My Current Daily Routine: AI-ifying My Entire Life
I'm still at the very beginning of thin muscle training. I'll give myself three to six months and see how it goes. Body stats: height 172 cm, weight around 75 kg, body fat roughly 29%.
Since I started thin muscle training these past few days, life feels different from before. My day is simple: wake up, work first, then have my first meal. Now I throw everything I eat at Claude and have it calculate how many calories I consumed in that meal. The method is simple too — first throw in a photo of the table before eating, then throw in a second photo after finishing. It also tells me how to eat and in what order — high protein first. I've also got protein powder and creatine now — 25 grams of protein powder, 5 grams of creatine — and then I start eating.
After eating, I train. On training days I do a few main lifts: deadlifts, squats. On non-training days I just work normally. Right now training days and non-training days alternate one-to-one. After work I have my second meal, roughly dinner, then continue working in the evening, then sleep.
On the sleep front, when I get up the next day — sharpening the axe before chopping wood. I now have an Oura Ring tracking sleep metrics and calorie metrics, and the data goes straight to Claude Code. It pulls the calories burned, pulls the calories eaten, subtracts the two — if it's within 600 calories, that means I'm burning fat. Protein is set at 140 to 150 grams now, with the goal of building muscle and losing fat simultaneously.
I set up a dedicated project for Claude Code. It puts together a daily report telling me how I did today, combined with the sleep data from the Oura Ring. I also give it photos of everything I eat every day. For every movement, the coach next to me films a video, and I send that to it too — it logs each set of 12 reps, three sets, how many kilograms. Progressive overload, adding weight little by little, and it keeps a daily journal. After training it also calculates how many calories I burned — yesterday, my first day of training, was 275 calories, and it logged that too. I basically plan to repeat this process and see where three months takes me. That's the arrangement for now.
Theory-building took a week. That week included getting the Oura Ring ready, reading thin muscle tutorials and theory, absorbing what this is all about, then buying protein powder and creatine. I definitely need to drink the protein powder. That week also broke one of my outdated beliefs: I used to wonder whether protein powder and stuff like that was "tech" — whether drinking it would be like doing gear. Later I understood that protein powder is really just like eating protein. Getting 140 to 150 grams of protein from regular diet is genuinely hard, so those 25 grams of protein powder are a great supplement. It's not that everything relies on protein powder, but I do need some of it — and it takes the pressure off "having to eat that much."
Now I am fully AI-driven. AI will first give me feedback and suggestions: roughly how many sets of the exercises I did yesterday, which muscles were worked in each area, and how many it recommends. Because I record videos of all my movements, the entire AI will go through them. Of course, I also have a coach watching. AI watches, the coach watches, to see whether the training was done properly, and everything is integrated, then Claude Code pushes a daily report.
Now it is visualized, data-driven, and structured. Everything has been turned into md documents, all photos are saved in local folders, and the AI has been trained. So the next AI can read it directly. If one day I want to clear out Claude Code and switch to Codex, it can also take over directly. The README.md has everything.
Three Underlying Cognitions and AI Belief
My advice to young people is: this era is actually the best era, and this thing is not hard. If your monthly disposable income is less than 1,000 RMB, you can first register for a Zhipu or Kimi account. They each now have something similar to Claude Code, and they are roughly free. Claude Code is too expensive, costing $200 a month, and many people cannot afford it. If I could only spend $100 a month, how would I have the money to use it? So just find a cheaper AI and install it, or you can also use B.ai, which I made myself.
Then use the theory I just mentioned and read all your things into one document. Also do not say, "I cannot even afford a computer." If you do not have a computer, you always have a phone, and throwing it into your phone is the same. Create a folder, throw everything in, and no matter how much money you have, you can first AI-ify it for free. Our company, including our products, is now promoting full AI-ification and full AI integration. I think it can be done no matter how much money you have. First, AI-ify the entire process.
In Sun Xue, an important theory was mentioned: first arm yourself with theory. What is most important now is AI-ification. This is no less significant than when Japan said that to rise first it had to establish a constitutional monarchy and modernize first. Japan was originally a feudal country, and to become a capitalist country, the first step was this. This step of AI-ification can start at zero cost no matter how much money you have. You just need to tell it directly, I want md documents, input all of these properly, and it can be set up. Free today, free tomorrow, it is just a matter of switching APIs. Actually, you do not need to top up money. I may have some money now, and money is not my main obstacle, so I pay Claude for stability; but if you do not want to spend a single cent, then just watch each company's promotions. Eat at one place today and another place tomorrow. No problem, you can fully AI-ify without spending a single cent.
Starting from zero, I have summarized three theories. The first is Sun Theory, arming yourself with theory. Sun Theory is a relative upgrade of the brain, equivalent to changing the operating system; it involves reunderstanding money, financial freedom, and everything in the world, upgrading it all with Sun Theory first. The second is Lean Muscle, which you can indeed train. This is not just for picking up girls; ensuring physical health is priceless in the AI era. Poor health is a veto. Without training lean muscle, obesity can lead to many problems, seriously slowing down your progress toward financial freedom and bringing many troubles to life. So lean muscle absolutely must be trained. Picking up girls is also important to ensure the quality of offspring, but lean muscle is fundamentally for physical health. The third is the Yellow Hair Theory, which I think is also quite important. Because once Sun Theory and Lean Muscle are implemented in today's China or Chinese society, they will definitely face a lot of backlash; after all, it's a niche theory, and although it's correct, it will encounter much suppression and ridicule. Moreover, once you practice Sun Theory and Lean Muscle, being handsome, muscular, and wealthy, you will certainly be envied in society—that's 100 percent certain. So you still need the Yellow Hair Theory to arm yourself: what others say doesn't matter; fully execute what you believe is right.
