After reaching $100 million in ARR, Meshy founder Hu Yuanming reflects on Founder mode, team management, AI and humans, strategy and execution, and personal principles.
The author of this article is Hu Yuanming (Ethan Yuanming Hu), founder and CEO of Meshy.ai.
This article was compiled from the original PDF, “48 Reflections on Reaching $100M ARR.” The original was first published on October 7, 2026, updated on October 8, 2026, and sourced from this WeChat article: original link. The
$100M ARRfigure is disclosed by the author.The main text follows the original. Images from the PDF and interface information from the WeChat article are not included. For background on the person, see the Hu Yuanming founder profile.
First, congratulations to the Meshy team on reaching a new milestone: under our strict accounting definition, annual recurring revenue (ARR) has surpassed $100M (one hundred million dollars). We have always advocated being pragmatic in business while pursuing “technological idealism,” so reaching this level of revenue on top of Meshy 7.1's leading foundation model is certainly cause for celebration. Every department in the company has put in tremendous effort.
Last year I wrote an article, “When a CEO Rereads a PhD: On Wisdom and Courage,” reflecting on and reviewing some of Meshy's experiences after finding PMF and just entering a period of rapid growth. As the company enters the new stage from 10 to 100, I think it is time for new reflections and to lay a foundation for the challenges ahead. These reflections revolve around three questions:
I started writing this article when Meshy was at $80M ARR and continued until it reached $100M ARR, over roughly one month, recording whatever came to mind. During a business trip over the National Day holiday, I spent another 5 days deleting 2/3 and refining it into 5 sections and 48 points.
The five sections are:
Put others before yourself; be a good person with principles.
All the education I encountered growing up asked me to “be kind to others” and “be a nice guy.” At the root of this was the fact that I met several exceptionally kind mentors at every stage. When I was doing OI in middle school, my competition coach always encouraged and cheered me on. Later, at MSRA, my mentor Steve Lin was also widely recognized across MSRA as the nicest mentor: he patiently taught me to revise papers bit by bit and strongly supported me in doing what I enjoyed. During my PhD at MIT, my two co-advisors Fredo and Bill both respected students' interests deeply, giving me the maximum time and freedom to work on the topic I loved, Taichi. Even at MIT there were advisors who would force students to do unwanted work such as writing grant proposals and package it as “I am preparing you to become a professor,” but Fredo and Bill did not do that; they handled those matters themselves. My parents also often told me, “Taking a loss is a blessing.” Before considering themselves, they considered others. On the surface this may look like taking a loss, but in practice it builds deep trust and lets the people around you perform to their full potential, producing better results for the entire team. This simple belief is also an important reason why we put “empathy” into our cultural principles when the company was founded.
The purpose of a school is to teach and educate, but the external environment of a business is very different: it must win battles to survive. Kindness in a company cannot lose its principles, or a person becomes a “people-pleaser,” or, more harshly, a compulsive do-gooder. It is of course good to “repay resentment with virtue” when conditions allow, but once a team grows, it inevitably has to answer the question, “How should virtue be repaid?” A saintly manager cannot manage a team or win battles; they may not even be able to work with principled people, and may drive excellent people away. So “being kind to others” cannot look only at one local moment; it must consider the whole team and a longer time horizon. That is how a “small favor” becomes great kindness to the entire team.
“Fairness” is a good principle, but it is extraordinarily difficult for a manager to implement fairly; it is a systems-engineering problem. A manager's laziness often makes honest people suffer. A common trap is that a colleague is diligent and willing to take on hard work, so the leader assigns all kinds of dirty work to them; out of responsibility, they do that dirty work well. The leader then gives them even more dirty work. In outcome terms, their sense of responsibility and diligence earned them more dirty work. This is essentially “punishing responsibility.” If the leader cannot compensate in other ways, it is certainly unfair and unsustainable. Similarly, we often feel more gratitude toward a parent who occasionally gives us a small favor than toward the parent who quietly does everything; employees who simply do their work may receive worse treatment than employees who make a fuss; or colleagues who work overtime are rewarded instead of colleagues who produce results more efficiently. All of these look only at a local part while ignoring the whole, and are unfair.
