[alexose]

As a huge OpenSCAD fan and everyday Cursor user, it seems obvious to me that there's a huge opportunity _if_ we can improve the baseline OpenSCAD code quality.

If the model could plan ahead well, set up good functions, pull from standard libraries, etc., it would be instantly better than most humans.

If it had a sense of real-world applications, physics, etc., well, it would be superhuman.

Is anyone working on this right now? If so I'd love to contribute.

[switchbak]

OpenSCAD has some fundamental issues with which folks are well aware. Build123d is a Python alternative that shows promise and seems more capable, and there's others around.

Hard to beat the mindshare of OpenSCAD at the moment though.