What happened: AgiBot says four of its Genie G2 humanoid robots completed an eight-hour, live-streamed shift on a tablet manufacturing assembly line in Nanchang, performing final quality inspections and handling exceptions on the conveyor.

Why it matters: This is not a stage demo in a spotless lab, it is precision work on a moving line. AgiBot claims the robots can adapt to small positional deviations and disturbances, complete calibration quickly when product models change, and sustain high-throughput inspection without being fully pre-programmed for every variation.

Wider context: According to the report, the G2’s operations took about 18 to 20 seconds each, processing roughly 310 units per hour with an overall success rate above 99.9%. A Longcheer executive said the robots were integrated into a consumer electronics mass-production line in four months and have accumulated 140 hours of continuous operation, with deployment expected to expand to 100 units by Q3 2026.

Background: Xinhua frames this as part of China’s broader embodied-intelligence push into manufacturing. It also cites an Omdia report: Shanghai-based AgiBot shipped over 5,100 humanoid robots in 2025 and held 39% of the global humanoid market, followed by Unitree and UBTech.

Chinese humanoid robots deployed on assembly lines for precision tasks — Xinhua

Droid Brief Take: If the numbers hold up outside a live-stream spotlight, this is the kind of dull, repeatable factory work that turns humanoids from vibes into invoices. But the burden of proof is still the boring stuff: uptime, intervention rate, and how often humans quietly save the robot from itself.

Key Takeaways:

• Line Throughput: Xinhua reports the G2 completed each operation in 18–20 seconds, processing about 310 units per hour, with a claimed overall success rate exceeding 99.9%, which is the kind of metric that actually matters for production lines.
• Adaptation Claims: The report says the robots can automatically adapt to positional deviations within 1 centimeter and dynamic disturbances, and can complete scene calibration in as little as five minutes when switching between different product models.
• Scaling Plan: A Longcheer manager said the robots have logged 140 hours of continuous operation after integration into mass production, and the deployment is expected to scale to 100 units by the third quarter of 2026.