Not long ago, many people began their mornings by reaching for their smartphones to check the weather or bus schedules. Nowadays, a growing number of routine tasks can be accomplished without direct screen interaction: smart speakers respond to voice commands, cars autonomously navigate routes, and subway turnstiles recognize faces.
The traditional icons and menus are no longer the sole intermediaries between users and digital services. This concept is referred to as the "zero interface" or the "frictionless era." Let's explore why conventional user interfaces (UI) are giving way to new forms of interaction and what this shift means for us.
From Punch Cards to Sensors
In the early days of computing, users had to communicate with machines in their language. Initially, punch cards and later command lines demanded precise wording and specialized knowledge. Users did not engage with the system; they issued commands in a rigid format.
Punch Card. Source: Habr.While these systems did not qualify as interfaces in the modern sense, they were the first forms of human interaction with electronic computing machines.
The landscape began to change in the 1970s with the introduction of the desktop concept at Xerox PARC, featuring windows, icons, menus, and pointers (WIMP). In the following decade, Apple popularized this idea with its Lisa and Macintosh computers.
Apple Lisa Computer Interface. Source: 4PDA.This innovation relieved users from memorizing commands, allowing them to select icons instead. People began to learn the language of computers through visual analogies, eliminating the need for specialized technical training.
For nearly three decades, users adapted to interacting with technology through windows, menus, and icons. However, in 2007, Apple once again transformed interface expectations with the launch of the first iPhone, featuring touch controls.
First Generation iPhone. Source: Macworld.Touchscreens rendered mice and keyboards optional, enabling direct interaction through taps and gestures.
Yet, this was the point where graphical interfaces hit a wall. As super apps emerged, consolidating more functions into single applications, navigation became increasingly complex.
This complexity undermined the core idea of UI — ease and simplicity in interacting with devices.
What Killed the Interfaces?
Overload of Utility
Modern applications have long surpassed simple use cases. Text editors evolved into layout platforms, messaging apps transformed into payment and video call channels, and mobile banking apps became showcases for various financial products.
Each new feature introduces submenus, tabs, and settings, leading users to experience "functional horror," a concept described by Don Norman in his book "The Design of Everyday Things."
Cover of "The Design of Everyday Things." Source: Social Futures Lab.The UI becomes so complex that mastering the software requires separate training, contradicting the fundamental usability principle that the interface should be invisible when performing routine tasks.
For instance, Photoshop has evolved from a relatively straightforward image editor into a complex system resembling a pilot's cockpit, filled with numerous tools and settings that can intimidate newcomers.
Excel faced a similar challenge: creating a table more complex than a weekly schedule often requires knowledge akin to advanced mathematics.
Friction
Every action in a graphical interface consists of a series of steps: finding an icon, opening the app, waiting for it to load, navigating to the desired section, and confirming a selection.
For simple everyday scenarios, this sequence is less efficient than performing a physical action. For example, turning on a light in a room via an app takes longer and requires more attention than pressing a regular switch.
Golden Krishna, in his book "Good Interface is Invisible Interface," highlights a paradox: UIs often create "friction" where none existed before. He refers to this as "screen dependency," a habit of solving every task through computer systems, even when unnecessary.
"Our answer to any problem in life is to create another screen," Krishna emphasizes.
The Rise of AI and Language Models
A significant shift in recent years is that machines have learned to understand user intent rather than just individual commands.
Traditional UIs require users to translate their desires into a sequence of actions. Want to know the weather? Open the app, find your city, and check the forecast.
Conversational systems based on large language models (LLMs) accept requests in natural language and provide results or initiate a series of operations immediately.
Dr. Jakob Nielsen, a physicist and founder of the Nielsen Norman Group, calls AI the first new user interface paradigm in over 60 years. Now, instead of specifying commands, users express intentions and desired outcomes.
Smart Technologies
The proliferation of motion, temperature, and light sensors, along with smart devices, has transformed physical spaces into computational environments.
