Meta's Super Intelligence Lab is quietly assembling a hardware team led by veteran product executive Xu Rui, who has extensive experience from his tenures at Xiaomi and ByteDance. While Meta's Reality Labs division is already well-known for its smart glasses and VR headsets, this latest development indicates the company is exploring novel AI devices that move beyond current form factors. To advance this strategy, some engineers from Reality Labs have been transferred to the Super Intelligence Lab.
Xu Rui is an '80s generation professional who grew up in Hefei. In 1999, he gained admission to Harbin Institute of Technology through the national college entrance examination. After completing his undergraduate studies, he pursued a master's degree at Seoul National University. Following graduation, Xu joined LG Electronics where he contributed to developing the world's first Linux-based mobile phone, a product that preceded Android phones by five years.
In 2008, Xu returned to China to join semiconductor startup C2 Microsystems as a founding engineer and software lead. After the company's acquisition, he moved to Intel as a software applications engineer, working on early Android systems, Google TV, and mobile/TV applications for Netflix. At age 30, Xu decided to shift his career focus to product management and enrolled in UCLA's Anderson School of Management for an MBA degree.
During his MBA studies, Xu joined Amazon's Lab 126, becoming one of the first product managers for Fire TV, which marked the beginning of his product management career. After completing his MBA, he briefly served as a product manager at AOL. Xu returned to China again in 2015, joining major technology companies including Xiaomi, Lenovo, ByteDance, and BOE, where he led or managed product development for Xiaomi Box, Lenovo smart displays, Smartisan Nut R2, Dali smart learning lamps, and BOE TCON chips.
In 2024, Xu resigned from his position as Senior Director of XR Products at Tencent to join Silicon Valley robotics startup K-Scale Labs as COO, leading robotics hardware development. However, the company ceased operations after just one year due to funding depletion. Subsequently, Xu joined AI agent startup Dreamer as hardware lead. Last month, Meta acquired Dreamer primarily for its talent, bringing Xu into the Meta organization.
The specific nature of Meta's new AI hardware remains unclear, but it's confirmed not to be an AI phone similar to "Bean Pod Phone" concepts. Xu characterizes such devices as essentially "macro automation tools" attached to large language models that clumsily mimic human finger operations at high speeds. He compares them to "nuclear-powered pencil sharpeners" - using AI-era "nuclear reactors" (large models) to power "old machinery" (app ecosystems) from the mobile 2.0 era, ultimately confined within traditional rectangular glass screens.
In his blog, Xu traces the evolution of computing platforms: the PC era operated on "mouse + menu" logic, while the mobile era followed "touch + app" principles. Each generation featured hardware and software paradigms that were highly self-consistent. The AI era, he argues, should have its own native logic. True AI hardware shouldn't simply incorporate AI for its own sake but should be capable of tool invocation, project execution, and solving physical world problems that previous hardware couldn't address.
In his latest writing, Xu proposes that "the answer for the AI era is dumb devices." He believes hardware capabilities themselves are far more important than built-in intelligence in the AI age. Conventional smart devices like intelligent rice cookers, smart speakers, and smart TVs, while labeled "smart," have their functions predetermined by product managers before leaving the factory, unable to truly understand user needs. For example, a high-quality rice cooker's heating element can precisely control temperatures from 30°C to 180°C, theoretically capable of making yogurt, slow-cooking steak, or fermenting dough, yet in practice it only cooks rice, soup, and porridge because product managers predetermined these limited functions.
AI integration will break these constraints, finally enabling "rice cookers to make yogurt." Future hardware excellence will depend not on preset firmware functions but on becoming extensions of AI agents in the physical world. In a February podcast, Meta Super Intelligence Lab lead Alexandr Wang stated they're developing cross-device personalized AI agents that will accompany users随时随地 in various forms, capable of "seeing what you see, hearing what you hear" while remaining constantly online. He added, "In the coming months, you'll see us advancing this at astonishing speed."
