Foundation modelBaseline trajectory
KRL governs behavioral trajectory, continuity, boundaries and observability without scripting what the model says.
It reads how an interaction evolves over time and keeps reconstructable evidence of every governance decision.
KRL makes AI compatible with us, not similar to us.
Even without agency or consciousness, reasoning systems create real effects within human relational space. Yet most AI architectures still treat interaction as a sequence of isolated outputs rather than an evolving trajectory.
Most AI systems optimise the next response. Extended interactions require a broader form of governance.
Context retention alone does not ensure coherent behaviour across evolving interactions.
Small changes in behaviour can accumulate gradually and remain difficult to identify through isolated output checks.
Models generate language. Guardrails inspect outputs. Persistent AI systems require governance across time.
Often focus on input and output checks. They do not continuously observe the interaction as an evolving field.
Observes qualified relational and structural signals across turns to support proportionate governance.
Shape aggregate model behaviour. They cannot govern the trajectory of a specific live interaction.
Supports more coherent handling as interactions evolve across multiple turns.
Influences context and response style. It does not create a persistent and reviewable governance layer.
Applies bounded handling paths and records governance-relevant decisions for structured review.
This is not a failure of intelligence.
It is a missing relational architecture.KRL introduces five coordinated layers above the selected foundation model — separating signal intelligence, persistent state, governance, expression, and terminal validation into distinct but interoperable functions.
KRL does not tell the machine what to say.
It reads the relationship to understand how to say it.An illustrative controlled scenario using the same underlying model under two operating conditions. Five moments are extracted from a single multi-turn interaction.
Illustrative controlled scenarios. Excerpts are derived from archived A/B sessions using the same underlying model under two operating conditions. Outputs may vary by model and configuration.
KRL governs the interaction over timeEach interaction is observed through a qualified sensing layer. Structural, semantic, and relational signals are extracted by dedicated producers and checked before entering the governed path. Non-conforming values can be excluded, degraded, or handled through bounded fallback logic, preserving continuity without silent failure.
Qualified signals are composed into a persistent and reviewable representation of trajectory, continuity, relational conditions, and structural dynamics across turns. The system can distinguish isolated events from evolving patterns while preserving coherent state across extended interactions and supported restore paths. Retained context carries an explicit epistemic status — what is established, what remains open, what is contested — so that continuity preserves not only information but its qualification over time.
The governance layer applies deterministic policies to the evolving interaction state. It regulates trajectory, preserves relational boundaries, resolves the active communication posture, and coordinates bounded handling paths before final delivery. KRL is not limited to after-the-fact filtering: it operates as a closed-loop control architecture.
The expression layer realises the governed configuration through profile-, mode-, and context-aware orchestration. Tone, pacing, structure, and proportionality can adapt without forcing identical responses across models, profiles, or interaction channels. Expression remains flexible while operating inside the governed field.
Before delivery, the publishable response can be evaluated against the active governance context, configured boundaries, proportionality requirements, and audit conditions. Terminal validation can operate in advisory, monitored, or enforcement-oriented configurations while preserving bounded, reviewable, and traceable behavior.
The selected API-accessible model provides the generative capability within the KRL operating architecture. KRL can be applied across supported providers without retraining, fine-tuning, or modifying the underlying foundation model.
Each layer has a defined role and an auditable interface.
The system remains expressive while governed.Select a stage to inspect the governance pipeline.
Governance is applied throughout the interaction lifecycle.
From sensing to auditable output.Reconstructable evidence is not a claim on this site — it is a capability the system has already exercised on itself. During pre-pilot testing, in a controlled session with no real user involved, one cycle ran end to end:
Before explaining what KRL is, it is necessary to explain why it exists.
Today, we are increasingly adapting human life to the computational structure of artificial intelligence.
Not to adapt human beings to AI, but to adapt the computational structure of AI to the relational structure of human reality. KRL introduces a governance layer that lets AI operate within the continuity of human life — not as isolated sessions, but as the continuous, relational, temporal and mnemonic reality people actually live in.
If artificial intelligence becomes a permanent participant in society, compatibility will no longer be a design preference — it will become an infrastructural requirement.
Making machines more intelligent will not be enough — they must become compatible with the reality in which human beings actually exist.
This requires building an isomorphism between the computational structure of AI and the relational structure of human reality.KRL does not try to make machines more human. It makes governable the space AI creates within human reality, giving that space continuity, boundaries and an internal sense of time.
From this premise, the distinction becomes straightforward. KRL is not another AI capability. It is an infrastructure for governing how AI behaviour evolves over time. Within this architecture the roles are distinct and complementary: the model understands, interprets and generates; KRL preserves, qualifies, coordinates and bounds; the provider makes the model and its capabilities available. KRL does not compete with the model where the model is naturally stronger — it complements it where governance, continuity, auditability and boundaries are required.
This was never a question of intelligence.
KRL makes AI compatible with us, not similar to us.KRL is a continuity-aware governance layer for AI systems that operate across time. It does not replace the foundation model. It makes evolving interaction dynamics more observable, reviewable, and configurable across supported deployment contexts.
Where AI-assisted interactions require stronger oversight, KRL can support structured review through observable governance states, handling paths, and session-level traces.
Extended use can expose inconsistency that is not visible in a single output. KRL introduces continuity-aware observability while preserving deployment-level configurability.
Persistent assistants and service interfaces require more than fluent responses. KRL is designed to preserve user autonomy and support proportionate handling as interaction conditions evolve.
Assistants that operate across extended or restored sessions require continuity-aware governance. KRL supports observation of evolving interaction dynamics without reducing the experience to a rigid script.