One way to help yourself make fair decisions is to “ask your conscience.” For example: “I am about to give A a raise and more options. Compensation is confidential, but if everyone's compensation were accidentally made public one day, would everyone think this arrangement was fair?” Any unfair treatment may solve a problem for a while, but in the long run it is always a hidden danger that will eventually explode.
“Objectivity and sincerity” is another good principle. If something is “good,” say it is good; if it is “not good,” say so. Do not call something “good” just because the other person finds “not good” hard to accept. You may comfort them today, but it will certainly be harmful over the long term. The way you express it can be tactful, but it must be grounded in reality. The result of disrespecting reality is only punishment by reality. Do not focus on who said what; look at what they did and whether they actually “solved the problem.” If there is a problem, say so. Even if the other person cannot take it and is about to leave the team, leaving today is better than leaving after a long delay. Dragging things along half-heartedly drains both sides.
“Think about the whole and weigh trade-offs” is also a good principle. Why can games make people addicted while the real world cannot? Because games offer immediate, certain rewards and mostly follow simple black-and-white logic. The real world is not so sharply divided, and it is extremely difficult to think through the pros and cons of something. A typical “seeing the trees but not the forest” error is to see a good person do one bad thing and conclude that they were bad all along, or see a bad person do one good thing and conclude that they are actually good. Managers cannot cling to a rigid rule; they must weigh trade-offs and consider multiple factors together: short term vs. long term, individual vs. team, fairness vs. efficiency, and so on. A landmine may be planted today and explode months or even years later. Even if there is no alternative, after planting a landmine today you must spend twice the effort tomorrow and the day after to dig it out; otherwise the harm is enormous.
The cognitive ceiling of the person at the top is the cognitive ceiling of the entire team. Time is the most important nonrenewable resource. The most important expression of the top person's cognition is therefore who they spend time with, what they do, and in what priority order.
For a Founder, “being kind to others” also has a limit. If you ignore your own feelings too much while considering other people, spend your energy on pointless matters, and consequently have no energy or attention left for the company's major decisions, that is irresponsible to the entire team. Again, it is a balance between “small kindness” and “great kindness.”
Entrepreneurship is a system of ordinary differential equations over time t. Internal and external factors of a company interact with one another. One of the most important jobs of the person at the top is to see the interactions among those factors and distinguish which are leading indicators and which are lagging indicators. Ray Dalio's The Changing World Order is worth consulting. This system of differential equations is essentially answering “how the world works”; finding that pattern can guide strategy and execution.
For example, a company's “value” has many layers, while the revenue floating at the top is only the tip of the iceberg. At the base are complex factors such as talent density, clarity of vision, technical strength, product strength, cultural values, morale, whether shareholders are aligned, and even the founder's physical health. These factors also influence and evolve with one another over time. If you judge a company's progress only through revenue and valuation, the view is extremely superficial, or at least very lagging. Two companies that look similar on external metrics may differ completely: one may still be accelerating upward with energy, while the other may already be rotting from the roots and about to descend.
A good friend in Silicon Valley taught me a way to calculate the cost of time: subtract today's valuation from the company's valuation one year from now, divide by 365 days in a year, and then divide by the working hours in a day. This is roughly the company's opportunity cost per hour. If we assume that 50% comes from decisions by the founding team and 50% from other people's execution, a founder can easily calculate their own opportunity cost per second. Once you have made this time cost clear, there is no need to agonize over many small matters, because the time spent agonizing has already caused a loss to the whole team: “failing to allocate the founder's time, which belongs to the team, properly” is irresponsible to the team. For example, if you consider who to meet during each hour and then consider that hour's opportunity cost, many people naturally become impossible to meet from the perspective of being responsible to the team. (Of course, this rule does not apply to spending time with family and friends, because that cannot be measured in money.)
Health is one of a Founder's most important resources. From my observations, during the startup stage, the best founders (or, in other words, everyone who reaches the top of an industry) in fact have no WLB at all. Jensen Huang puts it this way: “When I work, I am relaxing.” Perhaps a better term is “Work-life unification”: Work and Life are one. Objectively, health and output involve a trade-off. But even if you lie flat every day, your body will continue to deteriorate with age; nobody can live forever. All we can do is avoid meaningless damage to our health. Do not overeat or binge (for me, that means eating fewer spicy strips or McDonald's spicy chicken wings), try to stay up late less often, and set aside some time to exercise; from the perspective of the whole, that may be the better solution.
