Building before the door opens
The smart-glasses platforms aren't ready. Meta's developer toolkit is in preview. Public publishing is limited to a handful of hand-picked partners. The native assistant is walled off from developers. The whole thing runs tethered to a phone over a Bluetooth pipe. Every reasonable person's advice is the same: wait until it's sorted out.
We're not waiting. We're building a glasses agent now, in the open, while the ground is still moving. And it's worth saying plainly why — because it's the same reason behind everything else we do.
A lab that only builds on finished platforms is, by definition, always late.
Why early is the whole point
By the time a platform is comfortable — tools polished, publishing open, patterns obvious — the interesting decisions have already been made by someone else. The maturity you waited for is just the fossil record of choices other people got to make while you were being sensible.
The people who define a new interface era are never the ones who show up when it's safe. They're the ones with scar tissue from building when it was hard — who learned, first-hand and before anyone wrote it down, what a hands-free agent should actually feel like. That knowledge doesn't come from reading the docs when they're finished. It comes from fighting the constraints while they're still raw. The tethering, the bandwidth limits, the walled-off assistant — those aren't reasons to wait. They're the exact conditions that teach you what everyone building later will never have felt.
What we're actually building
A considered, hands-free agent that sees what you see and speaks only when it should — with our own intelligence behind it, not a wrapper on someone else's assistant.
That last part matters. The easy version of a glasses app is a thin skin over the platform's built-in AI. We're not doing that. The agent is ours: our model, our judgement about when to speak and when to stay silent, our decision about what earns a place in your ear and what doesn't. On glasses, restraint isn't a style choice — it's survival. Nobody wants a chatty thing on their face. The whole craft is knowing when to say nothing.
We're starting with accessibility. Not as a feel-good angle, but because it's the most demanding test there is: whether the agent can genuinely help a person navigate the world. Get that right and every hard question surfaces at once — trust, latency, restraint, and the discipline of silence. Solve it for the person who depends on it, and you've solved it for everyone.
We'll write it down as we go
This is the first entry, not the finished story. We'll publish the rest as it happens — what worked, what surprised us, what's still unsolved, what we killed. Not a case study buffed to a shine after the fact. The actual notes, while it's live and still going wrong in places.
That's what building in the open means. Not a launch. A logbook.
Why this is the bet
Every interface era rewards the same people: the ones who were already inside it, doing the work, before it was obvious it mattered. We think glasses and ambient computing are that next era, and we'd rather have scar tissue than a good seat when it arrives.
"Considered AI, shipped fast" was never only how we work for clients. It's how we bet on what's next.
The door isn't open yet. We're building anyway. We think that's exactly when you should.
cyber:cyber is a lab for AI products, AI features, and the agent infrastructure that runs them. We work with a small number of clients each year. The ones who can tell the difference. Email hello@cybercyber.ai.