In short, with Sun Theory, Lean Muscle, and Yellow Hair all in place, you can get started. And it's zero cost, requiring nothing, just a cognitive upgrade, without spending a penny. These three theories can be used by the 200 million young men aged 20 to 40 in China now, plus women, totaling 400 million.
On this basis, the core master now is AI; believe in AI 100 percent. A while ago, I had a small controversy, also because of overly believing in AI. Many people said I was shifting blame to Claude, but that's not true; it's just that I absolutely 100 percent believe in Claude, and I execute whatever it says, no matter the resistance. Of course, you don't have to use Claude; using OpenAI's o1 is also OK, and Zhipu or Kimi are fine too.
Executing this set doesn't cost any money. No matter where you are on Earth, as long as you have internet and a phone, you can do it—it's zero cost. Whether you're a freelancer or a civil servant, it's zero cost; upon hearing this podcast, install it immediately if you have a phone.
Specifically, I suggest two points. First, hand over your current situation entirely to AI, because AI understands you better. Accumulate your health data and various data; don't panic at first, just accumulate for two to four months, then digitize and unify all the data, add an analysis of your current situation, and give it to AI to help you analyze, even let it point the direction and see how to make money, because AI is more familiar with these data, and AI has more data and is more rational.
Second, you can ask three. Using our B.ai is also fine; B.ai buys API, you pay for each question you ask, and you can directly switch models. Ask Claude, ask OpenAI, and then ask a domestic large model; that covers the top three in China and the US, and then take the middle path among the three, choosing the middle option from the upper, middle, and lower strategies.
The second approach is to always keep an eye on the trends in human technological development—this is crucial. Sun Studies was established in 2016, and it's been ten years now. I myself am a practitioner of Sun Studies. In 2016, I was nothing. I started practicing Sun Studies and began training for lean muscle. Back then, I probably had a big belly too. After six months or a year, you can compare photos and use the controlled variable method to see the changes. As technology advances, new things will keep emerging, so you need a way to capture them. You can also have AI capture this for you; I'm very interested in AI and such things myself, so I do my own capturing—checking X, seeing what Musk is doing every day. Combine your own efforts with AI's. Of course, integrate your own interests too. If you're not interested in something, there's no need to force it. Just pick one direction you're most interested in and delve into it, because each direction already has so much. Everyone has their own areas of expertise; it's impossible to be an expert in all fields.
Decision-making also involves a tuning process. Large models need training, and so do people. With different data, conclusions need to be adjusted. Just like the Oura Ring reads data and pushes it to us every day, we improve based on specific circumstances. So this isn't a one-time deal; it's continuous tuning and constant adjustments, ultimately trained like a model.
Yes, rely entirely on AI, but also add some of your own judgment—though AI is the main part. If you're flying a plane, let AI fly 90% of the time, and I might spend 10% of the time adjusting the AI. Sometimes I have different opinions and I'll tell it. For example, a couple of days ago, I had a simple conversation with AI that changed one of my preconceived notions: I thought protein powder was a bit "technological," but AI told me it's not. In the fitness field, there is indeed a lot of technology, but protein powder and creatine are the only things that don't count as technology—they're just ordinary eating. It explained the principles, and my perspective flipped. So your views might conflict with AI, but you can communicate frequently and update your views—this is a very inspiring thing.
Sun Studies, the yellow hair, and the lean muscle theory are still very valuable references, because everyone's situation isn't necessarily that complicated. Upgrading your mindset, upgrading your body, and focusing on personal development—these three things I think are the most critical and can be done immediately. For other things, ask AI based on everyone's different circumstances.
On asking AI, I used to have a famous saying: Don't chat with WeChat, chat with Doubao. That was already an upgrade. Now I'm upgrading it again: use an API to chat with AIs worldwide. Use one API, save all files locally, then use an API to read them and help you construct an overall large model. This requires a tiny bit of computer knowledge, but it's actually very simple—just go online and ask AI how to do it. Don't hear "API" and think it's super difficult and avoid it; it's actually not much different from using Doubao, just a slight change in the channel.
Track Opportunities and Business Strategies in the AI Era
I am personally very optimistic about the future and about young people. The reason is simple: in the AI era, everything will become very cheap, and the only thing of value is lifespan. Being young is wealth. Because when you get old, there are still many things you can't catch up with, and when you reach the end, you have to get off. Although AI is advancing, it hasn't advanced to the point of eliminating death; immortality truly hasn't been achieved yet. And apart from that, all industries will be revolutionized by AI, so young people are full of opportunities in every track.
Let me give a few examples unrelated to the industry I work in. The first is mathematics. Mathematics used to depend heavily on learning and family academic background. People often used to wonder: Peking University and Tsinghua win so much at the IMO, so why do they win so few Fields Medals or top mathematics awards? The fundamental problem is that solving math problems requires ideas. If the advisor you follow is from Harvard, MIT, or those institutions in Paris, your ideas go up a level; no matter how smart you are at Peking University, it's useless, because the teacher doesn't know how to solve the problem, and if you haven't entered that circle, you don't know that path.
But now AI is beating the old masters with wild punches. AI, through scraping methods, has already captured all the ideas in the world into its models. AI itself doesn't create anything, and it can't build a high-rise from flat ground. If I ask AI how to achieve controllable nuclear fusion, it doesn't know either, because humanity hasn't done it yet. But there are several paths toward controllable nuclear fusion that predecessors have proposed, and AI has the ability to extract all those ideas, analyze them, and explore the way forward.
