Sherpa Robot North's advanced tactile dexterity.

The focus was productivity, autonomy and precise manipulation.

For a long time, delicate manual tasks were considered a natural limit of robotics - folding paper, lining up thin pieces, dealing cards or assembling a fragile object required something that machines simply did not have, sensitivity - but at CES 2026 that limit began to collapse in front of the public.

The robotics and artificial intelligence company Sharpa presented its first full-body humanoid robo, the North developed in Singapore with a clear objective, to go beyond rehearsed demonstrations and prove real usefulness in the physical world, no choreographed movements or artificial tasks.

What made this possible was not just the robot's body, but its hands.

North uses the SherpaWave, an anthropomorphic, human-scale robotic hand equipped with thousands of tactile sensors distributed across each finger; these sensors function as artificial nerve endings, allowing the robots to sense pressure, texture and resistance in real time.

It is not just about holding objects, but about interpreting what it is touching while the action occurs. During the demonstration, the robot dealt cards using vision and language in an integrated way, but the most impressive moment came later, North built a complete paper windmill, a task composed of more than 30 continuous steps, thin paper, rigid plastic piles and a rod demanding different levels of strength, precision and control.

An excessive squeeze would tear the paper, an imprecise movement would compromise the entire structure, yet the robot executed each step without pauses, adjusting the strength of the fingers and the position of the hands as it felt the behavior of the materials, according to Sharpa, it was the longest continuous manipulation sequence ever publicly demonstrated by a humanoid robot with real autonomy.

That performance does not come only from advanced hardware, it is born from the integration between touch sensors, anthropomorphic form and artificial intelligence models trained to interpret touch as cognitive information. North not only executes commands, it reacts to the physical world while acting by adapting decisions in real time, this profoundly alters the role of robberies Humanoids, repetitive, tiring and manual tasks no longer depend on human sensitivity as an exclusive requirement.