AnswersPublished25 sources
What AgiBot’s 24-Hour Robot Store Trial Showed—and What It Didn’t
AgiBot reported that six G2 robots completed 1,730 tasks during a livestreamed 24 hour Shanghai convenience store trial, which drew more than 2 million online views. The robots handled retail work including customer greeting, restocking, coffee and food service, and cleaning, across four shifts.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- AgiBot reported that six G2 robots completed 1,730 tasks during a livestreamed 24 hour Shanghai convenience store trial, which drew more than 2 million online views.
- The robots handled retail work including customer greeting, restocking, coffee and food service, and cleaning, across four shifts.
- Genie Master was described as coordinating tasks and using execution feedback; the available reporting does not establish that the full trial ran without human supervision.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What happened during AgiBot’s livestreamed 24-hour trial of six G2 humanoid robots in an operating convenience store in Shanghai’s Xuhui dis. Article summary: AgiBot’s six G2 robots completed a livestreamed 24-hour convenience-store trial, with the company reporting 1,730 completed tasks and more than 2 million online views. The available evidence supports those headline resul. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
AgiBot’s six G2 robots spent 24 hours working in a Shanghai convenience store during a livestreamed trial on November 2–3, 2026. The company reported 1,730 completed tasks and more than 2 million online views. Those headline figures show the scale of the event, but the available reports do not provide a category-by-category task count or enough performance data to judge how reliably the robots worked. 1
13
31
What the six robots did
The trial ran in a real store in Shanghai’s Xuhui district and was presented under the title “Robots Never Clock Out.” Coverage describes four shifts spanning busy daytime periods, overnight work and breakfast preparation. 6
17
19
The robots’ reported duties included greeting customers, restocking and organizing shelves, checking aisles, preparing coffee and hot food, handling checkout or service-counter work, and cleaning. One report describes two robots working as a restocking pair, with others assigned to aisle patrol, coffee, checkout and hot food, and table cleaning. 18
19
A livestream gave viewers a longer look at robots working around customers and store equipment than a brief demonstration would. But watching a shift unfold is not, by itself, evidence of uninterrupted autonomy or deployment-ready reliability. 4
11
21
What the 1,730-task total does—and doesn’t—show
AgiBot’s reported total is a count of completed tasks. The available coverage does not explain the counting rules or show how many tasks fell into each category, such as restocking, food service or cleaning. 1
31
The reports also do not establish a task success rate, failure rate, number of human interventions, or takeover rate for the completed trial. An earlier summary of the framework noted a lack of published success rates and error counts; the materials available here do not fill in those gaps. 8
11
26
So the task total is useful as an activity figure, but it cannot answer how often the robots completed tasks correctly, how much help they needed, or how their performance varied by job.
Genie Master and the limits of “no engineer in the loop”
The trial was described as using AgiBot’s GO-3 foundation model alongside Genie Master, an embodied-agent framework. A completed-trial summary says Genie Master coordinated task decisions and feedback among the six G2 robots. 1
Before the trial, reports described Genie Master as a system that could inspect execution logs, identify failures, adjust a robot’s actions or strategy, and retain fixes or lessons for later tasks. 3
5
7 That describes the intended learning approach; the available reporting does not give enough detail to verify the trial’s specific learning cycles, memory implementation or any quiet-hour practice.
The phrase “without an engineer in the loop” refers to the reported failure-diagnosis and adjustment process. It should not be read as proof that no person supervised, assisted or took over during the full store shift. The supplied reports do not provide a takeover count or rate. 5
7
11
What remains unverified
The available evidence does not establish a breakdown of the 1,730 tasks, probation results or evaluation criteria. It also does not document how the robots responded to specific drink-selection requests, deliberately disordered shelves or obstacles, nor does it detail particular fine-manipulation failures and recoveries.
The event does show that six G2 robots were put through a varied, livestreamed retail shift using GO-3 and Genie Master. It does not, on the reported figures alone, establish how accurately or independently they handled each task. 1
4
11
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "What AgiBot’s 24-Hour Robot Store Trial Showed—and What It Didn’t"?
AgiBot reported that six G2 robots completed 1,730 tasks during a livestreamed 24 hour Shanghai convenience store trial, which drew more than 2 million online views.
What are the key points to validate first?
AgiBot reported that six G2 robots completed 1,730 tasks during a livestreamed 24 hour Shanghai convenience store trial, which drew more than 2 million online views. The robots handled retail work including customer greeting, restocking, coffee and food service, and cleaning, across four shifts.
What should I do next in practice?
Genie Master was described as coordinating tasks and using execution feedback; the available reporting does not establish that the full trial ran without human supervision.
Sources
korthosrobotics.comAgiBot — Humanoid robotics Company Profile — Korthos
humanoidsdaily.comEverything we've written about AGIBOT (Zhiyuan Robotics)
robothot.comChina - RobotHOT
en.wedoany.comChina's AgiBot Launches 24-Hour Convenience Store Operation Test with 6 Robots