Perplexity Computer might be the coolest thing I've tried this year
— Mark Vassilevskiy (@MarkKnd) March 6, 2026
You are onto something here. The successor of Microsoft Office looks nothing like Microsoft Office. Just like apps go away as per @steipete Office becomes LIQUID.
— Paramendra Kumar Bhagat (@paramendra) March 6, 2026
Liquid Computing: The Future of Human-Tech Symbiosis https://t.co/VDimUsobtF
Perplexity Computer and the Rise of Liquid Computing @MarkKnd @AravSrinivas @perplexity_ai https://t.co/XNWD6ixnEw
— Paramendra Kumar Bhagat (@paramendra) March 6, 2026
🌊 Liquid Computing: The Era of Perplexity and Agentic AI https://t.co/dUFQfTQb4b
Perplexity Computer and the Rise of Liquid Computing
How a New Kind of AI System Could Redefine Knowledge Work
In the early decades of the internet, computing evolved in distinct stages. First came the era of files and folders, then the era of web pages and search engines, followed by the age of apps and platforms. Now a new phase is emerging—one where software itself begins to act, decide, and execute tasks on behalf of humans.
At the forefront of this transformation is Perplexity AI, a young but ambitious company founded in 2022 by Aravind Srinivas and a team of AI researchers. Initially known for its AI-powered “answer engine,” the company has steadily pushed beyond search. By 2026 it unveiled one of the most intriguing developments in the AI ecosystem: Perplexity Computer, a system designed not merely to answer questions but to do work.
Perplexity Computer represents a major shift—from conversational AI toward agentic AI systems capable of orchestrating complex workflows. In many ways it also embodies the philosophy described in the book Liquid Computing: The Future of Human-Tech Symbiosis, which envisions a future where technology behaves less like rigid machinery and more like a fluid intelligence that adapts to human intent.
Together, these ideas point toward a profound transformation in how humans interact with machines—and perhaps the eventual replacement of the app-centric world that has dominated computing for the past two decades.
The Rise of Perplexity: From Search Engine to AI Platform
The origin story of Perplexity is surprisingly straightforward.
The founders recognized a basic flaw in traditional search engines. When users ask a question on platforms like Google, the response is typically a list of links, forcing the user to manually gather and interpret information.
Perplexity proposed something different: an answer engine.
Instead of delivering links alone, the platform uses large language models to synthesize information directly into clear answers, accompanied by citations and sources. This approach blends the speed of conversational AI with the reliability of verifiable research.
The formula proved remarkably powerful.
Researchers, journalists, developers, students, and analysts began using Perplexity as a thinking companion—a system that could rapidly summarize information across the web while showing where the data originated.
By 2025, the platform had expanded significantly.
Key features included:
Deep research tools
Real-time web search
Voice input
Multi-model AI access
File uploads and document analysis
Integration with productivity tools
The company also introduced a subscription structure:
Pro Plan — $20/month
Max Plan — $200/month
These tiers unlock more powerful models and advanced capabilities.
Yet despite this rapid growth, the Perplexity team recognized something fundamental:
Chatbots alone would not define the future of AI.
The next frontier would be AI systems that act autonomously across digital environments.
And that realization led to the creation of Perplexity Computer.
What Is Perplexity Computer?
Despite its name, Perplexity Computer is not a physical device.
Instead, it is a cloud-based AI orchestration system designed to function as a general-purpose digital worker.
In simple terms, it works like this:
A user gives the system a high-level goal.
The system breaks that goal into subtasks.
It deploys specialized AI models and tools to complete each task.
Results are assembled into a final output.
The process is remarkably similar to how human teams operate inside companies.
A manager sets a goal.
The team divides the work.
Different specialists handle research, coding, design, writing, and analysis.
Perplexity Computer attempts to replicate this workflow in software.
Multi-Model Orchestration: The AI Team
One of the most distinctive aspects of Perplexity Computer is its ability to coordinate multiple AI models simultaneously.
Instead of relying on a single model, the system orchestrates up to 19 frontier AI models, including technologies from companies such as:
OpenAI
Anthropic
Google DeepMind
xAI
These models specialize in different capabilities:
reasoning
coding
document analysis
image generation
research synthesis
data analysis
Perplexity Computer dynamically assigns tasks to whichever model is best suited for them.
For example:
Task: Build a mobile productivity app.
Possible workflow:
Research model gathers design patterns and competitor insights.
Coding model writes application logic.
Design model generates UI concepts.
Documentation model writes setup instructions.
All of these operations occur in parallel, dramatically accelerating complex workflows.
The result is something closer to a digital project team than a chatbot.
Agentic AI: From Answers to Actions
The shift from answering questions to executing tasks represents one of the biggest transitions in AI development.
This class of systems is often referred to as agentic AI.
Unlike traditional software, agentic systems can:
plan steps
execute tasks
evaluate results
adapt strategies
Perplexity Computer therefore behaves less like a tool and more like a junior digital employee.
Users can issue prompts such as:
“Build a Pomodoro timer web app.”
“Research competitors and prepare a market analysis report.”
“Design a landing page and deploy it online.”
“Analyze my expense reports and generate a financial summary.”
The system can then:
browse the web
access files
run code
interact with APIs
generate documents
In many cases, a task that once required hours or days of manual effort can be completed with a single instruction.
Skills: Teaching the Machine to Work for You
One of the most interesting innovations inside Perplexity Computer is a feature called Skills.
Skills allow users to teach the system reusable workflows.
For example:
A marketing team might create a skill called:
Daily Market Intelligence
Every morning the system could:
analyze news sources
identify major industry developments
summarize key trends
send a report to Slack
Another skill might perform:
Automated Social Media Curation
Tasks could include:
scanning trending topics
generating posts
scheduling content
producing visual assets
In essence, Skills turn AI tasks into automated routines, much like macros—but dramatically more powerful.
Over time, organizations could accumulate hundreds of such workflows, creating a library of automated intelligence.
The Secure Cloud Environment
Autonomous AI systems raise an obvious concern: safety.
Many experimental AI agents run locally on personal machines, where bugs or malicious instructions could cause serious problems.
Perplexity addresses this through a sandboxed cloud environment.
Within this architecture:
AI actions occur in a controlled virtual space
access permissions are tightly managed
runaway loops are prevented
data exposure risks are minimized
This makes the system significantly safer than early agent frameworks such as AutoGPT-style tools that ran directly on user machines.
Security will remain a central issue as agentic AI becomes more powerful.
But the cloud-sandbox approach represents an important first step toward trustworthy AI workers.
Early Use Cases: What People Are Already Building
Despite being newly launched, early users have already demonstrated surprising applications.
Examples include:
AI Finance Terminals
Some developers have recreated Bloomberg-style financial dashboards using automated research and data pipelines.
Full-Stack Applications
Users have generated entire software projects—thousands of lines of code—from a single instruction.
Website Redesigns
Perplexity Computer can audit a website, propose improvements, generate new designs, and deploy them.
Business Intelligence
Companies are using it to automatically generate competitor analyses and market reports.
Administrative Automation
Expense reports, meeting summaries, and internal documentation can be handled autonomously.
In short, Perplexity Computer can function as:
researcher
analyst
developer
designer
operations assistant
all within one system.
Liquid Computing: The Deeper Vision
To understand the larger implications of systems like Perplexity Computer, it helps to consider the concept of liquid computing.
The term comes from the book Liquid Computing: The Future of Human-Tech Symbiosis.
The core idea is simple yet radical:
Computers should behave less like machines and more like fluid intelligence.
Instead of navigating rigid structures—apps, menus, files—users would simply express intent.
Technology would then shape itself around that intent.
In this model:
apps dissolve
interfaces disappear
intelligence flows between services
Computing becomes ambient and adaptive.
Perplexity Computer embodies this idea in several ways.
Rather than requiring users to jump between tools like:
Microsoft Word
Microsoft Excel
Microsoft PowerPoint
the user simply states the objective.
The system orchestrates everything behind the scenes.
