Web pages are shutting down or simplifying, while apps are becoming more complex

Image source | Internet (Please contact for deletion if infringement occurs) Partially generated by AI

Have you ever noticed that on PC browsers, the homepages that once carried information and services are becoming increasingly 'empty'? Websites no longer display content in exhaustive detail but instead adopt a minimalist, 'window-like' presentation.

Meanwhile, mobile apps are rapidly expanding. WeChat has evolved from a simple chat tool into an aircraft carrier loaded with mini-programs, video channels, and payment systems. Alipay has transformed from a payment tool into a service hub covering transportation, government services, and utility bill payments.

App installation packages have ballooned from under 10MB a decade ago to hundreds of megabytes today, with feature lists that seem endless.

Interestingly, while apps are becoming 'heavier,' user operations are getting 'lighter.'

Ordering food delivery with one sentence, recharging phone credit with a swipe—operations that once required navigating multiple menus are now folded into a single dialog box.

Although PC content has decreased, the experience isn't necessarily simpler. You still need to open a browser, search, click, navigate, and even deal with intrusive AI pop-ups.

One side slimming down, the other expanding; one becoming simpler, the other more complex. Behind this seemingly contradictory phenomenon lies a clear logic: internet 'portals' are being redefined, and the rise of AI Agents is the core force driving this shift.

Browsers are evolving into 'Agents'

Let's first examine what's happening on the PC side.

If you've recently opened Chrome or Edge, you've probably noticed the changes: an AI button next to the address bar, summary functions crammed into the toolbar, and a sidebar constantly awaiting your prompts.

These features aren't without value, but they're making browsers resemble increasingly cluttered houses—like renovations becoming more luxurious while usable space shrinks.

One user described the experience: 'I just wanted to look up a cake recipe, but the browser insisted on rewriting, analyzing, and summarizing it for me first.'

This 'AI-ification' trend is evident in the 15 Chrome updates announced at Google I/O 2026.

The most noteworthy is WebMCP, an open protocol that allows websites to expose their functionalities to AI Agents through standardized interfaces.

While technical in nature, it simply means: websites are designing interfaces for AI, not just pages for humans.

360 founder Zhou Hongyi predicted that websites and apps might split into two systems: one for humans, one for intelligent agents.

This judgment is gradually becoming reality. When your AI assistant can directly call a website's backend APIs to complete tasks, carefully designed graphic layouts, promotional pop-ups, and guidance buttons become mere noise to Agents.

What websites truly need to provide Agents are clean, structured, and executable interfaces.

Microsoft has even begun to 'regret' its past strategies on Windows 11, admitting failure in its push for PWA (Progressive Web Apps) and rebuilding teams to develop native apps. The reason is simple: PWAs perform 'slower, more sluggishly, and consume more memory' for users, while flashy AI features further annoy PC users.

Thus, the 'simplification' of PC web pages isn't voluntary weight loss but a strategic adjustment forced by circumstances.

As Agents increasingly handle information filtering and task execution, human-oriented web content naturally trends toward simplification—because Agents have already 'read' the redundant information for you.

Why are apps getting 'heavier?'

If PC slimming is passive, mobile expansion results from deliberate choices by tech giants, driven by sophisticated business logic.

The expansion of super apps essentially represents a battle for 'user time.' After the end of mobile internet's growth era, daily smartphone usage has peaked.

WeChat boasts 1.4 billion monthly active users, while Alipay's domestic user base has long surpassed 1 billion, with significant overlap between them.

This means every minute of competition is zero-sum. Packing more services into a single app, enabling users to accomplish more without switching, becomes the most direct competitive strategy.

Thus, WeChat developed its mini-program ecosystem, while Alipay extended from a payment tool to offer hundreds of services including transportation, healthcare, and government services.

The heavier the app, the higher the user migration cost, and the deeper the platform's moat.

What truly makes users feel 'operations getting lighter' is the intervene of another variable: AI is transforming apps' 'feature shelves' into 'task portals.'

This transformation reached a turning point in summer 2026. WeChat and Alipay nearly simultaneously launched AI upgrades, introducing AI assistants 'Xiaowei' and 'Abao.'

The core change wasn't adding another chat window but flipping the interaction logic. Previously, to recharge phone credit on Alipay, you needed to open the app → find 'Utilities' → select 'Phone Recharge' → enter number → confirm amount → pay.

Now you simply tell 'Abao,' 'Recharge my phone with 100 yuan,' and the AI understands your intent, matches the service, auto-fills information, and only requires you to confirm payment.

WeChat operates similarly. 'Xiaowei' can call mini-programs to complete tasks like ordering food, booking hotels, or paying bills, without manual page navigation.

Behind this logic, WeChat transformed its 4 million mini-programs into 'skill packages' callable by AI. Users just express needs, while AI handles task decomposition and service orchestration.

In other words, apps become 'heavier' to house sufficient service capabilities; AI makes operations 'lighter' because Agents complete the 'searching through heavy content' for users.

The 'last mile' of Agent implementation

But for Agents to truly evolve from 'answering questions' to 'completing tasks,' AI-ification within super apps alone isn't enough.

