Silicon Valley’s hottest livestream this week is a humanoid robot clocking in for a warehouse shift.
It began Wednesday, when Figure AI CEO Brett Adcock set out to prove to skeptics that his robots could complete an eight-hour stretch of autonomous labor. Within hours, Figure AI had a film crew at its San Jose headquarters and was streaming a humanoid doing one of the dullest tasks imaginable: sorting packages.
The internet was captivated. Millions tuned in to watch the robot pick up small packages and place them on a conveyor belt with the barcode facing down. Two humanoids stood on chargers in the background, ready to sub in when the working robot ran low on battery. One viewer called the feed “surprisingly addicting” and asked for a 24/7 livestream, while investor Jason Calacanis wrote that “robotic ASMR is bizarrely comforting.” As the livestream climbed past 1.5 million views on X over its first eight hours, some viewers gave the three robots names: Bob, Frank, and Gary.
Figure AI reached its goal of running the robot for eight hours with “zero failures,” Adcock said, and decided to keep going. By the 24-hour mark on Thursday morning, the humanoids had sorted more than 30,000 packages, with more than 3 million cumulative views.
The viral stream is more than a robotics stunt. For Figure AI, a startup valued near $40 billion, it is a public audition for a future in which humanoids can work long shifts in warehouses, factories, and eventually homes. The demo gave investors and potential customers a rare look at whether the company’s robots can perform repetitive labor reliably.
It also exposed the gap between spectacle and commercial readiness: Figure’s humanoids may be getting closer to human speed, but experts say they still have a long way to go before they can handle the messy reality of a logistics center.
‘A whole new economy’
Many tasks that feel mindless to humans remain challenging for robots, requiring dexterity, perception, balance, and judgment people barely notice themselves using. That is part of what makes videos of humanoids doing routine work so compelling.
Last week, Figure AI released a video of two of its humanoids making a bed together, while French startup Genesis AI, backed by former Google CEO Eric Schmidt, showed a robot cracking an egg and playing the piano.
Among the viewers of this week’s livestream were Figure investors and board members Jesse Coors-Blankenship and Gregg Hill of Parkway Venture Capital, who said they were watching from their New York office and planned to celebrate the milestone later.
Watch a team of humanoid robots running a full 8-hr shift at human performance levels. This is fully autonomous running Helix-02 https://t.co/IdZR0T1F5I
— Brett Adcock (@adcock_brett) May 13, 2026
Coors-Blankenship said the conveyor belt was a large loop, with the same packages cycling through repeatedly. The point, he said, was to show potential customers that Figure AI’s humanoids can work reliably for long stretches, including 24-hour shifts.
“It’s just something that’s never been done before, except for maybe in a movie,” Coors-Blankenship said. “It’s so captivating because everyone’s realizing we’re moving into a whole new economy.”
The livestream raises the stakes in the race to develop commercially useful humanoids. Figure AI faces stiff competition from Tesla, Agility Robotics, and China’s Unitree.
“I bet you that 50,000 of the viewers are Tesla investors,” Hill said.
‘More like a science project’
When asked about Figure AI’s livestream at an event in San Francisco on Wednesday night, Agility Robotics cofounder Jonathan Hurst replied: “Congratulations. We did that two years ago.” The Oregon-based startup has deployed its humanoid robot, Digit, with customers including Amazon, Schaeffler Group, and GXO, a logistics company.
Last year, Figure AI came under scrutiny after Fortune reported that Adcock appeared to overstate the company’s work with its marquee customer, BMW. Adcock disputed the report, and the company has said its previous humanoid model spent 11 months at BMW’s Spartanburg plant, where it ran 10-hour weekday shifts and helped produce more than 30,000 X3 vehicles.
Figure AI’s livestream did not have any major glitches, but the robots did have some prolonged pauses and eccentric gestures like touching an arm to its helmet-like head. The moments fueled speculation that the humanoids were getting help from a remote human operator. Adcock stressed that the robots are fully autonomous and decide what to do based on what they see through their cameras.
Adcock said that when a robot gets stuck, its AI model triggers an automatic reset, which viewers occasionally saw during the stream. And if a robot has a software or hardware issue, he added, it can autonomously leave for maintenance while another robot takes over.
“We haven’t had a failure yet, but statistically we probably will at some point,” Adcock wrote on X.
Human package sorters average about three seconds per package, according to Adcock, and Figure AI’s robots are now near human parity. But roboticist Ayanna Howard, dean of Ohio State University’s College of Engineering, said speed is only one measure of readiness.
Howard said the livestream was impressive because the robot appeared to operate for so long without a failure. Still, she said the humanoids looked more like a “science project” than machines ready for deployment, citing accuracy issues she saw during the stream, including packages placed on the conveyor belt with the barcode facing the wrong way and one package being knocked off the belt.
“It’s not ready for prime time,” Howard said, adding that the robot was performing only one small part of the package-sorting process.
“We’re a long way away from a fully autonomous humanoid in a logistics center.”
By Thursday evening, the livestream was still running, and the humanoids had logged 30 hours of continuous work. Adcock also introduced a new member to the robot crew: Rose.
Read the original article on Business Insider
The post Silicon Valley’s latest binge-watch is a humanoid warehouse worker appeared first on Business Insider.
