AGIBOT used APC 2026 to put a full embodied AI stack on the table: 5 robot platforms, 8 AI products, and a data system for training physical models. The notable part is scale context: the company says it rolled out its 10,000th robot in March 2026.
Physical Intelligence says π0.7 shows early compositional generalization, following new language commands and performing tasks not seen in training. In laundry folding, it matched expert teleoperators’ zero-shot success on a UR5e setup without task data for that robot.
Google DeepMind and Boston Dynamics are showing a clearer bridge between foundation models and robot APIs. The demo gives Spot tools for movement, photos and grasping, then lets Gemini Robotics plan from plain-English tasks.
HN focused less on the model drop and more on the hard robotics question: how fast does reasoning need to be before it is useful in the physical world? Google DeepMind frames Gemini Robotics-ER 1.6 around spatial reasoning, multi-view understanding, success detection, and instrument reading, while commenters zoomed in on gauge-reading demos, latency, and deployment reality.
Generalist says GEN-1 crosses a commercial threshold for simple physical tasks by combining higher success rates, faster execution, and lower task-specific robot data requirements.
Google DeepMind announced SIMA 2 on November 13, 2025 as a generalist foundation model for virtual 3D environments. The system is designed to play and reason alongside humans, with in-context learning that can improve behavior from examples.
At China's 2026 CCTV Spring Festival Gala on February 17, humanoid robots from Unitree, MagicLab, Noetix, and Beijing Galbot performed martial arts, acrobatics, and household tasks, showcasing rapid advances in motion control and embodied AI.
Google DeepMind announced Gemini Robotics-ER 2 on January 8, 2026, highlighting improved data efficiency and real-world action performance. The update targets a core robotics bottleneck: reliable generalization from training to physical environments.