Preprints.org logo
Preprint
Concept Paper

This version is not peer-reviewed.

Model G: Geometric Formalization of Information Spaces with Intrinsic Coherence

Submitted:

12 July 2026

Posted:

14 July 2026

You are already at the latest version

Abstract
Enterprise management systems have historically served as an externalized Master (in McGilchrist's sense): they hold the memory of business acts and their relational logic, but through a procedural paradigm dating to Pacioli (1494) marked by redundancy, rigidity, vulnerability, and reliance on external verification. Until now, humans have supplied the missing Emissary function, correcting inconsistencies through judgment the system itself lacks. Generative AI represents the first externalization of the Emissary, but lacking access to a Master, it invents data to produce coherent output. Rather than equipping AI agents to reconstruct an internal World Model at each query — a computationally expensive approach — this work proposes building an explicit, external Master accessible to generative AI. We formalize this as Modelo G, an algebraic framework separating a declarative Concrete System, which encodes business entities, documents, data, and rules as a relational structure, from an Abstract System, the generative AI, which acts as a grammatical translator interpreting natural language and documents against that structure without needing domain-specific training. Because logic resides in the Concrete System rather than in the model or the prompt, governance, operation, and audit collapse into a single continuous cycle, reversing the usual growth of the gap between formal rules and actual practice. Developed and validated in the domain of enterprise management systems, the model generalizes to other domains of structured information, and achieves through external declarative structure what architectures such as DeepSeek's Engram pursue internally: separating static knowledge from reasoning to avoid its constant reconstruction.
Keywords: 
database theory
;  
intrinsic coherence
;  
geometric models
;  
formal verification
;  
information spaces
Subject: 
Computer Science and Mathematics  -   Data Structures, Algorithms and Complexity

INTRODUCTION

1. Origin of the Management Paradigm

Part 1. A Look Back. Where We Come from Determines Where We Are Going

  • 1. Birth of the Documentary Accounting Model 
At the end of the 15th century, at a time of intense banking expansion and thanks to the appearance of the printing press, Friar Luca Pacioli, in his 'Summa de Arithmetica' (1494), Tractatus XI, systematized double-entry bookkeeping —already used in the accounting books of the Massari di Genova in 1340— and its principles of Journal, Ledger and trial balance through a process of sequential entries. In a period of low calculation capacity, propagating the data item to other records and accumulated states made the immediate, or almost immediate, consultation of information possible. Two characteristics should be highlighted: duplication of the data item and rigidity in the propagation flow.

2. Intensification with the Industrial Revolution 

The Industrial Revolution multiplies the scale of production, exchange and internal organization of companies. The increase in labor, materials, operations and administrative movements makes direct or intuitive control unviable. In this context, the accumulative model of recording and synthesis becomes indispensable, since direct calculation from the original documents was too costly in time and effort. For this reason, the model of Journal, Ledger and accumulated records extends to the other areas of the company.

3. Organizational Formalization in Modern Management 

In this context, the great theories of modern organization appear. Taylor proposes the decomposition, measurement and standardization of work; Weber formulates rational bureaucratic organization based on hierarchy, rules, defined competences and persistent documentation. Both approaches respond to the need to govern increasingly large and complex business structures.

4. Consolidation of the Propagative Paradigm 

Given that the available technology was limited in calculation capacity and immediate information processing, the efficient model consisted of propagating economic and administrative facts from the original documents toward accumulative records and synthesis states through predefined administrative procedures.

Conclusion 

In this work we call this management model the Accumulative Administrative Paradigm Based on Eager Propagation, or simply Eager Propagation. By eager propagation model we mean one in which each recorded fact drives the anticipated update of derived states, so that a later query does not calculate from origin, but accesses previously accumulated results, making the information available within the required time frame.

Part 2. The Traits of the Model

The Taylor-Weber Model, or Eager Propagation, presents four chained structural traits: redundancy, predefinition, vulnerability and the need for verification.

1. Redundancy 

Each recorded fact does not remain only in its document of origin, but is propagated toward successive documents, records, master files and accumulative files. Relevant information therefore ceases to exist at a single point and becomes repeated across multiple representations. This redundancy is not accidental, but a direct consequence of the historical need to obtain results without continuously recalculating from origin.

2. Predefinition 

Propagation is not possible unless the procedures and routes that carry the data item from its document of origin toward the next document or toward the corresponding accumulative file have been previously defined. The effectiveness of the paradigm depends not only on the accumulation of information, but also on the prior determination of the channels through which it must circulate. This turns it into a rigid system, costly to modify and difficult to adapt when a need arises that was not contemplated in its initial design.

3. Vulnerability to Error and Its Correction 

A transcription or latency error propagates throughout the system partially, absolutely or within accumulated records, making it difficult to identify and to detect information incoherence. Correcting an error means propagating the correction outside the formulated administrative flow, which means duplicating parallel processes that, if they do not adjust to the processes established in the administrative flow, generate more incoherence among the data.

4. Need for Verification 

As a consequence of redundancy, predefinition and vulnerability to corrections, the system requires permanent mechanisms of reconciliation, audit and control. Coherence no longer depends only on the correct recording of the original fact, but also on the correspondence among its different documentary and accumulative materializations. Verification thus appears not as an essential property of the model, but as the organizational response required to contain and minimize its structural incoherence.

Conclusion 

Consequently, eager propagation accelerates the availability of information, but transfers the structural cost to maintenance, reconciliation, audit and correction. Redundancy, which was originally a technically necessary response, also becomes the main source of rigidity and risk of incoherence in the paradigm.

Part 3. Digitalization

  • 1. Calculation Center: The Automation of Labor 
Before the 1950s, calculation centers operated strictly under the Taylor-Weber model, where large teams of operators, called "calculators", executed fragmented and specialized processing tasks. In this environment, management was already an analog algorithm: each operator received a data item, applied a predefined rule and propagated the result to the next link in the bureaucratic chain.
The initial computerization process was limited to replacing human "calculators" with machines specialized in two basic functions: computing (counting) and sorting (filing).

2. Taylor-Weber and Codd's Relational Model 

Edgar F. Codd in 1970 designed his model with the needs of data processing in mind, in environments where economic facts must be recorded, propagated in a controlled way and maintained in accumulated states for efficient consultation.
In this way, the relational model provided a formal and powerful language for implementing the logic of Eager Propagation: definition of structures (tables), predefined relations between them and integrity mechanisms that guarantee coherence during propagation. Far from breaking with the Taylor-Weber paradigm, it offered it a more solid mathematical and computational basis for its automation.

3. Microcomputers (CP/M, MS-DOS) 

Microcomputers, later called PCs, democratized physical access to computing. Their storage and processing limitations left no room for recalculations or sophisticated internal verifications. Taylor-Weber survived because it was the only model that fit into 64 KB of RAM and into low-cost economic storage systems (floppies), thus consolidating itself as the universal solution in digital management.