A founder must have an Ego, but the boundary between Ego and arrogance is extremely difficult to define. A founder may be right or wrong, but must start from the team's interests and vision, not from “proving that I am right.” Putting “whether I am right” before “what the facts are” is extremely dangerous. On this basis, “independent thinking” is very important, because only you have 100% of the context; your subordinates may have 50% of the context, and outsiders may have 1%. Similarly, you can safely ignore criticism from someone with only 1% of the context, because they are not criticizing you, but the “you” they imagine. The CEO is the try {...} catch {...} inside the main function; there is no upper layer to which an exception can be raised. Everything that happens to the company is ultimately your responsibility. If you let unreliable advice or criticism drive the company into a ditch (exit(-1)), you are still the one who pays.
A Founder may look glamorous on the surface but is actually extremely lonely inside. There are many truths they cannot casually say, and other people do not face the complex context they face every day, so it is difficult to have a deep conversation on the same level unless the other person is another founder. But founders are human too, so they can easily fall into the trap of hoping that subordinates will provide the emotional value ordinary people can readily obtain but a Founder lacks. This is a swamp: if a founder subconsciously needs subordinates to affirm them, it will inevitably leak into daily behavior, or lead to positive feedback for flattering employees. More employees will then start managing upward, eventually hiding the founder inside a carefully woven information bubble, or even psychologically manipulating the founder, blocking objective and fair judgment.
But a Founder is ultimately still a person, and people have the full range of emotions and desires. If emotional value cannot come from the approval of subordinates, where can it come from? For me, there are three sources. The more worldly one is that, as the founder of a company with a reasonably good business, my financial foundation lets me buy things I have looked forward to for a long time with my eyes closed—for example, a Hasselblad. I promised myself that if we closed our Series B, I would reward myself with one; the photos in this blog were taken with that camera. Building this company has already brought me considerable material and spiritual abundance, happiness, and growth as a founder. I am extremely lucky in this respect, because the returns from most startups are 0. The second source is cognitive growth: working and talking every day with more capable people lets me learn a great deal and makes me happy. A slightly more noble source of emotional value is seeing the people who build the business with me grow. I am sincerely happy when colleagues who struggled alongside me go on to achieve outstanding results and live good lives; I regard their happiness and growth as the success of my own work.
After the company created the 3D AI category and entered a period of rapid growth after reaching PMF, I have generally been quite happy. If the opposite of “happy” is “pain,” then for me the opposite of “happy” is probably “bored.” I feel that “whether one is happy” and “whether one is fulfilled” are orthogonal. Over the past two years I have experienced “happy fulfillment” and “painful fulfillment” (“satisfyingly painful”?), but as long as there is growth, I do not feel “bored.” I see every challenge I experience as an “opportunity for growth.”
Sometimes AI is the army of thousands that I command. At other times, I am a “carbon-based pet” raised by AI: AI “feeds” my curiosity with organized knowledge.
Without solid fundamentals, using AI only amplifies confusion and produces plausible-sounding conclusions that are laughable to professionals. My high-school math teacher often stressed, “If the foundation is unstable, the earth shakes”; I think this matters even more in the AI era. Sometimes I tell colleagues, “Sorry, I can't understand this report written by AI. Please rewrite it manually.” Or I express my protest against others trying to fob off AI-generated content on me more tactfully: I truly cannot understand what logic the “not this, but that” constructions in AI-written articles are supposed to express, and I understand even less which rhyming but incomprehensible metaphors they contain. After producing a stream of output, AI may even point out “One honest caveat” (“To be frank, my method still has a flaw”), making it appear to have thought deeply. But can AI-written articles really withstand scrutiny? Before AI, “form” and “content” still had some relationship; after AI, they have become completely decoupled, and “speaking nonsense with a serious face” has become normal.