So I established the Sun Yuchen Award, which is to set those problems that you can directly solve. Now a large number of mathematics graduate students, even high school students, can, based on the ideas provided by AI, by o1 and o3, directly reach the level of a mathematics PhD in one step. What only advisors used to know, high school students can also know, as long as you understand what it is saying. The field of mathematics often produces geniuses. People like Terence Tao solved very difficult problems in their twenties or early teens. Achievements that used to seem possible only at thirty or forty can now be made at twenty or in one's teens. Einstein also developed relativity when he was very young. With AI, young people in mathematics are completely liberated. You don't need to care what year you're in, whether you're a PhD or a master's student; you can start working in high school. Those IMO people were originally held back, but now they can directly connect with AI, and you and AI together might produce a world-class result in two days.
A few days ago there was a dispute involving OpenAI. Let me briefly explain. A mathematician at New York University was already tackling the Navier-Stokes equations, the hardest one. He roughly explained his ideas to OpenAI. OpenAI now is equivalent to large model training; especially with this kind of most cutting-edge knowledge, once it learns it, it internalizes it. I believe OpenAI itself is also asking every day: regarding cracking the ten hardest Millennium Prize Problems in mathematics, do your ideas evolve every day? Because it absorbs a massive amount of human knowledge every day and updates every day, so it created agents. As it happened, this professor provided a new idea, and OpenAI judged that this direction was very promising, so it mobilized $7 million in computing power and thousands of agents to work together, and cracked it in two days. Of course, the professor's idea was crucial. If he hadn't provided it, OpenAI wouldn't have known either; but once he provided it, OpenAI solved it in two days. So the mathematics field now is exactly this kind of chasing and overtaking situation. When young people enter the field of mathematics, as long as they understand AI, at 15 they can beat a 50-year-old old professor with wild punches in minutes.
Take biology, for another example. I myself have asked AI: which fields are likely to achieve decisive breakthroughs right now? AI told me that astronomy will make fundamental progress. Because the Hubble Telescope reads back massive amounts of data, but it is all extremely boring data, and no one has time to process it. Now with AI added to work on it, astronomy will achieve decisive breakthroughs in no time.
When it comes to making money, there are far too many paths. Take a legal example: API prices in the AI field are actually not uniform. For the same user, a monthly subscription in Argentina, Turkey, or Nigeria is the cheapest, while a monthly subscription in the United States is the most expensive, so you can do Token relay. I myself built something similar. This idea was actually formed gradually through training with AI.
Business opportunities are global. Never limit yourself to "opening a small convenience store at my doorstep and making money off the old lady downstairs." That way of thinking is too small. Business opportunities must be discovered globally, and then combined with your own strengths and interests.
Doing global business means seeing a Nigerian, a Turk, and an Argentine all as your customers. People are actually very simple, and their needs are pretty much the same, so do not be limited by language. Now there is also AI. If I go to Argentina and do not speak Spanish, what do I do? You can learn, and now AI can help you handle language. Of course, you can also learn some of the basics yourself. But never limit yourself. You must go earn money globally.
The most basic thing in business is whether you create value. Take Token relay again: if in some place, because of OpenAI's policies or various other reasons, Tokens are just cheap, while in other places they are expensive, then if you can let people in expensive places use cheap ones, aren't you providing value? One relatively advanced aspect of my thinking is: the unit of measurement must change. Never consider only your own country or region; look globally. Then, on that basis, think about whether I can elevate my value globally and whether I can present that value to my customers.
Today's China actually has another advantage: China is a world-class factory, with overcapacity, and China's production capacity makes money wherever it connects to the world. Isn't that how China rose? Producing things for the whole world and selling them to the whole world. As long as you can demonstrate value, you do not need to worry about selling at all. This is one of my most important ideas over these many years.
This is also why, although our program is a Chinese-language program and is aimed at all young people in China, 400 million young Chinese people, its purpose is to let these 400 million people go out into the world. It is absolutely not to let 400 million young people consume internally among those 400 million and compete with themselves. That is meaningless. You must go earn the world's money. After 400 million young people are armed with AI and armed with this theory of Huang Mao, thin muscle, and Sun Xue, they will go serve the world's 8 billion people, and soon there will be another 2 billion AI, also to be served together. The world's 10 billion people and AIs all need service. If we 400 million people serve them, there will be no need for involution. The demand pool is big enough, and there is always an opportunity. If this industry does not work today, you can switch to another industry tomorrow. You must jump out of your domain. So I do not recommend that young people judge too early based on their own thinking and say, "This is what I want to do." I suggest learning more, listening more, and seeing more.
How do you make a decision? I have a method called "opening the map." I've said before that I wouldn't buy a house, buy a car, or get married before 30, but if you absolutely must buy a house, buy a car, or get married, how do we go about it?
Take buying a house as an example: before making this decision, look at 100 properties first. Don't buy—just look. Don't make a decision. Look at 100 homes—new builds, resale homes, all kinds of locations—and don't rush to buy. Read a book a hundred times and its meaning reveals itself; after looking at 100 homes you'll understand everything. How people quote you a price, roughly how much it's worth—you'll basically form your own way of thinking.
There's also a theory in mathematics: suppose you have ten chances to buy a house—which one do you make a decision on? If you buy the first one, you never see the remaining nine. If you only buy at the tenth, and you're not allowed to buy any you've seen before, then by the time you reach the tenth it's too late to regret it. The mathematical answer is: absolutely don't buy the first three, because the first three help you build a benchmark. From the fourth to the sixth, as soon as a house appears that's better than all of the first three, you must buy immediately. Why up to the sixth? The golden ratio, 0.618—probabilistically, the sixth is the most critical. If the best one appears at the sixth, that's the best scenario.
