For me, a good side project needs to strike the right balance between challenging and achievable. On the one hand, if it doesn't stretch my abilities and force me to learn new things, I'll quickly get bored and move on. On the other hand, if it pushes me a little too far, to the point where I can't make progress during the time I have available, I'll probably stop coming back to it.

I think that tinkering with procedural generation could be a good candidate, as something that's easy to start simply and iteratively bump up in complexity. I have grand ideas of building a sandboxy gardening game, complete with procedurally generated plants in a little simulated ecosystem.

My plan is to start off by making a simple plant generator along the lines of an L-System and see where I go from there.