4. Distributed Architectures and Adaptation Layers 

The Taylor-Weber model requires a continuous flow where decision, calculation and recording alternate heterogeneously in an indivisible sequence. Distributed architectures, by contrast, fragment that continuity by nature.
To manage that friction, layers are created that keep communications with distributed systems open for when the Taylor-Weber sequence requires them.
Each layer adds its own mechanisms of state, memory and synchronization, which require energy and resources. A system must support both the management system and the entire environment surrounding it, where the requirements for the system are often greater than those for what is purely management.

Conclusion  

Digitalization does not represent a rupture with the eager propagation paradigm, but its automation and expansion. The effort is directed toward making new technologies fit into the Taylor-Weber model, adding layers to manage the friction between a model conceived for linear environments and infrastructures that are not linear.

Part 4. Externalization.

  • 1. Construction and Adaptation of the Management System 
In an administrative environment prior to computerization, when a change in management was required, the new rule was transmitted through the chain of command and each operator adjusted the part of the procedure that corresponded to them. Management knew and decided the procedures and their operation, was sovereign in its decisions and did not need to know many of the base operations in order to govern management.

2. Digitalization 

With digitalization, the company turns to external agents, even when it has an IT department. These agents do not only provide technology. They also determine how management is formalized, which processes are possible, which data must be entered, which routes information must follow and at what moments the system can be modified. The company no longer simply changes an internal rule: it must request, parameterize, develop, integrate, validate or wait for a software adaptation.
3. The GAP 
This externalization introduces a structural distance between real management and computerized management. The system is usually designed from the account given by management, area managers or consultants, but not always from the real base operation.
Management knows the objectives, indicators and expected results, but may not know many concrete details of how daily work is executed. Operators, for their part, know the real practice, but rarely have direct capacity to modify the system they must use.
From there the GAP emerges: the difference between what the system believes is happening and what is actually happening in the company's operation. It is not a one-off implementation error, but a structural consequence of externalizing the formalization of management.

4. Overcosts, Adaptation and Invisible Work 

A significant part of the cost of implementation, maintenance and evolution of an ERP comes from trying to reduce that original GAP. Parameterizations, custom developments, integrations, consulting, migrations, training, validations and tests seek to bring the computerized system closer to operational reality.
However, even after that effort, an accepted GAP usually remains. The company learns to live with it. Often, that cost falls on the operators, who must enter fields they do not understand, repeat information, adapt their way of working to the software or manually resolve what the system does not contemplate.
This invisible work does not appear as a formal part of the system, but it sustains its daily operation. The organization believes it has digitalized management, when in reality it has shifted part of the friction onto the people operating at the base.

Conclusion 

The externalization of the management system generates a structural paradox: the company delegates to third parties not only the technological tool, but also the formalization of its own operational logic. The result is a permanent gap between the reality of daily work and its computerized representation.
This gap is not an implementation defect, but an inherent consequence of the model. The invisible work of operators, constant adaptations and high maintenance costs emerge as necessary responses to sustain the operation of the system within the Eager Propagation paradigm.

Part 5. The Entropic System 

We call an entropic system one in which a growing proportion of resources (computational, economic, human and organizational) is devoted to sustaining, maintaining, controlling and self-correcting the system itself, instead of being applied directly to its main purpose.

1. Internal Origins of Entropy 

This dynamic arises from the structural traits of the Eager Propagation model itself:
  • Redundancy multiplies the possible points of incoherence.
  • Predefinition generates rigidity in the face of changes in operational reality.
  • Vulnerability to errors and corrections requires continuous verification mechanisms.
  • Distributed architectures force the creation of multiple layers of adaptation and synchronization.

2. How AI Accelerates Entropy 

The combination of generative AI controlled by agentic AI intensifies entropy. In the current model, AI must identify, among multiple redundant versions of the same fact, which one is "valid", always operating in a probabilistic domain. Although agentic AI improves precision with respect to generative AI, it still does not guarantee the exactness that data management requires.
This generates new burdens: greater human supervision, additional audit logs, complex guardrails and greater complexity in traceability and responsibility.

3. External Entropy 

As internal complexity grows, external mechanisms arise to contain incoherences: data governance, Master Data Management (MDM) platforms, independent rule engines and external validations. These parallel systems consume additional resources to compensate for the limitations of the main model.

4. The Entropy of Time 

The model was born to provide rapid information through anticipated propagation. However, in current practice, much of the time is consumed by reconciliations, synchronizations and corrections, diluting the immediacy it originally sought.

Conclusion 

New technologies such as microservices, which consist of dividing the process without ceasing to be propagative, add new layers of complexity, vulnerability and consumption. Event-sourcing systems, although they contemplate accumulated states, are in reality triggers of propagation.
AI dives into a sea of incongruent data, which must previously be externally cleaned and reconciled with whatever auxiliary operational systems are known. It ends up forcing the adjustment of data through "its own judgment" or hallucinations. It is then controlled through agentic AI, which increases the consumption of a generative AI tenfold and already forces us to consider the CO₂ footprint.

Epilogue 

Should we ask ourselves whether we accept forever a pre-computational model that becomes more expensive to sustain every day? Or whether engineers and IT architects, or perhaps a multidisciplinary platform, should propose a new paradigm that means a rupture with the Taylor-Weber model?
Because, in reality, our management systems do not manage information, documents or data: their objective is to replace the human process established by Taylor-Weber.

2. The Transactional Documents of Companies 

Documents 

A business transactional document is the formal record and proof of a specific business action, carried out or planned, over its active and passive resources, which modifies or fixes the relational state of the company.
Entities are the management resources of the business action: active resources —Persons— with which one acts, and passive resources —Elements— on which one acts.
The role acquired by an entity belongs to the transactional document in which it appears; it does not constitute a state of the entity. Therefore, the role emerges from the relation between entities and documents.
The remaining data contained in a document correspond to the circumstances in which the business action is carried out or will be carried out.
By establishing the role as a relation between entities and documents, the role acquires a circumstantial dimension: temporal, modal, cardinal or contextual, according to the conditions under which the business action is carried out or will be carried out.
Business rules establish relations between the document being formalized and the documentary antecedents of the entities, determining, restricting or deriving their circumstantial data.
Therefore, a document must not be understood as a simple digitalized element inside a management system, but as the formalization of a business action that incorporates within itself the process that makes it possible.
In this way, the load of business rules ceases to fall on an external procedural system and is instead resolved through a relational system among entities, documents and documentary antecedents.
The data item, as a circumstantial element of the document, acquires its own dimension: it ceases to be a value subjected to procedural checks and becomes part of the relational structure that determines the documented business action.
For this, the data item must acquire its own structure based on four properties:
Semantics: it does not only define the type of value, but the meaning it contributes to the document and whether it fulfills the function for which it exists.
Morphology: determines the form or format of the data item, both for its input and for its presentation.
Context: establishes the conditions of origin of the data item: who originates it, where it originates and under which operational environment it is produced.
Syntax: establishes the documentary function of the data item within the formalized business action.
Consequently, the business transactional document does not act as a passive container of data, but as a formal structure where entities, roles, circumstances, business rules and data acquire operational meaning through relations. Management ceases to depend on external procedures that verify states and begins to organize itself through documentary relations capable of determining the business action from its own structure.

