This is a list of ideas that I don't have time to work on right now, but I might work on later or contribute to if someone else started up the project. Naturally, all of these assume that the result is based on free and open-source software, uses open standards and protocols, leverages federation if appropriate, etc. Shoot me a message on Mastodon if you want to pick my brain on any of these ideas, or to let me know they exist.
A social networking site which is designed to enrich IRL relationships. Your instance would probably be run by someone in your neighborhood, and if you wanted to hang out it'd match you up with people with similar interests and help you hang out together IRL, be it having drinks at a bar or playing video games or board games for an evening, or whatever else. Instead of "social networking" sites which try to keep you stuck staring at their pages for as long as possible, the prime directive of this site would be to quickly get you off of your phone and into IRL face-time with other people.
A marriage of git and bittorrent, for the purpose of tracking large blobs. Like git-lfs, but decentralized. Instead of storing the URL to fetch files from like git-lfs does, it stores the infohash. git push should block until 100% of the changes are replicated in the swarm.
A vector graphics display with open hardware, an open standard for driving it from a PC, and upstream kernel drivers implementing that interface. Could be CRT-style or a laser+mirror+projector kind of system, or both. Kernel interface should be fairly general to encourage manufacturers to write drivers for it, and should be paired with a nice userspace library for driving the ioctls, like DRM but for vector graphics. Relevant resources: Hackaday, Arduino RTOS.
A Wayland port of xscreensaver based on layer-shell.
A new kernel for z80 calculators. KnightOS is far from the mark in terms of POSIX support and general Unixisms, but it'd probably be a good start. There's no reason a POSIX-compatible operating system wouldn't work on these devices. Particularly egregious failures in KnightOS include the file descriptor design and the lack of a TTY subsystem, though depending on who you ask that last one isn't so bad.
A simple replacement for gas/nasm and binutils, with less GNU and more modularity. Ideally should pair well with cproc/qbe.
An ethical engineering index which defines a set of criteria by which tech companies are evaluated for ethics and assigns each a score. Factors include their approach to military contracting, their treatment of their employees, attitude towards open source, etc. Above a certain rating, companies would be allowed to post job listings for free (under that level they wouldn't be allowed to post at all). There's no business model here, ping me for free hosting.