Reportedly: AgiBot releases embodied Agent framework Genie Master: robots can evolve while working
Reported by IT Home · not yet confirmed by the company or a second independent outlet. We update this page when it is.
AgiBot released the embodied Agent framework Genie Master on October 27, 2026, enabling robots to self-improve during real-world tasks. The Genie G2 robot using the framework demonstrated continuous improvement in convenience store tasks without human intervention. A 24-hour live test with six Genie G2 robots is scheduled for November 2, IT Home reported.
China context
- Original name
- 智元机器人
- Outside China
- Not stated in the sources yet
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers outside China can study the framework's approach to real-world recursive self-improvement, but no API or open weights are mentioned; verify if AgiBot plans to release technical details or code.
- For investors
- Investors should monitor the November 2 live test results as a concrete signal of the framework's commercial viability and AgiBot's ability to scale autonomous robot deployment.
Translated from Chinese. Quotes and facts link to the original sources.
AgiBot officially released the embodied Agent framework Genie Master, which allows robots to acquire new skills while on the job. The Genie G2 robot equipped with this framework, during real convenience store operations, independently identified failure causes, adjusted strategies, and accumulated experience without human demonstration or debugging, achieving continuous self-improvement across multiple complex fine-grained tasks. This marks the industry's first validation of Physical RSI (embodied recursive self-improvement) in a real business scenario, moving beyond simulation environments and laboratories. AgiBot also announced that six Genie G2 robots will conduct a continuous 24-hour convenience store operation live test on November 2.
Genie Master is an embodied Agent framework that enables robots to perform recursive self-improvement (Physical RSI) in real-world settings. The Genie G2 robot demonstrated the ability to autonomously diagnose failures, adjust strategies, and accumulate experience without human intervention, achieving continuous improvement in tasks such as cooked food packaging, item retrieval, and restocking in a convenience store. This represents a shift from simulation-based learning to real-world iterative learning for embodied agents.
AgiBot's Genie Master framework allows robots to improve autonomously in real business environments, reducing the need for human demonstrations and debugging. This could lower deployment costs and accelerate adoption of service robots in retail and logistics for companies outside China, while also raising the competitive bar for other robotics firms that still rely on simulation or manual tuning.
For businesses, Genie Master offers the potential to deploy robots that require less initial programming and ongoing human oversight, reducing integration costs and time-to-value. The framework's ability to adapt to new tasks through experience could enable robots to handle a wider variety of operations in dynamic environments like convenience stores, warehouses, and production lines.
The scheduled 24-hour live test on November 2 with six Genie G2 robots will provide observable evidence of the framework's reliability and scalability in continuous operation. Key signals to watch include the robots' task success rates, error recovery times, and whether the self-improvement claims hold under extended real-world conditions.