It's all there, or almost. The skills contest at Shanghai's Global Developer Pioneers Summit (GDPS) 2025 bills itself as the world's first skills competition in embodied intelligence built on the criteria of the WorldSkills Competition, according to the Global Times. The trappings are those of a serious assessment: six broad tracks, seventeen specific events, per Yicai Global. The judging panel, for its part, reportedly brought together for the first time experts in embodied AI, AI professionals and trade veterans from the WorldSkills Competition, the better to assess the consistency and reliability of the robots in the manner of master craftsmen, reports City News Service.
On paper, the ambition is clear: to measure, to compare, to separate the contenders. Except that, reading through the write-ups, two hallmarks of a genuine competition are nowhere to be found.
Not a single performance figure
The first absence, and the most glaring: no results. No scores, no rankings, no winner, no success or failure rate emerges from any of the reports. We know which events took place; we have no idea who passed them. The domestic-services event, for instance, asked robots to fold laundry and put away crockery in a household setting, while the emergency-rescue event set tasks on a punishing 10-by-30-metre course, again per Yicai Global. How many robots actually completed their task? How many seized up? On this point, the silence is total.
A skills competition that publishes no quantified skill remains, by definition, an empty shell. You can see the shape of the assessment, never its substance.
When the organiser supplies the critical kit
The second absence is subtler but decisive. According to NCSTI, the organisers supplied the teams on site with dexterous robotic arms, advanced sensors and high-precision gearboxes. Yet in a contest between manufacturers, the kit each one brings is part of what is being judged.
If the critical hardware is pooled, whatever each team contributes of its own fades into the background. The question becomes: what is actually being tested? The teams' software and AI, or a shared set of hardware that flattens precisely the differences meant to separate the contenders? With no published scoring scheme, there's no way to tell.
The event's true register: the stage
Where the event is awash with detail is in its showbiz dimension. At the opening ceremony on 13 December 2025, Unitree's G1 humanoid robot, kitted out in white Tai Chi garb, performed Tai Chi moves alongside human martial artists, reports Yicai Global. AgiBot's Yuanzheng A2, meanwhile, played the drums to lead a dance troupe of Lingxi X2 units, according to DigiTimes. The same outlet notes that AgiBot (Zhiyuan) and Unitree took to the stage together for the first time, their first joint humanoid performance.
The choreography came with landmark announcements: the start of construction on the Zhangjiang AI Innovation Town was announced at GDPS 2025, in tandem with this joint appearance, says Futu. In the background, Shanghai is aiming to grow the core of its embodied-AI industry to more than 50 billion yuan by 2027, according to Yicai Global. Which is to say a sizeable chunk of the event belongs to the marketing showcase, not to the graded trial.
The demo as proof: an industry-wide pattern
The unease sharpens once you set this arrangement against the very tools the sector likes to flaunt. Unitree's developer platform reportedly lets users control humanoid robots remotely from a smartphone, using the phone's camera to guide their movements, and is said to offer pre-loaded action routines including martial arts, a The Twist dance and ballet sequences, reports Interesting Engineering.
In other words, the impressive flourish you see on stage could perfectly well be teleoperated or pre-recorded. Nothing suggests that's the case for the GDPS demonstrations – but that is exactly what the absence of any protocol and any scores makes impossible to rule out.
Staking out the standard-setter's seat
The real point of GDPS is probably not to measure who folds laundry best. It's to stake out a position: that of the standard-setter. To get a benchmark for assessing embodied AI rubber-stamped by the prestige of WorldSkills – one whose figures, for now, nobody has actually seen. You certify a piece of choreography, and you install yourself as the player who will set tomorrow's rules of the game. The form of the competition is there for exactly that. The function is still waiting on its results.