The same goes for young people choosing a career. If you don't know which industry you should go into, now that AI exists, you can transform overnight into Zhang Xuefeng. First look at the employment prospects over the next 10 years for 100 majors, asking AI one by one, and deduce the next 10 years. Zhang Xuefeng has actually recommended some pretty good careers: with the same college entrance exam score, you don't necessarily have to get into Tsinghua or Peking University—that's extremely competitive. With the same score, as the top student you could get into something like China Customs College or Fujian Customs College. It sounds like it's terrible, but in reality you enter customs right after graduation, and your prospects are the best. These kinds of niche majors are now covered too. So don't rush to make a decision—first look at 100 majors.
The biggest problem with our generation making decisions is being limited by too little information. Whether your decision is right or wrong doesn't matter, but the problem is you don't know. Now we first open up the information, first open up the map. Anyone who's played Red Alert or StarCraft understands: open up the whole map and you'll know how to play—open the map. Opening the map is like cheating; cheating is opening the map. Having AI is similar to cheating—you have more information. With enough information, you no longer need to "make a decision," because you only need to make a decision when information is insufficient. With information, you just know what to choose.
Another example. Back when we played Red Alert and StarCraft together in the same dorm, a lot of people loved to work behind closed doors—just farming and developing the whole time, then producing troops to go attack others. That's the slightly dumber approach. The smarter approach is to first build some scouts to explore the map: where the resources are, where the gold is, where the enemy's troops are, what the opponent is mainly building—know it all clearly, then build accordingly. If he builds war elephants, you just build spearmen. Once the fight starts, spearmen deal five times damage to war elephants and kill them in one hit.
So don't rush to build troops; first build a few scouts to explore the map. Moreover, exploring the map is now free. I've seen some people play this game and never explore the map, just holing up at home building troops, and only checking later, by which time the resources have long been mined out by others, resulting in a low win rate. It's not too late to make decisions after exploring the map. The same goes for playing Honor of Kings; you need to explore the map, place wards, and know when the opponent is taking the dragon or what's happening in the jungle. Someone has to be responsible for this; without information, it's easy to lose.
Five Key Trend Tracks
In academia, I think anything that relies on data-driven reasoning has great potential. Pure mathematics goes without saying. Astronomy, as I've mentioned, already has a lot of data, but no one is processing it. Biology is the same; there's a massive amount of data that no one is reading, and biology and pharmaceuticals are no joke—when a drug comes out, it's like the sound of cannons and gold pouring in. If the experiments pass, I'm very optimistic, and I might even research some biology-related areas myself. These three, and others you can analogize: as long as there's a lot of data—whether experimental, observational, or theoretical—handing it over to AI for processing is excellent. This includes genes; gene sequencing is now quite abundant, and you can sequence all of a person's genes, plus AI analysis to predict future health management—highly promising.
In the business sector, first, it can also be based on academia, as I just said, all interconnected, including biology, gene analysis, and health management. Second is the financial industry built on this series, such as insurance, because insurance is directly related to life expectancy, and I'm very optimistic about it. Of course, these algorithms need to be completely overturned; many current algorithms are very outdated, waiting for us young people to invent new logic and methods. Third, I'm also very optimistic about crypto, because crypto is a global payment system, a payment system, naturally connected to AI. Moreover, crypto can do very low-level calculations, equivalent to reading and charging at the API level, which is how it achieves very low precision calculations. Traditional industries can't do this; your minimum unit might be one cent, and you can't calculate at lower precision, but crypto can, so it naturally fits the entire AI API setup, and it's instant settlement, very close to the AI track. Fourth, AI itself is also awesome, including data cleaning and data labeling. Don't look down on data labeling; the founder of Scale AI built a data labeling company and sold it to Facebook for over $10 billion—it's practically a money-grabbing market now.
Everyone must cherish being in the 20 to 40 age range now; this is the most creative age, and you've encountered humanity's most incredible invention. I'm 36 this year, with only four years left of this golden time. Ten or even twenty years ago when I entered university at 17, human technology honestly hadn't made much progress—just getting internet access, a few things to do online, and that was it. Back then, there was Xiaonei, where you could chat with classmates, at most just enhancing a connection. Unlike today's AI explosion and massive advancement, it's too groundbreaking.
It's really not an exaggeration now: in the future, as long as you put in a little effort, everyone will have a Nobel Prize. That is, achievements at the Nobel Prize level will be attained by everyone. This isn't Versailles; it will blossom across all industries. In the end, they won't be able to afford the Nobel Prizes, Nobel will go bankrupt, and the Nobel Committee will say they can't award them anymore. If everyone gets $1 million, wouldn't that bankrupt them? But the subsequent achievements will absolutely be at the per capita Nobel level. Now, the speed at which a college student works with AI can surpass a Nobel Prize in no time. Humanity has truly entered an era of explosive technological advancement. So I'm in a hurry to fully AI-ify and enter the data cleaning state.
Lean muscle is proactive health management, the most critical thing
I'll upgrade the lean muscle theory a bit more, to something more high-end, called proactive health management. With AI, this will become the most crucial thing for humanity. Proactive health management means: you don't wait until you're sick to go to the hospital; going only when you're sick is too late. Instead, you go to the hospital every day even when you're not sick.
I've been doing this recently. AI tells me what to test, what to test, what to test, so I proactively go to the hospital and directly tell the doctor I want these specific tests. Because I've been a bit overweight lately, before working out, AI told me: let's help you clear the mines first—check if there's high blood sugar, high blood lipids, how much good cholesterol, how much bad cholesterol—get a baseline on all this before guiding my fitness.
