Inspired by this post. I also mentioned this idea there and they've decided to borrow it.

I hadn't really thought about optimizing scripts for ink-efficiency before, but it's kind of interesting. I noticed that any linear script could (probably) be simplified by removing the interiors of the lines, leaving only the endpoints, and at that point you basically just have a dot script, which, fortunately, is to a certain extent already in Unicode in the form of Braille, at least if the dots are aligned to a 2x4 grid. It's also accessible to blind people.

We can make this kind of system featural by using certain parts of the dot grid to indicate features, such as the first dot to indicate voice. For non-binary distinctions (with more than two options), we can use multiple dot positions, for example in a 4-way classification we can use two dot positions with the first option being 00 (no dot), the second 10 (one dot in the first), the third 01, and the fourth 11.

Here's a sample (definitely random) phoneme inventory: /kgtdpb xɣɕʑθ̠ð̠szfv ɴnm uoaei/. There is a clear 4-way classification for the manner of articulation. We want to avoid any letters having no dots, so since we know voiced letters have at least one, we can safely indicate vowels (which are all voiced) with a 00 manner-of-articulation pattern, and don't have to worry about this anymore since the other manners of articulation have dots in their manner pattern.

Hopefully that explanation is involved enough to see how the rest of the system would work. Here's the end result; the first dot is voice, the next two dots are manner, and the last three are place:

• uoaei 100001 100000 100100 100010 100110
• xɕθ̠sf 010000 010100 010010 010110 010001
• ktp 001000 001010 001001
• ɴnm 111000 111010 111001

/kgtdpb xɣɕʑθ̠ð̠szfv ɴnm uoaei/ = <⠄⠅⠔⠕⠤⠥⠀⠂⠃⠊⠋⠒⠓⠚⠛⠢⠣⠀⠇⠗⠧⠀⠡⠁⠉⠑⠙>