Let's see:

• Back in 2007 or 2008 I attempted to create a CPU architecture that directly uses Brainfuck as its instruction set. I had to put it on hold before it was completed because I had a custom FPGA development board with really bad documentation but if I ever get my hands on an affordable FPGA, it will get done eventually.
• I've created a nonogram that solves to a rickroll QR code. I had to rely on the error correction because the exact pattern didn't result in a well-defined solution but I've recently learned about some more parameters that you can tweak on a QR code. So now I just need to acquire or more likely build a QR code generator that lets me manually control those parameters and an automatic nonogram solver so I don't have to manually solve a bunch of 25x25 nonograms to confirm they have a single solution.
• My plan for tonight is to start porting a 22-year-old handheld game to a ~35-year-old home console. I've acquired a C compiler but will probably have to learn assembly for a CPU architecture that was barely used for anything else. There is no chance to ever share the resulting game without getting sued to hell and back again.
• I've made chainmail bikinis for a couple of friends.
• Edit: One more because it might be my magnum opus. Have you ever played KJumpingCube? That doesn't only work on grids but on arbitrary graphs. My friends and I chose a Risk board. Not a digital one. A real life physical Risk board with actual dice on every country that need to be turned by hand. A single game took us about 6-7 hours with the winning move alone taking up the last hour.

That's just what I comes to mind at the moment. I'm sure if I spend some time thinking or digging around old hard drives, I can find more.