The Data Item 

The data item is the minimum structural and functional unit of the business transactional document, defined to contain a valid value according to its semantic, morphological, contextual and syntactic properties.
The data item may contain absolute values, which are entered directly in the document, or relative values, whose value is obtained through a calculation relation with other data.
When a relative value is obtained from some data item contained in another document, the origin document from which the values required for its calculation will come must be identified exactly.
This identification constitutes the relational connection, where an identification indicator points to the identifier of the foreign document. The identification indicator and the identifier are not part of the documentary content.
Relational connections form the structure of the system and define the vectors through which documents relate to one another and to entities, establishing the global logic of information.
The properties of the data item are acquired through the attribute that defines it. The attribute determines its semantic, morphological, contextual and syntactic properties, establishing the conditions that the value occupying it must satisfy.
Semantic Property: establishes the proper meaning of the data item through a conceptual domain and a verification algorithm that determines whether a value must be accepted as content of the data item.
Morphological Property: establishes the way in which a value must be entered and the format in which it is presented.
Contextual property: establishes the conditions of origin and operability over the data item: who generates it, where it is produced, in which environment it is recorded and under which conditions the value can be accepted.
Syntactic property: establishes the circumstantial function of the data item within the transactional document and allows it to be related to natural language without needing to evaluate its value.

Forms 

A form is a total, partial or composite view of documents in which the data maintain their relational connections.
Forms are the means through which values are collected or presented, independently of the support or channel used.
Noteworthy characteristics:
Because information is organized through transactional documents, each document incorporates its own temporal reference. This makes it possible to open temporal query windows over the relations among documents, entities and data without altering the structure of the system.
As a whole, documents, data and forms cease to be passive elements managed by external procedures and become self-managing declarative structures, capable of preserving their relations, validating their values and operating independently of the channel used.
By not depending on external procedural processes, the system avoids the need for integrations and reconciliations between modules, generating holistic management in which all information shares the same relational structure.
This model eliminates the traits of eager propagation:
Uniqueness of the data item. By organizing information through transactional documents with their own relational structure, each data item exists only once in its document of origin. Relative values are obtained through calculation relations at query time, eliminating redundancy and the incoherence derived from duplication.
Absence of pre-established processes. Management logic is incorporated into the relational structure of documents, data and forms. It is not necessary to design or maintain external procedural processes that propagate or verify states.
Ease of error correction. Since each data item exists at a single point of origin, correcting an error affects only that point. The relative values that depend on it are automatically recalculated in the next query, without additional intervention.
Linear audit. The temporal reference of each document and the traceability of relational connections allow the state of the information to be reconstructed at any past moment, without ambiguity and without additional audit processes.
Company sovereignty. Business rules are defined in the system's own structure by the company, without depending on external parameterizations or technical intermediaries. The company establishes its own management conditions and can modify them without outside intervention.

3. Artificial Intelligence in Information Management 

1. Balance in an Intelligent System 

Iain McGilchrist argues that the human brain is not simply divided between a "logical" hemisphere and a "creative" one, as the popular version usually claims. His thesis is deeper: the two hemispheres express two different, and often incompatible, ways of attending to the world and relating to it.
Iain McGilchrist formulates this relation through the metaphor of the Master and the Emissary. The right hemisphere, associated with the figure of the Master, maintains broad, open, vigilant and contextual attention. It does not limit itself to isolating objects, but captures relations, nuances, metaphors, ambiguities, presences and transformations. Its world is alive, complex, changing and interconnected.
The left hemisphere, represented by the Emissary, operates differently. Its attention is narrow, focused and oriented toward manipulation. It fragments, abstracts, classifies, calculates and turns reality into elements available for use. It is the hemisphere of the tool, of number, of explicit representation and of instrumental language. It does not see the world so much in its living presence as in its operational utility.
The problem appears when the Emissary ceases to serve the Master and intends to replace him: when abstraction, classification and instrumental manipulation lose their subordination to a broader, contextual and relational understanding of reality.
Michael Gazzaniga observed, in his studies on split-brain patients, that the left hemisphere can act as an "interpreter": when it does not have complete access to information, it does not necessarily stop, but constructs an explanation coherent with the available data.
This phenomenon is especially relevant for understanding hallucination in generative artificial intelligence. When the interpretive system does not have a concrete, contextual and traceable base, it can produce language, explanation and apparent coherence without being truly anchored in the reality it describes.
In functional terms, generative AI behaves like an Emissary without a Master: it interprets, formulates and justifies, but lacks a concrete system that preserves memory, determines operational logic and validates its responses against a structured reality.

2. Reliable Data 

The WHO, Sittig and Singh in JAMA, and Howell in BMJ Quality & Safety point out that generative AI applied to health can produce false, inaccurate, biased or incomplete responses, including hallucinations that are difficult to detect and capable of harming patients. The problem is not only the error, but the appearance of reliability of the error.
Asgari et al., in npj Digital Medicine, show that LLMs can generate medical summaries with omissions and hallucinations, as well as non-factual text or text not faithful to the source information. This evidences the need for a reliable connection with the clinical facts that must be summarized.
Garcia-Fernandez et al., in Trustworthy AI for Medicine, propose CHECK, a framework that integrates LLMs with structured clinical databases to detect factual and reasoning hallucinations. In other words, they identify that the reliability of medical AI requires a concrete source of data against which to verify its responses.
In medicine, a false or incomplete response can have irreparable consequences; for this reason, several specialized publications have pointed out the need for clinical AI to be connected to reliable, structured and verifiable sources.
In business management systems, the situation appears differently. Errors, duplications, incoherences, manual corrections and later reconciliations do not usually appear as exceptional facts, but as an ordinary part of system operation and not as a crisis. Their correction falls within the field of governance or audits.
In reality, business management systems have the same need as clinical systems: reliable, structured, verifiable and traceable data; the difference is that, in the company, that need is usually hidden under the normality of daily operation and displaced to the field of audits and governance.