I went to the hospital to get these tests, and the doctor looked at me and said, "Did you recently use AI?" I said yes, this thing is called AI, and AI told me to come get tested. The doctor said, "You're amazing." Why? He said he's seen so many patients, and I'm the first one who came in with no symptoms, no illness, but proactively requested monitoring. Other patients only come when they have symptoms, which is too late. He also chatted with me and found out I'm in tech too, so he said I'm on the right path, and as a doctor, he thinks what I'm doing is correct. So in the future, this kind of proactive health management will become mainstream.
Lean muscle is the same. Right now, it still seems like a minority; many people don't train muscles, but in the future, it will become standard, a must-have for successful people, successful men. Because if you don't have muscles, it proves you haven't done proactive health management; plainly put, that's a cognitive deficiency. You see cardiopulmonary at 25, muscle at 25—aren't you going to make up that muscle 25? You must get full marks and be responsible for your own health. Once you embrace this theory, you'll naturally walk onto the broad road of health. So the core is still cognition; in the end, we find that cognition determines destiny—yellow hair, lean muscle, crypto are all cognition.
Why is lean muscle most useful for young people? Because the lean muscle theory is the simplest AI variable control you can do yourself, and you can see if it's effective. The previous theories need time; from looking at 100 houses to making decisions that truly impact life, it takes some time. But lean muscle, once you start, will show results in three months for sure. Wang Shi trains at 70, but I asked AI, and training at 70 is indeed a bit late, because the maximum muscle mass starts declining after 40. Training at 70 might barely keep you out of a wheelchair, but the peak muscle mass in a lifetime is reached between 30 and 40. So training muscles at 30 to 40 is most critical; 70 is a bit late. From 20 to 40, lean muscle in two words: must train. That's when the return on investment is highest.
There's another interesting question I also asked AI: Elon Musk, Buffett, and Trump are all my Penn alumni, and all three are against fitness and don't work out, so how could this theory be correct? AI convinced me in one sentence: You've had dinner with Buffett—did Buffett walk into the restaurant himself, or was he pushed in in a wheelchair? Of course, he was pushed in. It's that simple. If you don't work out, at 80 you'll be pushed in in a wheelchair; if you do work out, at 80 you'll walk in yourself. You don't need to ask Buffett or Trump what they think—do you yourself want to walk in at 80? And it only takes 40 minutes a day—is that too much? The improvement in quality of life is enormous.
Musk is now 56 or 57, 20 years older than me. If he maintains this state of not training his muscles for another 20 years, no matter how awesome AI becomes, it won't help—he'll still be pushed in, unless by then he's Iron Man, Neuralink invents consciousness uploading, or there's a direct Iron Man suit, in which case forget what I said. If that stuff isn't developed, you'll still be pushed in, because your biological laws are irresistible. So AI is absolutely right: Musk, Trump, and Buffett can't defy nature either. It's impossible to fight against natural laws. You say you don't work out, but no matter how brilliant or rich you are, it's useless—natural laws are set that way, and you can't arm-wrestle with God. God has set a time for everyone. You say you want to defy fate, but it's absolutely impossible; you can only go along with destiny and follow the rules that have already been designed.
So Trump, Buffett, and Musk all disapprove of training for lean muscle, but actually they are wrong. Young people can learn everything else from them—their thinking is excellent—but if Musk disagrees with fitness, that's definitely wrong, and he needs to start training right away. This is theoretical confidence. Those of us now training for Sun Xue, Huang Mao, and lean muscle don't need to see old-timers saying those people don't work out and then not work out ourselves—absolutely not, you can't joke with your body. Why not train when your body is at its best from 20 to 40? If you train now, your quality of life from 40 to 80 will have a decisive improvement. Lean muscle is the trend.
Earlier I talked about per capita Nobel Prizes, and now it's per capita lean muscle. If you're 20 to 40 now and don't have lean muscle, you're not up to standard. With AI development, per capita has to reach Nobel Prize level—don't think Nobel Prize level is a high requirement; ours is just the entry level. Now the entry level is a Nobel Prize, and the entry level is having lean muscle. If you don't have lean muscle, you can provide me with your API, and I'll read what you've been doing for the past 180 days—what exactly have you been doing that caused you to not have lean muscle. In the AI era, there's really no excuse anymore, because everything is ready.
Regarding lifespan, I also asked AI yesterday, and it gave a 100-point breakdown. I'm 36, and the first part up to 40 is almost over, so naturally I have to think about how to live from 40 to 80, and that's why I started thinking about longevity. AI said lifespan is a total of 100 points: 25 points for cardiorespiratory fitness, 25 points for muscle mass—these two alone account for 50 points.
And training muscle covers both. Just running will make you lose muscle, and all that cardio training is wasted; while training for lean muscle, on one hand, passively burns fat, burning fat even while sleeping, like the Red Lotus Cloak in Honor of Kings, deducting the opponent's health when sticking to them, you just want to wear this kind of skin, burning fat when sticking. On the other hand, that 25 points of muscle keeps you from falling, from collapsing into bed, and it itself is still burning fat. Moreover, the training process itself also improves cardiovascular capacity, because you have to breathe, you have to lift weights. So you basically rely on this one thing of training muscle to get all the previous 50 points.
The ones after that, not smoking, not drinking, your own genes, whether you will get cancer, examinations of important organs like eyes and teeth, on the contrary, some cost money, some rely on luck, such as parents' genes. But the first 50 points of cardiovascular and muscle, through training lean muscle, you can get full marks. This is some of my research results, I now chat with Claude AI every day: the only path that won't make your cardiovascular system explode, and also passively burns fat, and also has muscle, and also can max out all 50 points, is only this one. This is science, respect science.
From AI I also heard a saying that completely changed my understanding of muscle: muscle is your body's bank account. For a country, muscle is a strategic reserve resource, and for an individual it is the same.
