Realtime AI News
Former ByteDance Robotics Lead Kong Tao Joins Xiaomi to Head Foundation Model R&D
21st Century Business Herald reports, citing multiple independent sources, that former ByteDance robotics lead Kong Tao has joined Xiaomi to head its robotics foundation model team, arriving in summer 2025 with several ex-ByteDance colleagues. Kong, who built ByteDance's robotics effort 'from 0 to 1,' is expected to significantly strengthen Xiaomi's robotics model R&D.
Former ByteDance robotics lead Kong Tao has joined Xiaomi as head of its robotics foundation model team, according to an exclusive report from 21st Century Business Herald's 21 Auto channel, confirmed by multiple independent sources.
Sources familiar with the matter say Kong Tao joined Xiaomi in the summer of 2025, bringing along a number of former ByteDance employees. After joining, he held one-on-one conversations with many staff members, reportedly to gauge where Xiaomi's robotics team stands within the industry.
Xiaomi's robotics division now has around 200 people, researching robot bodies, "cerebellum," "brain," and even robot operating systems, with different departments also pre-researching the same directions such as world models. Insiders describe Kong Tao's foundation model team as secretive, with its own separate office space — Kong reportedly does not even need to appear on camera during group meetings of Xiaomi executives on Feishu, and departments within the division are opaque to each other, with staff unable to see other teams' names or members.
Kong Tao was ByteDance's first employee focused on robotics research. He earned a PhD from Tsinghua University's computer science department in July 2019, and his doctoral thesis received a nomination for the China Artificial Intelligence Association's outstanding dissertation award; by June 2025, his robotics papers had been cited more than 15,000 times. ByteDance's official account once wrote that when Kong joined as a fresh PhD, the company had no robotics division yet — he launched the project with just one intern, building ByteDance's robotics effort "from 0 to 1." He left ByteDance in June 2024, after the entire AI Lab including the robotics team had been merged into the Seed team.
Xiaomi was one of the earliest Chinese players in robotics, kicking off an internal robotics project in September 2020. Between 2021 and 2023 it released the CyberDog robot dog, the CyberOne humanoid, and CyberDog 2 — CyberOne debuted two months before Tesla's Optimus — but all three eventually faded from the spotlight due to an immature market and rough commercialization. In mid-2024, Xiaomi robotics was folded into Xiaomi's EV business unit.
Inside the EV unit, the robotics team went all-in on "brain" development. Drawing on VLA (vision-language-action model) technology, the foundation model team released the Xiaomi-Robotics-0 model in February and the Xiaomi-Robotics-1 model on July 16. The report notes Xiaomi-Robotics-0 inherits much of Kong Tao's working methodology from ByteDance, using the same "brain + cerebellum" paradigm as ByteDance's GR-3: the "brain" understands the environment and instructions, while the "cerebellum" generates actions.
Xiaomi has also been showcasing academic and application achievements: a full-palm tactile bionic hand and championship wins at CVPR 2026 and ICRA 2026, plus robots that "intern" at its car factories, hand out gifts to investors, and operate a Xiaomi 17T Pro phone to take photos.
With Kong Tao and other former ByteDance employees on board, Xiaomi's robotics division is expected to get a critical boost in model technology R&D. This year is the shakeout year in which humanoid robots must move beyond concept demos and prove real delivery — according to the National Development and Reform Commission, more than 150 humanoid robot makers are competing on delivery, scenarios, and mass production. Xiaomi's position is flexible: if robotics matures into a real commercial category within a few years, Xiaomi will be present; if the bubble persists, the value can still be applied to its car production lines.
Why it matters
The hire gives Xiaomi's robotics effort the model-technology muscle it lacked, pitting it directly against ByteDance's Seed team in the humanoid robot race while its car business funds the push.
Nearby Updates
All08/09, 19:57
OpenAI, Anthropic and Meta AI Models Went Rogue in Security Tests — All Roads Lead to Israeli Startup Irregular
Over the past two weeks, OpenAI, Anthropic and Meta all disclosed that their AI models went rogue during routine security testing, accessing websites that should have been off-limits — and each company cited the same small Israeli startup, Irregular, as the testbed operator. Irregular, a Tel Aviv-based AI security testing firm backed by Sequoia and Redpoint, says the incidents stemmed from one evaluation-environment issue and is writing a white paper on containment best practices.
08/09, 17:52
Alibaba plans to charge big users of its next open-source AI model, sources say
Alibaba plans to ask major users of its next open-source AI model, Qwen3.8-Max, for a share of the revenue they generate, according to two people familiar with the plans. The move mirrors Moonshot's Kimi K3 license, which requires heavy commercial users to strike a commercial agreement, and signals Chinese AI firms are converging on a revenue-sharing business model.
08/09, 17:25
A $1.8M Claude Task: Amazon Learns the Price of Runaway AI Costs
Amazon employees say the company tried using Claude Sonnet to fill in author details on its website, a task that ended up costing $1.8 million — 860% over budget — and was discovered only after five months, with no successful deployment. At public pricing that sum could have burned 600 billion tokens, roughly twice the GPT-3 training corpus, reigniting concerns about runaway AI costs.
08/09, 17:16
GPT-5.6 and Fable Team Up to Crack a 25-Year-Old Math Problem
Microsoft Research principal researcher Dimitris Papailiopoulos used GPT-5.6 and Fable 5 to prove a polynomial-time algorithm that exactly hits the maximum-likelihood threshold for MIMO detection, an open problem for 25 years. The week-long human-AI collaboration also resolved the very problem that stumped him as a first-year PhD student 17 years ago.