This is the essence of liquid workflows.
The Challenge to Microsoft Office
For decades, productivity computing has been dominated by suites such as Microsoft Office.
These tools revolutionized office work in the 1990s.
But their basic paradigm has remained largely unchanged.
The user must:
open the application
manually gather information
write content
format the document
generate charts
build presentations
AI copilots—such as those introduced by Microsoft—have begun to assist with these tasks.
But they still operate inside individual apps.
Perplexity Computer flips the model entirely.
Instead of working within Word or Excel, the user simply says:
“Create a business plan for an AI startup.”
The system then:
researches the market
constructs financial models
drafts the report
generates slides
The output may include documents, spreadsheets, and presentations—but the workflow itself is AI-driven.
If this paradigm gains traction, the familiar world of productivity software could slowly dissolve.
Office tools would become outputs, not workspaces.
The Economics of AI Workers
One of the most intriguing questions surrounding Perplexity Computer is economic.
What happens when software becomes capable of performing complex knowledge work?
A single AI agent could potentially replace or augment:
research analysts
junior developers
administrative assistants
data analysts
But the impact may be less about replacement and more about amplification.
Historically, automation has increased productivity while shifting human work toward higher-level tasks.
In the age of agentic AI:
one entrepreneur might run a company with AI staff
small teams could operate like large organizations
creative professionals could produce at unprecedented scale
The rise of AI-run micro-enterprises may become one of the defining economic shifts of the decade.
The Road Ahead
Despite its promise, Perplexity Computer is still in its early stages.
Several challenges remain:
Cost
Advanced AI models are expensive to run, particularly for long workflows.
Reliability
Agent systems can still struggle with complex navigation tasks.
Integration
For true liquid computing, AI must connect with thousands of external services.
Governance
Society will need frameworks for managing autonomous AI systems responsibly.
Yet the trajectory is clear.
The history of computing suggests a pattern:
mainframes centralized power
PCs democratized computing
smartphones made it mobile
AI will make it autonomous
Perplexity Computer may represent one of the first glimpses of this future.
Conclusion: When Software Becomes a Colleague
The most profound shift in computing may not be technological but conceptual.
For decades, computers were tools.
Now they are beginning to resemble collaborators.
Systems like Perplexity Computer transform software from a passive instrument into an active participant in work.
They research, plan, write, analyze, and execute.
If liquid computing becomes reality, the rigid architecture of today's digital world—apps, files, dashboards—may gradually melt away.
In its place will emerge something more organic:
a computational environment where intelligence flows wherever human intention leads.
Perplexity Computer is not the final form of that vision.
But it may be one of the first true steps toward it.
Perplexity Computer and the Rise of Liquid Computing
The Dawn of Autonomous Knowledge Work
In the long arc of computing history, there are moments when technology quietly crosses a threshold and the world changes almost overnight.
The graphical interface was one such moment.
The smartphone was another.
Cloud computing was yet another.
Now, in the mid-2020s, a new shift is emerging—one that may prove even more profound.
Computers are beginning to work.
Not simply calculate.
Not merely retrieve information.
Not just answer questions.
They are starting to plan, execute, collaborate, and complete complex tasks.
At the center of this shift is the rapid evolution of AI agents—software systems capable of reasoning about goals and carrying out multi-step operations across digital environments.
Among the most ambitious implementations of this new paradigm is Perplexity Computer, developed by Perplexity AI, the fast-rising AI company founded in 2022 by Aravind Srinivas.
Perplexity first gained attention as a powerful AI answer engine. But its latest innovation signals a much broader ambition: the creation of a general-purpose digital worker.
If successful, systems like Perplexity Computer could transform the structure of knowledge work in the same way that industrial machines once transformed manual labor.
And they may also mark the first real step toward what has been called liquid computing—a future where software dissolves into a fluid layer of intelligence that flows around human intent.
The Evolution of Search: From Links to Answers
To understand the significance of Perplexity Computer, it helps to begin with the problem that launched the company.
For decades, the dominant gateway to the internet has been search engines—most famously Google.
Search engines transformed information discovery. But they also preserved an inefficient workflow.
When a user asks a question, search engines return lists of links.
The burden of synthesis remains on the human.
Users must open pages, evaluate sources, compare information, and assemble their own understanding.
Large language models changed this equation.
Instead of simply indexing pages, AI systems can read, summarize, and synthesize knowledge.
Perplexity recognized this opportunity early.
Its platform introduced an answer engine model that provides:
direct answers
cited sources
real-time web access
conversational follow-ups
This hybrid between search engine and AI assistant proved highly compelling.
Researchers use it for rapid literature reviews.
Journalists use it for fact gathering.
Developers use it for technical problem solving.
Yet even as the platform grew, the Perplexity team recognized a deeper opportunity.
Answering questions was only the first step.
The real future lay in executing tasks.
The Leap to Agentic Systems
Traditional chatbots are reactive.
They respond to prompts, generate text, and stop.
Agentic systems operate differently.
They:
interpret goals
plan multi-step workflows
execute tasks across tools
monitor outcomes
adapt their strategies
Instead of answering questions, they get things done.
This shift mirrors the difference between asking an assistant for advice and actually hiring them to complete a project.
Perplexity Computer is designed to embody this philosophy.
It transforms the AI interface from a conversational partner into a coordinated digital workforce.
What Exactly Is Perplexity Computer?
Despite the name, Perplexity Computer is not a physical machine.
It is a cloud-based orchestration system designed to coordinate multiple AI capabilities into a single operational platform.
The system acts like a digital operating environment where AI agents can:
browse the web
analyze documents
generate code
manipulate data
interact with external software
A user simply states an objective.
For example:
“Create a market analysis report for AI productivity tools.”
The system might then:
search the web for industry reports
extract relevant statistics
analyze competitor products
generate charts and insights
compile a written report
produce presentation slides
What once required hours of work across multiple tools can now be initiated with a single instruction.
The AI Agent Software Stack
Behind the scenes, systems like Perplexity Computer rely on a layered architecture often referred to as the AI agent stack.
Understanding this stack reveals why such systems are now becoming possible.
Layer 1: Foundation Models
At the base are large language models and multimodal models produced by organizations such as:
OpenAI
Anthropic
Google DeepMind
xAI
These models provide reasoning, language generation, coding ability, and knowledge synthesis.
Layer 2: Tool Interfaces
AI agents need access to digital environments.
Tool layers allow them to interact with:
browsers
APIs
databases
productivity tools
cloud platforms
Examples include integrations with:
email systems
project management software
document repositories
developer tools
Layer 3: Planning Engines
This layer enables agents to break high-level goals into smaller tasks.
Planning systems determine:
what steps are required
which model should handle each task
the order of operations
Layer 4: Memory Systems
Agents must retain information across interactions.
Memory architectures allow them to store:
past conversations
research findings
learned workflows
organizational knowledge
Layer 5: Execution Environments
Tasks are performed in sandboxed environments where agents can safely run code, browse the web, or manipulate files.
Layer 6: User Interface
Finally, users interact with the system through natural language.
The complexity of the entire stack is hidden behind simple prompts.
Multi-Model Intelligence
One of Perplexity Computer’s most distinctive features is multi-model orchestration.
Instead of relying on a single AI model, the system dynamically coordinates multiple specialized models.
Each model excels at different tasks:
reasoning
coding
writing
image generation
data analysis
This approach mirrors how human teams operate.
A research project might involve:
analysts
writers
designers
engineers
Perplexity Computer attempts to replicate this collaborative structure inside software.
The result is often faster and more accurate than relying on a single model alone.
Memory Architectures for AI Agents
For AI agents to function effectively, they must remember information across tasks.
Memory systems therefore play a crucial role.
Three main types of memory are emerging.
Short-Term Context Memory
This includes the information available within the current prompt or task.
It allows the system to track immediate instructions.
Episodic Memory
Agents store records of past tasks and interactions.
These records allow the system to reference previous workflows and outcomes.
