Robocurve is a Public Benefit Corporation, backed by Y Combinator, that builds open-source tools and independent benchmarks for measuring how capable AI models are at real-world robotics tasks. Its site reports frontier-model scores on robot task suites such as RoboDojo-RC and refusal benchmarks such as RoboHarm, along with cost-per-attempt and inference-speed comparisons across model configurations.
The site of Prometheus, an AI startup co-founded in November 2025 by Jeff Bezos and Vik Bajaj (co-CEOs) that is building AI tools to speed up engineering and manufacturing of physical products. It raised $12 billion at a $41 billion valuation in June 2026, per CNBC.
Microsoft Research blog post on running the heavy AI thinking for robots on remote computers instead of on the robot itself, and what that means for speed, cost and reliability in real-world robotics.
Why I recommend it: Free to read. Written by Microsoft researchers, so it presents their own approach favorably — Microsoft also sells the cloud computing this kind of setup relies on.
The manufacturer's own pages for the humanoid and four-legged robots you keep seeing in viral videos — H2, G1, R1, A2, B2, Go2 and the L2 lidar unit. Each model page lists the specifications, dimensions, sensors and intended uses, and the news section posts the demonstration videos. This is the clearest free way to see what commercially available 'physical AI' hardware actually is, rather than what a clip implies.
Why I recommend it: Free to browse, and useful as a reality check when someone tells you robots are about to take every job — read the specifications, the battery life and the intended use. Two flags: this is a company selling its own products, so the videos are marketing and not independent testing, and the robots are expensive hardware, not something you can try. The site itself warns users not to modify the robots or use them in dangerous ways.
