Agility’s bipedal humanoid Digit just learned to dead-lift a 65-lb (29.5 kg) object with dynamic whole-body balance. No new actuators, no mechanical upgrades. The entire policy was trained in simulation, then dropped onto the real robot. Keep in mind — Digit is the only humanoid currently on a payroll, already moving totes at GXO Logistics and Amazon.
How they pulled it off
Agility embedded target payloads directly into the sim loop and let the policy solve for load distribution, grip forces, and a shifting center of mass. Transfer to hardware pushed the existing joints and actuators to their limits — but nothing broke, nothing got redesigned. Pure sim-to-real, heavier loads.
The API angle
Digit ships with Agility Arc, a Digit Cloud plus SDK stack exposing REST and WebSocket endpoints. A WMS or AI agent can enqueue tote-handling missions, stream live robot telemetry, and trigger mission-level primitives across an entire fleet. That is how GXO is running Digit in production today. Heavy-load manipulation plugs into the same interface — no new integration work for existing customers.
Why it is trending this week
IEEE Spectrum’s April Video Friday led with the dead-lift clip, and robotics X and Bluesky have been chewing on it for days. The subtext everyone landed on: warehouse labor math just moved.
You Might Also Like
- Amazon Paid 50000 per Unit for This 3 5 Foot Humanoid Fauna Robotics Sprout is now an Amazon Robot
- Google Search Live now Covers 200 Countries and it Changes What Searching Means
- Sanctuary ai Phoenix Zero Shot Dexterous Hands 10 Cube Rotations Zero Real World Training
- Zeroth Robotics m1 w1 a 2899 Humanoid and a 4999 Cargo bot Ship This Week
- Bmws Leipzig Factory now has a Humanoid Coworker Hexagon Robotics Aeon
Top AI Product 
Leave a comment