One Dock, Two Modes
One gesture should be enough to choose between speaking and typing. We used to have a button that changed shape and a swipe you had to guess; we've replaced them with two small nodes connected by a thread of light.
A button that transformed, a gesture you had to guess
The switch between voice and text mode used to rely on a bar that morphed from one state to another, complemented by a horizontal swipe to toggle. On paper, it fit into one sentence. In hand, it was less clear: a gesture you had to know to trigger it intentionally, and which could also trigger accidentally at the wrong moment: a thumb slipping while holding the phone was sometimes enough to switch the screen unintentionally. It was hard to trust a button that didn't look the same twice in a row, and a gesture that sometimes triggered itself.
Two nodes, one thread
We started from scratch with a dock: a large node for the microphone, and next to it, a small "Aa" satellite, connected by a thin synaptic line where a point of light constantly travels from one node to the other. Touch the satellite, and the keyboard opens. It's no longer a button that changes identity, but two constantly visible destinations, and the luminous journey between them subtly reminds you that the two are just one and the same conversation, simply in another language. This dock is now the gateway to all voice or written interaction with XNeuronal.
The way back, always visible
In text mode, the keyboard occupies the screen and the dock disappears with it: so an equally obvious way back was needed. A small microphone button, alone, sits just above the keyboard, accompanied by a discreet caption: "tap to speak." A tap on it, and you're back in voice mode. We deliberately removed the previous text indication for this return: the small button, by always being in the same place, carries this promise all by itself, without needing to explain it in writing every time.
The top of the screen, tidied up
This project was also an opportunity to review the top bar, which had accumulated icons without any real logic. It is now read from left to right like a sentence: new conversation then history on the left, your neuron counter in the center, audio settings and parameters on the right. The "new conversation" button is a direct addition: it cuts off the current listening, properly closes the previous exchange on the server side, and starts a fresh thread. Useful when you want to completely change the subject without XNeuronal piecing together a conversation that's no longer relevant. The assistant's name takes advantage of the change to finally adopt its true visual identity, "xn" in blue followed by "euronal" in ink, instead of a simple uniform word.
The real gain from this project isn't visual, it's cognitive: freeing the horizontal swipe from its mode-switching burden leaves us with a completely clear path for other future uses, without the gesture having two meanings at once.
We also accept the compromise: tapping a small satellite requires a bit more precision than a swipe that could be triggered anywhere on the screen with a broad thumb gesture. We prefer this slightly more demanding precision to an ambiguous gesture that could be triggered without truly intending it. A dock you can see is better than a gesture you have to guess.