Thoughts

A Life Filled With LLMs

I spent 20 billion tokens building projects.

That sentence sounds exaggerated, but it is close to how this period of my life has actually felt. LLMs have entered my workflow deeply: writing code, changing code, reading documentation, looking up problems, explaining errors, generating plans, refactoring projects, writing tests, writing docs. Their shadow is almost everywhere.

The strange thing is that, in theory, I should feel lighter. When such a powerful tool enters the production process, it should reduce my workload, reduce my working hours, and reduce the pain of facing complex systems. Reality has been exactly the opposite.

My workload has not gone down.

My working hours have not gone down either.

Even worse, my brain feels more exhausted than before.

It Did Not Do the Work for Me. It Changed How I Work.

Many people imagine LLMs like this: you hand a task to the model, and it hands the result back to you. It sounds like a new kind of automation. But after actually using it to build projects, I increasingly feel that LLMs have not done my work for me. They have only transformed the shape of my work into something else.

In the past, when I built projects, I faced code, documentation, compilers, and debuggers. The problems were complicated, but the context was continuous. If a bug blocked me, I would follow the call chain, waveforms, logs, and source code all the way down. It was painful, but my brain stayed inside the same problem space.

Now it is different.

I have to constantly describe problems to an LLM, organize context, filter information, judge whether its answers are reliable, and then map its suggestions back into the project. When it writes a piece of code, I cannot trust it directly. When it gives an analysis, I cannot accept it directly either. I have to check whether it misunderstood the architecture, invented an interface, missed an edge case, or turned a local optimization into a global disaster.

So the continuous thinking I used to have gets cut into many fragments.

On one side is the actual context of the project; on the other is the LLM’s understanding of the project. On one side is the real structure in my head; on the other is the structure it temporarily assembles from prompts. I have to keep switching between these two worlds.

This is not relief.

It is a new kind of mental labor.

Context Switching in the Brain

What really tires me out is not writing code. It is context switching.

A complex project already requires holding a large amount of state in your head: relationships between modules, interface contracts, historical design decisions, unresolved issues, attempts you just made, paths already proven not to work, and bugs that look like bugs but are actually expected behavior.

After LLMs enter the process, these states do not disappear. They multiply.

I not only have to remember the project context. I also have to remember “what I just told the LLM,” “what it currently knows,” “what it does not know,” “what it may have misunderstood,” “whether the conclusion from the previous conversation is still usable,” and “whether the code generated in this round conflicts with the previous round.”

It feels like I already had a heavy system running in my head, and now I have launched a huge collaborative process beside it. Sometimes it speeds things up. Sometimes it consumes memory like crazy.

The scariest part is that it does not truly remember for you.

It looks as if it knows everything, but in reality it easily forgets, confuses, and drifts. You have to keep feeding it context, correcting it, and reminding it. In the end, you realize that the person carrying the long-term memory is still you.

The LLM generates. You take responsibility.

Memory Overflow

I have become more and more aware of a kind of “memory overflow.”

Before, when I worked on projects, I was tired too, but the tiredness was purer. Today I debugged this module, tomorrow I debugged that module, and there was still a relatively clear thread in my mind.

Now, LLMs make tasks move faster. In one day, you may read more code, try more approaches, generate more patches, open more branches, and discuss more possibilities. On the surface, this looks like improved efficiency.

But human working memory has not been upgraded along with it.

The faster tasks move, the more unclosed contexts accumulate in the brain. This approach was tried but did not completely fail. That module was changed but not fully verified. A sentence in one answer may be useful but has not been confirmed yet. A bug may have been fixed, or it may only have been hidden.

All of these things hang in the mind like memory that has not been released.

In the end, a person becomes extremely tired. Not physically sleepy, but mentally overflowed: sticky, deep, and impossible to organize.

LLMs make information production too easy, while the human ability to digest information has not improved at the same pace.

Faster Does Not Mean Less

This may be my greatest disillusionment with LLMs: they make many things faster, but they do not make things fewer.

Code is written faster, so requirements increase.

Documents are generated faster, so documents increase too.

Plans arrive faster, so the number of plans that need comparison also increases.

Prototypes are built faster, so people begin to expect you to produce in one day what used to take a week.

