London-based startup Generalist has developed an AI "brain" that allows humanoid robots to learn new physical skills in a matter of days [1].
This advancement could remove a primary bottleneck in industrial automation by drastically reducing the time required to program robots for complex warehouse and factory tasks. If robots can be retrained rapidly, companies can deploy them in dynamic environments where job requirements change frequently.
Founded by former DeepMind researcher Dr. Alex Smith, Generalist focuses on accelerating the deployment of humanoid robots to automate physical labor [1, 2]. The company demonstrated the technology publicly in March 2026, showcasing robots performing seven different manipulation tasks [2].
Traditional robotics training often takes months of precise programming or simulated data. The new system reduces this timeline to typically three to five days [1].
"Our AI brain can teach a robot a new physical skill in just a few days, cutting the training time from months to a matter of weeks," Smith said [1].
Jensen Huang, CEO of Nvidia, said the technology is the robotics equivalent of ChatGPT — a general-purpose learning engine that can be applied to any robot platform [2].
Smith said the company envisions a future where a single robot can pick up a new task on the fly, similar to how a human worker operates [3].
Generalist expects to ship its AI brain to industrial partners by the fourth quarter of 2026 [2]. However, some industry analysts predict that widespread adoption of humanoid robots will not occur until 2028 [2].
“"Our AI brain can teach a robot a new physical skill in just a few days."”
The transition from specialized, single-task robotics to general-purpose learning engines represents a shift toward flexible automation. By reducing training cycles from months to days, Generalist is attempting to make humanoid robots economically viable for small-to-medium enterprises that cannot afford lengthy downtime for custom programming.
