the demo of the APIless method of orchestration

Thank you Daniel

I remember asking several times to see the demo of the APIless
orchestration you mentioned, because I was genuinely interested
although not related to KR or this CG, it is much needed.  I want to see
such a thing!

But I do not remember seeing a demo, it could be that it got lost in
threads *I stop following threads when I realise that go off some tangent
as I am in desperate need to get on with my work

When you share it, please mark it clearly in the subject line *demo of
APIless method for LLM orchestration;  even if not
related to this CG

Do not forget to link it in the wiki and post it here, together with any
other record/pointer/link to contribution you make to the CG
But please do label it in such a way that we know it is what we have been
expecting to see for some time  *apologies if you have already shared it
and I have missed it

I  was so looking forward to it, and even created a space in the wiki for
your architecture, because you said you never edited a wiki before
 I then figured out that
*that I could not process  the architecture it within my framework *and
that it was out of scope of the CG etc

You can continue to edit  the wiki page to see fit explaining how your
architecture aligns
with the CG *or not . You can create a page for all your contributions, but
please explain if/how they relate to the scope of work
I understand that wiki s are difficult to edit and why you prefer another
platform, Just point to it in the wiki

There is a thread 'how to contribute to the CG on this list,  to guide
contributions that are in scope.

The demo I gave at Tpac on 14 Nov, during a breakout that you attended is
not related to your orchestration method at all
It has absolutely nothing to do with it. I find the need to continually
explain the obvious too straining. Thus I must avoid engaging
in continued exchanges.

During that meeting I asked you to demo the orchestration method  *the zoom
meeting is recorded
but you said it was not ready yet, nor it was on 17th Nov meeting.  I also
said multiple times 'show it any time it is ready'

There is nothing in the project  AI driven WEB standards that you link to
in your email below t related to LLM orchestration whatsoever!
My proposal does not leverage a cross- multi model approach

I have however  used a multimodel approach for a long time in my work
manually *by feeding the output of one model
as input for another as a form of refinement.

I have been conducting cross model evaluations and a bunch of other things
'manually'

This is why I remain interested in your method that bypasses APIs, if you
could share it
because it is APIs that break the multimodel ecosystem AND are expensive

But not related to KR/CG

I  forked Karpahty's  LLM Council's work
https://github.com/karpathy/llm-council/fork
and developed a web based version of it on colab with some difficulty,
which I have just debugged
it would break because of APIs. I am busy working on this and not ready to
share it but since you are twisting my arm

I opted for a workaround,  and it now runs *using Google free API that can
access only Google models *API only breaks a little here and there but
produced valid outputs I obtained two meaningful outputs *deliberations, I
pasted them on HF somewhere yesterday
https://colab.research.google.com/drive/1exmNTcMTDDcFmUfBRYHShDGJlr7LJntU?usp=sharing

The problem seems to be that APIs break  orchestrations. So we look forward
to your method

This LLM council thing is not related to KR or this CG
But please when you find an APIless method that works, please do share it
here too and do link it in the wiki

I really need to get back to work! I may continue not to read emails
without an explicit subject line

I take the opportunity to note that we are having this tangential
discussion about other projects under the subject line of a thread
that was discussing another topic. It is good practice to start a new
thread with an appropriate subject line when changing the topic

Have deadlines

P




On Thu, Dec 4, 2025 at 8:00 AM Daniel Campos Ramos <
danielcamposramos.68@gmail.com> wrote:

> Thanks, David — your clarification is exactly how I understood it too.
>
> Paola, a few concrete points so we’re all aligned:
>
>
>    - The NotebookLM already contains the full disclaimers, links back to
>    every CG source, and you’ve had editor access since the day it was created.
>    it’s not “my” repo or W3C’s: it’s a study aid so newcomers don’t have to
>    chase PDFs across dozens of threads.
>
>
>    - The CG routinely cites external references (Quine, ontology PDFs,
>    etc.). Consolidating them doesn’t change scope; it’s the only way to keep
>    pace with the materials you share. If anything needs updating, you can edit
>    directly or ping me — I’ll do it immediately.
>
>
>    - On the “scope” question: I shared the MVCIC orchestration method +
>    specs on Nov 10. A week later (Nov 17) the “AI-driven Web Standards
>    Specification” CG was proposed and by Nov 28 it was launched. Both posts (
>    https://www.w3.org/community/blog/2025/11/17/proposed-group-ai-driven-web-standards-specification-community-group/
>    and https://www.w3.org/community/aiwss/) cite the exact mandate we’ve
>    already been discussing here. If we’re going to school each other on
>    attribution, let’s at least acknowledge when work already in the AI-KR
>    threads is used to justify a new CG. That’s the fair way to keep provenance
>    intact.
>
>
>    - If there are specific terms or KR artefacts you want captured inside
>    the CG tools, tell me exactly what format you prefer, and I’ll put them
>    there with the same references already logged in the repo.
>
> In short: everything is fully attributed, open, and cross-linked. Let’s
> keep the energy on the substance, not on where the files sit.
>
> Daniel.
>