Technological progress does not naturally bring rest. More often, it raises the standard. In the past, you could only do one thing in a day. Now you can do three, so three things become the new normal workload.

So LLMs have not liberated people from work. They have only compressed work at a higher density into the same amount of time.

That is also why, after using so many LLMs, I increasingly miss the state before they appeared.

Back then things were slower, but the slowness had order.

Electronics Have Been Pulled In Too

LLMs are not only changing software development. They are also changing the whole environment of technology consumption.

Now every electronic product seems to need AI. Phones need AI. Computers need AI. Headphones need AI. Even keyboards and mice seem eager to have AI. Manufacturers have finally found a new reason to raise prices: stronger NPUs, larger memory, higher compute, smarter systems.

But as a user, do I really need this much AI?

Most of the time, I only want a stable, durable, cheap, quiet tool. But the market will not stop for my plain needs. AI has become a new marketing entrance, and also a new way to pass costs on to users. You may not need those features, but you still have to pay for them.

The more ironic part is that products have not become easier to use because of it. Many features are only wrapped in another layer: more buttons, more pop-ups, more cloud services, more subscription entrances. They look smarter, but the actual experience becomes more complicated.

Technology was supposed to return tools to their nature as tools. Instead, tools increasingly feel like shopping malls that keep trying to sell themselves to the user.

Media Quality Is Declining

Another problem LLMs bring is that the threshold for producing content has dropped sharply.

Articles can be generated automatically. Scripts can be generated automatically. Video copy can be generated automatically. Thumbnails and titles can be optimized automatically. Even comment sections can be filled automatically. There is more and more content, but the amount of content truly worth watching has not increased in proportion.

On short-video platforms, more and more content has a familiar taste: the title is provocative, the structure is complete, the tone is smooth, but after watching it, nothing remains. It feels as if it has been polished by the same machine: smooth, full, and hollow.

This is frightening.

In the past, low-quality content at least exposed a person’s roughness. You could see a person’s limitations, and through those limitations you could also see something real. Now, low-quality content is polished by LLMs until it looks respectable. It no longer appears crude. It may even look professional. But at its core, it still has no thought, no experience, and no real judgment.

When everything becomes more like content, truly valuable content becomes harder to see.

Information pollution does not happen simply because there is too much information. It happens because too much information looks real.

I Wish It Had Never Appeared

If I look at it only rationally, of course I know LLMs are powerful. They can indeed improve efficiency, help with learning, and lower the barrier to many things. Without them, I probably could not have advanced some projects so quickly, or encountered so many tools and methods in such a short time.

But human feelings do not always obey reason.

When I realize that my workload has not decreased, my working hours have not decreased, and yet my brain has become more tired; when I realize that every day I am switching context, organizing context, feeding context to models, and then rescuing context from model output; when I realize that technology products are becoming more expensive and media content increasingly resembles industrial waste, a very strong thought appears in me:

I wish LLMs had never existed.

This is not because I hate technology.

On the contrary, it is because I love technology too much.

The technology I love is the kind that makes people freer, clearer, and closer to the essence of things. Not the kind that pushes everyone toward higher speed, higher density, and louder noise.

Maybe LLMs will eventually become true productivity tools.

But at least right now, what they bring me does not feel like liberation. It feels like an accelerated predicament.

They have made the world faster.

But I am not sure the world has become better because of it.

Exams

In the life of modern Chinese people, there are probably three to four important exams. The first is the high school entrance exam, and the second is the gaokao. If you do not take the graduate entrance exam, the third is what I call the “social exam”: it decides whether you can receive a job offer from society when you graduate from university. If you choose to take the graduate entrance exam, then the third becomes that exam, and the “social exam” is postponed to become the fourth.

In other words, many exams before the “social exam” ultimately serve it. Finding a job has almost become the final goal on this path.

When thinking about problems, I am used to first asking about the essence, then looking for solutions according to first principles. So what is finding a job for? To make money. If so, why not study around “how to make money” from the beginning?

Many people fall into a misconception here: they believe only studying, and only finding a job through studying, can make money. This is actually a major misunderstanding.

