This is a working document. It will be wrong in places, and we will update it as the team and the field learn. We are publishing it early because we believe the strongest collaborators self-select on a clearly stated bet.
The premise
Every dominant computing medium has had a roughly forty-year life.
The mainframe defined computing from the early 1950s into the 1980s. The personal computer from the 1980s into the 2010s. The smartphone owned the 2010s and most of this decade — but its replacement is already taking shape.
What replaces a medium is rarely a better version of the medium. The PC did not replace the mainframe by being a smaller mainframe; it replaced it by introducing personal context. The smartphone did not replace the PC by being a portable PC; it replaced it by introducing constant connectivity and a touch surface.
What replaces the smartphone will not be a smaller smartphone. It will be something that addresses a constraint the smartphone cannot remove — the rectangle separating you from the world.
What we are betting on
Three forces are crossing at once, and we believe they only resolve into a product if a single team takes all three.
First — optics. Waveguide, MicroLED, and freeform optical tooling have crossed into manufacturable tolerances. The unit of bottleneck has shifted from physics to integration.
Second — on-device intelligence. The frontier of multimodal models is now small enough to fit on a wearable's compute budget. Not the frontier of last year, but a competitive class that runs locally.
Third — a new operating layer. The grammar of software is being rewritten by agents and intent. The rectangular OS abstractions — windows, apps, files — are starting to feel like a translation overhead between what the user means and what the system does.
Each of these is a serious bet on its own. The thesis is that all three only become a product if shipped together. Specialization in one layer locks you into the other two being someone else's compromise.
What we are deliberately not doing
Negative space matters as much as direction.
We are not building a phone accessory. The most popular AR posture today — "glasses that show your phone notifications" — is a 2020 idea pretending to be the future. It is a feature, not a medium.
We are not building entertainment-first hardware. Gaming-and-media VR is a beautiful, well-funded category — and a separate one. Our wearer is awake, mobile, and using the device while engaged with the physical world.
We are not building a research platform without a path to a product. Spatial computing has a long history of seductive prototypes that never reach the wearer. The roadmap is structured to ship a real, used product — even if smaller in scope than the long-term vision.
What we do not yet know
An honest list of open questions.
We do not know whether the right FOV for the first product is 25° or 40°. We have arguments for each, and the prototype will decide.
We do not know whether eye-tracking belongs in the first wearable or the second. The cost in weight and power is non-trivial; the gain in interface fluency may justify it.
We do not know which vertical pilot becomes the anchor customer. We have three serious candidates; the technology will route us toward the one with the most direct path to scale.
We are publishing these openly because pretending to have answered them would invite the wrong team and the wrong partners.
Who this is for
A short, specific invitation.
If you have shipped an optical engine — waveguide, microdisplay, projector — and you have a strong opinion about the next generation, write to us.
If you have worked on on-device multimodal inference at sub-50ms latency on constrained silicon, write to us.
If you have built an operating system, a window manager, or a shell, and the rectangle has started to feel like a constraint to you, write to us.
If you are a potential customer who thinks the next interface for your industry is spatial — procedural medicine, field engineering, complex assembly, fleet operations — write to us.
We respond to every serious note within forty-eight hours.
Lumira · Fuqing, Fujian · May 2026