Sure I can explain more. My process amounts to trial & error, guess & check, with some eureka moments, but I am trying to solve problems in a mostly straightforward manner.
First some background. Right around the time I learned TP (5 years ago) I was also investigating shorthand and always thought it would be cool to make one for TP. But shorthand was hard, unlike TP, and not really straightforward, so the whole idea took a backseat to the basics of both. I learned Teeline shorthand and after highschool I learned Gregg. Teeline was relatively slow, not aesthetically pleasing enough for me, and its rules are as convoluted as English grammar's so I didn't think an adaptation of it would catch on with the TP crowd (which as I understood it had many people from the Esperanto ranks). shorthand was pretty cool; it was essentially a synthesis of a 1700-1800s English shorthand called Pitman and the continental shorthands, many of which are German in origin such as Gabelsberger. These continentals are based on simplified forms of the cursive Roman alphabet. Pitman is based on simple geometric shapes, essentially straight lines and arcs of circles. The clincher for Gregg and Malone (the creators) is that the forms of Pitman are really not ergonomic and are simplified or rounded further and frequently blend in ways that aren't part of the theory per se; so Gregg takes a cursive approach like the continentals, and the cursive forms of Pitman, and works them into the theory. He also provides a geometric background for the theory in the form of the ellipse, which takes the place of the circle and the derived arcs in Pitman. So you really have the best of both worlds in my opinion; I am a Gregghead so you can consider me biased.
Learning Gregg for me was also about improving my longhand. Since it provided a ready model to emulate it appealed more to me than Pitman, and I could use it in place of longhand when I wanted to, allowing me to focus on making the letters pretty when I wanted to and using shorthand when I wanted to write fast and essentially relieving a lot of writing frustration, which would further aid my writing. Writing longhand would be for presentation and other people and Gregg would be for me (and no kidding, since it's basically a secret code too).
Japanese shorthand piqued my interest since it looks so crazy. Gregg as you pointed out, looks rather like Arabic except written in the wrong direction and composed of Copperplate-esque flourishes. Japanese shorthand looks like the hardest English shorthand Cross's Eclectic except, unlike Cross people actually use it! Compare: with ,, and ; I think Waseda even looks a little crazier than Cross with how it loops. They're also very large and have straight downstrokes which are nonexistent in Gregg and much cursive Western writing, but are very common in Japanese. The fastest writer of Gregg shorthand had a large writing style and his writing is very easy to read because of it and japanese shorthand looks like it is as much native japanese shorthand script (aka Sousho) as it is a Western-based system derived from Gregg, Pitman, or Cross. So this is what I mean when I say I want it to look like Japanese shorthand.
On the other hand, Japanese systems usually use stroke length and vowel/oval size to distinguish syllables or features, depending on the system. I have it in my head that the system has to be largely scaleable so that it function as a form of calligraphy or graffiti or what-have-you for toki pona, and if I use stroke length it will be difficult, because people will get very used to distinguishing and writing certain sizes of strokes that will be difficult to enlarge or exaggerate without hindering read-back. Toki pona is also mostly short words unlike Japanese so there is an inherent limit on how novel it will look written out. Which is where personalization would come in. I expect people to write this shorthand and because it's so simple they will be able to make the angles on their first or second try writing it, and be able to run with it without any fears of being unable to read it back. There are also so few words in TP that I expect once someone has something down they will stop reading the strokes letter-by-letter. This is something I hoped to thwart to a certain degree in creating the contracted syllabary but I don't think it worked; people will learn how to spell syllabically all the words and then convert their shape to memory; even reducing the number of syllables, I don't think there are enough minimal pairs to make people distinguish features. In a typical English shorthand many of the outlines will have multiple readings since there are too many vowels in the language for them all to be represented in a equally functional, quick way, so Gregg combines long and short vowels. Voicing of consonants is indicated partly by length (i.o.w. 2 sizes of the same stroke indicate p and b; but the third indicates s), although length also indicates other things. In TP, even using the contracted syllabary, there is nothing to pressure a person into being as consistent as would a writer of Gregg have to be, but I wanted to give a person that option by not specifying length as a feature. By only specifying the angle of stroke I would also allow other people to look at and get used to different ways of writing the same thing. A person can be as sloppy or tidy in their writing as they see fit without harming legibility (theoretically). Of course, people will get set in their own ways and writing and that's OK, so long as they are quick. I wanted to give people a lot of leeway which they don't have in any other kind of shorthand while still giving them a decent model to follow, choosing to follow it as loosely or closely as they like, so long as they are comfortable in their execution. Gregg, when I first began learning, looked unapproachable, what with all the beautiful writing that was available, but I have really gotten used to a lot of it, and then some of it I am still learning and improving upon; but I can see that the more relaxed you are while doing it the easier it becomes. In some ways, writing execution is like pronunciation of a language that you continue working on, if that makes sense.
