Hi Kurt,
Thank youbthatbs a very interesting perspective, and I appreciate the kind
words.
What youbre describing with Markdown and Databooks feels very much aligned
with the architectural direction I was trying to capture, but emerging through
a different path. In particular, the idea that Markdown is becoming a de facto
interface layerbespecially in the context of AI systemsbis something I
hadnbt fully considered, but it makes a lot of sense.
Your point about minimizing the need for large JavaScript layers also
resonates strongly. In the paper, I framed the issue as moving away from
imperative orchestration toward declarative specifications (pipelines,
transformations, queries). What youbre describing seems to push in the same
direction, but with documentsbMarkdown + metadata + embedded logicbserving
as the organizing structure.
The Databooks concept is especially interesting. The idea that a document can
encapsulate data, metadata, and executable logic starts to look very similar
to a declarative application model, just expressed in a different medium. In a
way, it feels like a modern, lightweight counterpart to what we were trying to
do with XML + XSLT + pipelines, but adapted to todaybs tooling and
workflows.
I also find it notable that you can embed XML and XSLT within that model. That
reinforces the idea that these technologies still have a role to play, even as
the surrounding ecosystem evolves.
Ibd be very interested in seeing where you take this next, especially around
the bsemantic executionb idea. It feels like therebs a broader
convergence happening between data-driven, document-driven, and declarative
architectures.
Thanks again for sharing thisbitbs a very useful perspective and a nice
complement to the discussion.
Best,
Roger