The Canadian startup smartARM is working on a prototype bionic prosthetic arm that relies on computer vision to automatically select the appropriate way to grasp everyday objects. The company says this approach aims to reduce the need to manually switch between grip patterns, a step typically required when moving, for example, from picking up a cup to using a spoon.
The project, presented by Meta on September 16, 2026, combines artificial intelligence hardware with open-source software. The arm uses a camera integrated into the palm to monitor the environment, while smartARM uses Meta’s open-source DINOv2 vision model to recognize objects based on a limited number of reference images.
How Does the Grip-Selection Mechanism Work?
Rather than relying on prior knowledge of every object or separate programming for each item, the system attempts to interpret the object’s visual features and then link them to the appropriate grip pattern. According to the material, the arm can adapt to objects it has been trained on using only a few images, in a process that smartARM says could reduce a task that might otherwise take weeks.
The intended result is not merely to increase the number of grip patterns, but to make operating the prosthetic limb more similar to natural interaction. The user is not expected to focus on selecting a software mode before every movement, but rather on the task they want to perform. The material describes the prototype as capable of working on the first attempt and without training, a result attributed to the project’s trial rather than to independent data on broad clinical performance.
The Role of Meta AI Glasses
Meta AI glasses serve as an optional addition to the camera in the arm. Through the Meta Wearables Device Access Toolkit, the glasses provide the system with a perspective from the user’s angle, which may add context about the object with which they are attempting to interact. The phone application also allows new objects to be added to the system, enabling users to customize the list of items according to their daily needs.
This integration shows that smartARM does not treat the prosthetic limb solely as an independent device, but as a system combining local sensing with data coming from a wearable device. However, the material does not specify the camera’s specifications or response time in figures, nor does it present clinical-test results, information about regulatory approval, or a commercial-availability date.
Why Does This Development Matter?
The project’s importance lies in addressing the everyday use of the prosthetic limb, not in adding an artificial-intelligence component in isolation from the user experience. If the recognition and grip mechanism succeeds under different conditions, it could reduce the operational burden on some users and give them a more direct way to interact with objects. However, the available material remains a presentation by Meta and smartARM of a prototype, so it alone is insufficient to assess durability, safety, or accuracy beyond the objects and environments in which it was tested.
Hamayal Choudhry, smartARM’s founder and chief executive officer, said that building a capable hand is only part of the challenge and that ease of use is equally important. Shaquem Griffin, a longtime product user and former NFL player, also described scalability and accessibility as the most important aspects of the technology for him. Open questions remain tied to how ready the model is for widespread use and how it will handle cases of misrecognition or unfamiliar objects.