Actions speak louder than words

Imagine you’re reading something complicated. You reach a sentence that makes little sense. You pause, furrow your brow and read it again. Suddenly, you understand. A faint smile crosses your face.

You haven’t said a word.

Yet your face has revealed something about your experience, moving from confusion to understanding, from effort to satisfaction.

Now imagine an AI assistant that recognises these moments. One that adapts when you’re struggling, learns what makes you laugh and understands what captures your interest.

This is where personal computing is heading. Artificial intelligence will increasingly learn from our behaviour, including the experiences we never put into words.

And the most natural interface for that intelligence is a pair of glasses.

Every Computing Revolution Begins With an Interface

The history of computing is a history of changing interfaces. Mainframes brought computing into large organisations. Personal computers moved it onto our desks through keyboards and mice. Smartphones made the internet continuously accessible through touchscreens. Each transition created entirely new industries.

Today, the tech industry is investing billions in increasingly intelligent AI models. But another competition is taking shape: the race to become the interface through which people experience AI.

For most people, AI still lives behind a screen. We open an application, ask a question and receive an answer. That model is giving way to intelligence that accompanies us throughout the day, understands our circumstances and helps when we need it.

I believe smart glasses will become the defining interface of this new era, because of where they sit on the body and what they can learn about our interactions with the world.

The Problem With Voice

Voice has emerged as a powerful interface for AI, which explains the growing interest in earbuds as a home for personal assistants. But consider how much time we actually spend speaking.

Research into everyday speech found that, by 2019, people in the studies spoke about 12,700 words a day on average, roughly 100 minutes of speech.

We spend most of our waking hours in silence. We observe, think, evaluate and decide. We struggle to understand an explanation. We enjoy a subtle joke. We discover a meal is unexpectedly delicious.

Most of these experiences require no speech. Yet they contain some of the most valuable information about our preferences, interests and needs. Voice-based AI hears only what we choose to say.

Earbuds also create a social obstacle. Wearing them all day tells others, “Do not disturb.” Glasses already belong in everyday social life, worn while working, eating, travelling and speaking face to face.

Why Glasses Are Different

Glasses sit alongside our eyes and ears, the principal senses through which we experience the world. Cameras capture what is in front of us. Microphones detect speech and surrounding sounds. Motion sensors track our activity and orientation.

The real breakthrough comes from inward-facing sensors.

Optical sensors built into the frame can continuously measure subtle movements around the eyes, eyebrows, cheeks and jaw. These signals reveal patterns of facial activity associated with everyday behaviours, from blinking and chewing to smiling, concentrating and reacting to unexpected events.

Consider someone watching a film. A quiet smile appears during a humorous scene. Later, their facial activity shifts at a moment of tension. Or a shopper who repeatedly examines one product, then chooses another.

These small moments repeat thousands of times a day. Their meaning emerges from context. A furrowed brow can accompany concentration, confusion, bright sunlight or physical exertion. Interpreting facial activity requires information about the environment, the sequence of events and the individual’s characteristic responses.

This is the foundation of what we call Facial Activity Intelligence: transforming continuous facial measurements into contextualised information about human behaviour.

From Responding to Understanding

Today’s AI assistants depend on explicit instructions. We ask a question, receive an answer and provide feedback if something is wrong.

Go back to that complicated document. A furrowed brow, a pause and a return to an earlier paragraph suggest the explanation has become difficult to follow. The assistant recognises the problem and offers clarification.

Now extend that principle across everyday life. While shopping, AI learns which products attract sustained attention. During meals, it learns about eating behaviour and enjoyment. During entertainment, it identifies what provokes amusement, interest or disengagement.

Over time, these observations combine with explicit feedback, previous decisions and personal history into a detailed model of the individual’s preferences, behaviour and needs.

This matters most as AI moves from answering questions to taking action.

AI Agents Need to Understand Us

The next generation of AI agents will organise travel, purchase products, manage schedules and support decisions about health and wellbeing. Their usefulness will depend on understanding the people they represent.

Ask an AI agent to plan a holiday. It can already compare flights, hotels and prices, and follow instructions on destination and budget. But our real preferences are often difficult to articulate.

We might enjoy lively cities while preferring quiet restaurants. We might love visiting museums, yet become tired after an hour. We might repeatedly choose familiar experiences despite enjoying something different when we try it.

These preferences emerge through behaviour and can’t be found in chat logs or emails. An agent that learns from our experiences develops a deeper understanding of what constitutes a good outcome.

Digital platforms already understand parts of our behaviour through clicks, purchases and browsing histories. AI glasses extend that understanding into the physical world, connecting what we encounter with how we respond.

That requires strong privacy protections. People must control how their behavioural information is collected and used. Trust will be fundamental to adoption.

A New Computing Platform Is Emerging

According to IDC, approximately 611.5 million wearable devices shipped worldwide in 2025. Smart glasses are among the fastest-growing categories, and IDC forecasts 13.6 million display-less units will ship in 2026.

As the hardware becomes lighter and more capable, intelligence will move from smartphones into devices we wear all day. Glasses combine capabilities that previously existed separately: visual context, sound, movement, communication and facial sensing.

The opportunity is a shared intelligence layer for entertainment, education, productivity, health and everyday decisions, one that grows more useful with experience.

Computing is moving from devices that wait for instructions towards personal intelligence that understands context and acts on our behalf.

The Future Is Personal

AI glasses will bring personalised intelligence into the fabric of everyday life. For the first time, AI will continuously measure aspects of our facial behaviour alongside what we are doing and experiencing. It will learn from our conversations, our choices and the many moments when we remain silent.

This is the future we are building towards with Sense: an interface that supports the individual throughout daily life by learning preferences and uses that understanding to make AI genuinely personal.

Because understanding what we say is only the beginning. It’s true that (facial) actions can speak louder than words.

Sources

IDC, Worldwide Wearable Device Market Insights, March 2026 update. https://idc.com/promo/wearablevendor

IDC, Smart Glasses Surge: The XR Market Is Rewriting Its Own Rules, 15 June 2026. https://www.idc.com/resource-center/blog/smart-glasses-surge-the-xr-market-is-rewriting-its-own-rules/

Pfeifer, V. A. and Mehl, M. R., Sliding Into Silence? We Are Speaking 300 Daily Words Fewer Every Year, Perspectives on Psychological Science 21(4), 2026. https://doi.org/10.1177/17456916261425131

Goel, S., Jara-Ettinger, J., Ong, D. C. and Gendron, M., Face and Context Integration in Emotion Inference Is Limited and Variable Across Categories and Individuals, Nature Communications 15, 2443 (2024). https://doi.org/10.1038/s41467-024-46670-5