The AI era is polarizing: the rich get richer, and the poor get poorer.

By jiayi

AI has already changed our daily habits—this is an undeniable fact.

We use AI to write emails, create PowerPoint presentations, search for information, and even draft WeChat Moments posts. We’ve grown so accustomed to AI’s presence that it feels as natural as WiFi.

Yet few of us pause to ask a critical question: Is the AI you’re using truly the same thing as the AI someone else is using?

“Fairness” in the AI Era Is the Greatest Illusion

Silicon Valley loves telling a story: AI gives everyone a super-powered assistant; knowledge is no longer the privilege of the few; true equality is within reach.

It sounds beautiful—but the reality is far harsher. At its core, AI is fundamentally *unfair*. What it actually measures is financial power.

From chips to compute, from model training to token consumption—every stage of AI burns money.

An NVIDIA H100 chip costs over $25,000. Training a GPT-4–level model costs more than $100 million. Every question you ask an AI consumes tokens—and tokens have real prices.

Claude Opus charges $5 per million input tokens and $25 per million output tokens. ChatGPT Pro costs $200/month. Add Perplexity, Cursor, Midjourney… A heavy AI user easily spends over $500/month on tools.

Some people spend $5,000/month leveraging AI to build competitive moats. Others believe they’re keeping up with the times by using the free version of ChatGPT.

They’re not on the same track—or even playing the same game.

At the National Level: Structural Gaps Are Now Irreversible

This logic becomes even starker at the national level.

The AI arms race rests on three pillars: chips, compute, and talent—all requiring massive capital investment.

The United States alone commands over 70% of global AI compute. China is racing to catch up—but chip export restrictions have choked off its supply. As for most developing countries—entry-level broadband costs consume 40% of monthly income in 46 emerging markets.

When a young Nigerian struggles to afford stable internet access, what does “AI equality” even mean?

In high-income countries, 94% of people have internet access; in low-income countries, only 23%. In high-income countries, 84% enjoy 5G coverage; in low-income countries, just 4%.

For many countries in the Global South, the starting line in the AI era isn’t merely behind—it doesn’t exist at all.

This structural gap cannot be closed through effort alone.

At the Individual Level: Your Ceiling Is Being Redefined by AI

The same logic applies to each of us as individuals.

A line I once wrote in my Twitter bio: “An individual’s ceiling = worldview + cognition + practical ability.”

So what has AI done to these three elements?

▶️ First, AI solves vast swathes of practical efficiency problems.

Where it used to take a week to produce an industry report, now it takes one day. Where coding once started from scratch, AI now scaffolds the entire framework. On the efficiency front, AI truly is leveling the field.

▶️ But second, AI massively amplifies cognitive disparities.

Using the exact same AI tool, your outcomes depend entirely on what you ask, how you ask it, and—most critically—whether you can judge whether AI’s answers are correct or flawed. All of this hinges on your preexisting cognitive foundation.

A person with deep domain knowledge uses Claude for research: they know which questions to ask, how to follow up, and which answers contain gaps requiring verification. AI saves them 80% of execution time—time they then invest in deeper thinking.

What about someone with shallow cognition? They dump their question into AI and accept whatever answer comes out—no reflection, no scrutiny, immediate delivery. Over time, they stop thinking altogether. AI doesn’t make them smarter—it makes them lazier, and ultimately, dumber.

▶️ Third, the quality gap in deliverables will widen exponentially.

The depth, accuracy, and timeliness of AI-generated outputs scale directly with your original cognitive foundation. Using the same Claude Opus, one person produces sharp, insightful analysis; another produces verbose but hollow content that merely *sounds* plausible.

A particularly revealing study from Finland’s Aalto University found that the more people use AI, the more they tend to overestimate their own abilities. AI makes you *feel* stronger—your outputs look professional and fluent. But without the capacity to discern quality, you’re merely producing “polished mediocrity.”

So the gaps across worldview, cognition, and practical ability aren’t shrinking—they’re exploding in the AI era.

The smart get smarter; the cognitively rich grow richer in insight; the wealthy leverage superior tools to pull further ahead. Meanwhile, those on the other end grow lazier, shallower, and poorer—even under AI’s “help.”

Cost × Cognition: Dual Divides Reinforcing Each Other

Here’s a causal chain many people overlook:

Money determines which tier of AI you can access → AI tier determines the quality and depth of information you receive → Information quality defines your cognitive boundaries → Cognitive boundaries shape your decision-making quality → Decision quality dictates how much money you earn.

It’s a self-reinforcing loop. The rich get richer; the poor get poorer.

The hallucination rate of the free version of ChatGPT hovers near 40%—meaning four out of every ten answers are fabricated. By contrast, the paid GPT-4 version clocks in at 28%, and the latest iteration cuts that figure by another 45%.

Decisions made using the free version versus those made using Claude Opus accumulate over time into two entirely divergent life trajectories.

Massive information asymmetries have always existed. AI hasn’t erased them—it has simply gated them behind paywalls.

Those Who “Climb the Wall” and Those Who Don’t Already Live in Separate Worlds

Let me share a personal observation that deeply unsettles me.

You’re reading this article right now—most likely because you can “climb the wall,” and browse Twitter freely.

But pause for a moment: How many people in your immediate circle *can’t* climb the wall? When you talk with them, don’t you already sense a stark divergence in cognitive alignment?

This isn’t an IQ gap. It’s a long-term cognitive divergence caused by radically different information environments.

One person consumes cutting-edge global insights, participates in deep analytical discussions, and follows world-class content creators. Another receives algorithmically curated short videos and filtered information feeds.

After five or ten years, their thought patterns, judgment capabilities, and worldviews become utterly incompatible.

The AI era magnifies this divide yet again. Those who climb the wall use Claude, Perplexity, and the world’s best AI tools. Those who don’t—facing bans on ChatGPT and Claude in China—must rely on localized alternatives or resort to costly proxy purchases.

The “wall” in the AI era isn’t just a physical firewall. There’s also the language barrier—cutting-edge AI models are vastly more optimized for English than for other languages. There’s the paywall. And there’s the algorithmic filter bubble. Each wall carves humanity into separate worlds.

Stanford University research shows non-English users consume five times more tokens to process the same content via AI. That means, for the same dollar spent, you receive less information—and lower-quality information.

The Most Terrifying Truth: You’ve Already Fallen Behind—And You Don’t Know It

This is the single point I most want readers to grasp.

Free-tier AI still answers questions. It still helps you write. It still helps you search. So users of free-tier tools feel—“I’m using AI too—I’m not falling behind.”

But free-tier AI reasons more shallowly, hallucinates more frequently, and relies on older data. Its answers *look* correct—but are riddled with subtle, plausible-sounding falsehoods.

It’s like two people both “running.” One sprints forward on open ground; the other jogs in place on a treadmill. Both feel like they’re running—but only one is actually moving ahead.

Psychology has a term for this: the Dunning-Kruger effect—the less someone knows, the more confident they mistakenly feel in their knowledge. AI multiplies this effect tenfold: the more you lean on AI, the more competent you *feel*. Yet you’ve already lost the capacity for independent thought—you just don’t realize it yet.

That is the cruelest truth of the AI era.

It’s not that AI will replace you. It’s that people wielding better AI—and possessing deeper cognition—will leave you impossibly far behind. And you might not even understand how or when you fell behind—until you’re replaced.