For me, the interesting case is smaller low-poly or voxel scenes where loading a full 3D stack may be overkill, and where keeping the scene in DOM/CSS gives you easier integration with normal layout, styling, events, etc. Once you have the HTML, you don't even need to load the library to render a static model.
Also, part of the experiment is testing the browser’s limits and getting a clearer sense of where this approach works, where it breaks down, and what the tradeoffs are.
Cheers!
Would you recommend this for hacking around or not?
Whereas THREE.js or webgl is purpose-built for realtime animated 3d scenes.