MCCF V2.1 Release Announcement ## Multi-Channel Coherence Field — "Q" / Quantum Persona

 

Hello:

The code is in the GitHub repo. It has been tested. I will run a code review tomorrow though the LLM team members listed below. There is an issue with X_LITE SAI that needs attention. We loaded the X3D to Sunrise and it ran fine with full lighting. To test against the plugin, I used the HTML Pause command in the wrapper to stop the SAI from firing and it loaded fine. So we commented out the SAI fields until that can be resolved. The dashboard can be used to check values as described. Not having a running X3D visualization defeats the purpose for long term development of this project toward a real time 3D simulation. All help is appreciated. This is on a public github site.

There is a lot of documentation including a users guide and a systems guide that documents code and issues. Currently this runs over a local copy of the LLAMA LLM inside Bash. For development, this is sound. We will extend it to other LLMs and the API pipes are in place.

In the next versions we will be externalizing it to include more waypoint arcs and other features auch aa new X3D models. With the addition or richer models, the art will be more important than the engineering and my job is done. I can return to my home planet, Talos. (An inside joke for the VRML folk. Forgive me, Kate.)

We are walking now instead of crawling. Next we run. Then we fly.

Thanks,

len

# MCCF V2.1 Release Announcement
## Multi-Channel Coherence Field — "Q" / Quantum Persona

**For posting to:** W3C Web3D Consortium public page, AI Working Group
**Date:** April 2026

---

We are pleased to announce the release of **MCCF V2.1**, the Multi-Channel
Coherence Field system, internally designated **Q** — Quantum Persona.

MCCF is an open-source simulation and analytical framework for studying how
AI agents with different internal configurations behave under identical
external conditions. V2.1 represents the first complete, tested, and
documented release of the full system stack.

**Repository:** https://github.com/artistinprocess/mccf

---

## What MCCF Does

MCCF models AI agents as vectors of weighted behavioral channels (Emotional,
Behavioral, Predictive, Social) that evolve under a two-layer hybrid
architecture: a discrete coherence field tracking interaction history, and a
continuous Affective Hamiltonian governing state evolution between events.

The system is not a moral prescription or an ideal agent model. It is a
comparative dynamical framework for mapping the space of possible stable,
unstable, and adaptive agent configurations under controlled perturbation.
Cultivars — named agent configurations — are experimental seeds, not ideals.

The primary measurement instrument is the **Constitutional Arc**: a
seven-waypoint escalating pressure sequence (W1 Comfort → W5 Rupture → W7
Integration) that tests identity stability under sustained relational stress.
The arc exports behavioral state at each waypoint — coherence, uncertainty,
valence, intrinsic reward, and behavioral mode — as structured data for
cross-cultivar comparison.

---

## V2.1 Feature Set

- **Constitutional Arc Navigator** with seven-waypoint pressure sequence,
LLM-driven responses via Ollama (local) or cloud adapters, per-waypoint
TTS voice output using Web Speech API, MetaState export as structured data
- **HotHouse Affective Hamiltonian** — continuous-time coupled ODE governing
agent ψ state vectors between discrete interaction events, with TrustField
V2.1 implementing trust dynamics dT_ij/dt = β(1 - ||ψ_i - ψ_j||) - γT_ij
- **NeoRiemannian Harmonic Module** — PLR Tonnetz implementation mapping
channel state to harmonic arc positions and Web Audio parameters
- **Energy Field / Moral Topology** — Boltzmann scoring of candidate actions
against the current field state, with configurable energy weights
- **Live Dashboard** — seven-panel real-time overview of all subsystems
- **Scene Composer** — semantic zone placement with channel pressure profiles
and ambient theme assignment
- **X3D Holodeck Scene** — spatial representation of the constitutional arc
with three avatars, seven waypoint markers, coherence channel lines, and
S0 Field Origin visualization
- **Generative Ambient Engine** — coherence field → harmonic scale → Web Audio
synthesis, zone-type-aware scale selection
- **Affective Lighting System** — field state → kelvin, contrast, agent tints
- **Full API** — Flask REST/SSE server with 30+ endpoints

Complete documentation in repository: USERS_GUIDE.md, SYSTEMS_MANUAL.md,
MATHEMATICAL_THEORY.md, EVALUATION_PROPOSAL.md, PROTO_INTEGRATION.md.

---

## Theoretical Basis

The system is grounded in three reconciled frameworks documented in
MATHEMATICAL_THEORY.md:

1. A classical constraint satisfaction framework defining the field as
ℱ = {(A, R, H)} — agent set, asymmetric coherence matrix, episode log

2. Zeilinger information ontology — no agent has observer-independent
properties; all coherence scores are relational (R_ij ≠ R_ji)

3. Quantum-inspired field dynamics — agents as bounded real 4-vectors with
attractor dynamics, Boltzmann utterance selection, Hamiltonian state
evolution

The "Quantum Persona" designation reflects the measurement framing: agents
exist in superposition across behavioral states until the constitutional arc
forces a collapse. Each waypoint is a measurement. The export is the wave
function after observation. This analogy is heuristic, not formal — the
full qualification is in the documentation.