When I was studying and taking exams in the Yao Class, the teaching assistant set a rule: a correct proof earned full marks, a wrong proof earned 0, while writing “I don't know” directly earned half marks. From a mechanism-design perspective, this made the strategy “I don't know, but I will write a long proof so the grader has to spend a lot of effort finding the flaw and may give me the points anyway” ineffective. In our work habits, the equivalent is probably “if you cannot explain it clearly, it is better not to say it.” Clear communication is the most basic requirement at our company. Words must be accurate, concise, and easy to understand; do not speak “correct nonsense.” Lack of clarity is often not a communication problem, but a sign that the thinking is not clear enough.
On the one hand, almost everyone in the company has nearly unlimited AI-token usage; see “I Work for 10 Claude Codes.” On the other hand, in interviews we place great value on the fundamental skill of old-school programming. In the hundreds of interviews I have conducted, I found that candidates whose logic is clear when writing code by hand often have stronger problem-solving ability. The emergence of AI has not changed this correlation at all. Moreover, if the fundamentals are weak, anything made with AI is a castle in the air with an extremely unstable foundation. In this sense, AI has not leveled the software barrier; it has instead made the strong stronger and the weak weaker.
In the fields described above, humans still temporarily need to understand and guide AI. But in more and more professional fields, I think we must acknowledge that AI's development has exceeded our imagination. These fields share several characteristics: they can be closed-looped in the software world (such as AI coding, which does not involve “moving atoms”), their results can be verified (such as AI for math, where Lean can verify them), and they do not require specialized knowledge from a vertical field (such as law).
Although this has not kept me awake at night, anxiety about humanity's future lingers in my mind every day. This wave of AI revolution is very different from previous industrial revolutions (steam, electricity, computers, and the internet). The earlier revolutions merely gave humans better tools that obeyed human commands; humans remained the ones giving orders. The frightening thing about this wave of AI technology is that “intelligence,” which humans consider uniquely theirs and take the greatest pride in, has already been surpassed by AI in many fields.
Take research in mathematics as an example: AI will soon replace humans as the main force proving mathematical theorems, just as OpenAI's recent progress on the Navier–Stokes problem suggests. Mathematical proof is a frightening opening because it can be completely closed-looped in the software world, and tools such as Lean can mechanically verify that the result is correct. The future may evolve like this: AI will give mathematical proofs like a god; human mathematicians will know that Lean can verify the proof, so the result must be correct, but they will not understand its internal logic or learn from it. Scott Aaronson's “The Age of Wonders and Terrors” is a good article, and Logan Graves's “Priest, Monk, and Mathematician” is also interesting and frightening. “The duty of future mathematicians will not be to prove things, but to interpret the ‘revelations’ given by AI like ‘priests’; soon even the priests' work may disappear, because AI can interpret its own proofs better.” For me, this has already happened: there are too many things about the world that I cannot understand, but asking AI questions and asking AI to write visualization tools lets me learn quickly. In these fields, AI is like a deity, turning truths obvious to it into forms that my limited intelligence can understand, thereby “feeding” my curiosity.
So should we try to understand AI after all? In the specific context of Coding Agents, if you have no understanding at all of what AI is doing, the practical problem you encounter is that you cannot write the next prompt that will help AI make effective progress. Conversely, if humans can keep writing prompts that drive a problem toward resolution, perhaps not understanding AI does not matter much; the problem gets solved in the end. In fact, today we still need some understanding of what AI actually did, especially when writing production code. I do not know how long this situation will last. The underlying question is probably “As carbon-based organisms with inferior intelligence, should humans make room for more intelligent silicon-based organisms?” This is not discussed here. In a sense, AGI has already been achieved; it is simply distributed unevenly across fields.
When you encounter a problem, it is very important to consult someone with no vested interest. When someone tells you that you are “sick,” it may not be because you are actually sick; they may simply want to sell you some medicine. From this perspective, AI still has great value: at least it can be basically “free of vested interests.”
A team also resembles quantum mechanics: an individual's ability matters, but the interactions among individuals have an equally profound effect on the team's overall output. When hiring someone, we should look not only at their own ability, but also at what change their arrival will bring to the team's “energy field.” We therefore examine both whether their fundamentals are solid and whether they are a culture fit: do they identify with the team's core values, and can they stimulate and bring out the best in one another? A team of a few people can be managed through brotherhood; a team of dozens requires the person at the top to understand business details deeply; a team of hundreds requires a set of principles and values that are genuinely practiced, along with a high-density talent pipeline. As the team grows, the leader must turn trust, collaboration, and shared aspiration among people into the force with which the organization keeps creating value.
