It is the same advice you hear for any other project - solve a problem.
The projects that work are the ones that solve people's problems. The ones that fail are the ones that try to create solutions for problems people do not have.
To add to this, advice from people successful in this area really recommend solving a problem *you* have and avoid trying to solve problems you know exist but aren't a problem for you.
This is good advice. Problems you have yourself you understand better and have probably already had some thoughts on how to solve them. Even idle thoughts in the past can be really handy when it comes time to think deeper about solving it in code.
The projects that work are the ones that solve people's problems. The ones that fail are the ones that try to create solutions for problems people do not have.