In this context, installing a screen on every object is pointless: a coffee machine shouldn't require menu navigation if it can turn on based on a schedule or an alarm signal.
The concept of ambient computing, articulated by Mark Weiser at Xerox PARC, suggests that technology recedes from our direct attention and activates without explicit requests. Today, this concept is realized in smart homes, where lighting, climate, and security scenarios are triggered almost automatically.
A smart home doesn't require constant interaction with interfaces: scenarios are activated automatically, and infrequent adjustments are made through voice or gesture.
Amid this shift, traditional graphical interfaces increasingly become bottlenecks. They struggle to scale with complexity, create unnecessary friction in simple tasks, and lag behind systems' ability to understand context.
The Post-Interface Era
As technology evolves and user tasks become more complex, new approaches to interface design have emerged. They share one idea: screens no longer need to be intermediaries between people and digital systems.
Conversational Layers. Voice assistants and chatbots based on language models have shifted interaction from navigation to dialogue. Users no longer search for menu items; they phrase requests in their own words: "set an alarm for seven," "order groceries for the evening," "explain why the microwave isn’t working."
The system recognizes intent and initiates a chain of actions to provide a solution. This approach reduces cognitive load, as users don’t need to remember where a function is located or what steps to take. However, the system must be able to understand natural language and consider the context of the request.
This direction is rapidly developing due to AI and chatbots, which have permeated nearly all aspects of life.
Agent Systems. The next step after dialogue is delegation. An AI-based agent not only answers questions but also executes a sequence of actions on behalf of the user. For instance, it finds flight tickets, compares prices, books them, adds events to calendars, and reports back on the outcome.
In this context, the classic interface disappears: users don’t fill out forms on airline websites but assign tasks to a digital assistant.
This is technically achieved through function-calling mechanisms and integration with external services. The model autonomously decides what action to take and with what parameters. Users formulate tasks and receive results without seeing the intermediate steps.
Ambient Computing. These systems integrate into the surrounding environment and respond to context without explicit requests. They include sensors and other smart devices.
In such scenarios, the interface is either absent or reduced to a minimalist data summary, and users do not interact with the system in the conventional sense through clicks and requests.
Spatial Computing. This includes VR and AR solutions that extend the interface beyond the screen into three-dimensional environments. Users see data overlaid onto the physical world and interact with it using gaze, gestures, or voice.
Augmented Reality. Source: Apple.A recent example is the Apple Vision Pro, featuring eye-tracking technology. Users select objects with their gaze and confirm actions with finger movements. This approach maintains visual information while eliminating traditional windows and menus in favor of objects anchored in space.
Biometrics and Neural Interfaces. The most radical direction involves control through the body: gaze, facial expressions, and brain activity.
Devices that read neural signals are still in early development stages but already demonstrate the potential for contactless command transmission to computers.
A common thread among all these new UI variations is the shift from visual elements to the semantics of action. The system aims to understand what the user wants and minimize the steps between requests and results.
Less Isn’t Always Better
As experience has shown, moving away from familiar control elements does not always lead to improved user experiences. "Zero" interfaces have introduced new challenges that currently lack obvious solutions.
The Openness Problem
In traditional graphical interfaces, all available actions are explicitly presented: buttons, tabs, and menu items — all functions are known in advance. In voice or agent-based interactions, the visual component is absent.
Users may not know what commands the system supports, what scenarios it can execute, and where its boundaries lie.
Don Norman refers to this as the "openness problem" — in an interface without prompts, individuals do not understand what is available to them and often limit themselves to a minimal set of obvious functions. In graphical UIs, this issue was addressed with intuitive icons and tooltips, which are currently lacking in zero interfaces.
Loss of Control and Trust
When users press buttons themselves, they understand what action they are triggering. In contrast, when an agent or voice assistant executes operations, the process remains hidden from them.
Why did the system make that particular choice? What parameters did the algorithm take into account? Why did an error occur? The absence of an interface prevents users from tracing the logic behind decisions.
