This is a pretty interesting topic — it hits right where it hurts.

Let me use the field I know best as an example. What we algorithm engineers do every day, at its core, boils down to one thing: feeding information to users. You think the internet has leveled the information playing field? What I see is the exact opposite — in certain dimensions, the internet has wildly amplified the cognitive divide.

A lot of people have a misconception about algorithms. They think algorithms are designed to help you discover new worlds. They’re not. In most commercial scenarios, the objective function of an algorithm is retention time and click-through rate. What does that mean? It means the system will desperately pander to your animal instincts.

You say that top-tier ways of thinking and scientific concepts are everywhere, and that’s true. In this era, as long as you want to learn, you can find Harvard’s open courses, papers by Turing Award winners, post-mortems by industry heavyweights, and even plenty of paid courses — all accessible.

But the problem lies precisely in that word: want.

Let me drop a hot take right now: the vast majority of people don’t actually want so-called elite thinking or breakthroughs. What they want is the illusion of making an effort, or some kind of emotional comfort.

Let me unpack this.

First, the cost of access has plummeted, but the cost of filtering has risen exponentially.

In the past, information asymmetry was physical isolation. The village didn’t have internet, so you had no idea how city folks lived. Now the internet is connected. The problem isn’t too little information — it’s too much.

As people who work with data, we understand signal-to-noise ratio all too well. Today’s internet is a gigantic noise generator. Truly valuable, high-density information is like a needle tossed into the Pacific Ocean.

You think picking up your phone means you’re acquiring information, but you’re actually just having your attention fought over by all kinds of meticulously designed hooks.

Let me paint a concrete picture.

Say you want to learn distributed systems. Twenty years ago, you’d go to the library, dig up that classic English textbook, and grind through it. Now? You fire up a search engine or a video platform.

You type in “distributed systems.”

What do you get?

At the top, it might be an ad from some training institute, telling you that in 30 days you’ll master architecture and earn a million a year.

Or some blogger explaining things in extremely oversimplified language — it feels great to listen to, you think you understand it, but all the critical boundary conditions have been left out.

Or maybe it’s a flame war on some forum about which programming language is better for backend development.

The genuinely hardcore paper on the derivation of the Paxos algorithm, or those classic Google File System papers — they might have fewer than a hundred views, quietly sitting in a corner.

Most people will spin their wheels in that low-density, emotion-driven, or highly processed second- and third-hand information. Because that stuff is easy to digest.

This is exactly what algorithm engineers do every day: we optimize for click-through rate. Content with a high cognitive threshold naturally has a low click-through rate, and its completion rate is downright tragic. The recommendation system is a snob — it won’t push something to you just because it has long-term value for you. It only pushes things that can keep you here right now.

So the internet, which looks like it has erased the access barrier, has actually erected an invisible filtering barrier. That barrier is no longer money or connections — it’s your discernment and your ability to resist distraction. And that’s way harder to acquire than money.

Second, knowledge is a compressed archive, and experience is the decompression password.

I hold this view pretty firmly.

A lot of young people — including some fresh graduates I’ve mentored — absolutely love hoarding “substance” (ganhuo). See a good article? Bookmark it. See a good tutorial? Save it to cloud storage.

They think owning the file means owning the knowledge.

It’s like downloading a massive compressed archive but not having the decompression password.

Top-tier ways of thinking are usually highly abstract. Think of Charlie Munger’s mental models, or first-principles thinking. These terms have become completely cliché by now. Grab any random internet product manager and they can toss around phrases like entropy increase, entropy decrease, and flywheel effect.

But you’ll notice that when it actually comes time to do real work, their execution falls apart completely.

Why?

Because those concepts are compressed data extracted from a vast ocean of hands-on experience.

Let me tell you a true story from my own life.

About five or six years ago, I was working on ad recommendation algorithms at a unicorn company. I was young back then, had read a bunch of books on system robustness and antifragility, and thought I was hot shit. I was constantly going on about how we needed to build antifragile systems.

At the time, we needed to rebuild a core recall module. I looked at the old code — it was a mess, full of hard-coded rules and patches everywhere. I was utterly contemptuous. I thought, man, these predecessors of mine are so lame.