3. The Master as Data Manager 

If we analyze data management systems, such as ERP or CRM systems in business management, we can say that they are automations of human management based on the workflow described by Pacioli in the 15th century in his book "Summa de arithmetica, geometria, proportioni et proportionalità (1494)", in the treatise "Particularis de computis et scripturis".
Pacioli was the first to describe workflow, an English term used in computing to indicate a flow of work. That model organized information through a sequence of records: from the economic fact to the entry, from the entry to the Journal, from the Journal to the Ledger and from there to balances or accumulated states. In other words, the data item was duplicated and propagated through different books according to the state of the accounting process.
Taylor, with his division of labor, and Weber, with bureaucratization, extended this model beyond accounting, but in essence the model was the same: extending information through redundancy.
Although this workflow-based model is the easiest to observe and apparently the closest to the real process, in reality it formalizes only part of the work. By automating the human process, we have placed in systems the logic of calculation, recording and propagation of the data item, but not the logic of relation.
For example, an ERP may have a customer file linked to a person file. Technically, there is a relation between both tables. But that relation does not explain why that person is a customer, from which business act they acquired that role, under which conditions they keep it or which antecedents modify it. The system points to the data item, but does not contain the relational logic that determines it.
The Master, understood as a concrete system, is composed of memory and relational logic: memory preserves the data fixed at origin; logic determines how they are linked, which roles they generate and which states they produce.
A data item in memory cannot be understood as a simple stored value. For a data item to be information, it must contain, in addition to its value, its location within the system: its position, its relational link, its meaning, its form of expression, the context in which it is generated and its signifier within the relation.
All these elements are not accessories of the data item; they are part of the relational logic that allows the system to understand it, link it and operate on it.
The Concrete System must have two parts: the facts as immutable memory and the relational logic that links all information and generates the contained reality.
Memory is immutable because it preserves the record of facts as they were fixed at origin. What can evolve is the relational logic that interprets them, links them and determines the states they produce.
Logic is what evolves and adapts to new rules. It does not modify the recorded fact, but it can give rise to a new operational understanding of already existing data. The system does not rewrite memory; it updates the way in which that memory is related, interpreted and projected within the system. 

4. The Abstract System: AI 

McGilchrist describes a historical escalation of the Emissary over the Master throughout Western culture. In this work it is proposed that a possible cause of that escalation may be found in the progressive externalization of memory: first through writing, then through the printing press, radio, television and other media, and today through the Internet.
Each of these technologies has made it possible to preserve, reproduce and transmit information outside direct experience. This has enormously expanded the human capacity for recording, but it has also displaced a growing part of the relation with reality toward systems of representation, communication and abstraction.
Today we are witnessing a new phase of that same movement: the externalization of interpretation through AI. If writing externalized memory and media expanded representation, AI externalizes functions proper to the Emissary: interpreting, summarizing, explaining, comparing, inferring and proposing.
In the terms of this work, generative AI constitutes the Abstract System: a system capable of operating on language, intention and representation, but which needs to be connected to the Concrete System capable of preserving memory, relational logic and verifiable reality.
This need is clearly seen in the Capture System, the point where an external fact is fixed as data in the Concrete System. Sources whose schema is defined within the system itself —HTML forms, XML, JSON, CSV— do not require an interpreter: a parser is enough to extract their values and present them as candidates, because their structure is already declared.
Sources generated outside the system —a PDF, an order in natural language or any expression without an internally declared schema— require another treatment. Before they can be incorporated into the Concrete System, they must be analyzed, interpreted and transformed into a structural proposal: which document they represent, which entities intervene, which values they contain, which context surrounds them and which relations they seem to imply. That work belongs to the Abstract System.
AI does not directly fix the fact in memory; it interprets the external source and turns it into a candidate value. Afterwards, the Concrete System validates that candidate against its memory, its relational logic and its current rules. Only then can the fact be fixed as data within the system.

The Concrete System

Chapter 2. Formal Theoretical Framework. Model G

1. Global Information Space

Definition 2.1 (Global Space). Model G defines a triaxial information space where each unit of information occupies a unique position defined by three independent coordinates:
G = A   x   K   x   F
∀ g ∈ G , ∃ a ∈ A , ∃ k ∈ K , ∃ f ∈ F : g = a , k , f
Where:
G: Structural and infinite information space.
g: Data item as value container:
a: Attribute (a ∈ A) (Axis of Meaning).
k: Key (k ∈ K) (Axis of Location).
f: Foreign reference (f ∈ F) (Axis of Connection)
v = V a l g

1.1. General Expression

G   =   g     g   =   a ,   k ,   f     ∧   ∃ v   : v   =   V a l g }

2. Types of Points in G

The space G is partitioned into two mutually exclusive categories according to the origin of their values:
Definition 2.2 (Base Points). Base points contain externally assigned values, constituting the system's source data.
G b a s e ⊆ G
∀ g   ∈ G b a s e :   v a l g   is   by   direct   assignment
Definition 2.3 (Calculated Points): Calculated points derive their values from other points through dependency relations.
G c a l c ⊆ G
∀ g   ∈ G c a l c :   v a l g   is by reference to other g
Partition Property:
G   =   G b a s e   ∪   G c a l c ,   G b a s e   ∩   G c a l c   =   ⌀

3. Flat Space R and Coherence Operator

Definition 2.4 (Candidate Value). Any value input proposed to the system before being admitted as the value contained in a point g ∈ G.
v ↦ v ( g ) ⇔ φ ( v : g ) = 1
Definition 2.5 (Flat Data Space -R-). A relation R represents a two-dimensional view of the space, the space where axes A and K relate; it is analogous to a table in the relational model.
R = { A ,   K }
Definition 2.6 (Set of Rules - Ͷ-). The set of rules that determines the attributes and the way in which they relate:
Ͷ : set of rules applicable to G
Definition 2.7 (Coherent Space -Ω-). The subset of G where all its points comply with Ͷ
Definition 2.8 (Coherence Operator – Φ -). Evaluates whether a point satisfies all the rules:
Φ Ͷ : G   x   V → { 0,1 }
By construction, Ω is always perfectly coherent and finite in structure.
Ω = { g ∈ G ∣ Φ Ͷ V a l g , g = 1 }

4. Properties of g by Attribute

Definition 2.9 (Attribute Properties). Each a ∈ A incorporates four validation and/or functional attributes
Sem (Semantic): Validates the domain or conceptual universe of the data item.
Mof (Morphological): Validates the physical form: masks and/or regular expressions of format.
Ctx (Contextual): Establishes validation and operational restrictions according to environment, flow or state.
Sin (Syntactic): Establishes the functionality of the data item within the transactional document. This parameter is functional in nature and not structural validation.
A ⊆ Ͷ ,   A = S e m , M o f , C t x , S i n
φ g = φ S e m V a l g ∧ φ M o f V a l g ∧ φ C t x V a l g →   0,1
Φ G = ⋀ g ϵ G φ g = 1