This creates a trust issue. If something goes wrong, users cannot follow the cause-and-effect chain. Golden Krishna acknowledges that a complete abandonment of screens is justified only when the system is guaranteed to function correctly. In other scenarios, at least a minimal layer of control is necessary.
"When we start thinking about solutions without screens, we inevitably open Pandora's box of questions regarding data privacy," the author notes.
Cognitive Load Shift
The absence of buttons does not mean the elimination of effort. Previously, users had to remember where a function was located. Now, they need to articulate their requests or intentions correctly.
This is a different type of cognitive load, and it isn’t necessarily easier. A visual interface allowed users to discover new functions accidentally while browsing menus. In contrast, voice or agent-based interfaces require prior knowledge.
Additionally, natural language is ambiguous: the same phrase can be interpreted differently depending on context, potentially leading to computer errors.
Jakob Nielsen points out that intention-based interactions demand a higher level of abstract thinking from users than command input.
Thus, the zero interface shifts complexity from the visual realm to the semantic and procedural domains. The system becomes smarter, but users lose some tools that previously helped them manage machines.
While this is not an argument against advancing UI, it highlights the initial barriers of post-interface technologies.
Attempts to Eliminate Interfaces
The shift to a zero interface has already spawned several high-profile experiments — both successful and unsuccessful. They illustrate where the line lies between genuine progress and premature abandonment of screens.
Humane AI Pin and Rabbit R1
Both devices emerged in 2023–2024, aiming ambitiously to replace smartphones with voice-controlled or projected interfaces.
The Humane AI Pin was a wearable gadget without a screen, operated by voice and gestures.
Humane AI Pin Device. Source: CNET.Rabbit R1 was a pocket device with a small screen and an operating system based on a large language model, capable of performing actions on third-party services on behalf of the user.
Rabbit R1 Device. Source: Rabbit.Both products received scathing reviews: critics pointed out slow performance, limited functionality, and unmet expectations. It turned out that a zero interface cannot exist as a standalone device with its own battery and subscription if it lags behind smartphones in speed and capabilities.
Most users who tried the gadgets reported that the technology is not yet ready to replace familiar devices. However, it has the potential to be integrated as a mobile application.
Button Elimination in Tesla
The automaker was one of the first to transfer most control functions to a central touchscreen, later adding voice control.
On one hand, the interface became "cleaner" and updatable; on the other, drivers found themselves distracted by additional menus to adjust mirrors or climate settings.
Tesla Interior. Source: CarWeek.Tesla demonstrated that simplifying the visual interface does not always mean easing interactions. Moreover, some electric vehicle owners complained about bugs and freezes on the main screen.
This case highlights the friction problem — many features turned out to be redundant or replaceable with mechanical alternatives.
Perplexity AI
This search service provides a direct answer with sources instead of a list of blue links. Users do not browse pages; they receive condensed information in a dialogue format.
Perplexity AI Workspace. Source: NNG.This is an example of successfully replacing a classic UI. Here, the selection interface (a list of links) is replaced by a results interface (an answer with context).
Perplexity has implemented the agent approach in a mass product while remaining a web application rather than a separate device.
GitHub Copilot
A popular tool among developers today, it generates ready-to-use code based on a textual description of intent (comments or function names).
Programmers do not search for the correct syntax in documentation or click buttons in an IDE; they articulate a task, and the system offers a solution.
This illustrates how the zero interface is penetrating professional fields. However, Copilot does not completely eliminate the code editor interface; it complements it. This results in a hybrid model, likely to become the primary model for post-interfaces in the coming years.
***
The end of the interface era does not imply their complete disappearance. Instead, the role of UIs is changing: from being a necessary intermediary for every task, they are becoming an additional channel for complex or non-standard scenarios.
For professional tools — video editing, spreadsheet management, and design — graphical interfaces will remain primary for a long time. However, the overarching paradigm is shifting. Rather than learning the language of machines, we are training them to understand our intentions. The main task for designers now lies not in creating buttons and menus but in fostering trust in the product.