classification. Cohen's kappa or Fleiss' kappa can be used where category structure permits, supplemented by raw agreement and adjudication logs. ACF is strengthened if disagreements localise to declared evidence uncertainty; it is weakened if evaluators given the same declared conditions repeatedly reach incompatible classifications.

15.3 Comparative Baselines

ACF should be compared against simpler classifiers rather than evaluated only on internal consistency. Candidate baselines include persistence-only, feedback/regulation-only, and constituent-adaptation heuristics. The relevant question is whether ACF's additional structure improves out-of-sample classification stability or prediction enough to justify its complexity.

16. Discussion

The strongest result of ACF development is not that every difficult system receives a universally obvious label. It is that ambiguity has become increasingly localisable. Boundary, locus, relation, horizon, retention, and evidence can be disputed explicitly without forcing the ontology to expand for every counterexample.

The mathematical revision strengthens this property. Adaptive modification is now separated from adaptive success. A system can genuinely modify its future response organisation and still be maladaptive. The matched-counterfactual gain definition also prevents state history from being mistaken for policy quality. These distinctions allow ACF to represent failed learning, treatment resistance, short-term optimisation, and destructive subsystem persistence without contradiction.

The cross-substrate test further suggests that substrate independence is at least operationally plausible: bacteria, robots, and corporations can instantiate the same formal roles while using different observables. The claim remains provisional until real datasets are analysed under pre-registered definitions.

17. Conclusion

The Adaptive Coherence Framework provides a substrate-independent conceptual and mathematical architecture for analysing systems that preserve organisational coherence under pressure through retained changes in future response. Its frozen ontology integrates coherence, pressure, regulation/adaptation, latent capacity, boundary, adaptation topology, hierarchical coherence, temporal horizon, and relational architecture.

The final mathematical core separates existence of adaptive modification from its outcome:

Adaptive modification <=> Delta theta != 0 AND M > 0 AND Delta F != 0

G_A(k) = E[C(t+k)|theta_new,X0,P] - E[C(t+k)|theta_old,X0,P]

Positive G_A indicates coherence-enhancing adaptation over the declared horizon; negative G_A indicates maladaptation. This formulation is explicitly counterfactual and matched on starting state, pressure, and horizon.

ACF is not presented as an empirically validated law, a definitive definition of biological life, or a replacement for established systems theories. It is a converged conceptual/formal framework with a clear empirical programme. Its next decisive tests are real-data operationalisation, cross-substrate replication, blinded inter-rater evaluation, and comparative predictive performance.

Appendix A. Evaluation Outcome Categories

Appendix B. Scientific Status

ACF is presented here as a conceptual and formal framework with a specified validation programme. No numerical self-rating is part of the scientific claim. Empirical validity, predictive advantage, inter-rater reliability, and cross-substrate reproducibility remain to be established through external testing.

Appendix C. Minimal Formal Classification Worksheet

References

Disclaimer & Scope of Application

The Adaptive Coherence Framework (ACF) is a conceptual and formal framework for analysing adaptive organisation across biological, artificial, collective, organisational, and other substrates. It is not presented as an empirically validated scientific theory, a definitive definition of biological life, or a replacement for established systems, cybernetic, biological, or complexity-science traditions.

ACF classifications are conditional on the system boundary, adaptation locus, evidence, relational architecture, and temporal horizon declared for a particular analysis. A classification produced using ACF should therefore be understood as an analytical conclusion under specified conditions, not as a universal or final ontological determination about the system being examined. The framework requires evidential uncertainty to be stated explicitly.

The mathematical formalisation operationalises the framework’s conceptual architecture for analysis and empirical testing. Mathematical representation, persistence, regulation, history dependence, predictive improvement, or retained state alone do not establish adaptive modification. The complete evidential requirements specified by ACF must be satisfied before such a classification is made.

ACF’s adversarial cases, cross-domain classification exercises, synthetic demonstrations, and empirical applications should be interpreted within their stated evidential limits. Existing cross-domain trials support the development and testing of the framework’s internal consistency and reproducibility targets, but do not constitute general empirical validation or establish that every ACF classification is uniquely correct.

Copyright & Permitted Use

© 2026 Tristan J. Radford. All rights reserved.
Published by Riff Systems Ltd (Company No. 17419314).

This publication may be read, cited, quoted in reasonable extracts, discussed in academic contexts, and used for non-commercial research, provided that appropriate attribution and citation are given and the work is not misrepresented.

Commercial Use

Except where permitted by law, commercial reproduction, republication, adaptation of protected expression, incorporation of protected material into software, products or services, or use of substantial portions of this work for commercial AI or machine-learning training, fine-tuning, dataset construction, or similar commercial exploitation requires prior written permission from Riff Systems Ltd.

Rights Administration

Copyright remains with Tristan J. Radford. Riff Systems Ltd is the publisher and administers commercial licensing and permissions for this publication.

Nothing in this notice limits statutory copyright exceptions or other uses permitted by applicable law. No permission is implied for use of Riff Systems trade marks, logos, confidential material, or unpublished research.