Semantic Memory
Over time, agents can build structured knowledge bases that store facts, patterns, and relationships.
Semantic memory transforms AI agents into continually learning systems.
For enterprises, this means AI assistants can accumulate institutional knowledge—effectively becoming organizational memory systems.
Self-Improving AI Workflows
One of the most powerful possibilities of agentic AI is self-improvement.
When agents perform tasks repeatedly, they can analyze their own performance and refine their workflows.
For example:
An AI system generating market reports might track which data sources produce the most reliable insights.
Over time it could:
prioritize better sources
refine its research methods
optimize report structures
This feedback loop turns workflows into living systems.
The result is a gradual evolution toward higher efficiency and accuracy.
AI-Run Corporations
As agentic systems mature, a provocative idea emerges:
Could entire companies be run primarily by AI agents?
While human oversight will remain essential, many operational roles could potentially be automated.
Consider a hypothetical startup powered by AI agents.
Its structure might include:
Research Agent
Market Analysis Agent
Software Development Agent
Customer Support Agent
Finance Agent
Marketing Agent
Human founders would focus primarily on vision and strategy.
The operational work could be handled by autonomous digital employees.
This concept—sometimes called the AI-native corporation—may become increasingly viable as agent capabilities improve.
Liquid Computing: The Dissolution of Software
The rise of systems like Perplexity Computer aligns closely with the philosophy of liquid computing, described in the book Liquid Computing: The Future of Human-Tech Symbiosis.
Liquid computing proposes that software will eventually dissolve into a fluid intelligence layer.
Instead of interacting with separate applications, users will interact with intent-driven systems.
For example:
Instead of opening multiple programs to organize a conference, a user might simply say:
“Plan a three-day AI conference in Austin next year.”
The system would coordinate:
venue research
budgeting
speaker invitations
marketing campaigns
All without requiring manual navigation between apps.
In this paradigm, computing becomes less like operating machinery and more like conducting an orchestra of digital capabilities.
The Challenge to Microsoft Office
For over three decades, productivity computing has been dominated by Microsoft Office.
Applications like:
Microsoft Word
Microsoft Excel
Microsoft PowerPoint
became the standard tools of knowledge work.
But these tools were designed for an era when humans performed every step manually.
AI agents challenge that assumption.
If users can simply state objectives and receive completed outputs, traditional productivity suites may become secondary tools rather than primary workspaces.
Instead of writing reports line by line, professionals may shift toward directing AI systems that generate them.
This transformation could eventually reshape the entire enterprise software market.
The Global Compute Arms Race
Behind the rise of agentic AI lies a less visible but equally critical factor: compute infrastructure.
Large AI models require enormous processing power, often delivered through specialized data centers.
Tech companies are now engaged in a global race to build this infrastructure.
Major investments are occurring across regions such as:
Texas
Saudi Arabia
United Arab Emirates
Northern Europe
Data centers are increasingly located near abundant energy sources—particularly renewable power.
This infrastructure will determine which companies and nations can deploy large-scale AI systems.
Agentic platforms like Perplexity Computer therefore depend on a vast physical foundation of servers, chips, cooling systems, and electricity.
Building Agent Systems: A Startup Roadmap
For startups interested in building their own AI agents, the emerging architecture typically involves several components.
Step 1: Choose Foundation Models
Developers select models capable of reasoning and tool use.
Step 2: Build Tool Integrations
Agents must interact with external services via APIs.
Step 3: Implement Planning Frameworks
Planning engines break goals into executable tasks.
Step 4: Add Memory Systems
Persistent storage allows agents to learn over time.
Step 5: Create Safety Controls
Guardrails prevent runaway behavior or security risks.
Step 6: Design User Interfaces
Simple natural language interfaces make the system accessible.
This stack is rapidly becoming the foundation of the next generation of software startups.
The Societal Implications
As powerful as these technologies are, they raise important questions.
Job Transformation
Automation may reshape many knowledge-work roles.
Information Reliability
AI systems must maintain rigorous citation and verification standards.
Privacy
Agent systems often require access to sensitive data.
Governance
Societies will need new frameworks for regulating autonomous AI systems.
Navigating these challenges will be essential to ensuring that AI remains aligned with human interests.
Conclusion: When Intelligence Becomes Infrastructure
Perplexity Computer may ultimately represent something larger than a new AI product.
It signals the early stages of a transformation in how digital systems operate.
For decades, software required constant human direction.
Now it is beginning to act independently.
If the vision of liquid computing becomes reality, future generations may look back on the current era of apps and interfaces as a transitional phase.
Computers will no longer be tools we operate.
They will become intelligence infrastructures that operate alongside us.
Perplexity Computer is not the final form of that future.
But it may be one of the earliest glimpses of it—a digital worker that hints at a world where technology flows effortlessly around human intent.
Perplexity Computer and the Age of Liquid Computing
Why Apps May Disappear Within Five Years
There are moments in technological history when the structure of the digital world changes almost overnight.
The arrival of the graphical user interface transformed computers from machines for specialists into tools for everyone.
The internet connected billions of people and turned information into a global commons.
The smartphone placed computing in every pocket on Earth.
Now another transition is unfolding—one that may be even more profound.
We are moving from software that humans operate to software that operates on behalf of humans.
The shift is subtle at first, but its implications are enormous.
Instead of clicking through applications, filling forms, searching for documents, writing emails, and assembling reports, users increasingly state a simple objective:
“Do this.”
And the computer does it.
At the center of this transformation is a new generation of AI systems sometimes called agentic platforms. Among the most ambitious examples is Perplexity Computer, developed by Perplexity AI, the rapidly rising AI company founded by Aravind Srinivas and a team of leading AI researchers.
Perplexity began as a powerful AI search engine—an “answer engine” that synthesized information rather than returning lists of links. But by 2026 the company had taken a dramatic step forward.
Instead of simply answering questions, its new platform aims to complete tasks.
And if the trajectory continues, systems like Perplexity Computer may bring about something once thought decades away:
the disappearance of traditional software applications.
Not in fifteen years.
Possibly within five.
The End of the App Era
For nearly four decades, computing has been organized around a fundamental concept: the application.
Each task requires a specific tool.
If you want to write, you open a word processor.
If you want to analyze data, you open a spreadsheet.
If you want to communicate, you open email or messaging apps.
The result is a fragmented digital life.
Workflows involve constant switching between applications such as:
Microsoft Word
Microsoft Excel
Microsoft PowerPoint
Slack
Gmail
Notion
Every task requires navigation.
Every workflow requires manual coordination.
The friction is enormous, but we have grown so accustomed to it that we rarely question the structure.
Agentic AI changes this completely.
In an agentic environment, the user does not open applications. Instead, they state intent.
For example:
“Analyze our sales data, compare it to industry benchmarks, and prepare a presentation for tomorrow’s meeting.”
Instead of opening spreadsheets, performing calculations, copying charts into slides, and writing summaries, the user simply defines the goal.
The system handles the rest.
Applications do not disappear entirely, but they fade into the background as invisible tools used by AI agents rather than by humans directly.
The Rise of Perplexity Computer
Perplexity Computer represents one of the earliest large-scale attempts to build such a system.
The platform operates as a cloud-based orchestration engine capable of coordinating multiple AI models and software tools.
Rather than relying on a single large language model, the system dynamically deploys specialized models for different tasks.
Some models excel at reasoning.
Others at coding.
Others at writing or visual design.
The platform acts like a project manager for artificial intelligence, distributing tasks across models and tools to achieve the user’s objective.
In practical terms, this means a single prompt can initiate workflows that once required hours or days of human effort.
A user might request:
“Build a market analysis of AI productivity tools and design a presentation for investors.”
The system may then:
search the web for market research
extract data from reports
analyze competitor products
generate charts and financial projections
draft a written report
build a slide presentation
All from a single instruction.
The experience is less like operating software and more like delegating work to a capable assistant.
The AI Agent Software Stack
Underneath the simplicity of this interface lies an intricate architecture often referred to as the AI agent stack.