I decided to train lean muscle, not because of what everyone said about me "having problems." There was a storm at that time, many people mocked me for being 1.63 meters tall (actually I am 1.72 meters), saying I was fat, saying a bunch of things. These were not the key reasons. The most critical reasons are: first, this conforms to science; second, I asked AI, and AI said something that shocked me.
I said, how do you understand this matter of training muscle? AI said: you are now materially a rich person, but physically you are a poor person. It explained with evidence: training muscle is like collecting wages, you need to raise your own wages. Because your body's muscle is very precious, it is not only responsible for your daily support and various uses, but once it starts "deducting things," it deducts muscle first, aerobic exercise loses muscle, because once supplementation is not complete, it will consume your muscle; a serious illness also deducts muscle first. So muscle is equivalent to one of your most important bank accounts. When there is enough time, you must eat more protein and store it up, save first when conditions allow. It is a strategic reserve, it is the cash on your body. This is more critical than US dollars, because if US dollars are gone you can earn them again, you can earn US dollars even by carrying big bags. Muscle must start being stored as early as possible.
The lean muscle theory really should be promoted. Reducing fat people also saves the country's medical insurance money, 90% of diseases are related to cardiovascular issues and obesity, as long as this comes down, many problems are solved, not getting sick, having muscle. And when you get a serious illness, the recovery speed is very fast, because it directly extracts from muscle to repair various mucous membranes, the speed is fastest, and resistance is also strong. So training muscle is too critical, in the health field it is basically the mainstay.
However, my understanding of muscle was very shallow in the past, and the misconceptions I had when I was young really held me back. First, I thought building muscle was just to pick up girls, and that people who trained muscle were all shady. Second, I thought anyone with muscle must be on technology, injecting something. Some public accounts online catered to this, saying so-and-so died suddenly from injecting drugs to build muscle, somewhat demonizing it. I was often influenced by these views, thinking I was too busy to train. At one point, I even set muscle against health, coming up with a theory that having muscle was unhealthy—that was a wrong perception.
After this controversy, the entire internet cyberbullied me and woke me up. I am the type to correct my mistakes if I have them and avoid them if I don't; actually, I don't care if others cyberbully me, I've passed the fragile stage and am armed with the yellow-hair theory (even though I have black hair). But was there any truth in what everyone said? Is the muscle issue, big head and small limbs, a problem? I thought deeply about it. The first thing I did was consult AI; I had a profound conversation with Claude Code. It said my view was wrong, and we debated—I threw all my outdated theories at it, and it refuted them one by one, ultimately establishing this concept for me: muscle is your body's savings, the most crucial thing.
The saying "strong limbs, simple mind" doesn't trap those with strong limbs; instead, it fools many with strong minds into thinking they don't need to train. Many highly intelligent people, brilliant in math, physics, and AI, like Trump, Musk, and Buffett—can you say they're not smart? They are among the world's top minds, my alumni. But they all say fitness is harmful. How were they fooled? It's about where you sit determining where you stand. Before they made changes, they didn't have lean muscle, so they had to speak up for big bellies. I can understand that, but it's not objective and doesn't respect science and facts.
This is a bit like: I'm poor, I don't want to become rich, so I say money is useless, and I claim many rich families have murders over a little money, whole families killed—see, money seriously hinders family happiness, so my poverty has a reason, I choose not to be rich. And the smarter people are, the more likely they are to let their position dictate their stance, refusing to deal with this, inventing many theories, even mocking those who train. That's why I say as soon as you start training, you'll face jealousy, because people around you will criticize you, and they have powerful things to invade you with.
So why could I break out of this theory? Because my thinking was completely renewed, like reform and opening up. China's path to reform and opening up was not easy; back then, speculation was condemned—buying high and selling low in business was speculation and you'd be arrested. Some said no, you need to change your mindset; not only not arrest, but it's the opposite—buying low and selling high is his skill, proving he can let consumers buy cheaper things. If I buy something cheap in the US and sell it in China, I'm not only not making money off Chinese people, but bringing technology and new things to China. So not only not arrest, but elect him as a people's representative or CPPCC member. The same goes for lean muscle: recognize that muscle is not harmful to health but beneficial, and you should accumulate more muscle, not the opposite of not training and controlling muscle. This thinking is the key transition from isolation to reform and opening up.
So no one listening to this podcast needs to rush into building muscle. Go ask for yourself—don't just borrow our conclusions. First, go ask for yourself; second, figure out whether you should even train. Respect science. I used to have another line of thinking: specifically avoiding muscle. Many women, when they start working out, ask their trainer right away: "I don't want to build any muscle at all. I want to stay healthy but have zero muscle." I was influenced by that mindset too—more cardio, less weights, less muscle. Now it's suddenly clear to me: that's completely wrong. You can even skip cardio, because when you train with weights and do resistance work, you're already improving your heart and lungs—no need to do dedicated cardio. What you need is muscle. Cardio and muscle are 25 points each, and by training muscle you get all of it. If you train cardio, you not only train your muscle away, you also only get the 25 points for cardio, and the 25 points for muscle are still missing. I've been completely woken up and have fully switched to the "lean muscle" theory.
I highly recommend the "Lean Muscle Bible." I downloaded and saved that video, and even had AI go through it sentence by sentence—it's very useful. The only thing that could be improved is this: viewers of the "Lean Muscle Bible" must already recognize that the lean muscle theory is correct, otherwise they'll just scroll past it. I think we could also do an episode of "Lean Muscle Q&A," picking some very basic questions and the most common misconceptions to answer, to wake these people up.