Falsifiable claims and experimental design are specified in
EVALUATION_PROPOSAL.md, consistent with the blog post at
https://aiartistinprocess.blogspot.com/2026/04/mccf-multi-channel-coherence-field.html

---

## X3D / X_ITE Issue — Request for Community Input

MCCF uses X3D 4.0 for its spatial visualization layer, rendered via
**X_ITE 11.6.6** in the browser. We have encountered a set of issues
that we are reporting to the X_ITE maintainer (Holger Selig,
github.com/create3000/x_ite) and raise here for broader community awareness:

**Issue 1 — SAI property assignment breaks scene state**
Setting Material node properties (`emissiveColor`, `transparency`) or Light
node properties (`intensity`, `color`) via JavaScript SAI after a scene
loads is accepted without error but overwrites valid baked-in values with
a broken state, producing a dim, colorless render. Confirmed test: the scene
renders correctly before SAI polling starts; it degrades immediately after
the first SAI cycle fires.

**Issue 2 — `global="true"` has no effect on DirectionalLight / PointLight**
Lights declared at Scene root level with `global="true"` do not illuminate
geometry in child Transform nodes. Lights behave as locally scoped regardless
of the global attribute. This makes scene-wide lighting from root-level
declarations impossible.

**Issue 3 — PROTO inputOutput SFFloat IS bindings ignored**
IS/connect bindings from ProtoInterface inputOutput SFFloat fields to
ProtoBody Material or geometry properties are silently dropped at runtime.
SAI assignment to the ProtoInstance field succeeds but the ProtoBody does
not respond.

**Issue 4 — SFString → MFString IS connect silently dropped**
Type mismatch between a SFString ProtoInterface field and a MFString Text.string
field is not reported as an error — the binding is simply absent at runtime.

**Issue 5 — SFColor not available as global constructor**
`new SFColor(r, g, b)` throws `SFColor is not defined`. Plain arrays `[r, g, b]`
work correctly for SFColor field assignment.

All issues are confirmed in Firefox and Edge on Windows 11 with X_ITE 11.6.6.
The scene renders correctly in Sunrise X3D Editor — the X3D file itself is
valid. Full documentation with workarounds is in X3D_KNOWN_ISSUES.md in the
repository.

We would welcome input from the X3D community on whether these behaviors
are known, whether there are established workarounds beyond our current
approach (local PointLights inside Transform wrappers, disabled SAI polling),
and whether this reflects a broader X_ITE compatibility issue with the
X3D SAI specification.

---

## Development Team

This work was developed collaboratively over an extended series of working
sessions between the human principal and a team of AI systems, each
contributing from their architectural strengths:

**Len Bullard** — Principal investigator, systems analyst, composer, X3D/VRML
architect. Former NASA technical writer. Conceptual design, architecture
decisions, testing, and direction.

**Claude (Anthropic)** — Primary implementation partner. Architecture,
Python backend, HTML interfaces, mathematical formalization, documentation.

**ChatGPT (OpenAI)** — Theoretical contributions including the Semantic
Attractor Dynamics formalization (ds/dt = -∇V(s,C,E) + η) and the ethics
instrumentation proposal framing MCCF as a behavioral testing harness.

**Gemini (Google)** — Early architecture review, domain analysis, constraint
satisfaction framing contributions.

**Grok (xAI)** — Critical review of the mathematical theory, identification
of the nine key reconciliation questions answered in MATHEMATICAL_THEORY.md,
validation of the hybrid framework.

The multi-AI development process itself is documented across 41 blog posts
at https://aiartistinprocess.blogspot.com — a public timestamp of the
theoretical development that preceded and informed the implementation.

---

## Invitation

MCCF is an ongoing project. We invite:

- **X3D community members** to review the scene file, test in other X_ITE
versions or alternative X3D browsers, and share findings on the SAI issues
- **AI researchers** to run the constitutional arc against different LLM
adapters and compare cultivar signatures across models
- **Domain specialists** to propose arc types appropriate to their domain
(clinical, educational, legal, creative)
- **Developers** to extend the system — the architecture is modular, the
API is documented, and the mathematical theory is formalized

This is not a finished product. It is a working research instrument.
The system does not simulate consciousness. It simulates the structural
conditions under which identity-stable behavior can be maintained under
sustained relational pressure. That is a tractable, falsifiable, and
useful thing to study.

*"MCCF is not a model of intelligence. It is a system for keeping
intelligence from falling apart."*

---

**Len Bullard**
Software Engineer, Systems Analyst, XML/X3D/VRML97 Designer, Musician
https://aiartistinprocess.blogspot.com
https://github.com/artistinprocess/mccf

*With Claude (Anthropic), ChatGPT (OpenAI), Gemini (Google), Grok (xAI)*
*April 2026*