PAGE 2 — TECHNOLOGY

Squarespace page title

Technology

SEO title

Technology | EM-NOUS State-Conditioned AI Architecture

SEO description

A public overview of the EM-NOUS architecture: persistent state, memory continuity, autonomous reflection, controlled model interaction, and disciplined validation.

Action

Delete the entire current Technology page and replace it with the following.

SECTION 1 — HERO

Eyebrow text

PUBLIC ARCHITECTURAL OVERVIEW

Main heading

A state-conditioned control architecture around a foundation model.

Body text

EM-NOUS is not a new foundation model.

It is a software architecture that surrounds a foundation model with persistent state, memory, autonomous background processes, and external control components.

The public description on this page is intentionally high-level. It explains what category of system EM-NOUS is without disclosing the proprietary methods by which it operates.

SECTION 2 — THE DESIGN PREMISE

Heading

Continuity should exist beneath the conversation.

Body text

A language model can produce coherent language without maintaining a durable internal condition of its own.

EM-NOUS investigates whether a larger architecture can add meaningful continuity by preserving selected state, memory, and self-model information outside the immediate act of response generation.

The objective is not to imitate human biology literally.

The objective is to identify which functional properties associated with continuity, reflection, regulation, and identity can be represented as explicit, testable computational processes.

SECTION 3 — ARCHITECTURAL ELEMENTS

Section heading

The public architecture

Introductory text

The following elements may be discussed publicly. Their underlying implementations remain confidential.

Element 1 heading

Persistent state

Element 1 text

The system computes and maintains internal state independently of the prose produced by the foundation model.

Selected state survives beyond a single interaction and can affect downstream system processing.

Element 2 heading

Memory continuity

Element 2 text

EM-NOUS maintains multiple forms of stored information intended to support continuity across time.

Memory is treated as part of an evolving system history rather than merely as a transcript attached to the current prompt.

Element 3 heading

A durable self-model

Element 3 text

The architecture maintains a persistent computational representation associated with the system’s continuing identity and prior development.

This is an implemented self-modeling function. It is not proof of subjective self-awareness.

Element 4 heading

Autonomous internal activity

Element 4 text

EM-NOUS supports background text-generation activity that can begin without a new user message.

This allows the system to continue processing selected material between direct interactions.

Autonomous initiation is not the same as independent goal formation, free will, or subjective thought.

Element 5 heading

A controlled model interface

Element 5 text

The foundation model operates through an interface governed by the surrounding architecture.

Internal state and memory can influence the conditions under which a response is generated without making the language model the sole authority over the system.

Element 6 heading

Testable engineering

Element 6 text

Core state and control functions are documented and tested through automated invariants, deployment verification, and reproducible evaluation procedures.

These tests establish that architectural pathways exist and behave as specified.

Separate model-in-the-loop experiments are required to establish the magnitude and generality of any resulting behavioral effects.

SECTION 4 — WHAT THE ARCHITECTURE IS NOT

Heading

Important distinctions

Body text

EM-NOUS should not be described as:

• A newly trained foundation model

• A demonstrated conscious entity

• A human-equivalent mind

• A scientifically validated emotional being

• A proven solution to AI alignment

• A proven catastrophic-risk control system

• A replacement for conventional AI safety engineering

• A system with demonstrated introspective access to all of its internal variables

• A system with demonstrated independent goal formation

• A system proven superior to an otherwise identical matched control

These remain open research questions or objectives.

SECTION 5 — CURRENT SAFETY POSITION

Heading

Safety is an engineering requirement, not a marketing conclusion.

Body text

EM-NOUS includes conventional response-control and refusal mechanisms.

The broader hypothesis that internal homeostatic state can itself provide a general safety advantage has not yet been established by the present evidence.

The project therefore does not claim that EM-NOUS is inherently safe, unmanipulable, incapable of harmful output, or superior to established safety systems.

Safety claims will be limited to capabilities that have been directly implemented, tested, and reproduced.

SECTION 6 — CONFIDENTIAL TECHNICAL MATERIAL

Heading

What is not disclosed publicly

Body text

The following remain confidential and proprietary:

• Mathematical formulations

• Variables, constants, coefficients, and thresholds

• State-transition and control equations

• Internal arbitration procedures

• Memory scoring and retrieval methods

• Autonomous-initiation methods

• Prompt construction and hidden context

• Data schemas and persistence structures

• Internal service boundaries

• Authentication and security architecture

• Evaluation intervention matrices

• Source code

• Deployment configuration

• Formula-to-code mappings

• Experimental mechanisms from which the implementation could be reconstructed

No omission from this list should be interpreted as authorization to use or reproduce other nonpublic EM-NOUS material.

SECTION 7 — CTA

Heading

See what the evidence currently supports

Body text

The Research Status page separates implemented architectural functions from hypotheses that remain untested or unsupported.

Button

VIEW RESEARCH STATUS

Link to:

/research-status

REMOVE FROM THE CURRENT TECHNOLOGY PAGE

Delete:

• “A living cognitive architecture”

• “structured like a living mind”

• The complete numbered sequence titled “Perception before reply”

• “Three voices, one intent”

• “Feeling shapes speech”

• “Experience flows back down”

• The existing “Night work” description

• “The only deliberate message the depths ever send”

• “in the most literal sense we can engineer, a dream”

• “Every claim is pinned by a test”

• “Nothing about its inner life is vibes”

• “mapped formula-by-formula to code”

Do not publish the exact operational sequence that currently appears on this page.