You will find that whether a person has received a good education or not, they can understand the importance of money. Money has long been deeply embedded in our culture. Westerners say “Happy New Year” during the New Year; we say “wish you wealth.” Even relatives who have not seen you for years often first ask what job you do and how much you earn. They may not be malicious. They are just quickly calculating how to evaluate you and how to adjust their attitude toward you.

What is strange is that how to make money is treated like a taboo topic. If someone systematically explains how to make money in a livestream, the stream may quickly be shut down. Even if someone truly wants to teach you, you probably may not believe them. Because in our default cognition, no one will sincerely teach others such a good thing as “making money.” So once someone says they want to teach you how to make money, people’s first reaction is often: this person is probably a scammer.

As a result, we almost avoid talking about “how to make money.” On one hand, people do have a selfish side; on the other hand, we also assume by default that others must be selfish.

So when you tell others that you are preparing for the graduate entrance exam, studying 8 hours every day for 9 months straight, people will think you are hardworking and approve of your effort. This is understandable.

But if you tell others that you spend 4 hours every day studying stock trading theory, combining it with practice, and also systematically studying microeconomics, macroeconomics, social psychology, and related knowledge, also for 9 months, many people’s first reaction may be: did something happen to you? Are you starting to dream of getting rich?

But in fact, even many traders who have already achieved financial freedom may not have studied investment so systematically. In other words, if a person is willing to seriously study and continuously practice this trading knowledge, they will probably not be trapped by economic problems in the future. Going further, becoming wealthy is not completely impossible.

More importantly, this knowledge does not become invalid because of one failed trade. The time you invest in cognition, experience, and practice will not be reset to zero because of a loss. From this perspective, the risk of this kind of learning is far lower than many people imagine. By comparison, the opportunity cost and uncertainty carried by the graduate entrance exam are unacceptably high.

Then why do so few people do this?

This brings us to human instinct.

Over the long course of evolution, humans usually did not actively chase invisible possibilities in order to save energy, because that meant huge trial-and-error costs. If a primitive person saw an elephant, he would not rush up to fight it one-on-one, because he had never seen anyone do that and succeed. People naturally prefer to do things within their cognitive range.

This is also why few people put the same energy they use for the graduate entrance exam into learning trading, investment, and the ability to make money: they have not personally seen enough credible successful examples. This instinct allowed humanity to reproduce until today in a low-risk, low-return way.

But today is different. The internet has greatly reduced information gaps. If you are willing to actively understand, you will find that many people are trading: they may be chefs, security guards, programmers, teachers, or people from all kinds of ordinary professions. Of course, most people have not made money. The reason is simple: they are not professional.

Making money is such an important matter, yet it is often treated in an extremely unprofessional way. That is absurd in itself. After all, even a third-rate programmer needs basic professionalism, let alone someone directly operating money and risk.

I say this not to encourage everyone to learn trading, and certainly not to encourage everyone to speculate in stocks.

What I really want to express is this: as long as you are willing to seriously study “how to make money,” making money is not something so mysterious that it cannot be touched.

When you first graduate, even if you can only save 3,000 yuan a month, more would of course be better, you can first save it. When you have saved 50,000 yuan or more, if you really do not want to keep working for others, you can seriously consider the next step. What you need to do next is use the capital in your hands and the money-making knowledge you have learned to make the money grow. For someone who truly has the ability to make money, time itself is like an ATM.

Working for wages is essentially saving ransom money, not striving to become the emperor of the wage-labor order. A person should be worthy of the education they have received. The way to break the situation is actually here.

Ideology

People are often naive and childish. Whether they are university students in their twenties or adults in their forties or sixties, they may think some things are too simple and other things too complicated.

Failing a university course, they feel life is over; failing CET-4 or CET-6, they feel they are failures; getting into graduate school, they feel they have finally reached the shore, as if everything will be fine from then on.

From childhood to adulthood, many people receive a fixed narrative: study hard, get into a good university; after entering a good university, join a good company; after joining a good company, earn a lot of money; after earning a lot of money, buy a house; after buying a house, get married; after getting married, have children; after having children, make them study hard too. Around and around this loop goes. When does it end?

If your family background is good, life may be relatively easy no matter how you live. But if your family is ordinary or even difficult, and you strictly follow this track, life can easily become a disaster.

