A Chinese smart glasses maker, Rokid, used its 2026 Open Day this week to unveil AIOS, an "AI native" operating system meant to replace the phone era software stack that glasses currently inherit. The category argument is plausible.
Smart glasses still run software that was written for phones. China's Rokid used its 2026 Open Day this week to argue that mismatch is the category's real problem, and to unveil AIOS, an "AI-native" operating system it describes as the first OS designed from the ground up for face-worn AI.
The form factor, on Rokid's telling, leaves little room for a phone-style interface. A handset can hide a launcher behind a tap. A pair of glasses cannot, so the company is arguing that the device has to read the wearer's intent and answer it without ever pulling up an "app." That is a different set of operating-system assumptions than the homescreen-and-app-drawer model Android and iOS were built around, which is why Rokid is repositioning its existing YodaOS as a base layer and putting AIOS on top.
If the claim is real, the stakes are not just product. A glasses-native OS would, in theory, set the rules for app distribution, on-device AI assistants, and the developer-facing standard any rival has to follow or fork. The category is no longer experimental. EDN China's CES 2026 roundup and a Time Weekly market write-up both flag a wave of Chinese AI-glasses launches on the show floor, and a 2026 product catalog from VRtuoluo describes a crowded, multi-vendor field. With that many entrants, the software stack that wins could decide whether AI glasses look like an extension of Android or like a new platform.
Rokid's pitch is concrete enough to test. The company says AIOS bundles three interaction modes: an always-on listening layer for ambient prompts, a quick-in-and-out mode for one-off questions, and a richer content mode for longer tasks. It is positioning an open interaction standard called AIUI as the API (the contract programmers use to talk to the OS) that app makers would target in place of Android intents. Rokid also says it has shipped 20 over-the-air updates and roughly 500 features to its glasses since September 2025, and that a new AI Assistant 2.0 can carry an intent across multiple apps without dropping the thread. Those are product claims, not stage slides.
What is missing is any voice other than Rokid's. The QbitAI write-up that carries the announcement is itself a republication of the company's event presentation. It cites no independent analyst, no partner developer, no enterprise customer, and no benchmark. The market context has the same limitation. A forecast that China smart-terminal shipments will exceed 900 million units in 2026, with traditional AI-terminal hardware above 300 million, is attributed in the source to IDC but was not independently verified in the receipts. It should be read as analyst-cited outlook, not a settled base. Whether AIOS diverges from Android in any way a regular user, developer, or competitor would notice is the actual test, and the source does not address it.
The category will not wait for Rokid to set the answer. Meta, Apple, Samsung, and a stack of Chinese vendors are all reported to be racing into face-worn AI hardware and software. If even one of them ships a credible glasses-native interaction model first, "first face OS" stops being a category statement and becomes one of several bets.
Three watch items follow from the announcement. First, whether the AIUI standard pulls in any developer who is not already a Rokid partner. Second, whether the on-device intent engine keeps state across apps in the way Rokid claims, or quietly degrades into per-app assistants under load. Third, whether any analyst, competitor, or independent tester publishes a side-by-side of AIOS against a forked Android build running on the same hardware.
Until then, "AI glasses no longer need a phone" remains a vendor slogan: a real category question wearing a press-release costume.