Boston Dynamics Gives Atlas New Dexterous Hands and Drops the Pinky
Boston Dynamics has once again drawn industry attention, this time by giving its all-electric Atlas a pair of newly designed robotic hands. The change is more than a hardware refresh; Boston Dynamics frames it as a deeper consideration of how embodied intelligence should evolve.
The new hands raise the degree of freedom from seven in the previous generation to thirteen. That jump significantly improves dexterity, but much of the industry conversation has focused on a different detail: the finger the company chose to remove.
Boston Dynamics says the hand system was built for what it calls "Real Work." It can use a three-finger grasp to handle heavy tools and still perform precise pinching motions. According to the company's official blog, the entire system was designed in a high-fidelity simulation environment.
That simulation-first approach matters because reinforcement learning is becoming a common way to train robots. Designing the hand in simulation can greatly shorten the sim-to-real gap, helping policies transfer from virtual training to physical machines more reliably.
The implication is clear: this is not an expensive laboratory toy. Boston Dynamics appears to be aiming at mass production and industrial deployment, where reliability, cost, and maintainability matter as much as raw dexterity.
By prioritizing practical tasks over human-like appearance, the new hand also reflects a broader debate in robotics: should humanoids copy human anatomy exactly, or should they be optimized for the jobs they will actually perform? Boston Dynamics' answer seems to lean toward function over form.
The missing finger has become a case study in design trade-offs. To determine how many fingers a robot really needs, Boston Dynamics CTO Zack Jackowski devised an unusual test. He asked members of the R&D team to tape their pinkies and ring fingers together and go about their daily life and work for an entire day.
The experiment found that even without a pinky, the vast majority of tasks could be completed with almost no drop in performance. Based on that result, the team decided to leave out the pinky entirely.
That decision saved the complexity, power consumption, and weight of three actuators. In a humanoid robot, every actuator adds mass, wiring, control burden, and potential failure points, so removing one finger can improve the overall system.
Alberto Rodriguez, Atlas's director of robot behavior, later shared and confirmed the decision. He stressed the importance of trade-offs among dexterity, strength, and durability. In other words, the goal is not maximum finger count but the best balance for real-world use.
This minimalist, practical logic stands out in a humanoid robot race that often pursues extreme biomimicry. While many companies try to replicate the human hand as closely as possible, Boston Dynamics appears willing to simplify when the data supports it.