248 articles · Last covered
RoboCurve benchmarked OpenAI’s new GPT-6 Astra on real robot arms, revealing dramatic efficiency and spatial reasoning gains on gross manipulation alongside a persistent 10% plateau on millimeter-precision tasks.
NVIDIA has agreed to acquire Hugging Face for $12.93 billion, bringing the open-source hub into its physical AI orbit just as Pollen Robotics launches the viral $399 Microduck.
Figure AI has partnered with cloud infrastructure provider Nscale to deploy up to 100,000 NVIDIA Vera Rubin GPUs, committing an initial $3.5 billion to train its Helix robotics models.
In an appearance on the Sources podcast, OpenAI CEO Sam Altman unequivocally confirmed that the AI lab will build humanoid robots alongside specialized data center machines.
Nori Robotics has debuted the A3, a sub-$1,700 bimanual mobile manipulator aimed at consumer homes. But technical papers and newly opened SDKs reveal the divide between appliance marketing and experimental developer hardware.
San Francisco startup Lightberry has opened reservations for Lumi, a $39,990 bipedal robot pairing a Unitree mechanical baseline with custom American sensing, NVIDIA Thor compute, and a modular interaction engine.
Masayoshi Son''s investment powerhouse is reportedly negotiating a controlling interest in OpenAI-backed humanoid maker 1X Technologies, at a figure well under the valuation the company pursued last September.
At its 2026 CEO Investor Day, Hyundai put hard numbers behind Atlas for the first time: 30,000 units of annual US capacity from 2028, 25,000 already spoken for. It also went out of its way to insist robots are "in service of humans" — a line delivered 24 hours after its Korean union settled a walkout that demanded protection from automation.
Figure AI has exited stealth on Index, a crowdsourced data-collection platform that has gathered over 16 million video uploads across 108 countries to train its Helix neural network.
Skild AI has detailed S1, a foundation model that executes unseen, 10-minute manipulation tasks from a single video prompt without task-specific post-training.
Figure CEO Brett Adcock shared video of a Figure 03 humanoid autonomously climbing up and down a 15-foot ladder, confirming the vertical locomotion milestone maps directly to an active customer use case.
Generalist AI claims a GPT-3 moment for robotics, demonstrating a foundation model that learns new physical tasks in seconds from a single prompt.
