Six ways into physical AI.
A hub-and-spoke field guide — from first principles and the humanoid race to the stack, fleet operations, real deployments, and the market shaping all of it.
PHY-01Physical & Embodied AI FoundationsWhat separates physical AI from a chatbot — embodiment, world models, vision-language-action models, sim-to-real transfer, and the sensor stack that lets a machine act in the real world.HUM-02Humanoids: The Race for the Robot BodyWho is actually shipping — Figure, Tesla Optimus, Agility Digit, Unitree, 1X, Boston Dynamics — with dated, sourced status updates and the unit economics behind the hype.STK-03The Physical AI StackWhat you actually build and train on — Nvidia Isaac/GR00T, ROS 2, simulation and digital twins, teleoperation data collection, and the edge compute that runs it on the robot.OPS-04RobotOps: Running a Robot FleetDevOps for physical machines — fleet telemetry and observability, OTA updates, CI/CD for robot software, incident response, predictive maintenance, and the interop standards that keep mixed fleets running.APP-05Where Physical AI Works TodayWarehouses, manufacturing lines, eldercare, farms, construction sites and retail floors — real deployments with real payback numbers, not demo-day footage.MKT-06Market, Investment & RegulationMarket forecasts and players, funding rounds, robotics ETFs, the EU AI Act as it applies to physical machines, machinery-safety norms, and the labour-market question everyone asks.