[rvz]

For learning, pick one that solves a problem just for you.

For research, pick an existing project and make it efficient.

For building a startup, pick one that solves a problem for many people and makes money.

The last one doesn't come without problems and now has just gotten even harder. It used to be 90% of startups failing.

Now it most likely is going to be 98% of startups failing.