We support paying employees at the top of the market. One reason is that we believe in “being kind to others” and hope that colleagues who work hard together can live a little better; another is that our business has high gross margins, which enables us to do this. Put plainly, compensation and benefits cannot exceed the company's ability to make money, or they cannot be sustained over the long term. Human resources are an investment, not a cost. I strongly agree with Zhang Yiming's view: do not use the term “labor cost”; use “human-resources investment.” In an emerging field such as generative AI, almost every achievement comes from the contribution of excellent employees. Thinking this way makes us more generous with compensation. On the other hand, when employees have less pressure in their personal lives, they can naturally do better work; under our business model, the logic is self-consistent.
Compared with being short-staffed, the more frightening problem is probably the decline in talent density caused by overexpansion. It is an irreversible cancer of an organization. Great people attract great people, while weak leaders only recruit people who are weaker still. This leads to several basic hiring rules: everyone must be stronger than the people already here; quality is more important than filling a seat, so do not hire someone who leaves you hesitating—there is a high chance the problem will explode later; hire only self-driven people, because nobody wants to micromanage.
Drive the team through a vision. For a team building a ship, you should inspire its longing for the sea; otherwise everyone will feel every day that they are merely hammering nails. The founder must work out a vision and describe it precisely. You cannot borrow someone else's vision, because it lacks the deepest underlying thinking. If the process of arriving at the vision is not solid, the team can easily challenge it, or it may not even pass your own test.
Earlier I said that a founder must guard against being psychologically manipulated by subordinates. Conversely, managers should not manipulate subordinates, superiors, or peers either, for example by using unobjective negation (negging) to undermine their motivation and self-understanding. Treat everyone fairly and do not violate principles, but understand what each person wants, place them in the right position, and tailor the rewards and growth they need. A good relationship should feel positive, beneficial, and comfortable to both sides; principles and putting people first can coexist. Do not “put the team through the wringer.” For what PUA means and how to improve your resistance to it, see Robin Stern's Gaslighting: The Effect of Gaslighting. Stay away from toxic relationships.
A company and team in the startup stage must be an extension of the founder; otherwise the founder will inevitably feel awkward, and an internally conflicted founder cannot build a good company. So the team reflects the founder. Quoting Jensen Huang: “Becoming a CEO does not mean you have to change your personality or behavior. You can continue to be yourself. A long time ago, I learned a lesson—I cannot remember where I first saw it: as a CEO and founder, you are building a race car that you yourself will drive. You want to build a Formula 1 car, but you have to build it in a way you can handle. Make the car fit you.” Original: “You don’t have to change your personality or behavior when you become a CEO. You can continue to be yourself. One thing I learned a long time ago—I have no idea where I first came across it—is this: as a CEO and founder, you’re building a car that you are going to race. You want to build a Formula 1 car, but you have to build it in a way that you can drive it. Adapt the car to you. — Jensen Huang”
Give the team room and let it try more things. Unless there is an obvious logical problem, do not reject an idea immediately. Bottom-up innovation is very important. “Restraint” in management is sometimes the most effective approach. The difficulty is probably deciding when to exercise restraint and when to look a little more closely. Andy Grove's “Task-Relavent Maturity” is a good standard; see High Output Management.
A leader's thinking must lead the team, so a leader will probably be misunderstood. That is fine: a leader can earn the team's trust through their track record, reasonable milestones, or strong communication skills. Because the AI era often has technical people leading teams, and technical people are often young and lack management experience, helping these technical geniuses become good leaders should be a problem every AI company faces. I have put some effort into personally teaching an internal training series called “Management Lessons for Good Students,” because I found that high-achieving students and stronger individual contributors sometimes face even greater psychological barriers when they become managers. If people are interested, leave a message and I will organize the lecture notes.
