However, if it's good for that, I will probably be buying one.
It sounds like it relies heavily on autocomplete, which means it's unlikely to be useful for coding instead of just English text.
Like the gesture for python's "create a for loop" would align with the english word "repeat" and in functional languages it might map to some manipulation that sets up recursion with a shortened list.
Actually coming up with a list of such a list of abstractions for app developers to map actions onto sounds like a dissertation for some psychologist somewhere.
i wonder how a trained stenographer on a chorded keyboard would compare to a trained typist with a good english language model backed autocomplete and a ui built for speed on a classic qwerty keyboard.
my understanding is that classic stenography is actually phonetic. so the chords match up to phonemes or phoneme like chunks which are then postprocessed to reconstruct english text.
a new approach using english language could be more data driven. a simple mapping could be one chord to one word with words sorted by frequency and easier chords assigned first. more complicated approaches involve chunking up the words into frequently used chunks and then doing the same.
code is harder, there are frequently reused strings, but they change from technology to technology. overall there is far more entropy in computing than english on a character by character level, so designing a chording scheme that is more efficient and isn't tied too much to a specific domain is actually a really hard problem, especially once you consider that the effectiveness of a given scheme is not only a function of how well it fits the problem (how often the user is actually entering things the scheme was designed for) but also how well users are able to learn the vocabulary of chords and the dexterity to execute them quickly.