A large part of English shorthand is phrasing (joining together words without lifting the pen) and abbreviation. I didn't do much of that but that would be next on my list once I get the thing banged out. I am considering a "joining stroke" which would function like a written space (such a thing does not exist in any shorthand I know of). Disjunction and crossing are used in Japanese but I've ruled them out. Abbreviation is not something I am too keen on but it's very doable because of how few words there are given the number of features. It seems a little inelegant but might actually make it look cooler. It would be appealing to me if this were considered not just a shorthand but an alternate alphabet, and I think no abbreviation would be good for that. I don't think the Esperanto people would go for it either.
I am writing off the top of my head; if something I said doesn't apply to the shorthand systems above, it certainly does to the shorthand I am hashing out. These are some of my aesthetic and practical concerns. When I get to posting it I will include more of my thought processes. I have included a demo of where I am at. I will almost certainly change things around to get better connections. The strokes are adapted from Kotani shorthand. Jan Akesimun, is this what you would like to see in the way of a CC license?
EDIT (Tuesday evening):
To give you a better idea of my though process I will give you an example off the top of my head. Since yesterday I have played around a bit and I have a new opinion on some things.
For instance, the rotation of the circles for representing vowels is rather nice -- it's regular and schematic. But in practice it removes the flexibility that I liked when I first decided to use two curved forms for certain letters (i.o.w. the vowel determines the choice of curve in many cases -- you'll see what I mean if you try writing 'sa' and 'se' using the scheme above). Writing out several words joined together with the space-ligature looks rather nice -- it actually looks a lot like the japanese shorthand examples I posted. Unfortunately it looks rather monotonous and a little too scratchy, which is unfortunate because I like that it seems to encourage idiosyncracies, it encourages them a little too much. If someone writes in a uniform manner the result shouldn't be too difficult to read because it looks boring. A lot of that could be helped by being careful in matching consonants to their shorthand symbols, and that would probably work for TP pretty well because of the small corpus. On the other hand, I don't really feel like I could find it. I think it would be solved by removing the counterclockwise/clockwise scheme that I have for the vowels and allowing free choice of curves. I feel like it's not too much to include a vowel distinction based on oval size, so that 'a' is a big oval and 'e' is a little oval. Making the big and small ovals is pretty easy. It won't look so monotonous either.
As far as the mapping scheme is concerned, there are about 6 descenders to 4 others. That means that depending on the chosen layout and consonant distribution in TP the shorthand could have an overall downward and right direction. This is a particular concern if the words are to be joined rather than returning back to the line for each word. It's also troublesome if you think about the shorthand needing to be written horizontally and not vertically. That's why I think the symbol used for 'p' right now should be switched with the space-ligature. That symbol goes against most of the other symbols and would probably return the writer to the home line, if there aren't too many downstrokes in the word itself. This is something I will have to play with more.
Another change might be getting rid of the straight up and down curves for the space-ligature. I like them because they're in the japanese shorthands but they're really only easy-to-write and distinctive on a large scale (for me); otherwise they tend to merge with 's'. I didn't originally consider them for this reason, but creating the space-ligature required me to come up with another symbol. I am considering replacing them with a large U-shaped hoop and its upside-down counterpart.
In case you were wondering about the clumping together of w/u/o/-n, it's a neat little loophole that will work for any word that one desires to write in TP. I have chosen to choose a default vowel to go along with each consonant which is 'u'; this is what makes this an alphasyllabary rather than an alphabet, because the letters are considered from the perspective of forming syllables -- who needs a bare consonant all by itself? Why not use a default vowel to go with any letter and use the other vowels to "over-ride" it. (And it also means I don't need to find another letter for medial/final 'u', which allows to use some sort of two-way distinction on the oval to represent the other vowels, a and e/i.) TP phonotactics being what they are, 'w' must appear in a word between vowels ('a' and 'e/i' in the shorthand because 'u' is considered the default for a bare consonant). If it occurs before another consonant it must be interpreted as '-n'. It can only occur after another consonant if that consonant is itself, in which case it is interpreted as '-nw-' -- UNLESS that consonant starts the word, in which case it can be interpreted as 'un(C)' or 'uw(V)' depending on if the letter that follows is a vowel (V) or a consonant (C). (Reminder: Anywhere I wrote 'u' you can substitute a 'o'). I think '-nw-' and 'uw(V)-' are unattested but COULD happen, as is 'unw(V)-' which would require an extra-extra-long line.
Large versions of straight line symbols will occur when two come together back to back within a word (none so far, excluding the mentioned 'un'/'ww').
I have also considered diactritics, but forgot to include them in the scheme above: a dot for 'i', a little 'e' for 'e', an 'o' for 'o' and a 'u' for 'u' (no tail on the 'u'). I didn't want to lean toward any default diacritics since I haven't thought it through and I wouldn't really want to require their use anywhere, with the exception of ken and kin and new words.