AI coding agents often fail when building Rokid apps because they don't know the device.
So we made an agent skill that distills real dev experience and know-how into context: HUD, touchpad, camera/mic quirks, UX patterns, best practices, and a template.
https://t.co/5c7J4pPfVl
the crazy thing is that this whole graph is available to agents just by scanning a QR code.
Think of smartglass agents:
- @oneshot_ar
- @visionclaw (with a shipmemory adapter)
- @GlassKit_ai (with a shipmemory adapter)
Robots
- On the @Auki ecosysten
- Robots with shipmemory
I recorded roughly 13 hours of my day with smart glasses, then turned that footage into something I could browse and query with AI.
It feels like the beginning of what "personal AI with real-world context" could become.
I built a smart glasses app that tells me when to stop pouring.
This is a drink-making assistant that:
- sees the ingredients
- chooses a recipe
- shows the steps
- watches what I'm doing in real time
- proactively guides me through the task
- points out mistakes
- says "stop" at the right moment while I'm pouring
I don't have to ask AI anything. It notices what's happening and helps on the spot.
Tech stack:
- https://t.co/movNO2HHHV (@zakariaornot) for fast live VLM inference
- @OpenAI LLM
- @RokidGlobal glasses (planning to support other devices too)
- @GlassKit_ai, my smart glasses dev toolkit
Full code lives in: https://t.co/MzY5wW62mG