Amazing robot hands from Kyber Labs

In this talk

1. 6:38

   Why force control matters

   The founders argue that people guide their hands more by feel and applied force than by exact finger positions. Kyber’s fingers can move with external force, and the team estimates contact through motor behavior rather than packing each finger with force sensors. They distinguish this torque transparency from backdrivability and say their design uses no gears in the hand.

2. 11:18

   Start with useful, bounded tasks

   They see hardware as the first hurdle: capable hands have been costly and prone to breaking, while software can build on better hardware. Kyber is developing a stationary lab system for clinical testing, where a robot can execute a roughly 30-step protocol using reusable motion primitives. The founders say this constrained setup can deliver value without solving general-purpose autonomy first.

3. 24:22

   Build a full workflow and business

   Kyber plans to provide integrated systems, combining its hands with commercial robots, cameras, computers, and control software. The founders say automation must cover nearly all of a workflow, such as pipetting and sample preparation, or people still need to step in. They aim to build an advantage through software, operational data, and customer relationships rather than hardware alone.

4. 31:12

   Deploy, learn, and improve

   The founders say real deployments are needed to discover which capabilities matter and what fails in practice. They emphasize robustness, heat management, and cost, and describe making the hands replaceable with a small number of bolts and a power connector. Spare units can keep customer work going while Kyber studies failures and improves both hardware and software.