Finish everything you choose to start
To me, this means that every project you start should end in a reasoned and conclusive way. It does not mean that the initial goal must be accomplished, or that you should waste time completing a project that has no value just because you started it. Instead, to "finish" something in this context is to either (i) accomplish it (even if the initial project or idea morphs into something new), or (ii) stop working on it for a specific articulable reason that does not amount to lack of motivation.
Consider for a moment a product feature you decide would be super cool to build. You start sketching it out and eventually begin writing code. A week into the project, you realize that there is no market potential for you feature (perhaps through speaking to somebody else on the team or realizing that one of your key assumptions is flawed). You decide to stop working on the feature, because there is no market potential for it. In this case, you "finished" what you started by stating a reasoned and conclusive decision for ending the project.
In contrast, consider a blog you decide to write. You put together an outline, consult with some colleagues who provide encouragement, and begin writing the first paragraph. Shortly after starting, you end up getting sidetracked. A week passes, then two weeks, and eventually you abandon the idea. Unless you have a clear reason for stopping your work on the blog, you have not finished what you started.
The goal in thinking about things in this way is to get yourself to consider why you are taking on a particular project in the first place and, if you end up not following through, stating your decision to stop in a clear, reasoned and conclusive way.
Ideally, this will help you only start projects that make sense, and keep you from abandoning projects without evaluating your reason for doing so.