Architecture choices in wearables

When designing a wearable, we usually think of it as a matrix of inputs (microphone, camera) and outputs (speaker, display, haptic feedback, LEDs), linked together by a set of features. Intuition often pushes us toward a misleading shortcut: since the feature bridges input and output, the computing power (the compute) should logically live inside the device itself.

This is a major design mistake. There are two fundamental pitfalls to avoid when defining a wearable's architecture: confusing the symbolic location of a feature with the physical location of the processor that runs it, and mistaking the thought process that led to that feature for an architectural constraint.

Take Iyo, the start-up founded by Jason Rugolo. By offering an “audio computer” with 4 GB of RAM right on the ear, the company carries over semantic and technical vestiges of traditional computing. Putting that much computing power directly on the ear makes no architectural sense. Just as Apple moved the Vision Pro's battery off the headset for weight reasons, there is no reason to weigh down an earpiece until it becomes a bulky puck prone to overheating. A more rational approach would be to design a peripheral that is extremely well thought out for the interface, but that delegates compute and power to an external module already in our pocket: the smartphone.

Of course, this hybrid architecture raises a major technical challenge: latency. If the brain of the operation is in the pocket and the sensors are on the face or the ear, the Bluetooth or Ultra Wideband (UWB) link has to be flawless. For a voice AI to feel natural, its response time has to approach that of a human conversation, i.e. under 500 milliseconds. This is where micro-computing (Edge AI) comes into its own: the wearable doesn't need to understand the whole context, it just needs to be smart enough to clean the signal, detect a keyword, and pass a perfect data stream to the phone, which does the heavy lifting.