This society hopes everyone takes on a mortgage, then works hard, and gradually gives up other possibilities in the process. Then they raise a child and work even harder. This is the current life pattern of many ordinary people.

Can we change the whole society? It is difficult. But at least we can change ourselves.

After graduation, you can first earn money for a few years, but do not buy a house lightly. Once a house is bought, most of life’s possibilities are basically locked. If you are unlucky and encounter unemployment or falling housing prices, you may end up defaulting and seeing the house auctioned. After going through that loop, even a strong person will be badly damaged. Working hard for years, then losing hundreds of thousands: examples like this are countless. Even worse is buying a presale property, then having the house auctioned while you have never even seen what the house truly looks like.

Then does that mean not getting married? This question itself precisely reveals the existence of social ideology.

What is social ideology? It is the collection of consciousness, values, and default judgments commonly held by many people in a region or society.

For example, almost all of Chinese society believes people should go to university. This is a kind of social ideology. Basically, no high-performing high school student or parent seriously thinks, “Do I really need to go to university?” or “Does my child absolutely have to go to university?”

Another example: people should get married in this life. This is also a social ideology. But this set of ideas has already been questioned by young people. Many people no longer like marriage, so today there are quite a few people who do not get married.

We can continue discussing other social ideologies.

For example, young people should buy houses in cities. This may be one of the most ridiculous and tragic social ideologies. It has ruined countless young people who originally had bright futures. Chinese people have an obsession with “home,” and houses hit this weak point exactly. If this idea were imposed on Americans, they might find it unbelievable: why take on such high debt just to live in an apartment? Unless one is already rich enough, what exactly is the point of buying a house?

Another example: people should be obedient. This ideology is even more absurd. The key is that it is systematically cultivated starting from primary school, middle school, and high school. Writing this, I can hardly avoid feeling angry. It is almost a crude discipline imposed on thought. Due to limited space, I will not expand with specific examples here.

There is also the idea that good grades mean a good student, and bottom grades mean a bad student. This is even more ridiculous. I once watched a video in which a parent said to a teacher: “Although my child studies very poorly, he is still my beloved child. I hope he grows up happily.” To me, that sentence is true clarity in the human world.

Let me give another example from around me. My cousin spends all day dealing with game accounts and also does so-called game farming, earning around 8,000 yuan a month. But his parents always oppose it, thinking it is not a proper occupation. The so-called “proper occupation” is essentially part of social ideology too. What is a proper occupation? The concept itself is very suspicious.

Every joule of work that wage workers do on objective objects is ultimately for money. If money can be earned reasonably, why distinguish between proper and improper occupations?

So when facing these social ideologies, the most important method is to figure out what you truly want.

Do you really want a mortgage? Do you really want to live every day as if fighting a war? Do you really want to be anxious every day, afraid of being laid off? If you do not want these things, do not be pushed along by social ideology. Refusal is an ability.

Social ideology is terrifying. If one is slightly careless, it can destroy a person’s whole life. In the torrent, a person must repeatedly confirm: what exactly do I truly want?

On the Graduate Entrance Exam

Many people take the graduate entrance exam not because they truly want to keep digging deeper into a field they like, but because they hope to first get into graduate school and then use that to find a job. At least in computer science, this is roughly what I have seen. As for other majors, I do not know enough to judge casually.

In fact, if a computer science undergraduate truly wants to find a decent job by graduation, their undergraduate years will certainly be quite substantial. As long as the direction is clear and the investment is enough, they will probably end with a pretty good offer.

Graduate entrance exam scores have already come out, and the next group of candidates has begun preparing. Thinking of many students starting to prepare from March or April, I was actually touched quite a bit and thought a lot. I could not help asking myself: is all of this really worth it?

Take 11408, the common subject combination for computer science graduate exams, as an example. Preparation often requires relearning advanced mathematics, linear algebra, and probability theory from the beginning. Many people did not truly understand these subjects in their first year of university, so preparing for the exam is almost equivalent to learning them again. Studying nearly 8 hours a day, mostly listening to lectures, reading books, and doing problems, requires considerable time and energy.

After mathematics, there is the huge amount of knowledge in 408, including computer networks, computer organization, operating systems, and data structures.

