Idea 1: A note-taking or brainstorming interface for n-ary trees. Using visualization tools like D3.js, data and, more importantly, thoughts can be nicely organized into a word tree which resembles a JSON. I think that using the Soli sensor's swipe to navigate such a tree and p5.speech to record thoughts and ideas makes for a unique brainstorming interface. If the interface shows promise, the idea can be expanded by connecting directly to a MongoDB, which also stores data as a JSON.
Idea 2: This idea involves some kind of 3d creative toy using Three.js. In order to make the limited gestures meaningful, a sequence of gestures can be used to orient, transform, and place a cube into a world. I think of my ideas this one is the least elegant interface, because consecutive swipes feels like a lot of work for a simple transformation. Anyway, to control a given type of transformation on the cube, the user can swipe into a given control plane and swipe to make the transformations.
Idea 3: This idea builds off of the previous idea of creating an environment. In this case, Golan pointed me to Everest Pipkin's game tileset. I quickly made this sketch which allows users to place trees in the 'world.' For the Soli interaction, the user can swipe to edit the 'cursor' position, and speak certain commands to add certain tiles. This is a fun way to create quick compositions that look really nice (all due to the quality of the hand-drawn tiles). Because there are 281 tiles total and each has a relatively concise name, there is a broad range of creative possibilities. Another benefit of using the tiles is that they are isometric and aligned, which makes it very easy to mock up code like this.
In my experience with the Soli sensor, while tap is a promising gesture, I have not gotten it to work. Hence, most of my ideas only rely on swipe gestures and maybe voice recognition as well.