AI Is Getting Smarter, But I'm Losing Myself

Preface: This article was written on a night in Lhasa. At an altitude of over 3,000 meters, the air is thin and the stars are very close. One evening I stood in the courtyard, alone, and watched the moon in the sky for a long time. That moon doesn’t lose a bit of its light because of AI’s existence.

The moonlight is still truly beautiful — What do humans still possess in the AI era?

Introduction: The Room for the Unsayable

Natsume Soseki refused to translate “I love you” directly into Japanese. He said it should be translated as “The moonlight is beautiful tonight.”

That which cannot be said directly is the room for human interpretation.

I. Insomnia 🌙

For a period of time, I woke up almost every day at 3 AM.

Not because of anxiety, or not entirely. It was closer to some kind of cognitive noise — a shapeless question rolling around in my brain, disturbing my sleep if it didn’t get an answer.

That question was roughly this:

If AI can think for me, then what am “I” doing?

Specific versions of this question appeared many times:

• I used Claude to write an architecture document, and when I read it back, I found it clearer and more logical than what I would have written myself
• I used Cursor to complete a complex function, and not only did it run correctly, it also fixed a boundary case I hadn’t noticed
• I asked NotebookLM about my notes from three years ago, and it gave me an answer that went deeper than my understanding at the time

Efficiency improved, anxiety also improved.

At first I thought this anxiety was professional — the programmer’s moat was being filled in, technical skills were no longer the threshold, what should I do? But slowly I realized that professional anxiety was just the surface.

The deeper question was not “What can I do?”, but “Who am I?”

This article is my attempt to answer this question. Not the final answer, but the trajectory of thinking.

II. What Have We Lost? 💔

Let’s honestly acknowledge the losses.

2.1 The Satisfaction of Detailed Control

In the past, when writing an algorithm, the feeling of personally handling all boundary cases had texture. It was the satisfaction of a craftsman — not just because the function was implemented, but because you witnessed every step of its creation from nothing to something.

Now AI generates code, and often you don’t fully understand it, but it just works. You’re enjoying something you didn’t fully participate in creating.

2.2 The Identity from Technical Barriers

Programming was once a privileged ability. Mastering it meant there was a visible boundary between you and those who didn’t. Now this boundary is beginning to blur. Ten-year-old children and sixty-year-old elders can all make machines do things using natural language.

This is certainly a good thing, but for many people, that boundary was once part of their self-identity. When the boundary disappears, identity also wavers.

2.3 The Necessity of “Slow Accumulation”

Learning a language, mastering a skill, putting in ten thousand hours in a certain field — these were once the only paths to knowledge. Now AI compresses the accumulation cycle, and knowledge is being democratized. Anyone can access in five minutes information that used to take five years to reach.

What we lose is the means, not the end

But the more I think about it, these losses are not the most fundamental.

They are losses at the level of means, not losses at the level of purpose. Just as after the appearance of cars, humans lost the profession of coachmen, but didn’t lose the desire to “go to distant places.”

What really kept me tossing and turning was the deeper layer:

If AI can not only replace execution, but also replace thinking — then when “I” think, where is “I”?

III. The Shattering of Three Illusions 💭

On this question, we had three illusions, and now they are breaking one by one.

Illusion One: Human Advantage Lies in Mastering Knowledge

This illusion died earliest, and most thoroughly.

Knowledge has never been equal to wisdom, but we have long confused the two. Degrees, certificates, extensive reading, professional vocabulary — these things constitute an appearance of “knowing a lot,” and we take it as proof of intellectual ability.

But knowledge is essentially encodable, and anything encodable can be stored and retrieved more efficiently. An AI that can recite all medical literature completely crushes any doctor at the knowledge level.

This is not a threat, this is a fact.

Illusion Two: Human Advantage Lies in Logical Reasoning

This illusion died a bit slower, but the ending is the same.

We once believed that even if AI could store knowledge, “deriving conclusions from premises” still required human rational ability. Chess, coding, mathematical proofs — these were the jewels in the crown of human intellect.

Now they have all been taken away. Not barely taken, but easily crushed.

• AlphaGo won at Go in a way that human players couldn’t understand at all
• GPT-4 passed the bar and medical licensing exams
• DeepSeek’s performance on reasoning tasks is beginning to approach and surpass human experts

Illusion Three: Human Advantage Lies in Creativity

This illusion is in the process of breaking.

AI writes poetry, composes music, paints, writes novels. It doesn’t just imitate styles, but can produce many things that humans simply cannot distinguish as “machine works.” If you mix an AI-written poem with a human-written poem and have people blindly choose which is better, the result is often fifty-fifty, or even slightly in favor of AI.

Three illusions shatter consecutively, what remains for humanity?

The consecutive shattering of three illusions has plunged many people into a kind of nihilism:

If even creativity is not unique to humans, then what exactly does “uniquely human” refer to?

I spent a long time standing in front of this question, not knowing which way to go.

