Ant builds an OS for AI glasses and refuses to build glasses
Day two at Shanghai's Inclusion·Bund Conference, and the loudest argument of the morning was about who gets to run the agent sitting on your face.
Ant Group upgraded GPASS, its trusted-connection framework for smart terminals, into LingYing — an "agent-native operating system" and open platform for AI glasses, earbuds and smart rings — announced September 11 at the conference's "AI-driven new terminals, new services" forum, according to Leiphone and China Daily. LingYing covers the whole agent run, from interaction and orchestration through policy, execution, tools and memory, and hides the hardware underneath: camera, display, sensors and temple controls each get one interface, so a developer stops re-adapting the same app for every chip and every pair of frames. Security is wired into the execution path rather than bolted beside it — devices are identity-verified on connection, sensitive data is isolated locally, and every tool call an agent makes is authenticated, authorized and traceable. Ant is also building an evaluation lab with CAICT's Tair terminal lab to write the standards around it.
The tell is the business logic, not the architecture. After a year of pilots across glasses payments, city tours, shared cycling and parking fees, senior vice president Zhou Zhifeng said Ant will not make hardware and would rather be the one "paving the road," leaving products to vendors and developers. A payments company volunteering to run the layer where agents act on the real world is the trust problem finally getting an owner — and it is the same bet Ant has been making in public on both rails, in Visa, Mastercard and Ant want one trust layer for AI agents and in Ant's car agent now pays tolls and fills the tank in 16 million cars.
A forum run by Hangzhou's Yungu School and San Lian Life Weekly closed September 12 morning with a proposal aimed at classroom procurement: AI in schools should carry a usage manual like a medicine label — who may use it, what problem it solves, and above all when it must not be used. The session framed the question as "cognitive enhancement versus cognitive outsourcing." Tsinghua's Wang Min'an reached for Hegel's master-slave dialectic to argue that a species which outsources thinking eventually hands over its autonomy; Peking University's Wang Qiong countered that cognition has always partly lived outside the skull, and that the real enemy is cognitive laziness, not the tool. Alibaba senior advisor Lu Su, who sits on Yungu's governing committee, flipped it once more — reading books while refusing AI was, he said, his own version of outsourcing.
The most useful answer came from a graduating student who used AI only to check whether turning plastic bottles into 3D-printing filament was feasible, then built the machine herself. The questioning-and-revising loop, she argued, is the thing worth grading.
What to watch: whether LingYing's chip partners actually ship devices on it, or whether the "OS" turns out to be a payments SDK in a nicer coat.
If an agent could pay for things using your face and your voice, would you trust it more on Ant's rails or on a phone maker's? Tell us in the comments.