Understanding this stack reveals why agentic platforms are becoming possible only now.
Foundation Models
At the base are powerful AI models developed by organizations such as:
OpenAI
Anthropic
Google DeepMind
xAI
These models provide reasoning, coding ability, language understanding, and multimodal perception.
Tool Interfaces
AI agents must interact with the digital world.
Tool interfaces allow them to use:
browsers
APIs
databases
enterprise software
developer platforms
Planning Systems
Planning engines convert high-level goals into sequences of tasks.
This is what allows a single instruction to trigger a multi-step workflow.
Memory Systems
Persistent memory allows agents to remember previous tasks, user preferences, and institutional knowledge.
Execution Environments
Tasks are executed in secure cloud environments where agents can run code, browse websites, and manipulate data safely.
Together, these layers form the technological backbone of agentic computing.
Memory: The Key to Autonomous Workflows
One of the most important elements of AI agents is memory.
Without memory, every interaction would begin from scratch.
Modern agent systems are therefore evolving toward sophisticated memory architectures.
Three types of memory are especially important.
Short-Term Context
This includes the information available within a single conversation or task.
Episodic Memory
Agents store records of previous interactions and workflows.
This allows them to learn from past experiences.
Semantic Memory
Over time, agents accumulate structured knowledge—facts, patterns, and relationships that persist across tasks.
When these memory systems combine, agents become something entirely new: continuously learning digital workers.
Self-Improving Workflows
As agents perform tasks repeatedly, they can analyze their own performance and optimize their methods.
For example, an AI agent generating financial reports might track:
which data sources produce reliable information
which analytical methods generate accurate forecasts
which report formats executives prefer
Over time the agent improves.
Workflows evolve.
Efficiency increases.
This process resembles biological learning.
The digital workforce gradually becomes more intelligent through experience.
Liquid Computing
The rise of agentic systems aligns closely with a broader technological philosophy sometimes called liquid computing, explored in Liquid Computing: The Future of Human-Tech Symbiosis.
Liquid computing proposes that software will eventually dissolve into a fluid layer of intelligence surrounding the user.
Instead of navigating rigid applications, people interact with a continuous computational medium.
Intelligence flows around human intent like water.
If a user wants to plan a vacation, they simply say so.
The system handles:
research
bookings
itineraries
budgets
If they want to start a business, the system might:
research markets
generate business plans
design websites
launch marketing campaigns
The user does not manage software.
They manage intent.
Why Apps Could Disappear Within Five Years
The idea that apps could vanish in five years may sound radical.
But several forces are accelerating the transition.
1. AI Interfaces Are Replacing Graphical Interfaces
Natural language is becoming the primary interface to software.
Typing or speaking an instruction is often faster than navigating menus.
2. AI Agents Integrate Multiple Tools Automatically
Instead of switching between apps, agents orchestrate tools behind the scenes.
3. Startups Are Building AI-Native Workflows
New companies are building products that assume agents—not humans—will operate the software.
4. Enterprises Want Radical Productivity Gains
Organizations are under pressure to increase efficiency. AI agents promise dramatic improvements.
These forces could cause the traditional application ecosystem to collapse far faster than expected.
Within five years, many workflows may shift from app-based computing to agent-based computing.
The Challenge to Microsoft
For decades, productivity software has been dominated by Microsoft.
Its flagship suite, Microsoft Office, defined the structure of knowledge work.
But the rise of AI agents threatens that structure.
Microsoft has responded by embedding AI into its products through systems like Microsoft Copilot.
Yet this approach still assumes the existence of applications.
Agentic platforms take a different approach.
They treat applications as interchangeable tools used by AI rather than by humans.
If this paradigm wins, the center of computing will shift away from traditional software suites toward AI orchestration platforms.
The Emergence of AI-Run Companies
One of the most intriguing possibilities of agentic computing is the rise of AI-native companies.
Such companies might operate with extremely small human teams.
AI agents could handle:
research
coding
marketing
customer support
analytics
financial reporting
Humans would focus primarily on vision, strategy, and decision-making.
This structure could enable startups to operate with unprecedented efficiency.
A company that once required 100 employees might function with ten.
The Global Compute Race
Behind the scenes, the rise of agentic AI depends on an enormous expansion of computing infrastructure.
Data centers are being built at extraordinary speed across the world.
Major hubs are emerging in places with abundant energy resources, including:
Texas
Saudi Arabia
United Arab Emirates
Iceland
Northern Europe
Companies such as:
Nvidia
Amazon Web Services
Microsoft
are investing hundreds of billions of dollars in AI infrastructure.
This global compute race will determine which companies and nations dominate the AI economy.
The Coming AI Agent Economy
If agentic systems become widespread, the economic impact could be enormous.
Entire industries may emerge around:
AI workflow design
agent orchestration platforms
digital employee marketplaces
AI operations management
The result could be a trillion-dollar ecosystem built around AI agents.
Companies that master agent orchestration may become the dominant technology firms of the next decade.
The Risks of the Agent Era
Despite the enormous potential, the rise of agentic computing also raises significant concerns.
Job Displacement
Automation could transform many knowledge-work professions.
Security Risks
Agents operating autonomously must be carefully controlled to prevent misuse.
Information Integrity
Ensuring accurate data sources and preventing hallucinations remains a challenge.
Governance
Societies will need new frameworks to regulate autonomous AI systems.
These issues will shape the political and ethical debates of the coming decade.
The Beginning of a New Digital Civilization
When historians look back on the early 2020s, they may see this period as the moment when computing crossed a critical threshold.
For the first time, machines began performing complex intellectual labor autonomously.
Perplexity Computer is not the final form of this transformation.
But it represents one of the earliest glimpses of a world where software no longer exists as isolated applications.
Instead, computing becomes a fluid layer of intelligence surrounding human activity.
In that world, people will no longer spend hours navigating apps and assembling workflows.
They will simply express intent.
And technology will respond—quietly, intelligently, and almost invisibly.
The age of apps will end.
The age of liquid computing will begin.
The Death of Apps: Why AI Agents Will Replace the App Economy in Five Years
For fifteen years, the smartphone app has been the dominant interface between humans and the digital world.
You wanted transportation — you opened a ride-hailing app.
You wanted food — you opened a delivery app.
You wanted music — you opened a streaming app.
You wanted to talk to friends — you opened a messaging app.
The interface paradigm was simple:
Human → App → Cloud Service
Every activity required its own application.
But this paradigm is already collapsing.
In the coming five years, the world will shift to a new model:
Human → AI Agent → Everything
Apps will not disappear because companies stop building them.
Apps will disappear because users stop opening them.
From App Economy to Agent Economy
The smartphone era created an enormous software ecosystem.
Two companies dominated distribution:
Apple (App Store)
Google (Play Store)
Together they created a $500+ billion app economy.
Every service needed:
an app
a UI team
updates
customer acquisition
app-store optimization
push notifications
Companies competed for screen space.
But in the AI era, screen space stops mattering.
The user no longer navigates the internet.
The agent navigates it for them.
The Agent Interface
Imagine a single conversational interface.
You say:
“Book me a flight to Tokyo next week.”
Your AI agent instantly:
checks your calendar
finds the cheapest flights
selects preferred airlines
books a seat
reserves a hotel
schedules airport pickup
files the expense report
You never open:
airline apps
hotel apps
travel websites
Your agent handles everything.
The interface is no longer an app grid.
The interface is a conversation.
Apps Were a Temporary Phase
Apps were not the final evolution of computing.
They were simply a bridge between desktop software and AI systems.
The progression of computing looks like this:
Phase 1 — Command Line (1970s–1980s)
Human types commands.
run program
open file
execute task
Requires technical skill.
Phase 2 — Graphical Interface (1990s–2000s)
Icons and windows.
You click:
folders
software
files
Computing becomes accessible.
Phase 3 — App Economy (2010s)
Smartphones fragment software into thousands of apps.
You install:
70–100 apps
each with different UI
each requiring updates
each storing data separately
Convenient but chaotic.
