Wouldn't it be better to use curves (i.e., shapes) of equal areas for each of the nucleotides? Right now, A and C have areas of + and - 0.25 units respectively, while G and T have areas of + and - 0.5 units respectively.
In some applications, like data storage in DNA, it is likely that optimal encodings have an equal number of each of the nucleotides, and having balanced representations could make it easier to spot imbalances and thereby compare different encodings visually.
This won't eliminate any of the other benefits stated in the blog post, but I think it will broaden the application of this visualization.