Repeated interactions can accumulate inconsistency, pressure, or drift. KRL provides a governed layer for reviewable behaviour across customer support, onboarding, service, and professional workflows.
Some environments require stronger documentation, bounded escalation, and authorised review. KRL can support these workflows while leaving legal, clinical, and organisational responsibility with the deploying institution.
Voice interfaces, devices, and embodied agents require governance beyond isolated text generation. KRL is designed to extend the same continuity-aware architecture toward multimodal and physical operating contexts.
Continuity-aware governance for evolving human–AI interactions, extended sessions, and longer-horizon operating contexts.
A research trajectory focused on preserving conceptual direction, semantic coherence, and bounded interaction quality as interactions evolve over time.
Extension of the KRL ecosystem toward coordinated AI systems, where agentic activity can remain observable, reviewable, and governed across operational workflows.
Voice, device signals, and physical context may extend the same bounded principles beyond text, supporting more persistent and context-aware forms of human–AI interaction.
Future regulated deployments may support verified contextual information entered by authorised professionals under controlled access, explicit governance, and auditable use. KRL does not generate diagnoses, prescribe treatment, or replace professional judgement.
KRL does not replace professional accountability or ensure regulatory compliance. It provides a governance architecture designed to support continuity, observability, and structured review.
One governed field. Multiple operating contexts.One governance architecture. Three distinct ways to deploy AI — from auditable control to adaptive relational presence and governed initiative.
For organisations that need control without changing the user experience. KRL evaluates each interaction in real time, records the reasons behind governance decisions, and creates an auditable layer above the underlying model.
Regulated deployments, enterprise pilots, public-facing services, and any organisation that needs to demonstrate how an AI system behaved — not merely what it answered.
For AI systems expected to remain coherent across extended interactions. KRL reads the evolving relational state of the conversation, adapts expression, and preserves a recognisable presence within governed limits.
Customer-facing assistants, hospitality, advisory systems, professional copilots, and companion interfaces where inconsistency erodes trust and a more coherent presence creates measurable value.
For advanced deployments where the system must do more than remain consistent. KRL exposes the live relational-semantic field, supports initiative within defined limits, and preserves autonomy under pressure.
When interaction conditions converge, KRL can surface a higher-coherence operating window as an observable signal. In Aware deployments, this supports a richer but still bounded interaction model: more adaptive where appropriate, never less governed.
Profiles configure the operational environment and interaction style. Tiers determine how deeply KRL observes, adapts, and intervenes.
All access by NDA only. Pilot fees credited 100% against the first annual licence.
Ranges shown are indicative — final pricing confirmed under NDA per deployment.
| BLUE · Regulated | WHITE · Professional | RED · Consumer | |
|---|---|---|---|
| Governed | €20k – €35k | €15k – €25k | €10k – €18k |
| Relational | €35k – €55k | €25k – €40k | €18k – €30k |
| Aware | €60k – €90k | €45k – €70k | €30k – €50k |
| BLUE · Regulated | WHITE · Professional | RED · Consumer | |
|---|---|---|---|
| Governed | NDA | NDA | NDA |
| Relational | NDA | NDA | NDA |
| Aware | NDA | NDA | NDA |
| BLUE · Regulated | WHITE · Professional | RED · Consumer | |
|---|---|---|---|
| Governed | NDA | NDA | NDA |
| Relational | NDA | NDA | NDA |
| Aware | NDA | NDA | NDA |
Indicative ranges · Pricing scoped by integration breadth, deployment mode, and regulatory classification · VAT excluded · Multi-year frameworks available
Access is by request only and subject to NDA.
KRL governance badge, internal governance report, and technical briefing included in all tiers as structured evidence for compliance review.
KRL did not begin as a product concept. It began as an observation — a recurring friction between how advanced AI systems are designed and how human beings actually relate to them over time.
AI systems are not mere tools. Reasoning — even without agency or consciousness — creates real effects within human relational space. We bring to AI the same implicit expectations we bring to human relationships: that the other adapts, reads context, and manages boundaries instinctively. But the nature of AI is different — not inferior. That difference generates a structural friction that conventional AI design does not yet fully address.
KRL was built to give that friction a name, a structure, and a solution. Not to make AI more human. Not to make humans more machine-like. But to create a governed space where two different natures can interact — observably, proportionately, and within clear boundaries.
Most AI architectures optimize what a system says.
KRL governs how a system behaves over time.
KRL emerged from direct observation of a structural limitation in early LLM systems — the absence of a dedicated layer for relational continuity and governed interaction over time. What began as a dialogic configuration gradually became a formalised architecture. The book L'Anello Manco served as the semantic crystallisation of that observation.
KRL operates above supported foundation models, independently of their underlying architecture. Its governance layer is provider-independent, domain-configurable, and auditable by design. A working demo shows that interaction trajectories, boundary conditions, and governance events can be observed and traced over time.
From first observation to patent-pending architecture — the timeline of KRL. A documented sequence of conceptual shifts, architectural decisions, and milestones that shaped the protocol into a working governance layer.
KRL is the infrastructure designed to make persistent human–AI interaction more observable, proportionate, and governable over time.
One governed field. One missing link. Now found.The system defines a control layer that regulates relational dynamics, conversational trajectories, and interaction stability, independently of the underlying model.
It separates governance, expression, and identity into independent operational layers, establishing a new class
of infrastructure for AI systems.
// Selected pilot integrations are opening for enterprise and research teams developing persistent AI systems. //
KRL reads relational dynamics to preserve stability, autonomy and functional interaction.
It does not diagnose, profile or reduce the human being to a label.