In addition, there are politics and English. These two subjects are different from the previous ones and require more patience and long-term accumulation, since they are largely humanities-oriented and rely more on memory and understanding.

Note that I am saying the amount of knowledge is large, not that they are terribly difficult. Basically, as long as a person maintains more than 8 hours of high-quality study every day for 9 months, scoring 380 is not a low-probability event. Of course, many people in reality cannot reach this result, because the exam is affected by too many factors: self-discipline, efficiency, mental state, information channels. A problem in any link can affect the final score.

Graduate entrance exam questions are also different from gaokao questions. The gaokao focuses more on selection and often designs routine problems and traps; graduate exam questions usually have fewer twists and turns, mainly testing understanding and application of knowledge. As long as you truly understand the required knowledge, your score generally will not be too bad.

But the problem is that the risk of the graduate entrance exam is not low. Taking the computer science major at my university as an example, about 60% of students choose to take the exam. In the end, roughly 15% get in, and among those who get in, about 60% go to our own university. In other words, in a major of 200 people, about 120 choose the exam, about 30 get in, and around 20 of them still go to our own university.

This risk is even higher than opening a 100x leveraged all-in position in the cryptocurrency market.

The Challenges of the Exam

There are many challenges during exam preparation. The first is schedule.

Many candidates sleep at 2 a.m., wake up at 10:30 a.m., and then eat brunch, meaning breakfast plus lunch. Around 2 p.m. they start reviewing, study for a while, look at their phone for a while, read for a while and get sleepy, then pick up the phone again. The afternoon passes quickly. After dinner they continue scrolling for a while, study until after 10 p.m., and when calculated, the truly effective study time in one day may only be 3 to 4 hours.

At that pace, even the first round of review for one mathematics subject may not be finished by summer vacation, let alone 408 afterward.

Of course, there are also people around me who can truly keep going for 9 months: starting study at 8 a.m. every day and leaving after 11 p.m. But such people are very rare.

The second challenge is psychological pressure.

Everyone preparing for the exam knows that it is almost an all-in bet. Once you fail, even if you then look for work, it is hard to have any advantage, because the knowledge reviewed for the exam does not help much in recruitment. Some students who fail the exam go to spring recruitment and are asked by HR right away whether they participated in the graduate entrance exam; if they did, they may be rejected directly. Our grade director also mentioned a similar situation: students who failed the exam almost did not get offers, because companies clearly did not want people who might try the exam again.

Therefore, the psychological pressure brought by going all in must not be underestimated.

What is even more painful is that when classmates around you gradually get internships or autumn recruitment offers, while you are still solving problems that have almost no real meaning beyond the exam, strong self-doubt can easily arise. Especially in September and October, this contrast is magnified and can even cause serious anxiety.

The main challenges are probably these two. As for courses, school affairs, and other issues, they are comparatively minor.

Risks and Returns

Some risks have already been mentioned: failing often means preparing for a second attempt. In the current environment, the probability of finding a good job after failing the exam is not high.

A second attempt brings new challenges. For example, you need to rent a room and a study space; these can at least be solved with money. What is truly harder to solve is the deep loneliness, more severe self-doubt, and the mental consumption of seeing no results for a long time. Some people compare this process to a phoenix being reborn from fire. Just hearing the metaphor is enough to feel the pain inside it.

So what return does the graduate entrance exam bring? An admission letter that lets you continue studying for three years.

But after receiving the admission letter, what then? If the school is ordinary, such as a regular first-tier university, the next three years may only be an extension of the undergraduate years. Taking my university as an example, first-year graduate students in the OS direction spend their days sitting at their desks studying the Linux kernel and reading ancient Linux source code. Only a minority can truly produce results. Most graduate students eventually still go into Java backend, C++ backend, civil service exams, public institution exams, or even switch to frontend.

In other words, graduate education at ordinary schools is often essentially an extension of undergraduate education. Before candidates actually get in, it is generally overestimated.

By ordinary schools, I include a small number of 211 and 985 universities, and most non-Double-First-Class schools. Even some highly ranked 211 and 985 universities have very backward graduate education. I once talked with a PhD student from Xidian University. He mentioned some situations: the school selects a few topics that look impressive and assigns them directly to graduate students who do not even understand Linux and only know 408. After muddling through one or two years, many still go into Java or take public sector exams.