IV. What Is Agency? 🤔

One day I walked for eight hours on a mountain road in Nepal.

It was a road with no cell signal, at an altitude above 4,000 meters, and my body started giving signals to retreat continuously. I remember that on a particularly difficult scree slope, my brain started automatically generating various reasons not to continue:

• Too tired
• It’s getting dark
• The target destination today is actually not important…

Then I started silently reciting the six-character mantra.

Om Mani Padme Hum.

Not because of faith, but to cut off those thoughts.

Those thoughts were essentially “thinking” — analysis of the current situation, prediction of the future, management of the self. But at that moment, all thinking was consuming me, not helping me.

The mechanical repetition of the six-character mantra was forcibly pulling myself from “analysis mode” back to “existence mode” — I was not calculating how many kilometers were left, I was this step, this breath, this moment.

On a mountain path in Nepal, I understood what “presence” means

This thing made me figure something out.

Agency is never the ability to think.

Thinking is just a tool. Agency is the thing that is “who is using this tool.”

In philosophical language: Consciousness is the necessary premise of all experience, not the result of experience.

• You can feel pain, not because you understand the neurophysiological mechanism of pain, but because you are in that pain
• You can fall in love with someone, not because you analyze their pros and cons, but because something moved inside you

AI can describe pain, can analyze love, can generate ten thousand kinds of words about these two things.

But AI doesn’t have that position of “being in it.”

It has no body, no history, no “I” that sees the world from a unique, concrete perspective.

This is not to say AI is incapable. This is to say, AI and humans are doing different things.

V. Feelings Come Before Thoughts 💓

In my notes, there’s a distinction I’ve thought about for a long time: feelings and thoughts are two different things.

More importantly, the order: the body always reacts first, then the brain gives an explanation.

You see that person walk into the room, and somewhere in your chest something tightens slightly — this is a feeling.

Then the brain starts explaining: oh, it’s because of last time’s argument, it’s because of some memory, it’s because of… The explanation comes later, and is often wrong.

We often take the stories our brains fabricate as our real feelings.

AI is very good at the latter part — generating explanations, constructing narratives, providing frameworks. It can give you ten thousand kinds of theories about “why you feel uneasy.”

But it can’t do the former part.

It doesn’t know that moment of chest tightening, because it has no chest.

Feelings are the language of the body, thoughts are the brain’s translation

This thing seems small, but I think it points to the hardest core of human agency — embodiment.

We have bodies, have positions, have feelings, have death. We are not omniscient beings looking down on the world from a God’s eye view; we start from a concrete, limited, fragile body, stumbling through experiencing this world.

This limitation is precisely the source of meaning.

VI. Meaning Is Not Calculated ✨

“Meaning is a product of surprise.”

I’ve written this sentence in my notes several times, but it took longer to truly understand it.

We are accustomed to thinking that meaning is something that can be searched for — you make enough effort, try enough paths, and then you “find” the meaning of life.

But careful observation reveals that no one’s life operates this way.

Meaning is never the result of a search; it always pops up when you’re doing something else.

• You don’t decide to have children because you think “being a parent is meaningful,” but you have children, and then one morning watching their sleeping face, something indescribable hits you — that’s meaning, sudden, unexpected
• You don’t love a certain job because you calculate it’s “valuable,” but in the middle of the night when a project is halfway through, you suddenly realize you’ve forgotten time — that state is meaning

Meaning emerges unexpectedly, not as a result of calculation

AI is very good at calculating:

“According to your values and existing data, the following three directions are most meaningful for you…”

But AI has misunderstood the essence of meaning.

Meaning is not calculated; it emerges. It requires concrete people, concrete moments, concrete bodily feelings, and a kind of contingency that belongs only to that instant.

This resonates with something I’ve always found beautiful:

Natsume Soseki said “The moonlight is beautiful tonight,” not “I love you.”

That detour, that not saying it directly, is precisely what preserves that which cannot be spoken.

Truly profound feelings are often completed before they reach language.

VII. The Secret of Inspiration 💡

I once pondered a question: Can AI produce “inspiration”?

AI can produce good output, no doubt about it. It can generate things that humans judge as “creative.”

But is this the same as inspiration?

Inspiration, in my view, has two necessary conditions:

7.1 Randomness

Inspiration always appears when you least expect it.

• You think of the structure of a paper while showering
• You suddenly understand the meaning of an experience from three years ago while walking on the street
• You catch the first word of the sentence you’re going to write tomorrow in the vague consciousness before sleep

This “suddenness” is not the result of calculation; it’s more like the brain in some relaxed state, letting different neural circuits connect freely, accidentally producing a new contact point.

7.2 Emotional Surge

Moreover, inspiration is always accompanied by an emotional surge.

“This is it!”

This feeling is not just cognitive confirmation; it’s also a physical reaction — heart beats a little faster, some excitement. That’s a feeling, not just a judgment.

