17 articles · Last covered
Skild reports real-world soccer play after more than 140 years of simulated self-play. Its published account details human-referenced drills before the goal-driven self-play stage.
A planned 2027 program would combine robot AI research, a physical-data library and early postal and logistics trials. Its budget is still being finalized.
The reported RAI acquisition would add robotics research capacity as Hyundai pursues SoftBank’s Boston Dynamics stake and SoftBank explores other humanoid investments.
Faraday Future adds nine configurations across five robot models, while its Master Mini documentation points to Booster hardware behind the $9,990 education humanoid.
Reuters reports IFR counted around 7,000 humanoid sales for industrial and professional use in 2025, highlighting the distinction between robot purchases and productive deployment.
RoboHarm adds a safety test to Astra’s robotics results: 97 attempts and 60 completions across 100 unsafe-instruction trials, with only two safety refusals.
GPT-6 Astra leads specialist models on selected simulation tasks. New robot-arm demos add to the promise, while tougher tests expose its limits.
Galbot and researchers from Tsinghua University introduce RoboGesture, a streaming diffusion-based system that aligns speech with expressive, collision-free humanoid gestures.
A joint team from BIGAI and Unitree has unveiled OmniXtreme, a unified control policy that enables humanoid robots to perform diverse extreme motions—from backflips to breakdancing—without the need for task-specific overfitting.
Physical Intelligence (Pi) has unveiled Multi-Scale Embodied Memory (MEM), a hybrid architecture that combines short-term video encoding with long-term textual summarization to help robots master long-horizon tasks like kitchen cleaning and in-context error recovery.
NVIDIA researchers have revealed EgoScale, a framework that leverages a massive 20,854-hour egocentric human dataset to train robots in complex, fine-grained manipulation with minimal robot-in-the-loop data.
NVIDIA has released SONIC, a generalist humanoid controller trained on 100 million frames of motion data, aiming to replace manual reward engineering with a scalable "System 1" foundation for whole-body movement.
