惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
IT之家
IT之家
博客园_首页
博客园 - 【当耐特】
V
V2EX
Apple Machine Learning Research
Apple Machine Learning Research
G
Google Developers Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Recent Announcements
Recent Announcements
F
Fortinet All Blogs
GbyAI
GbyAI
腾讯CDC
H
Hackread – Cybersecurity News, Data Breaches, AI and More
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
I
InfoQ
H
Help Net Security
T
Tailwind CSS Blog
B
Blog RSS Feed
Martin Fowler
Martin Fowler
人人都是产品经理
人人都是产品经理
The Cloudflare Blog
博客园 - 叶小钗
雷峰网
雷峰网
量子位

Forbes - Innovation

Why Do Humans Have Fingerprints? Hint: It’s Not What You Think Booking.com Confirms Data Breach, Reservation PIN Codes Changed Why Major News Sites Are Blocking The Internet Archive’s Wayback Machine iPhone Fold Release Date: New Report Details Frustrating Apple News Comet Tracker: How To See Pan-STARRS And Three Planets On Wednesday NYT Mini Crossword Today: Tuesday, April 14 Hints And Answers Today’s NYT Strands Hints, Spangram, Answers: Tuesday, April 14 (It’s A Little Unclear) Today’s Wordle #1760 Hints And Answer For Tuesday, April 14 Most Of The Microplastics In Urban Air Come From Tires Today’s Wordle #1759 Hints And Answer For Monday, April 13 NYT Mini Crossword Today: Monday, April 13 Hints And Answers NYT Pips Today: Hints, Answers And Walkthrough For Monday, April 13 The YC Chief Who Codes 10,000 Lines A Day Has A Simple Secret Samsung Expands One UI 8.5 Beta To More Galaxy Owners Why You Should Stop Using Your iPhone If It’s On This List Chamath Says Firms That Treat AI As A Strategy Hand Rivals Their Edge 3 Unexpected Habits Of Secure Couples, By A Psychologist The First Lamp That Folds Your Clothes Samsung’s Disappointing Price Update For Galaxy Phone Buyers 3 Subtle Signs Someone Is Falling In Love With You, By A Psychologist Do Mantis Shrimp See More Colors Than Humans? A Biologist Explains NYT Connections Answers Explained For Monday, April 13 (#1,037) NYT Connections Hints Today: Monday, April 13 Clues And Answers (#1,037) LEGO Luigi & Mach 8 (72050) Review: 2026’s Best Set Yet? Marc Andreessen Says AI Productivity Will Trigger A Hiring Boom 3D Printing Is The Ultimate Hack To Reduce Household Spending Apple iPhone Fold: Striking Design Revealed In Leaked Photos Apple Smart Glasses: New Leak Reveals A Major Design Twist To Beat Meta Tested: The AI Coming To The Rivian R2 Quordle Hints Today: Monday, April 13 Clues And Answers
Humanoid Livestream: Robot Makers Rushing To Show Machine...
John Koetsier · 2026-06-23 · via Forbes - Innovation
Agibot G2 robots working on a Longcheer tablet production line. Screenshot from an Agibot promo video.

Agibot G2 robots working on a Longcheer tablet production line. Screenshot from an Agibot promo video.

John Koetsier

The new flex in humanoid robotics isn’t a backflip or a dance routine. It’s a webcam pointed at a factory floor, running for days, watching robots do actual work on actual production lines in actual factories. Starting today and continuing through June 28, Chinese embodied-AI company Agibot is livestreaming a fleet of its G2 humanoid robots with customized grippers working a real production line at Longcheer Technology’s factory in Nanchang. The robots are stationed in the quality-inspection section of a tablet mass-production line, operating alongside human workers and the existing industrial workflow. Anyone can watch on Agibot’s X and YouTube channels.

The company’s wheeled G2 robots have been working on the production line since at least April of this year. Using customized grippers for the specific job they’re doing, they’re picking up, moving and placing tablets for testing, among other things.

Humanoid makers are under increased pressure to show actual utility in their robots and cobots, and livestreams appear to be part of that plan.

If a multi-day factory livestream feels familiar, that’s because Figure AI just did something similar, turning it into a bit of a challenge and spectacle. Last month, Figure ran a 10-hour "Human vs. Robot" package-sorting challenge, pitting its Figure 03 humanoid against a hapless human intern named Aime. The human won, barely: Aime sorted 12,924 packages to the robot's 12,732, a margin of just 192.

Humanity’s last win, perhaps?

Of course, winning wasn’t really the point: the intern’s fingers and hands were sore enough to quit, while Figure was still working. In this iteration of the rabbit versus the tortoise, we won the sprint but the robot was built for the marathon.

And Figure leaned in farther, live-streaming the first eight hours, then 24, then well past 100. Viewers gave the robots names like Bob, Frank, Gary, Rose, and Jim, watched to see which would jam or drop a box and turned an intentionally tedious feed into something between a reality show and a sleep aid. Figure later pushed the format to a 200-hour continuous run, with Figure 03 processing roughly 249,560 packages — about 1,248 boxes an hour, or one every 2.88 seconds.

Of course, the tedium is the point. If humanoids are going to earn a place in our factories and warehouses, the test isn’t a single elegant motion, it’s doing the same unglamorous thing for hours and days on end without failing.

Humanoid makers racing onto the factory floor

The bigger story of course is that the entire humanoid robotics industry is trying to prove a point: humanoids are not just toys. They can do actual real work.

Figure’s earlier Figure 02 went into BMW’s plant in Spartanburg, South Carolina, placing sheet-metal parts into fixtures and Figure 3 is going to be working at retail holding company Catalyst Brands. Agility Robotics’ Digit broke the ice as the first humanoid with a real job back in late 2024 and Tesla’s Optimus has been demonstrating battery sorting and parts handling inside Tesla’s own factories. In China, the list is long: Ubtech’s Walker S has trained inside auto plants for NIO, Geely’s Zeekr, BYD and others and expanded to Foxconn and SF Express; and Xiaomi hired a humanoid for an "internship" in its own car factory, reporting a 90.2% success rate on bilateral nut-fastening over a three-hour autonomous run.

None of these machines is necessarily the optimal tool yet. For fixed, repetitive motions, a purpose-built robotic arm is still faster and more reliable, and skilled humans remain cheaper and far better at handling the unexpected like a jammed box, a tilted bin, a dropped part.

But things are changing fast, and humanoid models are getting better and better, with robotic hands that are increasingly capable.

The relatively recent ecosystem-level change, however, is the burden of proof for humanoid makers.

The credible flex is no longer a cool demo video or a robot dance. The flex is "watch our robot do it live, IRL, for days on end, and see if it breaks."

Agibot’s six-day broadcast is the latest entry in a contest the whole field is now starting to run in public, one livestream at a time.