Yes, it uses dynamic programming to find the best finger choices while minimising cost. Cost can be for example number of shifts done or the gap to shift. The specific name is „Viterbi Algorithm“.
The advantage to modern AI solutions is that here we have control over specific costs and can control the output. Eg „I want to have the simplest finger choices possible.“ or „Avoid specific fingers on specific strings for harmonic sound (on violin open string are sounding different)“
This looks awesome! The thing I'd want to know is how much of the configuration lives on firmware vs. in the macOS app? If the app is a thin config editor and the mappings remain on-device, that's very different than if the dial and haptics stop working when a future macOS release breaks the app. Also, is there any commitment on firmware being flashable/open if the company someday isn't around?
Everything lives on the keyboard itself. The app just configures files on the device, basically. The keyboard also works without the app of course. This approach has the added benefit of making your configuration portable. For example if you remap your dial, and move to a different computer, your remapped dial will persist.
There are a couple of small quality-of-life features that do require the app: for example if you want haptic feedback when you receive a notification. This is because the keyboard has no idea what is happening on your OS, and requires some middleware. But all the core functionality requires no app at all.
Both Altar I and Altar II firmware are based on the open-source ZMK (https://zmk.dev). We try to contribute back to ZMK whenever we can because we have benefitted so much from it. Two members of our team are also long-standing ZMK contributors.