Inspiration is the moment of being surprised by oneself

AI-generated “creativity” doesn’t have this process.

It’s not suddenly realizing something in some relaxed moment; it’s performing a lot of statistical inference after receiving a prompt. The output can be brilliant, but it’s not surprised by its own output.

And the most beautiful part of human inspiration is precisely that moment of being surprised by oneself.

VIII. Software 2.0 Eats Software 1.0, What About Humans? 🤖

Some say: Software 2.0 (AI-driven) is eating Software 1.0 (rule-driven).

This is a very accurate description.

• Software in the past was changing the world
• Software now is changing the software from before
• Code is learning to write code
• Tools are learning to make tools

So where do humans stand in this map?

I increasingly feel that humans should not position themselves as “better tools,” but should re-understand themselves as “the subject that uses tools.”

This is not a downgrade, but an upgrade, provided you truly accept this role transition.

But accepting this transition requires precisely what we’ve been discussing — clear agency.

You need to be clear enough about who you are, what you want, why you’re doing it, to truly “use” AI rather than “be used by AI.”

A person without direction, no matter how good the engine, will only get lost faster.

This is why I believe that the most important ability in the AI era is not technical ability, but self-awareness ability.

The deeper you understand yourself, the more accurately you use AI, and the more irreplaceable what you create becomes — because behind it is a real, unreplicable person.

Humans are direction definers, AI is a powerful engine

IX. Tolerance vs. Endurance, and the Daily Life of Agency ⚖️

Agency is not just a grand proposition; it also lives in the details of daily life.

There’s a distinction I think is a concrete test of whether agency is present:

Are you “tolerating” something, or “enduring” it?

The difference is not in behavior, but in internal state.

Your ability to distinguish which mode you’re in is itself an expression of agency.

Because:

• Endurance means something external is greater than your own judgment, it’s suppressing you
• Tolerance means you’re truly processing it, it passes through you, not around you

In the AI era, this distinction applies equally.

Many people’s use of AI is a kind of endurance — they don’t like this change, but have to accept it; they use the tools, but their heart is resisting. This state gives you nothing, neither efficiency nor creativity.

The use with true agency is tolerance — you truly understand the essence of AI, you accept what it can and cannot do, you find your real position on this basis, and then set off.

That feeling of “setting off” is relaxation.

Not slow, not lazy, but not being chased by internal pressure. You can be very engaged, but not driven by fear, but by genuine curiosity and intention.

The test for this state is simple:

How bad would it be if you did nothing today?

• If your answer is “the sky is falling,” you’re probably in anxiety mode
• If your answer is “a bit of a pity, but continue tomorrow,” you’re probably in relaxation mode

Relaxation is not laziness, but not being driven by fear

X. The Moonlight Is Still Truly Beautiful 🌕

Recently I lived in a highland city, over 3,000 meters above sea level, thin air, stars very close.

One evening I stood in the courtyard, alone, and watched the moon in the sky for a long time.

Not thinking about anything, just watching.

That moon doesn’t lose a bit of its light because of AI’s existence, and the quietness under that moonlight is not something any algorithm can generate.

The moonlight is still truly beautiful, go take a look

I suddenly understood why Natsume Soseki translated it that way.

• “I love you” is a proposition, a piece of information, that can be clearly transmitted, understood, archived
• “The moonlight is beautiful tonight” transmits an invitation — an invitation to enter the same experience, to stand under this moonlight with me, to feel the same thing that cannot be captured by language

This is the deepest use of language:

Not to transmit information, but to co-create an experience.

AI can write a more fluent “I love you” than humans, but it cannot write “The moonlight is beautiful tonight” — not because of lack of ability, but because it doesn’t have that moon, doesn’t have that thing that needs to be said indirectly, doesn’t have that inner richness that makes direct speech too rough.

The room for humans is in that detour.

In those things that cannot be said directly, can only be hinted at, can only be felt when you and I are together.

That’s not skill, not knowledge, not logic, not creativity.

That’s presence — a living, breathing, historical, limited, feeling, mortal life, here now, seeing, feeling.

This thing, AI cannot do.

Epilogue: Making Peace with Insomnia 🌃

I haven’t solved the insomnia problem, but I’ve made peace with it.

Waking up at 3 AM is no longer just noise; it can also be some kind of calling — reminding me that some questions haven’t been thought through, some feelings haven’t been taken seriously, some life hasn’t been truly lived.

AI can help me complete many things faster, this is true.

But “completing faster” is not equal to “truly living.”

That gap that concerns me is everything we’re discussing — agency, feelings, meaning, presence.

Technology accelerates, but human maturity never accelerates.

You still have to experience those things, digest those things, be hit by something on some late night, and then slowly, grow something new.

No one can complete this for you.

AI can’t either.

The moonlight is still truly beautiful.

Go take a look.

References

• March 2025 Thought Notes
• Metacognitive Transformation Review
• Flow State: Enhanced Focus and Happiness Guide