Tencent has introduced Gander, a system that combines real-time conversation with AI agent capabilities. While the user communicates with the system, a separate component performs more complex tasks in the background.
According to The Decoder's description, the conversational part is conditionally referred to as the 'cerebellum,' while the working part is a replaceable 'brain.' The user can interrupt the system's operation at any time. In tests, Gander interrupted users in 8% cases but lagged behind competitors in task execution accuracy.
The practical significance of this approach is an attempt to maintain natural dialogue while the agent is occupied with longer operations. However, available data is insufficient to compare the testing methodology, operating speed, or real-world application scenarios with competitors.
editorial commentary
Why it matters
A likely consequence is increased interest in agents that do not interrupt conversation while working. The next observable signal will be detailed test results or a demonstration of real usage scenarios. Significant uncertainty remains due to the lack of a complete comparison methodology and independent verification.