5. Axes of Location and Connection

We establish G as:
G = g   g = a , k , f   ∧ a , k ∈ R   ∧ a , f ∈ H
Definition 2.10 (Identifying Key K). Keys in K guarantee uniqueness within each relation R.
K   ⊆ N   ∀ k ∈ K , k ∈ R ⇒ ∄ k ' ∈ R   |   k ' ≠ k
Definition 2.11 (Connection Key F). Connections establish the way in which different relations (R) connect.
∀ f ∈ H k 0 , R e f ( f ) ≠ k 0
F ⊆ K f : R i → K R j ∀ f ∈ F R i , ∃ k ∈ K R j : f = k ∧ R i ≠ R j ∀ f 1 , f 2 ∈ F R i , R e f f 1 = R e f f 2 ⇒ f 1 = f 2
Within the same relation, different connections point to different destinations
∀ f 1 , f 2 ∈ F R i , R e f f 1 = R e f f 2 ⇒ f 1 = f 2
The function associates each connection with its destination key R e f f n

6. Dependency Graph H

Definition 2.12 (Graph Structure - H -). The structure of graph H considers only axes A and F, modeling dependencies among g.
H = { A , F }
Definition 2.13 (Reach Graph or instance of H - H(k₀)- ). H(k₀) contains all coherent points whose key is reachable from k₀ by following chains of foreign references of length n ≥ 0.
For a key k₀ ∈ K:
H k 0 = g ∈ Ω     k 0 , k g ∈ ⋃ n ≥ 0 F n }  

7. Inverse Graph H⁻¹

Definition 2.14 (Inverse Graph - H⁻¹ -). H⁻¹(k₀) contains the connections that directly reference k₀ but are not in its reach graph.
H − 1 k 0 = f ∉ H k 0 f = k 0 }

7.1. Formal Properties

No self-reference: f H k 0 ≠   k 0
No reciprocity: f H k 0 = k 1   ⇒     f H k 1 ≠   k 0
Strict partial order: f H k 2 = k 1   ∧ f H k 1 =   k 0     ⇒   k 2 ≺   k 0
Transitive inclusion: ∀ f H k n = k 0   ⇒ H k 0 ⊂   H k n
Decomposition: ∀ f H k n = k 0   ⇒ H k n = H − 1 k 0   ⋃   H k 0
External validation: ∃ Ω ∖ H k 0 Λ   H − 1 k 0 ⇒ ⋁ φ H − 1 k 0 = 0

8. Propagation Vector Π

Definition 2.15 (Propagation Vector). Π(g₀) indicates the calculated points within the reach of k₀ that depend on g₀ through two or more reference steps (n > 1).
For a point g₀ ∈ Ω:
Π g 0 = { g ∈ H k 0 ∩ G c a l c ∣ g 0 , g ∈ ⋃ n ≥ 1   ≺ n }

8.1. Dependency Properties (≺ ) Propagation Vector (Π)

Irreflexivity: ∀ g ∈ G :   g   /   →   g
Antisymmetry: ∀ g , g ' ∈ G ,   g ≠ g ' : g → g '   ⇒   g ' /   →   g
Acyclicity: ∀ g , g ' , g ″ ∈ G : g → g '   ∧   g ' → g ″   ⇒   g ″ / → g

9. Propagation Determinant δ

Definition 2.16 (Propagation Determinant). The function ψ(x) determines, for calculated points, whether they intervene in the propagation vector (Π).
∀ g ∈ Ω   : δ g ∈ 0,1
∀ g ∈ Ω b a s e ⇒ δ g = 1
∀ g ∈ Ω c a l c ⇒ δ g = ψ x ∣ ψ x ∈ 0,1
δ g = 1 ⇒ g   ∉ Π
δ g = 0 ⇒ g ∈ Π

Chapter 3. NORMAL FORMS. G-Model

The FN-G-Model define the design conditions that a conventional database must meet in order to materialize structures compatible with Model G.

1. Normal Form 1 (FN-G1). Semantic Grouping

Definition: FN-G1 establishes the requirements for intratabular semantic grouping, guaranteeing that each table represents a clear and complete entity or concept, without NULL values and with unitary semantic meaning.
Algebraic formulation:
Let there be a plane R (A x K):
F N G 1 R ⇔ R ∈ 5 F N c l á s i c a ∀ a ∈ A ,   ∀ k ∈ K :   ∃ V a l a , k ∃ E :   A ⊆ S e m E  

2. Normal Form 2 (FN-G2). Systemic Unification of Equivalent Entities

Definition: FN-G2 establishes that semantically equivalent entities are unified in common systemic structures through table identifiers, avoiding the duplication of parallel schemas that represent the same concept.
A single identifier in a single structure
Algebraic formulation:
Let   R i A i x K i   y   R j A j x K j
F N G 2     ⇔     S e m R i = S e m R j   ⇒   ∃ R :   R i , R j ↪ R  

3. Normal Form 3 (FN-G3). Semantic Hierarchization

Definition: FN-G3 establishes that the entities of the system are organized into semantic hierarchies, where higher categories capture common properties and specializations are managed through specific extensions, without altering the semantic core.
Algebraic formulation:
R b a s e A 0 , K   ≺   R s p e c A 0 ∪ A s , K ⇔ A 0 ⊆ A s p e c   y   S e m R s p e c ⊃ S e m R b a s e

4. Normal Form 4 (FN-G4). Emergent Roles

Definition: FN-G4 establishes that the roles of entities are not defined through static attributes, but emerge dynamically from the effective relations they maintain with other entities in the system.
Algebraic formulation:
Let there be an entity e and the set of relations Rel(e):
R o l e ⇔ ρ R e l e = 1
(* When the relation that determines a role is recorded in a document D, and that document incorporates its temporal moment t, the role of the entity can be determined temporally. *) R o l t e ⇔ ρ ( R e l e , t   = 1

5. Normal Form 5 (FN-G5). Temporal Semantic Coherence

Definition: FN-G5 establishes that system values maintain semantic coherence over time through algebraic persistence conditions, where the propagation of values is fixed or activated according to the state of relations, without imperative intervention.
Algebraic formulation:
Let: g ∈ Ω
δ g ∈ 0,1 δ g = 1 ⇒ g ∉ Π δ g = 0 ⇒ g ∈ Π
and the temporal evolution:
V a l t g   → Π , δ   V a l t + 1 g

Chapter 4. Theoretical Model of the Interface

1. Definition of space U (User Interface)

Definition 4.1 (Space U). Model G defines the UI space (U) as a recursive and hierarchical set of elements u:
U = { u }
Where each element u can simultaneously be:
u t : Atomic element (terminal).
u c : container of other elements (infinite recursion).

1.1. General Expression

∀ n ∈ N : U n = u t n ∪ u c n | u c n ∈ { U n + 1 , ∅ } , u t n ∩ u c n = ∅

1.2. Axioms

Axiom 1: No repetition of forms
∀ n ∈ N : u c n = U n + 1 ⇒ ∀ k ≤ n : U n + 1 ≠ U k
A new level can never be identical to a previous level. This prevents cycles in the hierarchy.
Axiom 2: No repetition of objects
∀n ∈ ℕ : = ⇒ ∀k ≤ n : ∩ = ∅ u c n U n + 1 U n + 1 U k
New levels do not share elements with previous levels. Each element u appears at a single level.

