robotics developers

Unitree Robotics staged what it describes as the world's largest-scale real-time livestream performance by full-size humanoid robots, drawing over two million likes. The Chinese robotics firm is known for its G1 and H1 humanoids, and the coordinated group display is being widely shared as a showcase of how far bipedal robot control has advanced.

The Yuelu Conference 2026 opened in Changsha, Hunan Province, China, on September 23–24, bringing together artificial intelligence companies, robotics developers, researchers and investors. The event featured an opening ceremony and ten themed 'AI+' forums, positioning Changsha as a growing hub for China's AI and robotics sectors.

Black Forest Labs has announced Flux 3 Action, a 7-billion-parameter open-weight world action model designed for robots, published on Hugging Face. The release extends the Flux family beyond image generation into robotics, giving developers an openly available model for training robot behaviour. Early reaction is concentrated among AI and robotics practitioners discussing the small parameter count and open weights.

An engineer has written about a hands-on crash course in robotics, arguing that a key lesson is knowing when to get out of the way — letting automated systems and hardware do their job rather than over-intervening. The piece walks through practical mistakes, tuning challenges and design decisions made while learning the field, and is drawing attention from developers and robotics enthusiasts debating the approach.

New humanoid robots developed in the United States are generating significant buzz online, with viewers calling the technology shocking. The discussion centers on how rapidly American robotics are advancing, with commentary focused on what these human-like machines signal for the future of work, automation and AI. Interest spans a global audience following developments in US robotics.

Researchers have developed a tiny jumping robot that can precisely control the height of each jump, a step forward for small-scale robotics. Precise leaping control could prove useful for robots navigating uneven terrain, disaster zones, or confined spaces where wheels or larger machines cannot go. The work is drawing attention from robotics and engineering communities interested in miniature locomotion.

A startup is launching a program that pays people who use prosthetic hands to help train robots, an approach aimed at solving one of robotics' hardest problems: replicating the dexterity of the human hand. The company argues that prosthetic users have unique expertise in how hands grip and manipulate objects, data that could accelerate robot development. It is a novel twist on data collection in the booming robotics field.

A new benchmark has been introduced to measure how well AI systems handle memory in household robots. The test evaluates whether robots can retain and use past experiences to perform everyday domestic tasks more reliably. Researchers and industry watchers say memory remains a major hurdle for home robotics, and such benchmarks help compare progress across systems.

Developers building voice-enabled products are being reminded that the text they write directly affects how natural AI-generated speech sounds. A new guide explains how to write copy that reads well when spoken aloud by text-to-speech systems, covering wording, punctuation and phrasing choices. The piece is circulating among programming and product communities interested in making voice interfaces sound less robotic and more inclusive.

A roboticist is warning that safety should be the top priority in the development of humanoid robots. As companies race to build human-like machines for homes and workplaces, the researcher argues that protecting people must come before speed, capability or commercial goals. The remarks add to a growing debate about how quickly humanoid robots should be deployed and what standards should govern their behaviour around humans.

Innodata has opened a motion-capture lab aimed at helping humanoid robots move more like people. The facility will capture detailed human movement data that can be used to train robotic systems to walk, reach and gesture with more natural motion. The news, reported by The Robot Report, adds to growing investment in robotics training infrastructure as humanoid development accelerates across the industry.

Researchers have introduced a new benchmark designed to measure how well robots powered by vision-language models retain and use memory across tasks. The tool aims to standardize evaluation of robotic memory, a key weakness in current systems, and could help developers compare progress and push robots toward more reliable real-world behaviour.