355
Audio & Video Production344
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps173
Writing & Content Creation203
Data & Analytics140
Design & Creative169
Customer Support130
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Operations & Admin87
Education & Learning131
Figure AI has streamed its Figure 03 humanoid robots sorting packages around the clock, including a 10 hour contest against a human intern.
In short: Figure AI has been running a near nonstop livestream of its humanoid robots sorting packages, and the stream has turned into a viral online event.
Figure AI, a robotics startup, started livestreaming its Figure 03 humanoid robots on May 13. The robots inspect barcodes on small packages like boxes and padded envelopes, then place them on a conveyor belt with the barcode facing down.
The company said the robots are working on their own without a person controlling them. The main change from earlier demos is time. Figure aimed for an eight hour run, then decided to keep the stream going 24/7. The robots rotate in and out because each one needs to recharge after about three to four hours, like swapping batteries during a long road trip.
Viewers began treating the stream like a show. Commenters named robots Bob, Frank, Gary, and Rose. Figure’s CEO also promoted robot themed merchandise. Some people even placed bets on Polymarket, a prediction market (a site where people wager on outcomes), about how long the robots would run and how many packages they would move.
A big moment came on May 17, when one robot competed against a human intern for 10 hours doing the same task. The intern sorted 12,924 packages. The robots sorted 12,732. The CEO posted that the intern averaged 2.79 seconds per package versus 2.83 seconds for the robots.
The stream shows real mistakes, like dropped packages, which can make it feel more trustworthy than a short edited video. Still, outside observers cannot easily confirm whether any hidden help is involved. Even if the demo is fully independent, it only proves the robots can do one repetitive job in one specific setup, not that they can handle the variety of tasks found in real workplaces.
Source: Arstechnica