2. Projection of Model G (Ω) into U

Definition 4.2 (Shadow of the Graph - U). The set resulting from projecting the structural plane A x F of the coherent space Ω onto the user interface. It represents the static structural schema (the empty containers and atomic components) and determines the navigation topology of the application:
U = u   u =   A   x   F ,   g ,   g ∈   Ω }
Where:
Axis A determines the existence and type of local atomic elements. Axis F determines the direction of continuation vectors and the hierarchy between containers U t n U c n
Definition 4.3. (Instance of the graph - ). The instance in the UI as the loading of U by assigning a value from axis K U ( k 0 )
U k 0 = u v u ∈ = U ∧ v = v a l g , g ∈ H k 0 }
The abstract structural schema becomes a living and concrete interface, linking to each element u the corresponding value ($v$) obtained through the Reach Graph. H k 0

3. Graphs in the UI

Definition 4.4 (Work Context - H(k₀)). The graph H(k₀) represents the active context:
H k 0 =   { g   ∈   Ω   |   ( k 0 ,   k g )   ∈   U n ≥ 0 F n }
H(k₀) contains all elements g reachable from k₀ by following relations F, including both g_base (δ=1) and g_calc (δ=0 or δ=1).

4. Types of UI Elements

U = U i n p u t ∪ U s p a n ∪ U s e l e c t ∪ U i n p u t ∪ U g r i d ∪ U n a v ∪
F4.1. Editing elements
Input and presentation of val(g) /. They allow the editing and/or visualization of the elements. U i n p u t   U s p a n   g b a s e ∈ H ( k 0 )  
Selection of the connection vector - F. Allows determining U s e l e c t   f   ∈ F
Collection elements. Render H⁻¹(k₀), all records whose foreign references point to k₀. H − 1 ( k 0 )

4.2. Fixed Elements, Control Panel Elements

Navigator U n a v
Change the work instance: N a v :   K n → K m
It is a unique element.
Control panel or button bar. Contains the decision elements through which the user orders execution of their decisions U B o o t B a r  

5. Coherence Function φ and Φ-

  • 5.1. Terminal Elements: u t
φ u t = φ S e m V a l u t ∧ φ M o f V a l u t ∧ φ C t x V a l u t = 1
Where is the candidate value for V a l ( u t ) u t

5.2. For Containers: u c

φ u c = Φ U n + 1 = ⋀ u ∈ U n + 1 φ ( u ) = 1

5.3. Global Coherence of the Form: Φ ( U )

Φ U = ⋀ u ∈ U φ u = 1

6. Hierarchical Transition

S u b m i t U n + 1 → H ( k U ) n + 1 T e m p H ( k U ) n + 1 T e m p → Ω ≡ U n → Ω C a n c e l U n ⇒ ∀ m > n : H ( k U ) m T e m p → ∅

7. Axioms of Independence

1. Autonomy of contents. Contents are autonomous spaces. They do not operate among themselves, nor over their containers, nor over their contents.
∀ ∈ U n : ⋀ ¬ u i → U n ∀ u j ∈ U n , j ≠ i : ¬ u i → u j ∧ ¬ u j → u i ∀ u j ∈ U n + 1 , ¬ u i → u j n + 1 ∧ ¬ u j n + 1 → u i
2. Genericity of containers. Containers operate generically with their contents through collections. A content does not care who its container is; the container does not care who its content is.
∀ u ∈ U n : ¬ ( U n ⇒ u ) ¬ ( u ⇒   U n )
3. Indirect communication. Communication between elements occurs through signals and general states, not through direct calls.
4. Emergent coherence. Contents reach their state of coherence (φ=1) autonomously. Each element evaluates its own coherence according to local rules without centralized coordination.

8. Syntax Property in u

∀ u ∈ U : Sintaxis ( u ) ∈ { adverbios } ∪ { ( preposici ó n + adverbios ) }
The Syntax property assigns to each field u a grammatical label formed by an adverb or by a preposition plus adverb, in order to identify the circumstantial function that said field fulfills within the documented action.

Chapter 5. Theoretical Model of Flow Architecture

Definition 5.1 (Flow Schema).
Every flow F is instantiated according to the schema:
F = N c o n f i g ​ , F o r m a t , Φ , T
Where:
N c o n f i g ∈ N is the specification of the flow in the metadata repository
F o r m a t is the data structure generated by on which it acts N c o n f i g Φ
Φ is the application of the coherence operator over the values of F o r m a t
T is the transport protocol.
Definition 5.2 (Fundamental Flows).
The system defines three fundamental flows. Each fundamental flow expresses a different form of circulation within the system:
F = { D , E , M }
Where:
𝒟is the Data Flow. It transports values
ℰis the Structure Flow. It transports empty structures prepared to receive values
ℳis the Message Flow. It preserves events generated by the coherence bifurcation.
Definition 5.3 (Data Flow).
𝒟is the bidirectional flow of exchange of values between layers:
D r e a d : request for values → Ω responds with calculated values
D w r i t e : sending of validated values → Ω responds with the resolved identifier (K₀)
Definition 5.4 (Structure Flow).
ℰtransports empty forms from N to the agent's receiver. Four subflows are defined:
E = { E I U , E I o T , E A P I , E I A }
E I U : a single form; the human agent enters values through the browser.
E I o T : two forms — F₁ transports configuration parameters to the device; F₂ receives the data generated by the device.
E A P I : two forms — F₁ transports the request to the external service; F₂ receives the response.
E I A : two forms — F₁ transports the document or context to the agent; F₂ receives the inferred values.
In the agent and the receiver coincide. In , and the agent is external to the system. E I U E I o T E A P I E I A
Definition 5.5 (Message Flow).
ℳis the persistent event store. Every message M∈ ℳis defined as:
M = ( a s u n t o , f o r m u l a r i o , τ )
subject identifies the event; form contains the associated values; and τ is the generation timestamp.
Axioms
Axiom F1 (Orthogonality of Flows).
∀ F ᵢ , F ⱼ ∈ F , i ≠ j : F ᵢ ∩ F ⱼ = ∅
Axiom F2 (Coherence Bifurcation).
∀ F ∈ F , ∀ d ∈ F o r m a t ( F ) : Φ ( d ) = 1 ⟹ d → D → 𝛺 Φ ( d ) = 0 ⟹ M ( d ) → M
Axiom F3 (Message Persistence).
∀ M ∈ M : M   e s   i n m u t a b l e .
The management state does not alter its content nor remove it from ℳ.
Axiom F4 (Universality of the Agent).
∀   E x ∈ E
the agent that fills F₂ can be:
  • human,
  • device,
  • external service
  • AI agent.
The flow structure is invariant with respect to the type of agent.
Axiom F5 (Completeness of configuration in N).
∀ F ∈ F
Where:
F   e s   i n s t a n c i a b l e ⟺ N c o n f i g ( F ) ∈ N

