Comment by mguerville

Start by identifying the problem you're trying to solve, then validate that the problem is worth solving and that your approach can get people wanting to solve it. Then figure out what's the quickest way to test this, and only then if all is validated start building something.