Physical AI moves from pilots to products, listings and factory contracts

A humanoid company goes public

Churchill Capital Corp XI shareholders have approved the previously announced business combination with Agility Robotics, with more than 84% of votes cast in favour, clearing the way for the company to enter public markets. The transaction and related financings are expected to raise approximately $465 million gross before transaction expenses, comprising about $164 million from Churchill XI's trust account, about $201 million from a common stock PIPE committed at $10 per share and led by Foxconn, and $100 million of equity funding already received after the combination was announced. m:…3533

The transaction is expected to close on November 4, 2026, subject to closing conditions, with the combined company operating as Agility Robotics and its stock and warrants expected to begin trading on Nasdaq under the symbols AGLT and AGLTW on November 5, 2026. Upon closing, Agility is expected to become the only US publicly listed pure-play humanoid robotics company with proven, active commercial deployments, the company says, and Churchill XI will change its jurisdiction of incorporation from the Cayman Islands to Delaware. m:…3533

The listing matters beyond one company because it gives public-market investors a direct vehicle for humanoid robotics, with Foxconn's role in the PIPE linking the deal to manufacturing scale. Whether deployments justify that framing will now be tested in quarterly disclosures rather than in press releases. m:…3533

Robots already at work

Tutor Intelligence co-founder and CEO Josh Gruenstein told the a16z podcast that successful robotics requires a full-stack approach integrating AI models, hardware manufacturing and direct operational deployment, arguing that robotics is not any one layer of the stack alone. The company manufactures its own hardware in Watertown, Massachusetts, and uses human teleoperators, called tutors, to teach robots new tasks in real factory environments, creating training data that improves robot capability. m:…6942

Tutor says its robots have helped manufacture and ship over a quarter of a billion products to American consumers, and that its average time from first call to a signed robot contract is 40 to 60 days. Most of the market is greenfield, the episode summary notes, since the vast majority of US factories and warehouses still rely on manual labour, and Gruenstein argues the industry's failure to scale comes from a lack of general-purpose intelligence and a fragmented market. m:…6942

Pudu Robotics is making a different bet on the same thesis: commercial robotics technology transferred to the home. Its TOMO S series of robotic mowers, unveiled at a product launch event, is built around the SharpLoop self-sharpening system, which automatically sharpens rotary blades at the charging station when needed, and Pudu's One Brain, Multiple Embodiments architecture, bringing together embodied navigation, manipulation and interaction, motion control and multi-robot coordination. m:…8864

The mowers come in three versions for lawns from 1,500 square metres to 5,000 square metres, with four independent high-torque hub motors providing traction on slopes up to 45 degrees, dual 636-watt motors driving blades at up to 6,000 RPM, and Triple-Sense Navigation combining 360-degree LiDAR, AI vision and network RTK for centimetre-level positioning without a local reference station. Sales are expected in early spring 2027, and Pudu says it has shipped over 130,000 robots across more than 85 countries and regions, with more than 1,900 patent applications worldwide. m:…8864

The model layer opens up

Archetype AI has introduced a 14-day free trial for Newton, its physical world model, on Archetype AI Cloud, allowing machine-learning engineers and data scientists to build agents for machines using their own sensor data. Newton is pretrained through self-supervised learning on billions of cross-modal sensor measurements, including vibration, pressure, temperature, electrical current and acoustics, the company says. m:…2963

The trial includes a trial organisation and Newton API key, three Newton agents covering anomaly discovery, operational state monitoring and rare event detection, Python support and agent skills for Claude Code, Cursor, Codex, Copilot and more than 20 other coding agents. Archetype says Newton reaches up to 90% accuracy out of the box in customer deployments without model training, and that 90 to 95% of use cases in customer deployments required no fine-tuning, and it cites a wind-energy deployment in which Newton analysed more than 40 sensors across 63 turbines with no labelled data and found nine previously unknown failure patterns. m:…2963

Taken together, the four stories describe a sector moving from capability demonstrations to commercial plumbing: a listing, a consumer launch date, factory contracts measured in weeks, and a trial that lets engineers test a pretrained physical model on their own machines. The next test is not whether robots work in a demo, but whether the deployments, the margins and the accuracy claims hold up in production. m:…3533 m:…8864 m:…6942 m:…2963