This paper by Vi Hart and Henry Segerman has some interesting hand-drawn images that cry out for a treatment with your favourite TeX-friendly package. I like the concept of throwing up one's hands and saying 'forget it; hand-drawn images it is', but it's just as cool to see such interesting figures done by code.
Figure 1 c) is easy (for those in the know), figure 2 (or some minor variation on it) is do-able, but the 3D model of the cube with the appropriately perspectivised images on faces could be challenging. Figures 3 and 4 should follow from figure 2.
To give a concrete question:
Can figure 5 from the linked paper be generated programmatically? It seems completely systematic to me.