assemblies and tolerances are where AI generation quickly gets hard. but, for simpler things, it can be surprisingly good, especially for people who has no CAD knowledge. it certainly might feel a bit like brute-forcing through the LLM - but this also applies to vibe-coding..

Sure! but, for 3d printing hobbyist who wants to create and print a cube, both are able to provide a printable cube.

do you know the WASM bundle size of Python and CadQuery? OpenSCAD has a lot of cons in terms of concept and CAD approach, but there are things where it is good - it is embeddable and sandboxable and fast. It gets the job done in simple models and that is the reason why it is popular. It reminds me a bit how everyone was bashing PHP because 'it is not a programming language' while wordpress ruled the world.

No, the comparison is not apt with PHP (which I use on the daily and have done for twenty years). It's a bit more like comparing CSS with a subset of JavaScript that can do the same tasks. OpenSCAD isn't, really, a proper CAD package. It's a CSG drawing library with a declarative syntax. It is like nested CSS for 3D things.

Code complexity and repetition balloons with part complexity. Your thing will not scale well because OpenSCAD does not scale well.

Small changes will lead to subtle flaws; changes will need to be respecified in minute detail because OpenSCAD doesn't support sketch constraints that would generate correct geometry, doesn't support edges and faces that can help make parametric designs flow.

OpenSCAD is fun for making small, trivial objects, it is a good way to learn the most basic CSG operations, and it really excels at 3D representations of mathematical formulae.

But it's entirely the wrong foundation for the more general CAD-replacement thing you are trying to do.

CadQuery is not at all as bad a foundation (I think it's quite neat, and I think it could work).

But IMO much of this article talks about "findings" that should have been obvious to you if you know enough about CAD to know what it is you are replacing.

If you are continuing in this non-bRep, fast calculation approach, have you looked at generating SDF instead?

"Downloading WASM: 40mb of Python and build123d" - this hurts :( This is exactly the good thing about OpenSCAD: its WASM build size is minuscule compared to Python build, and you can edit/customize a model right in the browser. https://modelrift.com/models/desktop-organizer-with-drawer

CAD for 3D printing is finally becoming feasible with Flash 3.7 and 3.8. Exciting times. https://github.com/ModelRift/openscad-skill/

https://modelrift.com/blog/openscad-llm-benchmark/ LLMs are still weak at spatial reasoning, but it gets better. Check out modelrift.com for another alternative

take a look at modelrift.com, it is built around annotating built models by basic pen and arrow tools, works fairly well ('smarter' model is significantly better)

check this out: https://modelrift.com/blog/openscad-llm-benchmark LLMs are getting pretty good at OpenSCAD, especially if they work in agentic mode and can inspect intermediate results.

these small functional prints are exactly where OpenSCAD and LLM generation shines

I agree, my main concern regarding Google AI products is this endless pain around the UX of login / billing / upgrades / product sunsets... but their LLM models are good and Antigravity 2.0 is not that bad either (unless you lost all you Antigravity 1.0 setup and projects - like many people did)