At this point, you may already understand: the risks, costs, and returns of the graduate entrance exam are not proportional.

Then why do many candidates still go all in before truly getting admitted?

The answer is simple: information asymmetry.

If you are an undergraduate planning to take the exam, you should seriously talk with graduate students at your target school who are close to your level. You can also actively seek information on various platforms and try to break this information gap as much as possible.

Many graduate students have said one sentence: graduate students are happy only at the moment they receive the admission letter. Because after getting in, they gradually see the truth, but they may not systematically summarize it, and may not clearly transmit that gap to later exam-preparing groups. So the kind of all-in described above keeps happening year after year.

In investing, all in is always an extremely dangerous strategy, unless you have very high certainty.

What I Did in November

November 2024 was definitely the most painful month of my year. In the ysyx SoC part, I was stuck for an entire month.

I did quite a few things that month: learned a bit of Rust, refactored npc, refactored the AXI4-Lite handshake mechanism, worked with FPGA, got npc running on it, and also gave a ysyx sharing session. Of course, the most important thing was still connecting npc to the SoC.

That integration process was full of twists and turns. To make npc connect smoothly, I refactored many modules. Fortunately, in the end, it finally connected.

The hardest part was the final period of checking handshake signals. At that time, several handshake signals did not match my expectations. I repeatedly checked the SoC’s MROM module, checked it many times, and printed some signals, but the signal passed to the master was still wrong.

For those few days, I even suspected that the SoC itself had a problem, suspected that computer science no longer existed, and even wondered whether the Trisolarans had launched a sophon into my computer specifically to interfere with my experiment.

Later, after thinking carefully, I wrote several pages describing the problem and sent them to the discussion group. Very soon, someone in the group reminded me: it might be a reset problem, and I could first let npc reset for 10 cycles.

The moment I heard that suggestion, I became excited. I could feel that there was a high probability the problem was exactly the missing reset.

Sure enough, after I added the reset, it finally worked.

This problem blocked me for a full 3 days. During those days, I spent almost more than 8 hours every day thinking about this handshake signal. I am truly grateful to the people in the group.

Although I was stuck for a long time, I did gain a lot. I learned many things beyond npc, and I felt more deeply the power of focus: as long as you keep staring at one problem and thinking about it, you will definitely gain something.

Rule Consciousness

Primary school, middle school, high school, and even university education all keep sending one subconscious message: as long as you are obedient, you will succeed.

If a primary school student does not do homework, their grades are “bad”; if a middle school student does not listen to the teacher, they will get into all kinds of trouble; if a high school student does not follow arrangements or attend evening self-study, they will be seen as a bad student, and may even be predicted to fail the gaokao. By university, if a student is not obedient enough, they may lose points in comprehensive evaluations. And students who cannot get scholarships are very likely to receive little approval from teachers.

In fact, “obedience” has always been an attribute society wants.

The stable output of society does not rely only on science and technology. It also depends on the stability brought by obedience. For a huge system, stability is often placed above everything else.

In a system that is operating normally, if someone actively proposes improvements to managers, they will probably not be encouraged immediately. Instead, they may first be criticized or rejected. The manager’s first reaction is often: you have ideas, which means you are not obedient. Whether the suggestion itself has value becomes something to consider only later.

This long-term subconscious training has seriously harmed the way Chinese people, especially students, think, as well as their ability to try to succeed.

Even now, in university, I still see many students doing completely meaningless things for only one reason: the counselor told them to.

Also, a person needs to find their real teachers. I have never believed that the teachers in university are naturally my real teachers. During working hours, they teach some theoretical knowledge, and that certainly has value. But beyond that, many of the things they say do not seem very nutritious to me, and sometimes even feel absurd.

How can a group of people who have never truly entered society for work, and who have stayed in school for half their lives, reliably give advice about society and work?

Real teachers are everywhere. Classmates around you, creators on Bilibili, YouTube creators, strangers on forums: as long as someone says something valuable and makes you suddenly understand, that person can become your teacher.

A teacher is not an identity. A teacher is someone who enlightens you.

Accepting Failure

Failure is actually an attempt launched toward success. Unfortunately, the word “failure” itself often carries a negative meaning, which makes people instinctively want to avoid it.

