I struggled with this for a long time, I think a lot of us do. I don't know how broadly applicable my story is, but: at a certain point I basically cracked and picked a project and said "I'm going to finish this if it kills me." It was one of the larger-ish projects that I'd made a bit of progress on and put aside. The project became an exercise in finishing things, which helped motivate me through the times when I wasn't really interested in the project itself.

It took me longer than I care to think about to complete it. I think by the end I was pretty clear with myself that I wasn't really that interested in the project itself anymore; the point was to learn to finish something.

The result is here:

https://github.com/zenhack/haskell-capnp

Somehow this actually worked, and I've found myself not having that much trouble sticking with projects when I decide I want to build something. Most of my 2019 hacking was on a new programming language, which is "close" to being ready to announce (the code is up there if you look for it, but I haven't been too loud about it since I don't know what I'd tell someone if they showed up and wanted to help), though Sandstorm got active again and that's diverted some of my time. ...and I've been hacking on Sandstorm pretty consistently since.

Keeping myself organized and on track is still a struggle, but something shifted and sticking with something long term doesn't seem as hard.