Three humanoid robots completed more than 24 hours of continuous autonomous operation sorting packages without human intervention. The machines, developed by California-based startup Figure AI, were supposed to run for eight hours but kept working through the night and into the next day. The robots sorted more than 28,000 packages during the extended test, performing at speeds close to human workers. Viewers watching the livestream started calling the robots Bob, Frank and Gary, and the company added visible name tags after the nicknames caught on online.
The test revealed how far humanoid robotics has advanced in warehouse settings. The robots operated independently, using onboard cameras and artificial intelligence to detect barcodes and place packages correctly on conveyor belts. No remote operators controlled the machines. Every movement came directly from the Helix-02 AI system developed by Figure AI.
AI system combines vision, touch and movement control
The Helix-02 neural network integrates multiple functions that humanoid robots need to handle warehouse tasks. The system combines vision processing, touch sensing, body awareness and movement control into a single platform. Robots using this technology must do more than simply move their arms. They need to maintain balance, adjust their grip on different package sizes, shift their posture and respond when objects land in awkward positions.
Figure AI built the system entirely in-house rather than licensing technology from other companies. The robots used onboard cameras to scan each package, identify the barcode location and determine the correct placement on the conveyor belt. The company emphasized that all reasoning happened within the robot’s own processing units. No external computer systems guided the machines through wireless connections.
The task itself sounds simple but requires consistent precision. Pick up a package. Find the barcode. Place the package on a conveyor belt with the barcode facing down. Then repeat. Warehouse operations depend on steady movement, quick decisions and the ability to keep going when small problems appear. A package might arrive damaged, a barcode might be obscured or equipment might shift slightly during operation.
Robots demonstrate automatic reset capability
One feature that distinguishes the Helix-02 system involves recovery from unexpected situations. Figure AI says the robots can trigger an automatic reset when they encounter a problem outside their expected behavior patterns. The machine pauses, evaluates the situation and resumes work without human assistance. This capability could prove crucial in real workplace environments.
- Robots that need frequent human help become operational burdens rather than productivity tools.
- Machines that can pause, reset and resume work offer more practical value.
- The system can direct a robot to leave the work floor for maintenance if software or hardware issues appear.
- Another robot can take over the task while maintenance happens.
The ability to swap robots in and out without stopping the entire operation addresses one of the main concerns companies have about automation. Traditional industrial robots often require complete line shutdowns when problems occur. A system that allows continuous operation through robot rotation could change how warehouses approach staffing and maintenance schedules.
Company faces competition from major tech firms
Figure AI is not alone in developing humanoid robots for industrial applications. Tesla, Agility Robotics and Apptronik are all working on similar machines designed for warehouses, factories and logistics operations. The competition reflects growing industry interest in automation that can handle tasks currently performed by human workers.
Figure AI has already tested its robots at BMW manufacturing facilities in South Carolina. That partnership suggests where this technology might appear first. Controlled industrial environments with repetitive tasks provide ideal testing grounds before robots move into more complex or unpredictable settings. Companies can monitor performance, identify failure points and refine the technology before broader deployment.
The package-sorting demonstration gives potential customers a concrete example of what these robots can do. A 24-hour test run provides more credible evidence than short promotional videos. However, one livestreamed task does not prove the robots can handle the full range of warehouse conditions. Real facilities present challenges that controlled tests may not capture.
Real-world conditions present additional challenges
Warehouse floors can get chaotic in ways a demonstration cannot fully replicate. Packages arrive in different shapes and sizes. Labels appear in odd places or become damaged. Conveyor belts can jam. People walk through work areas. Equipment fails. A robot that handles one carefully monitored task still needs to prove it can manage the messier version of the job.
Companies considering humanoid robots will want evidence beyond what one startup provides. They will ask how often the robots fail during extended use. They will want data on maintenance requirements and costs. They will need proof that robots can handle disruptions without slowing down entire operations. Independent verification from outside testing organizations could provide more confidence than company demonstrations.
The cost remains another major question. Figure AI has not disclosed pricing for its humanoid robots. Companies will compare the purchase price, maintenance costs and operational expenses against the wages and benefits they currently pay human workers. The calculation includes training time, error rates, speed and the ability to work unusual shifts. Robots that work 24 hours without breaks sound impressive, but only if the total cost makes financial sense.
Automation raises questions about workforce impact
The demonstration highlights broader concerns about automation and employment. A robot that can work continuously without breaks, sick days or vacation time appeals to companies looking to cut labor costs. For workers in warehouses and distribution centers, that capability raises obvious concerns about job security. Package sorting and similar repetitive tasks could become targets for automation as the technology improves.
Not every warehouse job will disappear immediately. Real workplaces are messy and unpredictable in ways that challenge current robotics. People still solve problems that demonstrations rarely show. Workers handle exceptions, deal with damaged goods, communicate with supervisors and adapt to changing conditions. However, the Figure AI test suggests humanoid robots are moving from short promotional clips toward longer workplace trials that could eventually lead to commercial deployment.
The impact of warehouse automation may show up in ways consumers notice. Faster package handling could affect delivery times. Companies might change how they staff overnight shifts or seasonal peak periods. Robots could fill roles that are physically demanding or hard to staff. The technology is moving from laboratory concepts toward practical applications that could reshape how distribution centers operate in the coming years.