Does that cycle connect to the question of why you're starting those projects in the first place? If it's "play" and experiencing joy, then quitting after the fun part would seem to do its job.

Part of the reason is wanting to make cool stuff. I'm not even half as brilliant as guys like Fabrice Bellard but I certainly would like to be.

My strongest reason to start a project though is very much along the "self-soothing" lines described in TFA. I do it to prove to myself I'm not insane for thinking that something is possible and that things could be different. If I can think of something, surely people much smarter than me would have done it already, right?

For example, I wanted to embed data into ELF executables and access it at runtime. The accepted solution was to add sections and have the program find, open, read and parse its own executable in order to read those sections. That just didn't seem right to me, I couldn't accept it and I didn't rest until I figured out the real way to do it.

https://www.matheusmoreira.com/articles/self-contained-lone-...

I got the Linux kernel to find, open, read and parse the executable for me. It memory maps the data before the program even starts. When it does, it just needs to follow a bunch of pointers to find it. Simpler and more robust. As far as I know, no one else has done this. At least one linker out there gained features just to make this easy and efficient.

My most painful free software development experience was when I tried to contribute one of these "insane" ideas and someone described it as schizophrenic. Pretty much just dropped it and never went back there again. Patches are still on the mailing list so who knows.