Global

TwelveLabs said its latest video-understanding model, Pegasus 1.6, can now analyze first-person footage, more accurately identify people and objects, and process still images through the same API used for video, broadening its use for robotics, media and security customers. The model builds on Pegasus’s core function: converting video into structured, searchable text such as summaries, tags and timestamps that applications can use. The company said the biggest addition is support for egocentric video — footage shot from the perspective of a person or machine, including body cameras and robot-mounted cameras. Such data is increasingly used by robotics and embodied-AI teams for labeling and training, a category TwelveLabs previously did not support well. Pegasus 1.6 also improves entity recognition, making it...

Boston Dynamics has tapped ex‑Amazon AI executive Rohit Prasad to fill its chief executive role, nearly nine months after Robert Playter stepped down from the top post. The robotics firm stated Prasad’s appointment will take effect the day following the announcement. Previously serving as senior vice‑president and lead scientist for Alexa and Amazon’s artificial general intelligence initiatives, Prasad is tasked with accelerating Boston Dynamics’ physical‑AI roadmap. The firm notes his expertise will fuse robotic hardware with cutting‑edge AI to scale commercial deployment of intelligent machines. “Boston Dynamics holds a distinctive edge in advancing physical AI, built on its world‑renowned robotics engineering talent,” Prasad commented in an official press release. “I’m privileged to join the business at this defining turning point...

Researchers at Caltech and Stanford have unveiled HomeBody, a system that gives GPT Astra persistent spatial memory and the ability to execute compound humanoid actions for long-horizon tasks. In a kitchen demo, a Unitree G1 robot guided by Astra cleaned the room and retrieved remembered items from ambiguous requests. The team said the system requires no environment-specific training or additional policy learning. HomeBody replaces the trained intermediate VLA layer in conventional three-part architectures—System2 VLM for vision and instructions, System1 VLA for control commands, and System0 for coordinated motion. As frontier models such as Astra improve, the researchers argue trained VLAs are no longer necessary and System2 can directly manage reusable motion skills. A plug-and-play VLM calls a composable skill library....

SiMa.ai, a leader in Physical AI, announced $150 million in Series C financing, bringing the total capital raised to $500 million and valuing the company at $1.45 billion. The round was co-led by Fidelity Management & Research Company and Amplify, with participation from Alter Venture Partners, Dell Technologies Capital, Maverick Capital, +ND Capital, Point72 and StepStone Group. AllianceBernstein, Baron Capital, J.P. Morgan, and the State of Michigan also joined the round as new investors. The global market for Physical AI devices, including robotics, automotive, and drones, is set to explode with cumulative shipments projected to hit 145 million units by 2035, according to Counterpoint Research. This surge marks AI’s transition from software applications into tangible, real-world consumer and industrial hardware,...

A robotics enthusiast accidentally unpacked Tesla Inc’s (TSLA.O) Android app package and discovered unreleased designs for the company’s third-generation Optimus humanoid robot, according to a now-deleted post on X. User @wholemars said in a post on Tuesday, which has since been deleted, that he extracted the material from Tesla’s app. The files included renderings labeled “gen3” and side-by-side images of Optimus 2.5 and Gen 3 to distinguish the two generations. The images were not concept sketches or fan renderings, but internal material Tesla embedded in a publicly distributed app package, according to the post. According to the leaked images, the Gen 3 Optimus appears more streamlined and designed for factory assembly lines rather than the laboratory. Compared with Optimus 2.5,...

Boston Dynamics has introduced a new four-fingered hand for its Atlas humanoid, a 13-degree-of-freedom, directly actuated design intended for mass production. The hand marks a strategic shift. Earlier Atlas hands had seven degrees of freedom and were designed to grasp a wide range of objects. The new version focuses on manipulating them. “We considered many, many designs,” said Alberto Rodriguez, director of robot behavior for Atlas. “It boils down to a trade-off of competing objectives.” The design omits a pinky and is tailored for high-fidelity simulation and sim-to-real reinforcement learning. It uses a single actuator type, fully encapsulated joints and no fragile cables crossing joints. Boston Dynamics said the hand can slide the thumb tip along and across the other...

Toyota Motor Corp is launching a sweeping global factory upgrade that will put 400,000 robots to work, including a semi-humanoid model designed to learn tasks alongside humans, as the automaker seeks to build plants where people and machines coexist. The push includes ELEY, short for Embodied Learning robot for Enhanced Yield, a Toyota-developed semi-humanoid robot. ELEY has two fingers on each robotic hand, weighs about 50 kg, runs on battery or external power and moves on a wheeled base, according to Toyota’s website. Toyota said ELEY’s joints are designed on the same principles as human joints, with all dimensions based on the average Japanese adult male. The aim is to use large-model-based physical AI to let the robot directly imitate...

Figure, the U.S. humanoid robot developer, has released Helix 2.5, its most advanced neural network model to date, and said it enabled zero-shot whole-body autonomy in 30 real homes. Figure called it the first demonstration at that scale of zero-shot whole-body behaviour generalisation for a humanoid robot. The company said a single pre-trained foundation model was applied to three types of behaviour: locomotion; manipulation of rigid and deformable objects; and bimanual coordination, as well as active perception. Figure said that, with task-specific data, architecture, training method and evaluation unchanged, pre-training on Index alone raised zero-shot recognition success to 56% from 9%. Index is Figure’s robot dataset project built for global scale. The model also validated a Human-to-Humanoid Robot Transfer Scaling...

Agility Robotics unveiled Digit 5, its latest humanoid robot, as the firm advances plans to go public via a proposed SPAC merger with Churchill Capital Corp. XI. Digit 5 features new legs, upgraded batteries and a more comprehensive safety architecture, and can work in close proximity to people without physical safety barriers, Agility said. CEO Peggy Johnson said“a major barrier to scaling humanoid robots in industrial environments,” adding that the market has responded with more than $300 million in multi-year orders across manufacturing, warehousing and logistics. Agility reported $1.8 million in net sales in 2025, according to a recent SEC filing. The company expects early access to Digit 5 in the first half of 2027 and general availability for manufacturing,...