The Interface Revolution

As our realities trend toward the algorithmic, we’re no longer just technology users — we are participants in an evolving dialogue between human intent and machine capability. We can watch from the sidelines as a new paradigm in human-computer interaction evolves or we can build it.  

At the core, this is about changing how we engage with information, each other, and the world around us. AI is reshaping how computer systems are designed to process, synthesize, and navigate information. We talk to our data. Cloud data centers generate minute-long videos from nothing more than a text prompt. At the same time, new forms and methods for interacting with applications and interfaces are emerging. Apple and Meta continue exploring the potential of spatial computing, and devices as small as a Raspberry Pi are now capable of running sophisticated AI models. For example, Llamafile (a Mozilla Builders project) runs at 14 tk/s generation speed with tinyllama-1.1B on the Pi 5.

From the first keyboards to today’s spatial interfaces, human-computer interaction is constantly evolving. These advances aren’t just about making computers more capable; they’re about creating more intuitive, natural ways for humans to express their intent and receive information. This is just the beginning of the age of AI — and this exchange will only become more fluid and sophisticated.

The Evolution of Input

Breakthroughs in computing often give birth to new interaction patterns. In the last decade, the hamburger menu emerged to tame cluttered mobile interfaces. Infinite scroll transformed how we consume social media. The simple act of swiping left or right has become a universal gesture for binary choice. These patterns aren’t just design solutions — they’re elements of a growing digital vocabulary, born from necessity and shaped by human intuition.

With new generative capabilities emerging from multimodal machine learning models, we face a new wave of hyper-personalized capabilities for our software. Development copilots write code alongside developers and AI-powered platforms can generate entire websites. What was once the domain of programmers (and soon expert programmers) is now attainable through natural language and prompting. 

Through this lens, the ambient nature of our relationship to data and information presents multidimensional opportunities for innovation across the user experience. What might the future hold in a world where AI allows us to generate software interfaces on the fly? 

M Eilo, an artist and former Principal Designer at Microsoft and Y Combinator Research, contrasts the popular chatbot interface with tools like Adobe Photoshop, which offer interfaces abstracted from the content (layering and compositing, for example) to allow users to create new media. While the output of generative AI systems today is largely one-shot — a prompt goes in, and new text, images, or video comes out — the idea of combining tools in a single interface that independently takes advantage of machine learning capabilities represents a far more powerful experience for interacting with information and models than what we have today. 

The history of computing is, in many ways, a history of abstraction. First there were punch cards, then command lines and graphical interfaces, and thousands of iterations on input and display devices — each abstraction layer has made computing more accessible while simultaneously masking its underlying complexity. The Apple I was an esoteric machine for hobbyists; it required knowledge and patience to operate. Babies use iPhones. This evolution of abstraction tracks with mass adoption — the more accessible technology becomes, the more deeply it weaves itself into the fabric of daily life.

AI presents a paradox in this progression. While natural language interfaces represent perhaps the highest level of abstraction yet — simply ask for what you want — they also create new complexities. The challenge of “prompt engineering” shows how expressing precise intent to AI systems often requires its own kind of technical literacy. Complete abstraction may not always be the answer; sometimes, users need meaningful ways to peek beneath the surface, to understand and guide the systems they’re interacting with. 

So what is the brave new world in user experience design? Is it one in which interfaces dynamically adjust their level of abstraction based on the user’s needs and expertise? Imagine AI interfaces that could shift seamlessly between natural conversation and precise technical control or gradually reveal their underlying capabilities as users become more sophisticated. The goal of the dynamic interface suggested above is not just to make technology easier to use but to make it more empowering — to give users the right level of control at exactly the right moment. 

Context is Key

As personal machines continue to develop an understanding of the environmental and personal contexts that we bring to a given interaction, more complex modes of engagement will present themselves. Imagine asking your home assistant for the weather and having it respond not just with the temperature, but with a recommendation of what to wear and pack based on knowledge of your itinerary, closet, and an innate understanding of how your context will change throughout the day.

In classic software development, the user interface is sometimes referred to as the view or the presentation layer of an application. These architectures vary in how they separate the logic of an application from how the user sees or interacts with the system, but the underlying principles of separating the data and controlling behaviors are well-practiced techniques grounded in computer science fundamentals. The metaphor carries forward within the context of agentic software, but with the added complexity of a device ecosystem interface compared to the classic graphical user interface. 

So far, the dominant paradigms for interacting with AI have mirrored the behaviors we use to communicate with each other. Conversations have become a new normal for engaging with highly capable applications across many different tasks, but chat interfaces are not the ultimate form factor for all tasks. While language is arguably the most effective form of compressing our thoughts into a transmissible format, there is a vast spectrum of feeling, sensation, emotion, perception, and cognition that cannot be directly translated into words. We must also remember that our digital lives exist as an ecosystem of devices and networks that surround us. 

Subsequently, we can begin to understand that the future of the user experience is multidimensional, taking into account our traditional views of software and the ambient sources of information that we engage with regularly. 

The anthropomorphization of our software presents both risks and opportunities. In a 2023 paper, researchers emphasized that trust in a system is gained when a user can relate to the behaviors of a human-like bot. Our virtual relationships risk clouding out the in-person variety. The increasing emotional impact seen in our relationship with our digital lives presents a deep need to consider the psychological impact of future interfaces. This is especially high-risk in AI and machine learning, where tracking, auditing, and understanding how an output is generated or controlled is still difficult.

Will the future bring more personal and more pervasive interfaces, systems that can generate custom interfaces for immediate needs while adapting to individual contexts? If so, one could imagine how these systems could learn from each interaction and continue to evolve, creating a feedback loop of increasingly sophisticated human-computer interaction. The capabilities of these future interfaces can both enhance and diminish human agency. We must ensure they do the former. 

Evolving Intent

Many opportunities exist even within the world of flat screens, but the proliferation of ambient, omnipresent devices will unlock further innovation capabilities for how users interact with computers. Spatial interfaces as a form factor are enabled by AI, utilizing computer vision and graphics to map the physical world for use in digital understanding. Designing user experiences that seamlessly draw from our spatial interactions allows us to carry our natural and intuitive behaviors into the digital realm. Just like a face-to-face conversation feels like a vast deviation from a Zoom call, volumetric and 3D forms of conversation will also shift our behaviors in intentional and imperceptible ways. 

As humans, we build on our existing models of the world and our experiences to interpret one’s communicative intent when we engage with another person. One of the challenges with LLMs and AI is that there is no fundamental communicative intent from the model itself — AI chatbots are massive statistical algorithms predicting text in a way that makes it look like human-generated text. Textual language, though, is only one facet of what we communicate.

When communicating with other humans, we draw upon shared experiences, emotional context, and unspoken understanding. Without true communicative intent from our computing systems, how do we ensure that the interfaces we design that act with simulated intent encourage positive psychological responses within our brains? 

The challenge extends beyond just making existing capabilities more accessible. Traditional interfaces were designed to navigate known features efficiently. AI interfaces, by contrast, must help users explore possibilities they might not know exist. We’re moving from a world of explicit commands to one of collaborative discovery, where the interface itself becomes a partner in problem-solving.

This shift demands a fundamental rethinking of user experience design. It is no longer enough to create intuitive paths through predetermined features. Instead, we must design systems that can adapt to emerging capabilities, understand contextual nuance, and bridge the gap between what users can articulate and what they truly need. Our interfaces must become more than just tools — they need to be interpreters of human intent, capable of understanding not just what users say, but what they mean.