User needs naturally span multiple apps—booking flights requires TravelSky, hailing rides needs Didi, and online meetings use Feishu. Having users issue separate commands to each app's AI assistant creates fragmented experiences.

This introduces the most noteworthy variable today: AI smartphones.

In September 2026, Nubia released the NaviX Ultra, positioned as the 'world's first mass-produced AI Agent smartphone.'

Its core breakthrough isn't hardware specifications but enabling cross-app 'automated operations'—users can issue a single voice command, and the phone automatically completes tasks across multiple apps, with end-to-end success rates exceeding 80%.

More importantly, Nubia simultaneously introduced the Screen Automation Operation Declaration Protocol (SAEP), delegating AI assistant operation boundaries to third-party developers: each app can precisely declare what content AI can read, modify, or delete within it, with undeclared apps remaining untouched by default.

This protocol's significance extends beyond technology. It attempts to answer an industry question yet to be clarified: When AI Agents operate your apps, where do app developers' rights begin and end?

If AI can freely simulate clicks and read data in any app, developers' investments in user experience and interface design become obsolete.

SAEP's emergence at least shows the industry recognizes the need for rules to balance Agent convenience with ecosystem sustainability.

Meanwhile, Huawei, Apple, Xiaomi, OPPO, vivo, and Samsung obtained filing for their on-device large models in July 2026, accelerating smartphones' transformation into 'AI Agents.'

This means one thing: Agents are no longer just chatbots in the cloud but becoming part of your phone's operating system, with scheduling rights over all your digital services.

Facing these changes, industry voices have grown bold. NVIDIA CEO Jensen Huang predicted traditional app forms might disappear, while the founder of open-source AI Agent OpenClaw claimed 80% of existing apps would vanish.

Experts at the Zhongguancun Forum also stated that Agents would gradually replace traditional SaaS and apps, reshaping the software industry.

However, I lean toward a milder judgment: Apps won't disappear, but their roles will fundamentally change—from 'destinations users actively open' to 'service nodes called by Agents in the background.'

Media analysis suggests: When people say Agents will end apps, it doesn't mean apps disappear but that feature-shelf apps are being rewritten as task portals. Users no longer need to remember which app does what or switch between them.

App value shifts from 'interfaces' to 'service capabilities.' Whoever provides better service interfaces for Agents to call will survive the Agent era.

But this doesn't mean smooth sailing. The Agent-app service model directly challenges existing business logic.

App monetization heavily relies on user attention and time spent in-app—splash ads, feed ads, and recommendation slots will lose value under Agent models.

An AI Agent ordering food delivery won't see splash ads or fall for 'limited-time offers.' It will simply compare prices, reviews, and delivery times before making the optimal choice. For platforms, this boosts efficiency but also delivers a dimensionality-reducing blow to business models.

Moreover, trust remains the biggest barrier to Agent adoption. Letting AI check the weather is easy, but entrusting it with payments, contract signings, or medical appointments raises different concerns.

As one social media user put it: 'I don't distrust AI; I distrust handing everything over to one AI.' Not to mention, AI in 2026 still makes basic errors.

A Stanford HAI report showed even the best models only achieve ~50% accuracy in tasks like reading analog clocks.

If I had to use a metaphor for what's happening, the internet of the past two decades resembled a massive shopping mall. Each app was a store, and users strolled from shop to shop while merchants competed for 'foot traffic' with promotions and displays.

The Agent-era internet resembles a personal concierge service. You don't shop anymore—you tell the concierge what you need, and they run errands, compare prices, and place orders for you.

Merchants no longer need to make their storefronts more attractive but must ensure concierges can easily access their inventory and pricing information.

This transformation presents different opportunities and risks for various players.

For super apps, the challenge lies in how brand recognition and user loyalty will be severely weakened when Agents become users' sole interaction interface.

Users might stop 'using WeChat' and instead 'use services provided through WeChat via an Agent.' This will force super apps to standardize and interface their capabilities, transforming from 'destinations' to 'infrastructure.'

For smartphone manufacturers, the opportunity arises if AI phones truly become Agent core carriers. Phone operating systems will evolve from 'app managers' to 'Agent orchestration centers.'

Apple's Siri, Huawei's Xiaoyi, and Xiaomi's Super Xiaoi are upgrading from voice assistants to system-level AI Agents. Whoever builds a mature Agent ecosystem first will gain an edge in the next upgrade cycle.

For small and medium developers, this could be the best or worst of times. The upside: Agent distribution no longer relies on app store rankings or user searches—a useful service interface can reach massive users through Agents.

The downside: When Agents 'choose' services for you, products without brand recognition or differentiated capabilities might never get seen.

Technology is shifting 'heaviness' from users to systems. Previously, complexity fell on users—you needed to learn each app, remember feature locations, and switch between interfaces.

Now, complexity is absorbed by Agents. While apps remain bloated, you don't need to scour every corner; while web pages exist, Agents read what you don't need to see.

This isn't victory for any single product but a paradigm shift in interaction. From 'people seeking functions' to 'functions finding people,' from 'opening apps' to 'stating needs,' the internet is undergoing a profound portal migration.

Where this migration ends, no one can say for certain.

The only certainty is that teams still debating 'whether apps need AI' may already lag behind those pondering 'what we can offer in the Agent era.'