I went at it boldly, rewriting everything according to the perfect architecture from the textbooks. I used the latest frameworks, threw in design patterns left and right.

On launch day, it crashed.

It wasn’t a logic error. What I had completely overlooked was the dirty data and extreme long-tail distribution in real traffic. Those ugly-looking if else blocks in the old code were actually patches that my predecessors had applied after stepping into countless pitfalls — they were the safety nets.

In that moment, I finally understood: the robustness the books talked about wasn’t about writing beautiful code — it was about being able to keep running even when you’re in a mud pit.

Without that experience — the full-scale rollback, getting chewed out by my boss until I was drenched in blood, debugging logs at 3 a.m. — antifragile would have remained nothing more than a pretty word to me. After that experience, the word truly became part of my cognition.

The internet today serves you the result on a silver platter, cutting out the process.

Most people look at those elite ways of thinking the way they watch someone else’s workout videos. Your eyes understand it, your brain feels like it’s got it, but your muscles have no memory of it. Real breakthroughs require you to throw your physical body against reality, to feel the pain, to get the feedback.

All elite thinking is, at its core, post-hoc summary. If you haven’t lived through the event, being handed the summary is useless.

Third, the brain’s biological instinct is to avoid thinking whenever possible.

We need to start from the hardware on this one. Our brains don’t make up a large proportion of body weight, but they consume an enormous amount of energy. Evolution has dictated that humans are, by instinct, cognitive misers.

If you can judge with intuition, never use logic. If you can solve it with experience, never try innovation.

Most content on the internet caters to this instinct.

Why are short videos so popular? Because they’re passive input — they don’t require you to actively allocate computing power.

Why are power-fantasy novels so addictive? Because their logic is simple and the feedback is instant.

But those scientific concepts and elite ways of thinking you mentioned — they’re usually counterintuitive.

Take Bayes’ theorem, for example. It requires you to continuously revise your prior probabilities, which goes against the human instinct for neat, deterministic linear causality.

Take delayed gratification — that goes against the biological instinct for immediate pleasure.

Learning these things is physiologically painful. It’s a literal brain burn — your brain is consuming glucose, forcing the creation of new synaptic connections.

The vast majority of people — myself included — when they’re dog-tired after work, will instinctively choose to scroll through funny videos rather than crack open The Art of Computer Programming.

So even though knowledge is everywhere, the pain of ingesting knowledge hasn’t decreased at all. In fact, because there are so many temptations, that pain has become even harder to endure.

Fourth, between knowing and doing lies an enormous emotional cost.

This is a point many people easily overlook.

I know what I should do, but I just can’t do it. This isn’t an IQ problem — it’s a problem of emotion and psychological energy.

For instance, everyone knows that if you want a technical breakthrough, you need to read source code, build wheels from scratch, and contribute to open source.

The path is crystal clear. GitHub is right there. The source code is public.

So why don’t people do it?

Because of fear. Fear that you won’t understand it. Fear of the frustration. Fear that you’ll invest time and get no return.

The internet has set the bar for success ridiculously high. The moment you open your phone, it’s all stories of million-dollar salaries, retiring at 35, and startup success. This invisibly manufactures massive anxiety.

When you see how easily others seem to have succeeded, while you’ve spent an entire evening grinding through source code and only understood two lines — that enormous gap can shatter your psychological defenses in an instant.

So you give up. You tell yourself, maybe I’m just not cut out for tech, maybe I have no talent.

That’s not really it.

The real breakthrough process is usually tedious, long, and full of self-doubt.

I once maintained an open-source project on GitHub that earned a decent number of stars. In the first six months, nobody paid attention at all. It was nothing but bugs. I questioned my life choices every single day. That kind of loneliness and helplessness — you can’t feel it reading success-porn articles.

Most people can’t break through because they can’t endure the loneliness of going a long time without positive feedback.

The internet hasn’t just provided knowledge — it’s also provided substitutes for instant gratification.

When work feels too hard to break through, you can carry the whole team in a video game.

When real-life socializing feels too difficult, you can be a keyboard warrior online and lecture the world.

These cheap dopamine hits perfectly intercept the energy that was supposed to be used for breaking through.