Musk plays a very bad role here. He himself is fat—not completely fat, but like me: thin limbs, with fat mainly concentrated in the torso, a big torso. Bryan Johnson, the anti-aging guy, also used to have thin limbs and a big torso—that's a common problem for us internet entrepreneurs. Later he followed the lean muscle theory and trained, and now, even if you don't call it lean muscle, at least his body is healthy and has muscle. But Musk left a comment under Bryan Johnson's tweet attacking the lean muscle theory, saying "You look better before." Bryan posted an old unhealthy photo next to a post-training photo to encourage people to work out, and Musk replied that he looked better before—meaning all that training was wasted. Where you stand depends on where you sit.
But the person who loses the most from Musk spreading this wrong idea is Musk himself. AI already told me: Musk is 20 years older than me—now 56 or 57. Every year he doesn't train, he loses another precious year, because muscle declines at 1% to 2% per year. If he waits until after 60, it'll be too late to train. Right now you can train on a two-day cycle; later, when the body's recovery ability is worse, it might take three or four days, or four or five days, for one cycle, because you need to recover. This bill will only get bigger and bigger. So you must train, and train early. Welcome to the Lean Muscle Club.
Mobility, Highly Liquid Assets, and a Ten-Year Strategy
I used to have another theory, also within Sun Studies: I really admire nomadic peoples. Their characteristic is simple—five words: follow the water and grass. Wherever there's water and grass, they go live there. Because nomads herd sheep, and sheep need water and grass, so wherever there's water and grass, the Xiongnu, Liao, and Jin would go.
The Liao, Jin, and Xiongnu have been heavily demonized in Chinese history. This is because when we read textbooks, we sit on the side of the Song; the Liao and Jin were constantly attacking us, so how could they be good people? But today, the Liao and Jin have also joined China and are members of the Chinese family. Why should they be considered inferior to the Song? All 56 ethnic groups are equal. Our history still has a tendency to demonize the Liao and Jin, so we haven't analyzed them. The Liao and Jin were so powerful and made the Song cry out; they must have had their own impressive aspects.
Falling to today, living by following water and grass first has no fixed answer; it's not about settling in one place for a lifetime and never daring to move. My personal approach to living by following water and grass in the new era is to go where I can maximize my role and engage in whichever industry. Whether I can earn enough money is a criterion, because earning more proves that you have greater value in that place. This can be seen as a very important feedback, treating money as data. If I can make money, it proves that the value I create is great, and others are willing to pay me, indicating that my industry has relatively great value for all humanity.
The same person working in the AI field earns more than at Xibei, which is almost going bankrupt these days. Doing the same in the catering business, making a braised noodle dish, compared to doing AI, which has greater value? To be honest, AI definitely has greater value. Liang Wenfeng worked on DeepSeek for a few years and it's about to go public, with talk of breaking a trillion dollars, a company worth several trillion RMB, only after a few years; Xibei has been doing it for over a decade and is about to go bankrupt. For the same person, the industry premium given by the state and society is indeed much greater in AI. So living by following water and grass: wherever I can maximize value, I will go there. I won't stubbornly stick to one place. This is also why the theories of Sun Xue, Huang Mao, and Bo Ji are very useful—mobility and flexibility.
So I don't recommend that everyone buy a car or a house, and marriage is also a point. Buying a house, a car, and getting married all mean settling down. Once settled, you are fixed in that place, and sometimes you aren't even aware of it.
Take the simplest example: tomorrow morning OpenAI says they've admitted you to work in San Francisco, but you've bought a house in Beijing—can you let me work in Beijing? Most likely they won't agree, and you'll lose the opportunity. If OpenAI has water, you go to SF. Young people buying houses, cars, and getting married too early will immediately pin you down.
I've seen many excellent talents. When we want them to come to big cities, they all say they have to discuss it with their wives, and later they all reject our offers. Because his wife earns 5,000 RMB a month but has the final say in the family; he earns 40,000 RMB a month but has lower status than his wife who earns 5,000. The wife says I can't quit; I want to stay in this job forever, and she won't allow him to take a job with a monthly salary of 40,000 elsewhere, so he can only earn 20,000 in place. Therefore, mobility must be strong. I myself rarely invest in careers with low mobility; mobility must be strong. Wherever I can maximize value and achieve the greatest returns, I will go there.
Flexibility is the most important thing, it comes first. Under this premise, then choose what suits you and fits you. Sun Theory, Yellow Hair, Thin Muscle Theory are also the best, all are flexible: system updates are flexible, the AI in the phone is also flexible, the phone can be carried away with a lift, no matter how much stuff is installed it doesn't matter.
There is also a bit of era characteristic here. People often ask me: you spent $6 million to buy a banana, why not spend $6 million to buy gold, buy diamonds, buy jewelry? Because this banana is conceptual art, it spreads on the internet, it is fluid, as long as I agree, in any museum in any country in the world, tomorrow a banana can be taped up there, the fluidity is very strong. Whereas gold and jewelry and these things are dead, once you fix them in one place you can't take them with you, and they are also easily stolen. This banana cannot be stolen, it is forever bound to my name. It is an investment in a digital asset.
Many people don't understand, but this is just like our theory, very cutting-edge, very new, the vast majority of Chinese people today cannot accept it, it's too new. But this is right. Our value as evangelists lies precisely here: if what I talk about is known to all Chinese people, and I talk about 1+1=2 every day, there is no meaning either. Everyone says you talking about 1+1=2 has nothing new, why talk about it? We all already know. So this theory being niche is actually a good thing, being niche proves it is a trend. Once it is mainstream, there is no meaning. Everyone not accepting it and cursing it a lot is instead good, because I am spreading it, spreading the correct theory demonstrates my value, this makes me more firm in promoting the theory.
