
Glydways vehicles run with no driver and no staff, so the interface has to do their job. Lumasis is the design system that makes that possible: one set of values the rider, the operator and the vehicle all read the same way.
The vehicles don't see the world the way people do — they see sensor data. The visual system is built from that machine perception, not layered on top of it afterward.
Over two years, that perception became one tokenized foundation — color, light, glass, depth, type, icons, messaging and accessibility defaults — shared by every surface a rider or operator touches.

The principles keep that machine-built language human. It grows from how the vehicles see, sense and decide, but it is built for the people who rely on them.
Rounded shapes, conversational micro-copy and organic motion instead of abrupt transitions.
Deep negative space and blur keep attention on what matters right now.
Components morph and adapt, taking the rider from booking to boarding without jarring context switches.
Predictive states surface actions, like opening doors or adjusting climate, exactly when the rider needs them.
Type and pictograms come from Metro and stay fixed on every surface. Color is different: with no driver on board, it carries what a driver would otherwise say — which line, what's next, what needs attention. So color is the one parameter built to adapt, and light, glass and depth all derive from it. Pick a color, a glass and a depth below to watch the whole stack follow.
The vehicles already speak in data, so the tokens had to as well. Every name is derived from its value — family from the HSL hue, step from the HSL lightness — so the token file reads the same to a designer, an engineer and a model. AI did the translation work: it turned the parameters into tokens and mapped them against ADA and WCAG requirements.
With no staff on board, nobody is there to help a rider who can't read a screen, so accessibility has to live in the interface itself.
A dark theme with glowing accents has to clear WCAG AA[1] on every primary text and control; each applied value sits 18 to 30 points of lightness above its core.
Frosted, Dark, Tinted and Prismatic all carry text at the same light values, so contrast doesn't reset depending on which surface a control lands on.


The control room is where the autonomy is supervised. An operator watches two hundred vehicles move across three circuit maps; the map carries position and nothing else, and everything a vehicle reports — and everything the system infers about it — opens in a window when the operator selects a dot. Status built into the dots was rejected: legible at five vehicles, unreadable at two hundred.

With no one on board, the interface is what explains the vehicle. A rider on a platform can tell what it is about to do, and an operator watching two hundred in motion can tell which one needs attention — both reading the same tokens, with nothing translated in between.









