Four robotics leaders: inspection dogs, home humanoids, and the factory of the future

Chapters

1. 0:00 AI and tech

   Opening at Machina in Paris

   Jay Calacanis opens from the Machina conference in Paris, framing it as an AI-in-the-real-world event. He teases a series of interviews with the leaders of the top robotics companies and immediately moves from the cold open into the first conversation. The chapter sets the stage for the episode’s core questions: where robots are deployed today, what the form factors are, and how quickly humanoids will move into homes and factories.

2. 1:02 AI and tech

   ANYbotics: four-legged inspectors for critical infrastructure

   Peter Funkhouser, co-founder and CEO of ANYbotics, says four-legged robots became the standard for industrial inspection because they climb stairs, keep their balance on slippery floors, and carry a wide, stable footprint. The robots are packed with thermal cameras, acoustic microphones, gas sensors, and GPUs; they cost in the low hundreds of thousands and run repeated missions, sometimes 40 times a day. The real value is avoiding downtime at oil, gas, and offshore wind assets that lose hundreds of thousands of dollars per hour. The company builds spark-proof units for explosive atmospheres and sources zero components from China. Funkhouser argues Chinese competitors make polished hardware but lack the autonomy, workflow integration, and data-trust certification needed to win industrial customers. He has no military program and joined Boston Dynamics and others in condemning weaponized robots.

3. 14:59 AI and tech

   1X Neo: home humanoid pre-orders and open platform

   Bert Borick, co-founder and CEO of 1X, says Neo pre-orders sold out the first 10,000 units in days, with about $500 per month pricing and first units shipping in 2026, initially rough around the edges. He frames Neo as an open platform: buyers can use the robot operating system, gather data with tactile-sensor gloves, fine-tune models, run teleoperation, or deploy third-party models from labs such as OpenAI or Claude. The company’s big bet is making Neo humanlike enough to pretrain on internet-scale video of humans, since robotics control data does not exist at that scale. Borick says the path to a fully autonomous home agent is close, and he predicts hard takeoff—robots building robots, data centers, and chip fabs—in under 10 years, with his current estimate at 3 years.

4. 33:50 Policy and regulation

   Boston Dynamics: Spot, Atlas, and US robotics strategy

   Amanda McMaster, interim CEO of Boston Dynamics, says the company is now a deployment business, not a research lab. Spot has more than 500 customers in 46 countries, with industrial inspection and security the largest use cases, mean time between intervention over 3,000 hours, and prices from roughly $100,000 to $300,000. Atlas will likely be sold as robot-as-a-service. She describes a two-brain architecture: physical control stays on the robot, while semantic reasoning can live in the cloud with partners such as Google DeepMind. She says 100% of robots are built in the United States, opposes Chinese humanoid robots in US critical infrastructure on security grounds, and calls for a national robotics strategy. On military work, the company has an anti-weaponization stance but does EOD and non-weaponization government work.

5. 47:26 AI and tech

   Why humanoids are real now: perception, data, and practice

   Jonathan Hurst, co-founder and chief robotic officer of Agility Robotics, explains why this humanoid moment is real after decades of false starts. Hardware that looks human is easy; hardware that does useful work in human spaces is hard. Large language models made perception nearly free, but robot control still lacks an internet-scale training set of torques and sensor inputs, so the field uses teleoperation, demonstration data, animation, world models, and sim-to-real together. There is no silver bullet; progress is a snowball of money and engineering effort. Humans need little data to learn a skill, while robots currently need much more, but robots can share learned skills over Wi-Fi once one figure out a task, which changes the scaling equation.

6. 55:43 AI and tech

   Digit V5, warehouse economics, and the future of work

   Hurst walks through Agility's Digit robot doing tote-moving and palletizing workflows. Digit V5, due later this year, is the first balancing humanoid that can leave a work cell without physical safety barriers, a milestone that unlocks wider warehouse adoption. The business model is capex or robot-as-a-service; robots run about 20 hours per day and accumulate roughly 40,000 hours over five years. Costs are already tens of thousands and should fall toward car-like prices once 100,000 units scale, undercutting $20–$40/hour human labor. Hurst says many factories are already majority-robot, with lights-out sortation coming, and humanoids will expand into retail, grocery, hospitals, construction, and last-mile delivery. He also highlights blue-collar careers in robot assembly, operation, and maintenance, and urges students to build engineering skills.