The underlying logic of Crypto: translating human civilization for machines and AI
What we do is very simple: TRON, crypto are actually doing one thing, translating currency into something machines and the internet can understand.
This is very similar to the Sun Yuchen Award I established. The Sun Yuchen Award encourages translating mathematical proofs into Lean formal proofs, that is, making mathematical theories and all derivations readable by machines. This of course also lays the groundwork for AI, because AI can only understand what machines can understand, if this thing of yours is not something machines can understand, it won't understand what you are saying.
Then are China Merchants Bank and PayPal currencies that machines can understand? No. CMB and PayPal are only the digitization of traditional currency, they themselves are not a machine-readable currency language, the two have an essential difference. The difference essentially depends on: finance has a ledger, is the management of this ledger centralized server management, or decentralized server management? Only a ledger under decentralized server management, because it operates entirely according to cryptographic mathematical methods, so machines can immediately understand what you are doing. This is also why TRON, Bitcoin and these are conceptual languages that machines can immediately understand, and all their APIs naturally integrate perfectly with these payment methods.
What we're doing is taking humanity's ancient currencies from the past few thousand years—currency being a financial communication tool invented by humans—and translating them into a language that AI and machines can understand. That's what blockchain and crypto are doing.
Of course, many people who don't understand it keep shouting "cutting leeks." It's like training for lean muscle: you're talking about health, but people love to demonize it. A key tactic of demonization is to frame your actually legitimate motives as selfish—like saying you train for lean muscle just to pick up girls, as if you're being opportunistic and sleazy. That's wrong, but many people fall for it because they don't look at what you actually do. You say training for lean muscle is good for my health, but they won't listen—they bring up something you can't even explain: "You're only doing it to pick up girls." Motives are subjective; others can hardly disprove or prove them. They reduce blockchain to "cutting leeks": if you do blockchain, you're cutting leeks. "Cutting leeks" means selling something worthless to others at a high price, leaving them with something worthless, and your motive is to do something bad. They don't talk about the benefits this technology and this thing bring to humanity. But what we're doing is the same as AI.
Another point that really strikes me: everything must be translated and put on the internet to become AI training data—this is crucial. Because if something doesn't become AI training data in the future, it's as if it never happened. A rather striking example: recently they've been working on the formalization of Fermat's Last Theorem, and in the future, the formalization of the Poincaré conjecture. Fermat's Last Theorem and the Poincaré conjecture have already been proven by humans, but now the problem is that machines can't read the proofs. If machines can't read them, they can't be absorbed by computers and AI worldwide, so the thing has no value—proving it was pointless. You proved it, but how many people in the world know your approach? Nobody knows. You have to formalize it, upload it to AI, and let computers understand it—only then is it truly digested.
So humanity is now in a transition from human to machine, from carbon-based to silicon-based. In between, you play a crucial role as a translator: translating mathematics for machines, translating currency for machines—all human inventions must be translated for machines. Once machines understand, they can hit the road.
Think about the current SWIFT system—we're doing the same thing: sending money to any corner of the world, whether it's dollars or anything else. SWIFT was invented around 1955, built on the Allied telegraph system after World War II. Back then there were no computers, no AI, and you couldn't expect it to be compatible with AI and computers. So it's a completely outdated, offline, heavy system. What we're doing is absolutely right, because we're translating it into something machines can read. In the future, if no one translates your thing for machines, your life itself will have no value.
Let's go back 500 years. Suppose there was a country with 50 million people. If these 50 million people had no historical records, no corpus records, and never got onto AI in the future, and all of them perished, then they effectively never existed on Earth, because no one would know they ever existed. You can't blame AI either, because AI would say you didn't feed me anything. So getting everything online is crucial now — you must get it online, let AI read this corpus. This is the opportunity to enter humanity's next civilization. If AI hasn't read your corpus, it's as if you never existed.
So there are many engineering projects I think are very meaningful: digitizing Chinese history, Chinese characters, Chinese poetry, and uploading them online for free. I believe DeepSeek and ByteDance may already be doing this. Chinese history, Chinese culture, all the books in Chinese libraries — they all need to be digitized and uploaded online, because if you don't upload them and don't digitize them, it's as if they don't exist.
The Fermat example also illustrates another thing: if a line of thinking isn't preserved, humanity has to take 100 extra years. If you only say it in private, if you don't record it, if you don't speak it out, then that flash of insight will very likely just pass.
Take another example — that professor from New York University. From the individual scientist's perspective, it might be a tragedy, but from humanity's perspective, it's a good thing. He had an idea, told OpenAI, and the idea was indeed taken by OpenAI and used to crack the problem. But from humanity's standpoint, humanity locked in the solution to the NS equation — don't underestimate this. Fermat's Last Theorem — when Fermat proposed it, he said he could solve it, but the margin was too small to write the proof. What happened? Maybe he could prove it, maybe he knew the approach, but he didn't share it with humanity, and it took humanity over 100 more years to find the next person who could prove it. Because Fermat didn't have time. If he'd had a bit of time — if it were the AI era, Fermat could have just told AI the approach, even via voice conversation, recorded the voice, AI remembered it and sent it to other mathematicians, and someone might have proven it — proven 100 years earlier. Just like the speed everyone competes at today — you speak the idea, and it's proven for you in two days. So from the individual scientist's perspective it might be a tragedy, but for all humanity's progress bar, it's accelerated. If you stand from the perspective of all humanity, then you welcome AI to "steal" more ideas.
Self-media is also worth mentioning. Why is it called self-media? Previously only the media could speak; if you had ideas, you couldn't speak for yourself. The media was just a few people within media organizations — the editor-in-
