From Smart Glasses to Agentic Robots: The Next Hardware Layer of the AI Stack (2026 Thesis)

2 mini robots with an agentic layer are in transit and expecting one more in August.

After smart glasses, this is my next prediction on hardware advancement before the full robotic invasion in the 2030s (one alongside self-driving cars and other IoT updates).

For smart glasses – I’ve been tracking this down since early Chinese models and cheap VR knockoffs, through the Metaverse Quest, to Snap Spectacles, the Meta Ray-Bans (which I wear weekly), and the Rokid AI/AR glasses.

Plugging more software into glasses makes a lot of sense. While Quest was too heavy on the vestibular and leading to actual diziness 90 minutes in, a pair of Metas can take a full day. Especially if glasses stay clear of dust and fingertips and don’t get blurry too early.

AI robots however, feel like the next transition of OpenClaws and Claude SDK agentic toolkits, and other GPT wrappers with a hardware twist. It’s also not wildly revolutionary when you think about devices like Alexa – whereas Echos have been sitting in each room here for many years now and the kids picked up games and quests and voice search there.

Interactive robots with a trained agent do feel like a personal assistant at home + a proxy between a laptop (fully fledged keyboard) and home assistance. It’s also an alerting tool for calls or specific emails or outages – which may be deployed in offices.

Each instrument has pros and cons. But agentic robots are one I’m exploring more thoroughly in 2026.