SINK, without looking away
Everything SINK knows during a film currently lives on a phone — which is exactly where you don't want to look while watching one. On smart glasses it sits in your peripheral vision instead: a line of text that answers the question you had, then gets out of the way.
What appears on the lens
Where the room is
The shared clock and whether it's playing. If sync drops, the lens says so instead of showing you a frozen time as if it were live.
The scene you're in
One line about the beat currently on screen, drawn from SINK's own scene knowledge. Never a scene you haven't reached — the display is gated on your playhead, not the room's.
The room reacting
When people react, you see the burst and it clears. Ambient enough to feel like company, brief enough to ignore.
What doesn't appear on it
Chat. It was the obvious thing to build and we decided against it. The display is three lines of monochrome text; a live conversation scrolling through it is the single most reliable way to stop watching the film you put on. If that changes, it'll be because a viewer asked for it, not because it was easy.
Anything ahead of you. A scene note only appears once you've reached that scene. The glasses have no way to spoil something you haven't watched, even when the rest of the room is further along.
A camera feed. The first device we're targeting has no camera, and nothing here needs one. Sync still comes from your phone, the way it does today.
Where it stands
The display layer is built and the stream runs. What doesn't exist yet is time on real hardware — everything so far has been verified against simulators, and a waveguide has its own opinions about legibility and lag that a monitor cannot tell us.
We're starting with the Even Realities G2, whose apps are ordinary web apps relayed to the lens by your phone. That first version is tethered to an iPhone, which is a real limitation and not one we can engineer around. The same stream is designed to drive MentraOS and Meta's display glasses, so the device choice isn't final.
There is no ship date. There's a working prototype and an honest list of what's left.
Tell us which pair you own
It genuinely decides what we build next.
Building something against this? The heads-up stream is an authenticated SSE endpoint that emits plain text frames, and it authorizes the same way the MCP server does — so any device that can hold a connection and draw three lines can render it.