We have already gained some experience with needing a networking library through our chess engine, as we created a "online" version of the app allowing to play remotely. However, that was quite a poor implementation and didn't have all the feature set that would be wanted in a proper networking library. Hence, it would be interesting to build a library that would allow use to use networking in all our projects. For example, have it work with the Game Engine to allow online games.
This would need to have all the necessary error-case features and allow for custom data packets and sizes to be sent, unlike our chess networking that had fixed sizes.
"Machine Learning" is quite a vast term due to there being so many options of what sort of Machine Learning / AI implementation you want, whether it is a image recognition algorithm, audio recognition, a large language model or a generation based game player. We have left it that way because we couldn't pick which one was more interesting to experiment with than the others.
This would be a compiled language, similar to C (Hence the name Uni-C). It seems like an interesting project to take on due the need to manage memory and have OS-Specific kernel interactions ensuring that we produce an executable that can be ran regardless of the device it was compiled on. To make it easier for ourselves it itself wouldn't produce machine code but rather assembly that can be properly compiled by an assembler as we can then keep some stuff cross-architecture.
Designing a filesystem will have a close difficulty to our own language, as there are a lot of test-cases you need to look out for where it could easily screw the whole thing up and all data stored will be gone. However, if functional would be a really cool thing to have, and of course port to our operating system.
Note: These are not in order of implementations