But failure has value.

A few days ago, because I had dug too large a hole for myself earlier, I struggled for a long time in one project. In the end, I gathered the courage to refactor the entire project. But the project was too complex, and during the refactor I even began to doubt my own judgment. Halfway through, I rolled the version back, because that refactor had failed.

It was a failed attempt, but I do not think it produced nothing.

After rolling back to the old version, I continued refactoring and launched a second charge. I went through many small failures along the way, but by today the project has finally made great progress and is almost close to success.

In fact, people should be willing to face failure. If everything you do can succeed on the first try, that only means what you are doing is too simple.

Educational Philosophy

Many people remain mediocre and cannot live the life they truly like. Apart from objective conditions, one important reason is that they are unwilling to fail, or rarely experience truly valuable failure.

Suppose a person’s probability of success on each attempt is only 1%. After 100 attempts, the probability of succeeding at least once is about 63.4%. This does not even count the improvement brought by reviewing failures and correcting strategy. If every failure is followed by serious reflection, perhaps after 20 attempts the person may already have a high probability of approaching success.

There are many university courses. Most of them are indeed practical and help improve knowledge, such as advanced mathematics, college physics, and introductory computer courses. But few courses truly teach students how to summarize experience from failure, and few people systematically teach a methodology for “how to succeed.”

I think methodology is very important. Through abstraction and refinement, it forms a set of principles for doing things, helping people approach their goals faster.

Notice that I said approach goals faster, not succeed faster. Doing things must follow the nature of the matter. For example, if you want to marry someone, the key is not to rack your brains over getting a marriage certificate, but to get closer and build affection. Similarly, if you want to accomplish something, you should apply force toward the core of the matter, instead of staring only at the final result.

The mindset of being overly cautious and always trying to avoid failure is, in essence, anti-educational.

Everyone needs to move toward success through failure.

Why Learn English

I often tell people around me: instead of learning 10 computer languages, it is better to truly master one human language, such as English or Japanese.

Today, programming languages seem to have become the new skill everyone chases. But if you look further ahead, you will find that mastering a human language is often more valuable than knowing a programming language.

If you are a programmer, you should understand that what truly matters is not how many programming languages you know, but whether you have solid programming thinking. Similarly, for people living in a non-English environment, the benefits of learning English well go far beyond “knowing one more language.” It affects access to knowledge, education for the next generation, your view of the future, and even your judgment and decisions in major events.

As technology develops rapidly, programming languages iterate very quickly. Some languages that were once popular, such as Pascal, Perl, PHP, and Objective-C, now have a much smaller scope of use in industry, and some have gradually been replaced by newer languages. Even the still important C language is seeing its market share change in certain fields.

If someone invests a large amount of time in one specific programming language, and that language gradually fades from the mainstream in the future, the time and energy already invested may depreciate significantly.

By contrast, English is a language that will not easily become obsolete. As a global bridge language, English occupies an extremely important position in international communication, scientific research, technical documentation, business cooperation, and education. Mastering English is an ability that will not quickly disappear with changes in technical trends.

As long as you keep using and improving your English, its value will keep accumulating, and one could even say it grows with compound interest.

Instead of spending a large amount of time learning a programming language that may be eliminated, it is better to use part of that time to improve English. After mastering English, you can access frontier information and high-quality resources more freely. In today’s age of information explosion, this is especially important.

Of course, mastering certain programming languages can indeed help you enter specific fields and find specific jobs. But the role of English is not limited to one industry. It is a key to many top companies. Whether you want to do software development, product management, marketing, data analysis, or any work that requires collaboration with international teams, fluent English may become a key advantage.

Especially in today’s globalized world, many companies want to hire employees who can communicate in English. This means you can deal with international clients, participate in global projects, and even open overseas markets. English ability not only makes career choices more diverse, but also significantly raises your salary ceiling.

More importantly, English ability is cumulative. Over time, the more you use it, the stronger it becomes; the stronger it becomes, the more information, opportunities, and worlds it opens to you.

Learning a new programming language, whether Python, Java, Rust, or Go, may help you find a job in the short term or solve certain technical problems. But it is more like a stage-specific transaction: once the technical trend changes, the original technical reserve may lose value.

