Work on the free software projects I keep wanting to then being too burned out from the work that pays the bills to do anything with. SBCL could use a good graphics toolkit and editing environment. Free graphics and audio production software all need more love, as does CAD. It still baffles me that the free 3d graphics package that won is Blender, when enough users want Houdini to make it a successful commercial product. A computer algebra system that could work with hand-written expressions, for tablets and those interactive whiteboard things, would be very handy. Axiom and Reduce would both be decent CASes to build on, are free software, and would benefit from having a decent environment built on them. I keep thinking it's a shame that these great communities form around extending videogames, then fall apart after a few years when they're no longer the cool new thing, and it would be nice to have a free platform that could be kept graphically modern for people who want to do things with games but don't want to build one from the ground up. There have been attempts to do things like that in the past, but they're either deliberately primitive to limit their scope or never reach a usable state because they're too ambitious for a hobby project.
FreeNet seems to be dead and without successors, which is a shame because it or something like it would be better than Tor hidden services for some applications.
I work (no, actually, I play) in Ableton and it's still my favorite piece of software in any vein that I've sampled. I run CAD software too, and I can tell you that the industry standard 3D CAD functionality doesn't even come close to what Ableton does in the audio world: overall application stability, plug-in capabilities, sleekness of GUI, recovery of crashed sessions, etc. Much respect for working in software, it's an under-appreciated corner of tech.audio production software
Proprietary software for scientists and engineers all tends to have the same problem; their users aren't (primarily) programmers, so they'll only write their own tools if it doesn't take too long or there's nothing close to what they need available, but they will figure out a way to solve their problems with whatever they can find. It's really, really hard to write software so shitty those guys won't be able to make it work, and as long as they can make it work they'll be content. That makes it hard to make a commercial case for doing any better than just good enough.
Yeah, buncha my coworkers have built up personal libraries that they are loath to share with anyone. Not that it would do you much good anyway... doesn't matter how well it's commented, you're going to end up trudging through the mud to figure out what's going on, and you might as well have just written your own program to begin with. Most ridiculous piece of software in my industry has to be SIMION, for that exact reason. C'mon, even the website is straight out of the 90's.That makes it hard to make a commercial case for doing any better than just good enough.