Figure AI's three humanoid robots have crossed a significant threshold: more than 24 hours of continuous, fully autonomous operation without human intervention or failure. The California-based robotics startup livestreamed the nonstop work session, where the robots, nicknamed Bob, Frank, and Gary by online viewers, sorted over 28,000 packages while maintaining speeds comparable to human workers.

What Makes This 24-Hour Run Different From Earlier Tests?

Figure AI originally planned an eight-hour autonomous test using its Helix-02 AI system. When the robots completed that initial target without any failures, the company made an unexpected decision: keep them running. The livestream drew significant online attention as viewers continuously tracked the robots' uptime and performance metrics as the operation extended far beyond its original scope.

"Our original goal was an 8-hour run. After zero failures yesterday, we decided to keep going. We're now over 24 hours of continuous autonomous operation without a failure. This is uncharted territory," stated Brett Adcock, founder and CEO of Figure AI.

Brett Adcock, Founder and CEO at Figure AI

The robots detect barcodes using onboard cameras, pick up packages, and place them barcode-down onto conveyor belts using AI reasoning. According to Adcock, the robots have achieved near-parity with human workers, processing packages in roughly three seconds each, the same speed as human employees.

How Does the Autonomous System Handle Problems Without Human Help?

One of the most significant aspects of this demonstration is that the robots operated with zero teleoperation, meaning no remote human control guided their actions. Instead, Helix-02, Figure AI's in-house neural network running entirely onboard each robot, made all decisions independently. The system combines vision, touch sensing, proprioception (awareness of body position), and whole-body control into a single unified AI model, unlike traditional industrial robots that separate movement and manipulation systems.

The robots also demonstrated autonomous problem-solving capabilities that could reshape how warehouses handle equipment downtime:

• Automatic Recovery: If a robot gets stuck or encounters an unfamiliar situation outside its training distribution, the Helix-02 system triggers an autonomous reset and resumes work without human intervention.
• Self-Maintenance Routing: When a robot detects software or hardware issues, it autonomously leaves the work floor for maintenance while another robot automatically takes over operations to maintain continuous uptime.
• Reasoning From Camera Pixels: The robots process raw visual information directly, reasoning about package placement and barcode orientation without pre-programmed instructions for every scenario.

This self-managing approach addresses one of the biggest operational challenges in warehouse automation: downtime. Rather than halting the entire operation when one unit needs service, the system redistributes work dynamically.

How to Evaluate Humanoid Robot Readiness for Your Warehouse

• Continuous Uptime Performance: Look for robots that can operate 24/7 without human intervention, as demonstrated by Figure AI's extended run, rather than systems requiring frequent manual resets or teleoperation.
• Speed Parity With Human Workers: Verify that robots can match or exceed human processing speeds (roughly three seconds per package in sorting tasks) to ensure they add genuine productivity gains rather than creating bottlenecks.
• Autonomous Problem-Solving: Assess whether robots can detect and recover from errors independently, route themselves for maintenance, and redistribute tasks across multiple units without human oversight.
• Onboard AI Processing: Confirm that decision-making happens entirely on the robot itself rather than relying on cloud connectivity or remote human operators, which reduces latency and improves reliability.

Figure AI's demonstration builds on earlier claims that its humanoid robots completed full eight-hour shifts autonomously using Helix-02. The company has also previously tested humanoid robots at BMW manufacturing facilities in South Carolina, providing real-world validation beyond controlled demonstrations.

The 24-hour milestone arrives as Figure AI competes with firms including Tesla, Agility Robotics, and Apptronik to commercialize humanoid robots for warehouse, factory, and logistics operations. While other companies have announced humanoid robot projects, few have publicly demonstrated this level of sustained autonomous operation without failures or human intervention.

The livestreamed nature of the test also served a secondary purpose: building public confidence in the technology. By allowing online viewers to monitor the robots continuously and even nickname them, Figure AI created transparency around a technology that many people remain skeptical about. The fact that the robots maintained performance over an extended period, rather than failing after a few hours, provides concrete evidence that the underlying AI system can handle real-world variability and unexpected situations.