[x3d-public] x3d-public Digest, Vol 205, Issue 73

[John Carlson]

Hi Andreas, I am considering X3DJSAIL and X3DJSONLD code generators and how
they could generate code more amenable to revision control.  Some use
incremental integers to create unique variable names and others create
hierarchies susceptible to rapid revision when modified using LLMs.  It
seems like we need specific tools for this rather than line by line diffing.

That said, xmldiff and jsondiff do a pretty good job, except on
JavaScript.  Now that numbers are being rounded in SAI, that’s good too.

Ideally, one would rely on XML being what changed, and you accurately state
x3dom.  But with HAnim, that might be too many tokens, and run up a lot of
money.  My hope is for non-mesh equations with parameters to generate
humanoids.  I saw something mentioned on that recently, so I’m coming back
to that.  If only 5% of DNA is used, are we getting in the LLM ballpark?

Ideally, indeed, something very high level, like English is desired,
controlling an existing humanoid.

I think you mention this:  To translate a humanoid from .x3d to .glb, I
don’t feed LLM humanoid.x3d and humanoid.glb, I tell it to convert basic
x3d nodes to .glb, and produce python to do the conversion, which I feed
back into LLM, enhancing the python.  The LLM can look up X3D and glb
(looking for python libraries) on its  own.

John

On Thu, Apr 30, 2026 at 8:09 AM Andreas Plesch via x3d-public <
x3d-public at web3d.org> wrote:

> Hi Len,
>
> I forked the repo as the easiest way to just take a quick look and try
> some things. My main interest was to see how the X3D interface works and
> how x3dom might be a good match. I think it could be that by using more
> direct DOM manipulation of the X3D scene which could be simpler and more
> 'understandable' to the LLMs generating the code. This would also mean
> moving logic out of x3d scripts and into an external bridge layer. Please
> proceed at full speed since I do not anticipate being able to contribute to
> the core MCCF idea and implementation.
>
> I noticed there is a non-LLM agent to choose from when running an arc.
> This could be useful for development purposes. Is there an expected impact
> on the x3d visualization doing so ?
>
> It may be nice to have the coding assistants generate more fine grained
> commits and PRs with explanations for human consumption, perhaps when
> things settle down.
>
> Andreas
>
>
>> Date: Tue, 28 Apr 2026 14:25:27 -0500
>> From: cbullard at hiwaay.net
>> To: X3D AI Working Group <ai at web3d.org>, x3d-public at web3d.org
>> Subject: [x3d-public] MCCF Implementation Specification V3
>> Message-ID: <147eb1b15cc3b2e324ae241a76bcc11e at hiwaay.net>
>> Content-Type: text/plain; charset=US-ASCII; format=flowed
>>
>> Good afternoon:
>>
>> This is the github link for the Multi Channel Coherence Field (MCCF)
>> version 3 implementation.   This was created after review of the draft
>> by all participating members (Claude, ChatGPT, Grok, Gemini and Len
>> Bullard).  Work on implementation begins tomorrow.
>>
>> Comments or suggestions are welcome but we are not adding additional
>> requirements. This version is where the relationship to the X3D standard
>> and implementations gets tighter.  I am will ask questions in this forum
>> to help clarify this.  We are currently running the X_Lite viewer and
>> sanity testing on the X_Lite editor.  I am happy to include other
>> viewers as possible.   The XML design is clean for exporting and
>> implementing scenes.  I am happy to provide any instance of these to
>> interested parties.  We will start keeping those in the GitHub.
>>
>> I note that one member of this community has forked the repo.  Thank you
>> for your interest.  Your contributions are welcome.  We are going fast
>> so refresh as you need.  A goal of the project is to keep as much
>> content contexts (application domains) as external as possible to ensure
>> the general applicability of the MCCF to different applications.  Some
>> of the original narrative basis has leaked into the design but we will
>> clear that out.
>>
>> Thanks,
>>
>> len
>>
>> https://github.com/artistinprocess/mccf/blob/master/MCCF_V3_SPEC.md
>>
>
>
>
> Andreas Plesch
> Waltham, MA 02453
> _______________________________________________
> 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/x3d-public_web3d.org/attachments/20260430/9ca59122/attachment.html>