Minglue's WRC keynote proposed a fleet level coordinator called Octo — one concrete answer to embodied AI's real world deployment problem. How the "organizational brain" framing differs from the broader "brain not body" industry debate.
At the 2026 World Robot Conference (WRC) in Beijing, which opened August 19, Minglue (2718.HK) and Hikvision Robotics shared a booth pitching "embodied AI" — robots that plan and act in messy, real-world settings rather than follow pre-programmed paths. The demos covered three scenes: restaurant cleaning, industrial logistics, and patrol. Wheeled humanoids and dual-arm units cleared tables, kept user belongings intact while removing waste — phones, keys, and files left in place — and ran a "high-rack pick to flat transport to box flow to fine picking" chain in a warehouse.
In a main-forum keynote, Minglue CEO Wu Minghui argued the next robot race will be won by the best "brain," not the best body. He split the brain into two layers: the per-robot intelligence stack (vision-language and vision-language-action models) and an "organizational brain" — a fleet-level coordinator, embodied in Minglue's Octo platform, that connects digital agents to physical machines.
That framing is one company's positioning, not yet an industry pattern. The source set has no independent analyst read, no named restaurant or warehouse customer, and no shipped-unit data. Whether "organizational brain" becomes a category label or stays a single-firm slogan is the open question.
The "organizational brain" pitch is distinct from the broader "brain not body" framing circulating in the embodied AI beat. Where that framing has become a shorthand for AI-software firms positioning themselves as the intelligence layer inside commercial robots, Minglue's proposal centers on a specific architectural layer — a fleet-level coordinator connecting digital agents to physical machines — rather than per-robot model improvements alone.