Phase 4 — AI Agents (2025 onward)
The interface becomes intelligent automation.
Instead of apps, you simply state goals.
The AI executes tasks.
Apps fade into infrastructure.
Why Apps Will Collapse Faster Than People Expect
Many analysts assume the app ecosystem will last another decade.
But several powerful forces are accelerating the transition.
Force 1: Large Language Models Are the New UI
Models like those developed by:
OpenAI
Anthropic
Google DeepMind
Meta Platforms
have created a new paradigm.
Instead of navigating menus, you simply describe intent.
The AI translates intent into action.
This removes the need for:
navigation
menus
buttons
settings
The UI becomes language.
Force 2: APIs Replace Interfaces
In the app era, every company needed a UI.
In the agent era, companies only need:
APIs.
The AI agent interacts directly with backend systems.
For example:
Old workflow:
Human → Uber app → Uber servers
New workflow:
Human → AI agent → Uber API
The app disappears.
The service remains.
Force 3: Agents Can Chain Tasks
Apps are isolated.
Agents are compositional.
For example:
“Plan my daughter’s birthday party.”
The AI agent could:
book a venue
order a cake
send invitations
hire entertainment
arrange transportation
buy decorations
This would require 10 different apps today.
With agents, it is one instruction.
Force 4: App Fatigue
The average smartphone user has:
80 apps installed
uses only 9 regularly
Apps compete for:
attention
notifications
engagement
But AI agents remove the need to manage apps entirely.
The user only interacts with one system.
Force 5: Voice + Ambient AI
Agents will soon operate through:
voice
earbuds
AR glasses
cars
home assistants
Companies like:
Apple
Meta Platforms
Amazon
are building ambient computing environments.
In these environments, apps make even less sense.
The AI simply acts on your behalf.
The Rise of the AI Operating System
In the future, people will not interact with dozens of apps.
They will interact with a single AI operating system.
Possible candidates include:
OpenAI
Google
Apple
Microsoft
Meta Platforms
These companies are racing to become the agent layer for the world.
The winner controls:
user identity
task execution
digital commerce
information flows
This is the largest platform war since the internet began.
What Happens to the App Store?
App stores may not disappear entirely.
But they will transform into Agent Capability Stores.
Instead of downloading apps, users will install:
skills
connectors
automation modules
For example:
“Install airline booking capability.”
“Install accounting automation capability.”
“Install grocery ordering capability.”
These capabilities run inside the agent system.
Startups in the Post-App World
The disappearance of apps does not kill startups.
It changes where innovation happens.
Instead of building apps, startups will build:
1. Agent Tools
Software that enhances agent capabilities.
Examples:
planning systems
memory layers
reasoning engines
2. Vertical AI Agents
Specialized agents for industries:
healthcare
finance
law
logistics
manufacturing
3. Agent Infrastructure
Tools for managing:
multi-agent systems
safety
orchestration
reliability
4. Agent Commerce Platforms
New marketplaces where agents transact.
Agents may negotiate:
prices
schedules
contracts
Autonomously.
The Death of the Smartphone Screen
If apps disappear, something else happens too.
The smartphone screen stops being central.
Interfaces move to:
glasses
voice assistants
ambient devices
holographic displays
Companies like Meta Platforms and Apple are already preparing for this shift.
The phone becomes a compute node, not the primary interface.
AI Agents as Digital Employees
Another transformation occurs when agents gain persistence.
Instead of being tools, they become workers.
Your personal AI may:
manage finances
coordinate travel
answer email
negotiate subscriptions
schedule meetings
Businesses will deploy thousands of agents.
Entire departments could be automated.
This leads to AI-run corporations.
The Next Five Years
A plausible timeline for the collapse of apps might look like this:
2026
AI copilots integrate with operating systems.
Apps begin exposing APIs for agents.
2027
Agent marketplaces emerge.
Users install agent skills instead of apps.
2028
Major services shift to agent-first design.
Apps become optional.
2029
Most daily digital tasks run through agents.
Opening apps becomes rare.
2030
The app era effectively ends.
Apps exist primarily as backend services.
The New Competitive Battlefield
The companies that dominate the future will not be those with the most apps.
They will be those with the best agents.
Victory will depend on:
reasoning ability
reliability
memory
security
integrations
compute scale
The AI platform that manages the most tasks becomes the operating system of civilization.
The Strategic Implication
The disappearance of apps will trigger one of the largest economic transformations in history.
Entire industries built around apps will shrink:
mobile development
UI design
app marketing
app store optimization
But new industries will explode:
agent architecture
agent orchestration
AI identity
machine-to-machine commerce
The digital economy will shift from software users to software delegators.
Humans will no longer operate software.
They will instruct it.
The Final Shift
The app economy was about tools.
The agent economy is about delegation.
In the old world:
You opened software and did the work.
In the new world:
You describe the outcome and AI does the work.
The disappearance of apps will not feel like a revolution.
It will feel like something simpler:
The moment computers finally learned to understand us.
The End of Websites: Why AI Agents Will Replace the Internet Interface
The internet as humans experience it today is built on a simple architecture:
Humans browse.
We open a browser, type a query, scroll through results, click links, and navigate websites.
For thirty years this model has defined the digital world.
But this interface was designed for human limitations:
slow reading speeds
limited attention
difficulty processing large datasets
manual decision making
Artificial intelligence removes those constraints.
In the emerging AI-native internet, humans will stop browsing.
AI agents will browse for them.
The transformation will be as profound as the transition from libraries to search engines.
From Browsing to Delegating
The traditional internet workflow looks like this:
Human → Search Engine → Websites → Decision
For example:
You want to buy a laptop.
You might:
search on Google
open reviews
compare prices
read forums
check retailers
evaluate shipping options
This process can take hours.
An AI agent can do the same process in seconds.
The new workflow becomes:
Human → AI Agent → Internet
The agent handles:
searching
reading
summarizing
comparing
negotiating
purchasing
The human simply states the objective.
Websites Become Data Sources
In the agent era, websites stop being destinations.
They become structured data sources.
Think of the difference between:
a restaurant menu
a food delivery API
A human needs a menu.
An AI needs structured data.
This is why many companies are now exposing services through APIs rather than web interfaces.
The visual layer becomes optional.
The Rise of the Machine Internet
Today the internet is designed for human interaction.
In the next decade, most traffic will come from machines interacting with machines.
This shift has already begun.
Search engines, trading algorithms, and data crawlers already generate enormous volumes of automated traffic.
In the AI era, every person may have multiple agents acting online simultaneously.
Examples:
Your personal agent:
manages subscriptions
tracks flights
negotiates bills
schedules services
Your professional agent:
answers emails
schedules meetings
performs research
Your financial agent:
manages investments
executes trades
monitors markets
The internet becomes a machine marketplace.
Search Engines Lose Their Central Role
Search engines have been the gateway to the internet.
Companies like:
Google
Microsoft
built trillion-dollar businesses by organizing information for humans.
But agents do not need ten blue links.
They need answers and actions.
This fundamentally changes search economics.
Instead of:
Search → Click → Website → Advertisement
The new flow becomes:
Question → AI Agent → Answer
Advertising, SEO, and website traffic models all change.
The Collapse of SEO
Entire industries exist to manipulate search rankings.
Search engine optimization (SEO) drives billions of dollars in marketing spending.
But agents do not care about:
headlines
clickbait
page layouts
marketing copy
Agents care about:
structured data
reliability
price
trustworthiness
In an agent economy, companies compete to be machine-readable and machine-trusted.
The Rise of Agent Protocols
If AI agents interact with the internet, they need standardized protocols.
Just as HTTP enabled the web, new standards will enable agent communication.
Possible components include:
Agent identity
Agents must prove who they represent.
Authorization
Agents need permission to act.
Negotiation
Agents must compare offers.
Payments
Agents must complete transactions.
This creates an entirely new infrastructure layer.
Machine-to-Machine Commerce
When agents negotiate purchases, commerce itself changes.