English learning is different. It is more like a long-term appreciating investment. Whether in school, work, or daily life, English can continue to bring you immeasurable returns.

How to Learn English

My understanding is simple: listen, and keep listening.

More precisely, keep doing comprehensible input.

A language expert once expressed a similar view: as long as a person receives a large amount of comprehensible listening input, they will gradually learn the language, including listening, speaking, reading, and other abilities.

I agree with this view because I have already felt it personally.

In a place without an English environment, it is certainly difficult for a person to master English. But after a large amount of listening input accumulates, the person slowly builds familiarity with English. When that familiarity accumulates to a certain point, a qualitative change happens.

I usually do not speak English much. I mainly listen and read. But some time ago, I tried to express my views in English, and the result exceeded my expectations. Although my word choice was not accurate enough, I was already able to express meaningful ideas. This shocked me, because before that I had always thought it would be very difficult for me to express complex thoughts in English.

So the summary is one sentence: listen, listen a lot; but do not listen blindly. Do comprehensible input.

A Regular Schedule Is Not the Same as a Healthy Schedule

My schedule has always been relatively regular. Previously, my study time was mostly concentrated from noon to before 11:30 p.m. Before 12 p.m., it was basically a relatively relaxed period: sleeping, attending classes, or doing things that did not consume much energy. After eating at 11:30, I would officially enter study mode.

This arrangement was indeed efficient in the short term, but after a long time, the body slowly began to send signals of fatigue.

Later I realized that a regular schedule is not necessarily a healthy schedule. Long-term late waking, late studying, and late sleeping may be stable in daily rhythm, but they are not necessarily friendly to the brain and body.

Therefore, I plan to adjust to a healthier rhythm: get up at 7:30 every day, start studying at 9, eat at 11:40, then rest until 2:30 in the afternoon, and continue studying until 10:30 at night. In any case, I must leave the lab before 10:30 to avoid keeping my body in a state of excessive fatigue for a long time.

Calculated this way, the total study time is about 10 hours and 40 minutes, a little more than the previous 10 hours and 30 minutes. Of course, this does not mean I must always maintain such long high-intensity study. If the truly effective study time reaches about 9 hours, that is already very good.

Energy Management Matters More Than Time Management

Some people seem to have a lot of time, but a large part of it is actually “garbage time.” During those hours, they do not have enough energy. Their minds are dull and their attention is scattered. Even if they sit there, it is difficult to produce high-quality learning results.

So I increasingly feel that energy management is more important than simple time management. Without energy, learning efficiency will definitely drop.

I have also seen many bloggers share methods of energy management, such as the Pomodoro technique, ensuring sleep, eating snacks to replenish energy, and so on. These certainly make some sense, but for me, their effect is not as great as imagined.

My own experience is that instead of constantly reminding myself “what I should do,” it is better to first clarify “what I should not do.” The most important things are: do not stay up late scrolling on the phone, do not frequently scroll on the phone during the day, and do not spend large amounts of time on games and short videos. They are almost double killers of energy and time.

As long as these sources of consumption are avoided, energy naturally becomes much more abundant. Combined with a long enough block of usable time, the efficiency of study and work improves significantly.

Therefore, I have almost quit scrolling on my phone before sleep. Human self-control is limited, so it is best not to bring the phone to bed. I can directly wear earphones and listen to an English podcast, then fall asleep before 11:40.

The same is true during the day. If I can avoid looking at the phone, I should avoid it as much as possible, especially during rest. Many people think, “I am resting now, so looking at my phone will help me relax.” But in reality, this is often a misunderstanding. Scrolling on the phone does not truly restore energy. Instead, it makes people more and more tired.

What If a Plan Cannot Move Forward?

Sometimes, a task can block a person for a long time, even several days.

This happens often not because the person is not working hard, but because the task itself has not been refined enough. I once tried breaking a complex task into 10 smaller tasks, and progress became much smoother.

This is a very valuable experience. It made me realize that solving problems and pushing tasks forward actually require a reusable paradigm.

Musk once spent more than 100,000 dollars hiring a lecturer to train employees. The core idea that lecturer kept emphasizing could be summarized in four Chinese characters: refine the task.