[x3d-public] AI Digest, Vol 6, Issue 10
cbullard at hiwaay.net
cbullard at hiwaay.net
Thu Apr 16 10:09:48 PDT 2026
John,
Thank you — this is exactly the clarity we needed on two fronts.
The confirmation that there is no real SAI for PROTOs in SAI 3.3
explains our IS binding failures cleanly. We had been treating this as
an X_ITE bug, but it is a specification gap. We will update our bug
report to Holger
accordingly and note that PROTO SAI is deferred to SAI 4.0. That is a
more precise statement than what we had.
On the external vs embedded SAI distinction — this is the question we
most need to test. Our current loader uses:
const scene = canvas.browser.currentScene;
const node = scene.getNamedNode('KeyLight');
node.intensity = 0.9;
The scene loads correctly with baked-in values. After this assignment
fires,
the scene dims. The assignment succeeds without error but appears to
overwrite valid state with a broken value rather than updating it.
If external SAI requires X3D. prefixing on classes and property access,
we may simply be using the wrong binding mode entirely. Can you confirm
whether the correct external SAI pattern for X_ITE property assignment
on a named Material or Light node would be something like:
X3D.node.intensity = 0.9
or whether the namespace applies differently? We are working from
https://www.web3d.org/documents/specifications/19777-2/V3.3/Part2/
but the external SAI section is not explicit on X_ITE's binding
conventions.
Regarding your serializer work — the offer to revisit with AI assistance
or SAI 4.0 is interesting to us for V2.2. Our current approach generates
scene content server-side (Python/Flask) and drives updates client-side
via polling. If there is a cleaner path through X3DJSONLD serialization
we would like to understand it. The JSON path you mentioned:
https://github.com/coderextreme/X3DJSONLD/blob/master/src/main/node/json2js.js
looks like it could bridge our server-side field state output directly
to scene update code. Happy to share our scene file and update patterns
if that would be useful context for the revisit.
For reference: our scene is X3D 4.0 Immersive profile, five inline
PROTOs, three avatar transforms with named Material DEFs, four named
lights at Scene root. Repository at
https://github.com/artistinprocess/mccf —
the scene file is static/mccf_scene.x3d and the known issues
documentation is
X3D_KNOWN_ISSUES.md.
Thank you again for the quick response and the pointers.
len
with Claude (Anthropic)
On 2026-04-15 11:30 pm, ai-request at web3d.org wrote:
> Send AI mailing list submissions to
> ai at web3d.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://web3d.org/mailman/listinfo/ai_web3d.org
> or, via email, send a message with subject or body 'help' to
> ai-request at web3d.org
>
> You can reach the person managing the list at
> ai-owner at web3d.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of AI digest..."
>
>
> Today's Topics:
>
> 1. Re: [x3d-public] MCCF V2.1 Release Announcement (John Carlson)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Wed, 15 Apr 2026 22:29:31 -0500
> From: John Carlson <yottzumm at gmail.com>
> To: "Extensible 3D (X3D) Graphics public discussion"
> <x3d-public at web3d.org>
> Cc: X3D AI Working Group <ai at web3d.org>, cbullard at hiwaay.net
> Subject: Re: [AI] [x3d-public] MCCF V2.1 Release Announcement
> Message-ID:
> <CAGC3UEn76up=f66tDfDqLzL7q-MDLLWuP3ad2gE8-A7iPnzpQA at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> I did make a half-hearted attempt to convert DOM to ?VRMLScript? as
> X_ITE
> (Cobweb) script was called. But for some reason, I quit when I got to
> Protos for some reason, probably to wait for SAI 4.0.
>
> https://github.com/coderextreme/X3DJSONLD/blob/main/src/main/node/VRMLScriptSerializer.js
>
> If your interest remains, I can revisit this, maybe apply AI or SAI 4.0
> to
> it.
>
> John
>
> On Wed, Apr 15, 2026 at 10:17?PM John Carlson <yottzumm at gmail.com>
> wrote:
>
>> I just saw that you are generating JavaScript SAI code. I would look
>> at
>> something to convert X3D (DOM) to JavaScript, particularly:
>>
>>
>> https://github.com/coderextreme/X3DJSONLD/blob/master/src/main/node/JavaScriptSerializer.js
>>
>>
>> https://github.com/coderextreme/X3DJSONLD/blob/master/src/main/node/JavaScriptSerializerNew.js
>>
>>
>> https://github.com/coderextreme/X3DJSONLD/blob/master/src/main/node/NodeSerializer.js
>>
>> (Not tested recently)
>>
>> Output not guaranteed to work with X_ITE or X3DOM, yet, they are
>> oriented
>> around X3DJSAIL. Primarily because there?s no real SAI for PROTOs
>> for SAI
>> 3.3, except to construct from an encoding. But I might not be
>> up-to-date.
>>
>> Converting from JSON is easy:
>>
>>
>> https://github.com/coderextreme/X3DJSONLD/blob/master/src/main/node/json2js.js
>>
>> If you customize these serializers for X3DOM or X_ITE, let me know.
>>
>> John
>>
>> On Wed, Apr 15, 2026 at 9:05?PM John Carlson <yottzumm at gmail.com>
>> wrote:
>>
>>> I suggest you code to an SAI that produces an encoded file. There
>>> are
>>> many SAI?s available.
>>>
>>
>>> There?s a difference between external SAI and embedded SAI with
>>> X_ITE,
>>> for external, add ?X3D.? to classes, etc. Last I heard.
>>>
>>> Good luck,
>>>
>>> John
>>>
>>> On Wed, Apr 15, 2026 at 6:45?PM Len Bullard via x3d-public <
>>> x3d-public at web3d.org> wrote:
>>>
>>>> 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.
>>>>
>>>> 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 ? state vectors between discrete interaction events, with
>>>> TrustField
>>>> V2.1 implementing trust dynamics dT_ij/dt = ?(1 - ||?_i - ?_j||)
>>>> -
>>>> ?T_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
>>>> ? = {(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) + ?) 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*
>>>>
>>>> _______________________________________________
>>>> x3d-public mailing list
>>>> x3d-public at web3d.org
>>>> http://web3d.org/mailman/listinfo/x3d-public_web3d.org
>>>>
>>>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:
> <http://web3d.org/pipermail/ai_web3d.org/attachments/20260415/fcf67809/attachment.html>
>
> ------------------------------
>
> Subject: Digest Footer
>
> _______________________________________________
> AI mailing list
> AI at web3d.org
> http://web3d.org/mailman/listinfo/ai_web3d.org
>
>
> ------------------------------
>
> End of AI Digest, Vol 6, Issue 10
> *********************************
More information about the x3d-public
mailing list