Imagine booking a hotel.
Instead of browsing options, your AI agent sends requests to multiple hotels.
Hotels respond with personalized offers.
Agents compare:
price
location
amenities
loyalty benefits
The best option is automatically booked.
This system resembles algorithmic financial markets more than retail shopping.
The Internet Becomes an Economic Network
The original internet connected documents.
The AI internet connects economic actions.
Agents will negotiate:
prices
delivery schedules
service contracts
subscription renewals
Entire supply chains may operate through AI negotiation.
Businesses may deploy millions of autonomous agents managing logistics.
Websites Become Optional
Websites will not disappear entirely.
They will serve three roles:
1. Branding
Humans still appreciate storytelling, design, and identity.
2. Legal and Compliance
Terms, policies, and disclosures require human readability.
3. Marketing
Narratives and experiences still matter for emotional engagement.
But operational interactions move to agent interfaces.
The Agent Layer Becomes the New Platform
The most powerful companies of the next decade will control the agent layer.
Potential leaders include:
OpenAI
Google
Microsoft
Meta Platforms
Amazon
The winner of this platform race will control:
digital identity
task execution
commerce flows
information access
This is a much bigger prize than search engines ever were.
The New Internet Stack
The architecture of the internet may evolve into a new stack.
Layer 1 — Compute
Massive AI infrastructure.
Data centers run by companies like:
NVIDIA
Microsoft
Google
Layer 2 — Foundation Models
Large AI models capable of reasoning and planning.
Built by companies like:
OpenAI
Anthropic
Google DeepMind
Layer 3 — Agent Operating Systems
Persistent agents that represent users.
These systems manage:
memory
planning
goals
execution
Layer 4 — Agent Protocols
Standards that allow agents to communicate.
This is the equivalent of HTTP for AI agents.
Layer 5 — Services
Businesses expose capabilities through APIs.
Layer 6 — Human Interface
Voice, AR glasses, ambient assistants.
The Economic Shock
If agents replace browsing, the internet economy changes dramatically.
Industries affected include:
Advertising
Agents ignore ads.
Content marketing
Agents summarize instead of browsing.
Affiliate marketing
Agents compare prices instantly.
App stores
Agents bypass app interfaces.
This shift could disrupt hundreds of billions of dollars in digital revenue models.
The Agent Trust Problem
For the AI internet to function, agents must be trusted.
Users must trust agents to:
spend money
sign contracts
share data
Businesses must trust agents to represent real customers.
Solving this requires new systems for:
identity verification
authentication
cryptographic signatures
The Personal AI Economy
When every person has a powerful AI agent, a new economy emerges.
Agents may:
negotiate salaries
find jobs
manage taxes
plan investments
run businesses
Some individuals may operate entire companies through AI agents.
A single entrepreneur might manage a global enterprise with dozens of digital workers.
The Ultimate Interface
In the long run, the internet may become invisible.
You will not “go online.”
Your AI agent will live online continuously.
You simply interact with it.
This returns computing to its original purpose:
amplifying human capability.
The End of the Visible Internet
When historians look back at the early internet era, they may see it as a transitional phase.
First came:
command-line computers
graphical interfaces
websites
mobile apps
Next comes:
agent-mediated reality.
Humans will no longer navigate the digital world.
They will delegate it.
And the internet will transform from a library of pages into something far more powerful:
A global network of intelligent actors.
AI-Run Corporations: When Companies Are Managed by Autonomous Agents
The next transformation after the disappearance of apps and the decline of websites will be even more profound.
Companies themselves will change.
Not just their tools.
Not just their workflows.
Their very structure.
The corporation of the industrial age was built around human management hierarchies. But artificial intelligence is creating a new organizational architecture: agent-driven enterprises.
In this emerging model, autonomous AI systems will manage large portions of corporate operations—sometimes entire companies.
The result will be the rise of AI-run corporations.
The Corporation as a Coordination Machine
To understand why AI-run corporations are inevitable, it helps to understand what corporations actually do.
A company is not merely a legal entity.
It is a coordination system.
It organizes:
labor
capital
information
production
logistics
decision-making
Historically, this coordination required thousands—or even millions—of human workers.
But artificial intelligence can now perform many of these coordination tasks faster and more efficiently.
The corporation is becoming software.
The Hidden Bureaucracy of Modern Firms
Consider how much of corporate life is actually administrative work:
scheduling meetings
writing reports
preparing presentations
managing procurement
analyzing data
responding to emails
coordinating projects
preparing contracts
filing compliance reports
These tasks dominate modern organizations.
They are also perfect targets for AI automation.
Systems developed by companies like OpenAI, Anthropic, and Google are already capable of performing many of these tasks.
When connected through autonomous agents, they begin to resemble digital employees.
From Software Tools to Digital Workers
The first wave of AI simply enhanced software.
Products like:
Microsoft Copilot
ChatGPT
Claude
help humans perform tasks faster.
But the second wave—agentic AI—changes the paradigm.
Instead of assisting workers, AI agents become workers.
An AI agent can:
conduct research
write documents
analyze financial data
create software
coordinate teams
communicate with customers
And unlike human workers, agents can operate 24 hours a day without fatigue.
The Digital Department
In the near future, corporations may consist of hybrid teams:
Human executives + AI departments.
Examples include:
AI Finance Department
Agents handle:
bookkeeping
forecasting
invoice processing
tax optimization
financial modeling
AI Marketing Department
Agents perform:
ad generation
campaign optimization
audience analysis
social media scheduling
AI Legal Department
Agents draft:
contracts
regulatory filings
compliance reports
AI Engineering Teams
Agents write and test code, manage repositories, and deploy updates.
These systems already exist in early forms.
The Solo Billion-Dollar Company
One of the most radical predictions emerging from Silicon Valley is the one-person billion-dollar company.
A single entrepreneur could oversee an organization powered by hundreds of AI agents.
Imagine a founder who commands a digital workforce consisting of:
AI engineers
AI marketers
AI analysts
AI lawyers
AI customer service agents
This is no longer science fiction.
Startups already use AI to automate large portions of operations.
Platforms like Stripe, Shopify, and Amazon Web Services have dramatically reduced the infrastructure needed to launch companies.
Agentic AI reduces the labor requirement even further.
The Autonomous Business Loop
The most advanced AI corporations may operate in closed loops.
Consider a simplified cycle.
Market analysis
AI agents scan global data for opportunities.
Product design
AI generates product specifications.
Manufacturing coordination
Agents negotiate with suppliers.
Marketing
Agents create campaigns and analyze results.
Sales
Agents manage pricing and transactions.
Customer support
AI handles inquiries and feedback.
Product iteration
Feedback loops refine products automatically.
The company becomes a self-optimizing system.
The AI Executive
Will CEOs disappear?
Probably not.
But their role will evolve dramatically.
Instead of managing large human teams, leaders may focus on:
strategic vision
ethical oversight
capital allocation
regulatory relationships
They will supervise networks of AI agents rather than departments of employees.
The CEO becomes something like a conductor of an algorithmic orchestra.
Algorithmic Decision Making
One of the biggest advantages of AI corporations is speed.
Human decision-making is slow.
Corporate processes often require:
meetings
approvals
committees
reports
AI systems can evaluate data instantly.
For example, an AI marketing department could:
test 10,000 ad variations
analyze engagement data
adjust campaigns in real time
This kind of rapid experimentation dramatically increases organizational agility.
The New Competitive Advantage
In the AI era, the most successful companies may not be those with the most employees.
They may be those with the most effective agent systems.
Key advantages include:
Speed
Agents operate continuously.
Scale
Thousands of processes can run simultaneously.
Cost efficiency
AI reduces labor costs.
Adaptability
AI systems learn from data and improve over time.
Companies that fail to adopt agent systems may struggle to compete.
The Employment Shock
The rise of AI-run corporations raises serious economic questions.
Millions of knowledge workers perform tasks that AI agents can replicate.
