cosenlab
Architecture

Systems atlas

Systems take shape before they are built.

Architecture is where principles become operational: authority is separated, evidence is preserved, boundaries are made explicit, and intelligence finds a governed place inside real systems.

8
Principles
6
Layers
8
Reusable patterns
5
Boundaries

Logical flow · ARCH-00A

Reference
  1. 01Human intent
  2. 02Context
  3. 03Memory
  4. 04Evidence · knowledge
  5. 05Intelligence
  6. 06Governance
  7. 07Human authority
Determinism where it mattersEvidence first
01Principles

How we think
about systems.

Architecture is not just choosing components. It is the deliberate placement of responsibility, evidence, authority, and failure.

  1. P-01

    Human authority

    Preparing a decision and having the authority to make it are two distinct architectural responsibilities.

  2. P-02

    Evidence and provenance

    Outputs that decisions depend on must retain a verifiable basis.

  3. P-03

    Explicit boundaries

    A system must state what it can do, what it cannot do, and where authority changes hands.

  4. P-04

    Separation of concerns

    Knowledge, intelligence, governance, experience, and systems of record should not be merged into a single layer.

  5. P-05

    Determinism where it matters

    Critical rules and governance paths must not depend on unnecessary probabilistic behavior.

  6. P-06

    Governance by design

    Control lives inside the architecture, not only in the policies written around it.

  7. P-07

    Interoperability

    New intelligence coexists with existing systems instead of demanding their replacement.

  8. P-08

    Visible failure

    Uncertainty, conflicts, and missing evidence stay visible rather than being silently normalized.

02Systems atlas · COSEN-ARCH-001

Six layers.
The same concepts in every product.

Products occupy different layers, but the same elements always recur: human intent, experience, intelligence, admitted knowledge, governance, and authoritative enterprise systems. Select a system.

  1. 01PeopleIntent · authority · accountabilityPeople
  2. 02ExperienceWorkspace · interaction · contextClabbitQuuipu
  3. 03IntelligenceAnalysis · models · decision preparationIRIS Core AIIRIS applications
  4. 04KnowledgeEvidence · rules · memoryIRIS DiscoverySempliGoClabbit
  5. 05GovernancePolicy · provenance · boundaries · controlIRIS Core AIIRIS Discovery
  6. 06Enterprise systemsSystems of record · twins · operational dataERP · AS400 · platformsIRIS applications

Primary architecture

Governed intelligence

Policy · evidence · authority

Layers occupied: Intelligence · Governance

Go to IRIS Core AI
  • 01 · Authority flows down only when it is explicit.
  • 02 · Evidence remains inspectable across layers.
  • 03 · Systems of record remain authoritative.
03Reference model · GI-01

Governed intelligence.

Intelligence prepares. Evidence supports. Governance constrains. Human authority decides.

  1. InputHuman intent
  2. InputContext
  3. ProcessingIntelligence
  4. ControlEvidence
  5. ControlGovernance
  6. OutputPrepared decision
  7. DecisionHuman authority

Design premise

More capable models must not erase the difference between recommendation and authority.

Input
Intent + context
Control
Evidence + policy
Output
Prepared decision

No layer silently inherits another layer’s authority.

04Patterns

Solutions we
reuse.

Recurring structural patterns across our systems, from cognitive memory to enterprise decision layers.

  1. PAT-01

    Read-only twin

    Replicate operational state for intelligence without transferring transactional authority.

  2. PAT-02

    Evidence chain

    Keep every output linked to its source, transformation, rule, and provenance.

  3. PAT-03

    Decision boundary

    Make the handoff from machine preparation to human authority explicit.

  4. PAT-04

    Knowledge admission

    Separate collected information from knowledge admitted for use.

  5. PAT-05

    Context preservation

    Maintain relevant context across sessions, systems, and decision cycles.

  6. PAT-06

    Fail-safe closure

    When critical evidence or authority is missing, constrain behavior instead of inventing certainty.

  7. PAT-07

    Separation of authority

    Identity, knowledge, policy, intelligence, and decision remain distinguishable.

  8. PAT-08

    Decision to action

    Turn recommendations into controlled, observable operational steps.

05Boundaries

Boundaries are
part of the architecture too.

Advanced systems become dangerous or misleading when closely related concepts are allowed to blur into one another.

  1. AI≠is notauthority
  2. Memory≠is nottruth
  3. A recommendation≠is nota decision
  4. Knowledge≠is notgenerated text
  5. The digital twin≠is notthe system of record
06Architecture register

Documents
under revision control.

Models, maps, notes, specifications, and decision records, each with its revision. These are internal working documents, available to partners and clients on request.

  1. REV 1.0

    Reference model · Governed AI

    Governed intelligence reference model

    A layered model that separates intelligence, evidence, policy, and human authority.

    On request
  2. REV 0.9

    System map · Cognitive systems

    Cognitive infrastructure map

    Memory, context, intent, and continuity as stable system capabilities.

    On request
  3. REV 1.1

    Architecture note · Knowledge

    Regulatory knowledge boundaries

    Why source authority, admitted knowledge, and generated interpretation must be kept separate.

    On request
  4. REV 1.0

    Specification · Experience

    Intent-driven experience model

    Actor, intent, context, workspace, and adaptive representation.

    On request
  5. ADR-014

    Decision record · Enterprise

    Read-only enterprise twin

    Why transactional authority stays in the source system.

    On request
Request an architecture document

Research → experiment → architecture

Architecture makes principles operational.

Research asks what might be possible. Architecture establishes what must remain true when that possibility becomes a system.

ResearchExperimentsDiscuss architecture