Model G. Algebra of the Multiverse

1. Fundamental Spaces

1.1. Global Space G

Definition 6.1 (Global Space G). (see Def. 2.1) G constitutes the abstract and infinite space where all possible information resides. In the distributed context, G is shared by all coherent universes.
G = A   x   K   x   F
Where:
  • G is infinite (it contains all possible information)
  • G is abstract (it has no single physical materialization)
  • ∀ information that exists: ∃ g ∈ G that represents it
Definition 6.2 (Coherent Universe Ωᵢ). Extension of (Def. 2.7) to the distributed case. Each Ωᵢ is a finite subset of G, managed by its own rules Ͷᵢ (Def. 2.7), materialized in a database DBᵢ, autonomous and independent.
Ω ᵢ = A Ͷ ᵢ × K Ͷ ᵢ × F Ͷ ᵢ Ω ᵢ = g ∈ G Φ Ͷ ᵢ g , g = 1 }
Definition 6.3 (Vectorial Point g). (see Def. 2.1) g is the Node or data item.
g = ( a , k , f ) ∈ G
Definition 6.4 (Point g in Ω). A point g belongs to Ω when it complies with the rules Ͷ.
g ∈ Ω ⟺ φ Ͷ ( V a l ( g ) , g ) = 1
Definition 6.5 (Multiverse G). Set of coherent universes that are autonomous and independent among themselves, all located within the same abstract space G
G = { Ω ᵢ | i ∈ I } = A   x   K   x   F
Where:
∪ A Ͷ ᵢ ⊆ A ∪ K Ͷ ᵢ ⊆ K ∪ F Ͷ ᵢ ⊆ F

2. Intersection, Publication and Subscription

Definition 6.6 (Published View Ͷ'ᵢ). The strict subset of rules Ͷᵢ over g ∈ Ωᵢ where a restricted space of points g available to other coherent spaces is declared.
Ͷ ' ᵢ ( g ) ⊂ Ͷ ᵢ ( g ) ∣ φ Ͷ ' ᵢ ( V a l ( g ) , g ) = 1
Definition 6.7 (Subscription Ͷⱼ→ᵢ). The subset of published rules Ͷ'ᵢ to which Ωⱼ subscribes in order to access the published space of Ωᵢ.
Ͷ ⱼ → ᵢ g ⊆ Ͷ ' ᵢ g ∣ φ Ͷ ⱼ → ᵢ ( V a l ( g ) , g ) = 1
Definition 6.8 (Intersection Ω'ᵢ). The subspace of Ωᵢ that becomes accessible to Ωⱼ.
Ͷ ⱼ → ᵢ ( g ) ⟺ g ∈ Ω ᵢ ∩ Ω ⱼ Ω ' ᵢ = { g ∈ G ∣ Ͷ ⱼ → ᵢ ( g ) } = Ω ᵢ ∩ Ω ⱼ

3. Distributed Coherence

Definition 6.9 (Distributed Coherence Operator Φᵢ). Extension of (Def. 2.8) to the distributed context.
Φ ᵢ : Ω ' ᵢ → { 0 , 1 } Φ ᵢ ( g ) = Φ _ Ͷ ⱼ ( V a l ( g ) , g ) s i   g ∈ Ω ᵢ Φ ᵢ ( g ) = Φ _ Ͷ ' ⱼ ( V a l ( g ) , g ) s i   g ∈ Ω ' ᵢ
Property 6.10 (Coherence Guarantee).
∀ g   v i s i b l e   e n Ω ᵢ : Φ ᵢ ( g ) = 1
Axiom M1 (Preserved Coherence).
∀ g ∈ Ω Ͷ ' ᵢ → ⱼ : Φ Ͷ ᵢ ( V a l ( g ) , g ) = 1

4. Distributed Determinant

Definition 6.11 (Distributed Determinant δᵢ). Extension of (Def. 2.16) to the distributed context.
δ ᵢ : Ω ᵢ → { 0 , 1 } ∀ g ∈ Ω ᵢ : δ ᵢ ( g ) = 1 ⇒ g ∈ G b a s e ( i n d e p e n d i e n t e ) ∀ g ∈ Ω ᵢ : δ ᵢ ( g ) = 0 ⇒ g ∈ G c a l c ( d e p e n d i e n t e )
Axiom M2 (Remote Dependency).
∀ g ∈ Ͷ ' ⱼ → ᵢ : δ j ( g ) = 0
Remote data are always calculated, because Ωᵢ does not control their origin.
Definition 6.12 (Switching Function Valᵢ).
V a l ᵢ ( g ) = V a l j ( g ) · δ ᵢ ( g ) + f ( g ' , g ) · ( 1 − δ ᵢ ( g ) )
Where:
Valⱼ(g) is the locally stored value and f(g', g) the function that calculates from the remote source g'.

5. Graph and Propagation

Definition 6.13 (Local Graph Hᵢ). Extension of (Def. 2.12) to the distributed context.
H ᵢ ( k ₀ ) = { g ∈ Ω ᵢ | ( k ₀ , k ( g ) ) ∈ ⋃ _ { n ≥ 0 } F ᵢ ⁿ }
Definition 6.14 (Global Graph H_G). Extension of the local graph incorporating accessible remote subgraphs.
H G ( k ₀ ) = H ᵢ ( k ₀ ) ∪ ( ⋃ { j ≠ i } H ⱼ ( k ₀ ) ∩ Ͷ ' ⱼ → ᵢ )
Definition 6.15 (Local Propagation Vector Πᵢ). Extension of (Def. 2.15) to the distributed context.
Π ᵢ ( g ₀ ) = { g ∈ H ᵢ ( k ₀ ) | δ ᵢ ( g ) = 0 ∧ g ₀ ≺ g }
Definition 6.16 (Distributed Propagation ΠG).
Π G ( g ₀ ) = Π j ( g ₀ ) ∪ ( ⋃ ⱼ Π ⱼ → ᵢ ( g ₀ ) )