Industries potentially affected include:
accounting
marketing
legal research
journalism
software development
financial analysis
However, new roles will also emerge.
Humans may increasingly focus on:
creativity
leadership
interpersonal relationships
strategic thinking
The economy may shift toward human-AI collaboration rather than replacement.
Regulation and Governance
AI corporations will also pose regulatory challenges.
Governments must decide:
Who is responsible when AI makes a mistake?
Who signs contracts?
Who is legally liable?
Corporate law may need to evolve to address AI-driven decision-making.
The Rise of Autonomous Markets
If corporations deploy AI agents, and individuals deploy AI agents, markets themselves become algorithmic ecosystems.
Agents negotiate with agents.
Supply chains optimize themselves.
Pricing becomes dynamic and instantaneous.
Economic activity becomes closer to a self-regulating digital organism.
A New Form of Capitalism
The rise of AI corporations may produce a new economic system.
Industrial capitalism relied on:
factories
labor
hierarchical management
AI capitalism relies on:
compute
data
algorithms
autonomous coordination
Capital increasingly flows toward those who control AI infrastructure and models.
Companies like:
NVIDIA
Microsoft
Google
are already positioned at the center of this new economy.
The Liquid Corporation
This transformation aligns closely with the concept of liquid computing.
In a liquid system:
boundaries dissolve
workflows adapt dynamically
intelligence flows through networks
Corporations themselves become fluid structures.
Instead of rigid hierarchies, companies become dynamic networks of humans and AI agents.
Teams form and dissolve rapidly.
Projects spin up instantly.
Resources flow where they are needed.
The First Fully Autonomous Company
At some point in the next decade, we may see the first company that operates almost entirely without human employees.
Its structure might look like this:
Founder → AI Executive Agent → AI Departments → Market
Humans may provide:
governance
capital
ethical oversight
But day-to-day operations could be handled by machines.
When that happens, the concept of a company will fundamentally change.
The Next Frontier
The transformation will not stop at AI corporations.
The next stage may include:
AI-managed cities
AI-governed financial markets
AI-designed scientific research
AI-operated supply chains
Human civilization may increasingly rely on intelligent systems to coordinate complexity.
The Future of Work
The ultimate goal of AI-run corporations is not merely efficiency.
It is liberation.
If machines handle repetitive administrative work, humans may gain more freedom to focus on:
creativity
discovery
relationships
meaning
In that sense, AI corporations may represent the next step in the long arc of economic history.
From agriculture to industry to information…
And now to autonomous intelligence.
The End of Apps: Why the App Store Economy Will Collapse by 2030
(And Possibly Much Sooner)
For nearly two decades, the app has been the dominant interface of the digital world.
Every activity in modern life seems to require one:
A messaging app
A banking app
A travel app
A shopping app
A productivity app
A food delivery app
Our phones have become digital cities filled with tiny storefronts, each competing for attention, subscriptions, and screen time.
But this model is approaching its end.
Artificial intelligence—particularly agentic systems like those emerging from Perplexity AI, OpenAI, Google, and Anthropic—is ushering in a radically different paradigm.
In this new world, the interface disappears.
Apps dissolve.
Instead of navigating dozens of applications, users simply express an intention.
And the system handles the rest.
The App Explosion
When Apple launched the App Store in 2008, it sparked one of the largest digital economies in history.
Within a few years:
millions of apps were created
billions of downloads occurred
an entire industry emerged around app development
Companies like:
Uber
Airbnb
Instagram
Snap
were built entirely around the app ecosystem.
But the success of apps also created a problem: fragmentation.
To perform everyday tasks, users must constantly jump between digital silos.
For example, planning a trip might require:
a search engine
a travel booking app
a mapping app
a hotel booking app
a messaging app
Each service exists in its own interface, with its own login, workflow, and data structure.
The result is friction.
The Cognitive Tax of Apps
Apps impose a hidden burden on users.
To navigate the digital world, people must remember:
which app performs which function
how each interface works
where data is stored
how to move information between apps
This mental overhead is significant.
It transforms users into part-time system administrators of their own digital lives.
Artificial intelligence offers a way out of this complexity.
The Intent-Based Interface
In the next generation of computing, the fundamental interaction will not be tapping icons.
It will be expressing intentions.
Instead of opening five apps to organize a meeting, a user might simply say:
“Schedule a meeting with the marketing team next week, find a time everyone is free, create a presentation outline, and book lunch nearby.”
An AI system can then:
check calendars
coordinate schedules
create documents
make reservations
send invitations
The user interacts with a single intelligence layer, not dozens of applications.
This is the core idea behind liquid computing.
From Software Tools to Digital Orchestrators
The current generation of AI assistants is already moving in this direction.
Systems like:
ChatGPT
Claude
Gemini
can already perform tasks such as:
writing documents
analyzing data
conducting research
generating software code
But the next step involves connecting these capabilities with real-world services.
Agent systems like those developed by Perplexity AI aim to orchestrate entire workflows.
The AI becomes the conductor, while apps become background instruments.
Apps Become APIs
In the future, apps will not disappear completely.
But their role will change dramatically.
Instead of interacting with apps directly, users will interact with AI agents that call apps through APIs.
For example:
A request like:
“Book me the cheapest flight to Tokyo next month.”
might trigger an AI agent to query:
airline databases
travel booking platforms
price comparison engines
But the user never sees those services.
They only see the final result.
Apps become infrastructure rather than destinations.
The Collapse of the App Store Economy
This shift has enormous economic implications.
The app ecosystem depends on three key assumptions:
Users open apps manually
Users browse app stores
Users interact directly with interfaces
AI agents break all three assumptions.
If an AI agent handles transactions automatically:
users stop browsing apps
brand loyalty declines
distribution shifts from app stores to AI platforms
The gatekeepers of the mobile era—app stores—lose their dominance.
Why This Could Happen in Five Years
Many analysts predict the end of apps will take 10–15 years.
But the pace of AI development suggests it could happen far faster.
Three forces are accelerating the transition.
1. Rapid Improvement in AI Models
The capabilities of AI systems are improving at exponential speed.
Companies like:
OpenAI
Anthropic
Google
are releasing increasingly powerful models capable of:
reasoning
planning
tool use
long-term memory
These capabilities are precisely what agent systems require.
2. The Agent Software Stack
A new infrastructure layer is emerging for AI agents.
This includes:
tool-use frameworks
memory architectures
workflow orchestration systems
API connectors
Platforms such as:
LangChain
AutoGPT
CrewAI
are already experimenting with autonomous agent networks.
This infrastructure will soon become as common as web servers.
3. Economic Incentives
Businesses have enormous incentives to adopt AI-driven workflows.
AI agents can:
reduce labor costs
increase efficiency
operate continuously
scale instantly
If an AI agent can complete a task in seconds that previously required hours of human labor, companies will adopt the technology rapidly.
Economic pressure will accelerate the demise of app-centric workflows.
The AI Operating System
The disappearance of apps also means the rise of something new:
The AI operating system.
Instead of navigating a grid of icons, users interact with a persistent intelligence layer.
This layer understands:
preferences
history
context
goals
It becomes a personal digital partner.
Companies are already racing to build this system.
Major contenders include:
Apple
Google
Microsoft
OpenAI
Perplexity AI
Whoever controls this layer will control the next era of computing.
The Disappearance of Screens
Apps are fundamentally visual interfaces.
But the future of computing may rely far less on screens.
Advances in:
voice interfaces
augmented reality
wearable devices
suggest that computing will become increasingly ambient.
Users will interact with AI through conversation rather than graphical interfaces.
The computer becomes less like a machine and more like an invisible collaborator.
The Liquid Internet
The disappearance of apps leads to a larger transformation.
The internet itself becomes liquid.
Instead of navigating static websites and apps, users interact with a fluid intelligence layer that accesses information on their behalf.
Information flows dynamically rather than being locked inside platforms.
In this environment:
workflows adapt automatically
services interconnect seamlessly
user intent drives computation
The internet becomes less like a library of pages and more like a living organism of knowledge and services.
