We are witnessing a fundamental shift in how humans interact with the digital world. Not long ago, the morning began with a smartphone: unlocking the screen, checking the weather, planning a route. Today, these operations are increasingly happening without a screen — a smart speaker responds by voice, a car builds a route on its own, and a metro turnstile recognizes faces. This is not just a trend, but the formation of a new paradigm called the "zero interface" or the "frictionless era."
Evolution: From Punch Cards to Sensors
The history of human-machine interaction is a journey from submitting to the computer's language to teaching machines to understand us. Punch cards and the command line required precise phrasing and specialized knowledge. The turning point came in the 1970s with the Xerox PARC desktop concept (windows, icons, menus, pointer), which Apple mass-adopted in the Lisa and Macintosh. Visual imagery replaced the need to memorize commands.
The next revolution came in 2007 with the arrival of the iPhone. The touchscreen made the mouse and keyboard optional, bringing control closer to direct manipulation. But it was here that the graphical interface reached its limit. Super-apps, overloaded with features, destroyed the core principle of UI — simplicity. Navigation became more complex, and the very idea of interaction convenience began to suffer.
Triggers of the UI's Decline
Feature overload. Modern apps have turned into monsters: a text editor became a layout platform, a messenger became a payment channel, a bank became a showcase of financial products. Each new feature adds submenus and settings. This is what Don Norman called "feature creep" or "functionalism." Photoshop today resembles a cockpit, and Excel requires advanced math knowledge for complex spreadsheets. The interface began to demand separate training, contradicting its basic purpose — to be invisible.
Friction. Any action in a UI is a chain of steps: find the icon, open the app, wait for it to load, navigate to a section, confirm. For simple tasks, like turning on a light, this loses to a physical switch. Golden Krishna aptly notes: "Our answer to any problem is creating another screen." This is the "screen dependency" — the habit of solving problems through interfaces, even when it is unjustified.
AI and language models. The key shift is that machines have learned to understand intent, not just commands. Classic UI requires the user to translate a desire into a sequence of actions themselves. LLM systems accept a request in natural form and immediately produce a result. Jakob Nielsen calls AI the first new UI paradigm in 60 years.
"Smart" technologies. Motion, light, and temperature sensors turn physical space into a computing environment. Mark Weiser's concept of ambient computing from Xerox PARC is realized in smart homes, where scenarios run automatically, without explicit requests. Hanging a screen on every object is pointless if the system can work on its own.
Post-Interface Forms
The new approaches share one idea: the screen does not have to be the intermediary. Conversational layer — voice assistants and chatbots turn interaction into dialogue. Agentic systems delegate task execution to an AI assistant that finds tickets, books, and reports on its own. Ambient computing embeds into the environment and reacts to context. Spatial computing (VR/AR) brings the interface into three-dimensional space, like the Apple Vision Pro with eye tracking. Biometrics and neural interfaces — radical control through the body and brain.
The Flip Side of "Zero UI"
Abandoning familiar elements has created new problems. Discoverability. In a graphical interface, all functions are visible. With voice interaction, the user does not know the boundaries of the system's capabilities. Loss of control. When an agent performs operations, the logic of decisions is hidden. Golden Krishna warns: "When we start thinking about solutions without screens, we open a Pandora's box of questions about data privacy." Cognitive load shifts: instead of remembering where a function is, you need to formulate a request correctly. This is a different kind of effort, not necessarily easier.
Experiments: Successes and Failures
The Humane AI Pin and Rabbit R1, which tried to replace the smartphone, received scathing reviews — they fell short in speed and functionality. Tesla, by removing buttons, created new problems with driver distraction. Successful examples include Perplexity AI, which replaced a list of links with a ready-made answer, and GitHub Copilot, which generates code from a description of intent. Notably, Copilot does not eliminate the editor but complements it — this is a hybrid model that will likely become dominant.
The end of the interface era is not their disappearance, but a change in role. For complex professional tasks, the graphical UI will remain primary. But the paradigm is shifting: instead of learning the language of machines, we are teaching them to understand our intentions. The task for designers now is not creating buttons, but building trust. And here lies the main challenge: until systems learn to work reliably without failures and explain their decisions, a complete abandonment of screens is premature. The post-interface era has arrived, but it will be hybrid.