Agility Robotics Unveils Digit 5, Aiming to Scale Safe, Affordable Humanoid Labor to Tens of Thousands

Agility Robotics, the company I helped launch out of my Oregon State University robotics lab in 2015, has introduced Digit 5—our latest attempt to turn humanoid mobility into a practical industrial tool. The new machine is built with a specific ambition: to become the first human-shaped robot that can perform genuinely useful work safely, affordably, and at large scale. If it succeeds, it could move humanoid robots from a handful of demonstrations into deployments numbering in the tens of thousands.

Digit 5 is engineered for warehouse labor. It can lift 23-kilogram bins and place them onto high shelving, a task designed to mirror the repetitive, physically demanding jobs that many employers struggle to staff. Its power system is tuned for fast turnaround: the robot can operate for about 90 minutes after a charge that takes roughly nine minutes. The goal is not an occasional pilot but sustained work—cycle after cycle, shift after shift—particularly in roles that are punishing to the body or simply too monotonous for most people to want.

Yet in an important way, Digit 5 represents a departure from the robot I once imagined building. It is substantial in mass and volume, and it does not have the fluid, lifelike motion that defined earlier machines in my career. For a commercial product, we made choices that prioritize reliability, safety, and manufacturability over elegance. The result is a robot that is closer to an industrial appliance than the graceful, physics-forward walker that first inspired me.

My long-term aim has always been to merge locomotion and manipulation so robots can traverse the same spaces people do and interact with the world in similarly versatile ways. When I founded Oregon State’s Dynamic Robotics Laboratory in 2008, our work leaned heavily on biomechanics. Decades of research suggested that a basic “spring-mass” abstraction—essentially, a body supported by springlike legs—could account for the range of gaits seen across animals, from walking to running.

We tested that idea by building ATRIAS, a bipedal platform whose name—“Assume The Robot Is A Sphere”—nods to the physicist’s habit of simplifying complex problems. ATRIAS was never meant to resemble a person. It looked more like a large box balanced on long legs, a physical proof-of-concept rather than a product. But when it found a stable gait, the motion looked strikingly natural, hinting that the underlying principles might scale beyond the lab.

After ATRIAS, the main obstacles no longer felt like unanswered science questions. Instead, they became engineering challenges and the practical realities of building an organization capable of transforming a prototype into something repeatable. That belief helped drive the decision to form a startup. I was convinced we could make robots that could go where people go—and that such mobility would be broadly valuable. Like many founders, I underestimated how difficult the path would be.

In hindsight, we didn’t follow the standard playbook of starting with a defined customer problem and then designing a solution. My cofounders, Damion Shelton and Mikhail Jones, and I largely trusted that the technology itself would open doors, and that the business model would emerge along the way. So we built Cassie: a dynamic biped intended for research customers. Cassie balanced through active, continuously adjusting motion rather than slow, deliberate stepping. It had long, birdlike legs—an outcome of designing for performance, not for imitation—and it lacked a torso, arms, or a head because commercial utility was never the point.

Research institutions did buy it. Cassie units went to groups at Harvard, Caltech, the University of Michigan, and the University of California, Berkeley, among others. But the market for $250,000 bipedal research platforms was small. After selling 10 robots, it was effectively saturated—setting the stage for the far more demanding transition to machines like Digit 5.

Why it matters

Warehouses and similar workplaces face persistent labor shortages and high injury risk from repetitive, heavy tasks—exactly the niche Digit 5 is designed to target. If Agility can deliver safe, economical humanoid robots at scale, it could reshape how physically taxing logistics work is staffed and performed. The next key questions are whether the robots can sustain real-world uptime and whether manufacturing can meet the promised volumes.