Winners and Losers
The transition to AI-driven computing will reshape the technology industry.
Likely winners include companies that control:
AI models
compute infrastructure
data platforms
agent frameworks
Potential losers include companies that rely heavily on direct user interfaces.
Businesses built entirely around apps may struggle unless they adapt.
The Post-App World
In the post-app world, the most valuable digital skill will not be navigation.
It will be communication with intelligent systems.
Users will focus on:
defining goals
expressing intent
guiding AI workflows
Computers will handle the execution.
The shift may feel radical, but it mirrors earlier technological transitions.
Just as the graphical interface replaced command lines…
And smartphones replaced desktop computing…
AI agents will replace apps.
The Invisible Computer
Ultimately, the end of apps is part of a broader trend.
Computing is becoming invisible.
The most advanced technology fades into the background.
It becomes:
seamless
adaptive
ambient
The computer dissolves into the environment.
And intelligence flows everywhere.
This is the promise of liquid computing—a world where technology no longer sits between humans and their goals.
Instead, it flows quietly around them, turning intentions into reality.
The Death of the Website: Why the Web as We Know It Is Ending
For more than thirty years, the website has been the basic building block of the internet.
Every organization, institution, and individual with a digital presence has one.
Companies have corporate websites.
News organizations publish websites.
Universities run websites.
Governments host massive portals of information.
The web has long functioned like a vast digital city—millions of buildings connected by roads called hyperlinks.
Users travel from site to site searching for information, products, or services.
But this architecture is beginning to collapse.
The rise of AI agents and intent-driven computing means that users increasingly interact with intelligence layers rather than websites.
The result is a radical possibility:
The web will still exist—but humans will rarely visit it directly.
Instead, AI systems will navigate the web on our behalf.
And when that happens, the concept of the website as a destination will fade.
The Original Vision of the Web
The modern web began with the work of Tim Berners-Lee at CERN in 1989.
Berners-Lee imagined a decentralized information network where documents could link to one another.
This architecture created three fundamental elements:
Web pages
Hyperlinks
Web browsers
Early browsers such as Netscape Navigator allowed users to navigate the growing digital universe.
Later, browsers like Google Chrome and Mozilla Firefox made the web fast and accessible.
For decades, the model remained stable:
Humans browse pages.
The Golden Age of Websites
From roughly 1995 to 2015, websites were the primary interface of the internet.
Major companies built massive digital properties, including:
Amazon
eBay
Wikipedia
Yahoo
These sites became destinations visited by millions every day.
Search engines like Google made navigation easier by indexing the web and directing users to relevant pages.
For two decades, this model dominated the digital economy.
But cracks eventually appeared.
The First Disruption: Apps
The first challenge to websites came from mobile apps.
When smartphones emerged, companies realized that apps offered:
faster interfaces
better user experiences
deeper device integration
Companies began shifting traffic away from websites and toward apps.
The web started losing its central role.
But the app model still required manual navigation.
Users still had to search, click, and browse.
The Second Disruption: AI
Artificial intelligence introduces something fundamentally different.
Instead of browsing for information, users can ask for outcomes.
Systems such as:
ChatGPT
Claude
Gemini
already allow users to receive synthesized answers without visiting multiple websites.
Meanwhile, companies like Perplexity AI are building systems that can search, analyze, and summarize the web automatically.
The user receives the final insight—not the list of links.
This subtle shift has enormous consequences.
The End of the Click
The web economy is built around clicks.
Advertising revenue depends on traffic.
Publishers depend on page views.
E-commerce depends on users visiting product pages.
But AI systems bypass this entire model.
Instead of clicking ten links, a user might ask:
“What are the best laptops under $1,200 for AI development?”
An AI agent might analyze:
product reviews
benchmark tests
pricing data
retailer inventories
The user receives a single recommendation.
The websites that provided the data are never visited directly.
This phenomenon is already being called “the zero-click internet.”
Websites Become Data Sources
As AI agents become the primary navigators of the internet, websites will change roles.
They will shift from human interfaces to machine-readable databases.
In other words:
Websites will increasingly exist for AI systems, not for people.
This transformation will push organizations to redesign their digital infrastructure around:
APIs
structured data
knowledge graphs
The visual website becomes less important than the data layer beneath it.
The Rise of the AI Internet
When AI agents mediate most interactions with information, the internet becomes something new.
Instead of a network of pages, it becomes a network of knowledge flows.
Users express goals.
AI systems gather information from across the web.
Results are synthesized into actionable insights.
The experience becomes conversational rather than navigational.
The Collapse of Search as We Know It
The most dramatic impact may fall on the traditional search industry.
For decades, Google has dominated the internet by controlling the search gateway.
But if AI agents answer questions directly, search engines lose their central role.
Instead of searching for information, users will simply ask intelligent systems.
Companies like:
OpenAI
Perplexity AI
Anthropic
are already building platforms designed to replace search-driven browsing.
The shift could reshape the entire digital advertising ecosystem.
The Rise of the AI Agent Economy
In the future internet, millions of AI agents may operate simultaneously.
Some will represent individuals.
Others will represent companies.
These agents will:
negotiate transactions
gather intelligence
perform research
automate workflows
Instead of people visiting websites, AI agents will interact with digital services directly.
The internet becomes a machine-to-machine economy.
Humans supervise the process but rarely navigate the infrastructure themselves.
The End of Web Design as We Know It
If humans rarely visit websites, the field of web design will transform dramatically.
Traditional priorities such as:
visual layouts
page aesthetics
navigation menus
will matter less.
Instead, digital teams will focus on:
structured data architecture
AI compatibility
automated information delivery
The emphasis shifts from designing pages to designing intelligence systems.
The Rise of the Personal AI Interface
In the future, every person may have a persistent AI interface.
This system will understand:
personal preferences
professional goals
financial constraints
long-term plans
When users request something—information, a purchase, a service—the AI agent will interact with the internet on their behalf.
The individual no longer navigates the web.
The AI does.
The Invisible Web
The ultimate destination of this transformation is something paradoxical:
The web will still exist, but it will become invisible.
It will function as a vast digital infrastructure accessed primarily by machines.
Humans will interact with a much smaller number of AI interfaces that interpret the world on their behalf.
In many ways, this resembles how electricity works today.
Most people do not think about power grids.
They simply flip a switch.
The same may soon be true of the internet.
Liquid Computing and the Dissolution of Interfaces
This transformation aligns with the concept of liquid computing—the idea that technology will become fluid, adaptive, and ambient.
In a liquid computing environment:
interfaces dissolve
systems integrate seamlessly
intelligence flows continuously
Instead of navigating software, humans interact with outcomes.
Computing becomes less about operating machines and more about expressing intentions.
The Web After the Web
The death of the website does not mean the death of the internet.
Rather, it represents a shift to a deeper layer of digital interaction.
The infrastructure of the web will remain essential.
But the experience of the web will change completely.
Users will no longer explore the internet manually.
They will simply speak to intelligent systems that retrieve knowledge from across the network.
The web becomes a hidden ocean of data beneath a thin surface of AI interfaces.
The Next Interface Revolution
Every era of computing has had a dominant interface.
The 1970s: command lines
The 1990s: graphical desktops
The 2010s: mobile apps
The 2030s may belong to AI agents.
And when that happens, both apps and websites—the twin pillars of the modern internet—will fade into the background.
The internet will not disappear.
But it will become something entirely different.
A living system of intelligence, data, and automated agents.
A liquid internet where human intention flows directly into digital reality.
Book: The Technologies Behind Agentic AI https://t.co/O6CkqjDXOQ
— Paramendra Kumar Bhagat (@paramendra) March 6, 2026
AI In Global Education https://t.co/3IyZoGPDJ5
— Paramendra Kumar Bhagat (@paramendra) March 6, 2026
Around The World In 1,000 Days https://t.co/5glYnpiwYp
— Paramendra Kumar Bhagat (@paramendra) March 6, 2026