6. Multiverse Theorems

Theorem M1 (Trans-Omega Propagation).
Let:
g₀ ∈ Ωᵢ with g₀ ∈ Ͷ'ᵢ→ⱼ, and g₁ ∈ Ωⱼ with δⱼ(g₁) = 0 and g₀ ≺ g₁.
Change in V a l ᵢ ( g ₀ ) ⇒ g ₁ ∈ Π ⱼ ( g ₀ )
Proof:
g ₀ ∈ Ͷ ' ᵢ → ⱼ ⇒ v i s i b l e   e n   Ω ⱼ δ ⱼ ( g ₁ ) = 0 ∧ g ₀ ≺ g ₁ ⇒ g ₁ ∈ Π ⱼ ( g ₀ )
Valⱼ(g₁) recalculates with new Valᵢ(g₀).
Theorem M2 (Isolation by Determinant).
Let:
g ∈ Ω ᵢ c o n g ∈ Π i ( g ' ) , g ' ∈ Ͷ ' ⱼ → ᵢ δ ᵢ ( g ) : 0 → 1 ⇒ g ∉ Π ᵢ ( g ' )
Proof:
δ ᵢ ( g ) = 0 ⇒ V a l ᵢ ( g ) = f ( g ' , . . . ) . δ ᵢ ( g ) = 1 ⇒ V a l ᵢ ( g ) = V a l j ( g ) .
Theorem M3 (Impossibility of Propagated Incoherence).
∀ g ∈ Ͷ ' ᵢ → ⱼ : Φ Ͷ ᵢ ( V a l ( g ) , g ) = 1
Proof (contradiction):
Assume with . By Def. 6.4, only g with . are published. Contradiction. If recalculates from g and , g' does not enter into . g   ∈   Ͷ ' ᵢ → ⱼ   Φ Ͷ ᵢ ( V a l ( g ) ,   g )   =   0 Φ   =   1 g '   ∈   Ω ⱼ Φ Ͷ ᵢ ( V a l ( g ' ) ,   g ' )   =   0 Ω ⱼ
Theorem M4 (No Trans-Omega Circularity).
∀ Ω ᵢ , Ω ⱼ : g ᵢ ≺ g ⱼ ⇒ g ⱼ ⊀ g ᵢ
Proof: Direct consequence of the acyclicity of ≺ (Def. 2.15). Propagation always terminates.
System improvement and learning cycle.
Before developing the cycle, we briefly recall the three elements on which it rests:
Ω (Omega) — the system's memory. It records facts as they were fixed at origin: business acts that have already occurred. It is immutable; it is not rewritten, only completed with new facts.
Ͷ — relational logic. It determines how data are linked, which roles they generate, which conditions must be met for a candidate to be accepted in Ω. Unlike Ω, Ͷ evolves: it changes the operational interpretation of the same facts, never the fact already recorded.
ℳ — the signal. It is activated when a candidate does not comply with the current Ͷ and, therefore, cannot be fixed in Ω. It does not interpret the reason for rejection; it only indicates it —like a warning light on a panel— leaving a traceable record that something did not fit.

1. The GAP as a Structural Problem

In the first chapters we identified the GAP as a structural consequence of MAPA: an administrative model that digitalizes successive technological waves without replacing its paradigm of origin. In MAPA, when the reality of the company departs from what the system contemplates, the response is external —parallel spreadsheets, corrective consulting, code rewriting— because the structure of the data item is fixed in the software itself. The GAP does not close: it is managed from outside, with growing costs.

2. Governance, Consulting and Audit as one Same Cycle

Model G proposes an alternative: structure (Ͷ) is not code, it is declarative data. This allows the three functions traditionally treated as separate phases —governance (definition of the data item), consulting (initial construction of Ω/Ͷ) and audit (review of what has already operated)— to behave as a single continuous cycle instead of three successive projects.
Governance defines Ͷ. Consulting builds Ω/Ͷ from that definition, document by document, until implementation. But once the system operates, each candidate that does not comply with Ͷ is recorded in ℳ, the signal that marks the distance between what the system contemplates and what the company actually does. That record is, in itself, continuous audit: not a process activated periodically, but an emergent property of the design, present in every operation.

3. Closing the Cycle

When ℳ accumulates a pattern —the same type of rejection repeating— the Abstract System can analyze it, at the user's request, and offer a concrete adjustment mechanism over Ͷ. The decision to accept it always remains human: AI interprets and proposes, never decides. But the time between detecting the incoherence and correcting Ͷ can be minimal, because there are no intermediate administrative phases to cross: governance, operation and adjustment occur in the same temporal space.
This reverses the dynamic of the GAP. In MAPA, the GAP grows with use, because every mismatch requires an external intervention that is late in arriving. In Model G, the GAP is reduced through use, because each mismatch detected by ℳ feeds directly back into Ͷ, without needing to rewrite code or launch a new consulting project.

4. Conclusion

The system is not finished after implementation: it continues to complete itself with each act it audits while operating. The separation between memory (Ω), logic (Ͷ) and execution mechanism —which was established in the previous chapters for reasons of traceability— is what allows governance, consulting and audit to cease being three different projects and become a single cycle that closes itself, in real time, through the company's daily work.

References

  1. Pacioli, L. Summa de arithmetica, geometria, proportioni et proportionalità; Tractatus XI: Particularis de computis et scripturis; Paganino de' Paganini: Venice.
  2. Taylor, F. W. The Principles of Scientific Management; Harper & Brothers: New York, 1911. [Google Scholar]
  3. Weber, M. Economy and Society: An Outline of Interpretive Sociology; Original work published posthumously in 1922; Roth, G., Wittich, C., Eds.; University of California Press: Berkeley, 1978. [Google Scholar]
  4. Codd, E. F. A relational model of data for large shared data banks. Commun. ACM 1970, 13(6), 377–387. [Google Scholar] [CrossRef]
  5. McGilchrist, I. The Master and His Emissary: The Divided Brain and the Making of the Western World; Yale University Press: New Haven, 2009. [Google Scholar]
  6. Gazzaniga, M. S. The split brain revisited. Sci. Am. 1998, 279(1), 50–55. [Google Scholar] [CrossRef] [PubMed]
  7. World Health Organization. Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models; World Health Organization: Geneva, 2025; ISBN 978-92-4-008475-9. [Google Scholar]
  8. Sittig, D. F.; Singh, H. Recommendations to ensure safety of AI in real-world clinical care. JAMA 2025, 333(6), 457–458. [Google Scholar] [CrossRef] [PubMed]
  9. Howell, M. D. Generative artificial intelligence, patient safety and healthcare quality: A review. BMJ Qual. Saf. 2024, 33(11), 748–754. [Google Scholar] [CrossRef] [PubMed]
  10. Asgari, E.; Montaña-Brown, N.; Dubois, M.; Khalil, S.; Balloch, J.; Au Yeung, J.; Pimenta, D. A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation. npj Digit. Med. 2025, 8, 274. [Google Scholar] [CrossRef] [PubMed]
  11. Garcia-Fernandez, C.; Felipe, L.; Shotande, M.; Zitu, M.; Tripathi, A.; Rasool, G.; El Naqa, I.; Rudrapatna, V.; Valdes, G. Trustworthy AI for Medicine: Continuous hallucination detection and elimination with CHECK. arXiv 2025, arXiv:2506.11129. [Google Scholar]
  12. Newman, S. Building Microservices: Designing Fine-Grained Systems; O'Reilly Media: Sebastopol, 2015. [Google Scholar]
  13. Fowler, M. Event Sourcing. martinfowler.com. 2005.
  14. Kleppmann, M. Designing Data-Intensive Applications; O'Reilly Media: Sebastopol, 2017. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.