Now here's something more original, that probably would only work with a virtual spirograph:
t2 += cos(t)*0.2;
Now the rotation of the second arm is dependent on that of the first.
And you get things like the attachment. I will experiment with sine and tangent and whatever else I can think of.