Just leaving this here — you are up to something. Karpathy:
> Products with extensive/rich UIs lots of sliders, switches, menus, with no scripting support, and built on opaque, custom, binary formats are ngmi in the era of heavy human+AI collaboration.
If an LLM can't read the underlying representations and manipulate them and all of the related settings via scripting, then it also can't co-pilot your product with existing professionals and it doesn't allow vibe coding for the 100X more aspiring prosumers.
Example high risk (binary objects/artifacts, no text DSL): every Adobe product, DAWs, CAD/3D Example medium-high risk (already partially text scriptable): Blender, Unity Example medium-low risk (mostly but not entirely text already, some automation/plugins ecosystem): Excel Example low risk (already just all text, lucky!): IDEs like VS Code, Figma, Jupyter, Obsidian, ...
AIs will get better and better at human UIUX (Operator and friends), but I suspect the products that attempt to exclusively wait for this future without trying to meet the technology halfway where it is today are not going to have a good time.
That really seems like it is written by someone without any experience using 3D CAD, particularly of the "make physical stuff" variety, rather than the artistic movie and video game asset variety.
English (or other human) language is pretty bad at describing 3D geometry. People have been using drawings and physical 3D models to describe designs going way back into antiquity. The idea you can just use text suitable for a current-design LLM to manipulate 3D technical data really seems like an idea from someone who has never tried.
But also, Solidworks and Inventor (and I think NX) have extensive API coverage that would allow an LLM to manipulate the geometry data, if the LLM could figure out what the data meant, and what it wants to do. If you can do it on screen with a mouse and keyboard, you can likely do it over the API. An add-in could certainly be written to accept requests for information and instructions on api calls to make from an LLM.
Maybe you will be a good judge of this, since I am not: try using claude to generate an openscad model by describing it. Open it up and iterate with claude until you get close to what you want. Did it work?
I've done this multiple times with great success, resulting in OpenSCAD projects that I feel comfortable maintaining, but would not have created without LLM assistance, and it has definitely been productive for me to start an OpenSCAD project this way, learn from the results, and refine things with ease.
Be sure to emphasise to the model that things should be well-commented, and be prepared to be amazed at the depth and breadth (pun not intended) of knowledge that can be obtained about the OpenSCAD ecosystem using this technique.