Fifth, the solidification of social circles creates hidden barriers of tacit knowledge.

Although explicit knowledge (books, code, formulas) is now public, tacit knowledge still only circulates within small circles.

What is tacit knowledge?

It’s the pitfalls — the specific scenarios where a certain technical solution will fail.

It’s the assessments — who in the industry is reliable and who isn’t.

It’s the signals — the cues before a particular wave truly takes off.

Nobody writes this stuff on public forums. Nobody turns it into short videos. It exists at offline dinner gatherings, in late-night private WeChat chats, in private GitHub repositories.

The internet has connected everyone, but it hasn’t broken down circles. On the contrary, because of the algorithm’s clustering effect, circles have become even more closed off.

You think you’re seeing the whole world, but you’re really only seeing the group of people whose profile resembles yours.

A person who scrolls through entertainment news every day — their recommendation feed will never surface cutting-edge deep learning papers. And even if one occasionally appears, they’ll swipe past it. The system will keenly detect that swipe-away action and never recommend it again.

This is the xinxijianfang (information cocoon / filter bubble) — not just created by the algorithm, but created by each of your own choices, one after another.

1. Set ruthlessly utilitarian learning goals.

Don’t give me vague words like “self-improvement.” What concrete problem are you trying to solve?

For example, I recently needed to tackle a performance bottleneck in large-scale vector retrieval. So I only looked at things related to that. Whether it was elite thinking or a third-rate blog post — as long as it helped me solve the problem, it was good stuff.

Don’t learn for the sake of learning. Learn for the sake of solving problems.

Only in the process of solving problems can that knowledge get welded into your brain.

2. Even if you’re just copying, go through it with your own hands.

For technical people, this is absolutely crucial.

You can look at architecture diagrams a hundred times, but it’s not as good as writing a demo and running it yourself once.

You can read a hundred business analyses, but it’s not as good as setting up a street stall yourself or trying to sell a small piece of software.

When you do it with your own hands, all the details you overlooked will surface. And the devil is in the details. Breakthroughs often emerge from mastery over those details.

3. Block out the noise. Actively reduce noise.

I now force myself to disconnect from the internet for a period every day. Or I only consume long-form content.

If you want to develop the ability to think deeply, you have to train yourself in reading long essays, tackling thick books, and sustaining focus for extended periods. This isn’t for showing off — it’s to physically rewire your brain circuits, to adapt them to deep processing mode rather than the shallow stimulus mode of short videos.

You need to learn to say no to the algorithm. Search more, scroll feeds less. Search is you actively seeking information; feeds are other people feeding you information. Who holds sovereignty makes a world of difference.

4. Find a circle of people who are actually doing real work.

Don’t hang around in groups where people just shoot the breeze. Find circles where people are genuinely building projects, writing code, and running businesses.

Even if you’re just watching from the sidelines — watching how they argue, how they discuss problems — the tacit knowledge you’ll absorb is worth more than reading ten “substance” articles.

5. Accept mediocrity, but keep your touch.

This last point might sound a bit bleak, but it’s very real.

Most of us, over the course of our entire lives, may never achieve that kind of earth-shattering breakthrough. We probably won’t become Steve Jobs, and we won’t win a Turing Award.

But that doesn’t mean we can’t get better.

Keeping your touch is a really important concept. It’s like a basketball player practicing their shot.

Write a bit of code every day. Read something difficult every day. Think about a complex problem every day.

Don’t count on sudden enlightenment. What we call breakthroughs are mostly quantitative changes tipping into qualitative ones.

The internet is just a tool. It gives you a hammer, and it also gives you a shovel, and it also gives you a pile of video game consoles and porn.

Most people choose the game console. A minority pick up the shovel and start digging. A tiny handful pick up the hammer and shatter the old world.

That’s pretty fair.

There’s no point pretending not to see it — this is itself a form of natural selection.

The elite stuff is just sitting there, and it comes with built-in encryption. The encryption isn’t a password — it’s boredom, difficulty, and going against human nature.

Those who can pass this trial will naturally receive the reward. Those who can’t, who hang around at the base of the mountain watching the spectacle and scrolling on their phones, are actually pretty happy, too. Aren’t they?