Thirty-Five Hundred Hours of Actually Using a Brain Implant

There's a number that should make every robotics company pause: 3,800 hours. That's how long Casey Harrell, paralyzed by ALS, has used a brain-computer interface to speak, work, and browse the web. Not in controlled demonstrations. Not in supervised trials. In his actual life, over nearly two years.

The robotics industry loves to showcase capabilities. We see humanoids doing backflips, warehouse robots moving at impressive speeds, and surgical systems performing delicate procedures. But Harrell's story reveals something the industry desperately needs more of: longitudinal data on systems that people actually use, day after day, when the cameras are off.

Consider what 3,800 hours represents. It's the equivalent of running a factory robot for nearly half a year of continuous operation. In industrial settings, we obsess over mean time between failures and duty cycles. Yet when it comes to assistive technology and human-robot interaction, we rarely get to see what happens after month six, let alone month twenty-two.

The UC Davis researchers who developed Harrell's system didn't just create a working brain-computer interface—they created one reliable enough that someone would choose to use it thousands of hours independently. That's a fundamentally different engineering challenge than building something that works impressively in a demo.

This matters for the broader robotics field because we're entering an era where robots are supposed to work alongside humans for extended periods. The humanoid robots Boston Dynamics plans to deploy, the collaborative robots ABB is integrating with dexterous grippers, the autonomous vehicles Gatik is putting in supply chains—all of these depend on sustained, reliable operation in uncontrolled environments.

Yet most robotics announcements focus on capabilities rather than durability. We hear about what a robot can do, rarely about how long it can keep doing it or how much human intervention it requires over time. Genesis AI's Eno robot and Kawasaki's RL030N platform both represent significant technical achievements, but their real test will come in the accumulated hours of actual deployment.

Harrell's experience also highlights something the industry often overlooks: the human side of human-robot interaction. After 3,800 hours, he's not just using the technology—he's living with it. He's discovered its quirks, worked around its limitations, and integrated it into his routines in ways the designers probably never anticipated. That kind of co-adaptation only happens with real, sustained use.

The physical AI partnerships we're seeing—Built Robotics with Penn, Kawasaki with its platform approach—are collecting operational data measured in thousands of hours. That's the right direction. But we need to get better at sharing not just success metrics, but the messy reality of long-term deployment: the unexpected failure modes, the maintenance requirements, the ways users adapt and systems drift.

One man using a brain implant for 3,800 hours has generated more useful data about sustained human-AI interaction than a thousand perfect demos ever could. The robotics industry should take note. The future isn't built on what works once in a lab. It's built on what keeps working, day after day, when real people depend on it.