“Founder mode” should not be only one founder's mode of working; it should be the way the entire team works. What kind of leader do we like? A leader must understand details to a certain degree and, when necessary, look two or three layers down, or even get their hands dirty. There are several main reasons: talented people want to learn from their leaders, so a leader who does not understand details will struggle to retain the best people and attract the best people to join; a leader who does not understand details cannot accurately evaluate each person's contribution. If you have never done something yourself, how can you judge whether a subordinate did it well or poorly? A technical leader who does not understand details can easily feel insecure and worry about being replaced by subordinates, leading to unrestrained interference in their work. One of a leader's most important jobs is to weigh trade-offs comprehensively, consider the constraints across talent, capital, brand, product, market, and other directions together, and then find the global optimum. Without understanding details, deep global thinking is impossible.
Many decisions are like the temperature of the air conditioner in an office: some men will find it too hot, some women too cold, and it is impossible to satisfy everyone. You have to accept that. I recommend that every new manager play it once.
Low performers drive high performers away. After low performers mess things up, they lick each other's wounds and “join hands” to lower the standard of work quality. Worse, because most people do not want to admit their own problems, some low performers blame other departments, damaging the team atmosphere. If you do not want to lose your best employees, say goodbye decisively to colleagues who are not a fit for the team. Of course, if you are also running a high-margin business, handle the parting with warmth so that both sides can leave well.
When I first heard the word “laodeng,” I resisted it: what happened to “respecting the elderly and caring for the young” as a traditional Chinese virtue? Later I found that, in the context of the AI era, the term does contain something and seems hard to replace. As I write this, I cannot help wondering: I am turning 30 this year; in the eyes of younger readers, am I also a laodeng, or perhaps a “middle-aged deng”? How can I keep myself from becoming one? It may help to analyze a few examples of people who are old but have not become “deng.” Three examples immediately come to mind: Nintendo (founded 137 years ago), Studio Ghibli (founded 41 years ago, with Hayao Miyazaki still creating at age 80), and Ray Dalio of Bridgewater Associates (77 this year).
What do they have in common? When Nintendo and Studio Ghibli create their work, they put players and audiences first, polishing every detail to perfection so that players enjoy games and audiences are moved by films. For example, the god of game design Shigeru Miyamoto would pull a cleaning colleague from the hallway to play a game he had made, then quietly observe the person's reaction from behind; Satoru Iwata called this “Mr. Miyamoto's over-the-shoulder view.” Readers may know more of Ray Dalio's books through the animated videos he has made on YouTube, and you can also feel that he sincerely wants his content to help others. Most important, Nintendo's players, Studio Ghibli's audiences, and Ray Dalio's readers can genuinely learn something and grow through their content. This returns us to the opening idea of being kind to others: put other people's lives going well first, rather than whether they listen to you or whether you feel involved. In the new era, being “not working” and creating no new value is the original sin. Create new value, and you naturally lose the “deng vibe.”
At this point, we can probably conclude that “laodeng” behavior is often driven by insecurity: fear of being abandoned by the times, fear that young people no longer listen, and therefore a need to prove oneself. In fact, every generation has its own mission, and being left behind by the times is normal. If one day I can no longer be state of the art at something I am doing, I will let someone else do it. I will do something else instead of giving directions from the sidelines. Set a good goal for them, just as in Breath of the Wild and Tears of the Kingdom: place the towers, shrines, Koroks, and Lynels in the right places, and players will naturally find ways to solve them; you do not need to micromanage exactly what players do. Of course, there is one problem: no player will rescue Princess Zelda. But as long as players have fun, perhaps not rescuing the princess is fine too.
That is what I say, but if you do not grow fast enough to become the best in your industry, it is indeed difficult to level up enough to set goals for the smartest young people of the new era. This is probably one source of the anxiety the AI era brings to everyone. Only by growing as quickly as possible can you continually shake off the “deng vibe” from your body. Once a person stops learning quickly, they are corroded by it. Some people are not old, but they have already become “deng”; some are old, but they have not.
Do difficult and correct things; the world should reward people who think deeply and execute solidly.
Most effort in the world is futile; entrepreneurship and new businesses are the same. Therefore, spending more time thinking about what matters is more effective than spending all your time executing with your head down. Also, if you are too afraid of losing, you cannot gain. Risk always exists. “Bet big to win big, and accept the loss if you make the bet.” “Be strong in your willingness to bet.”
I like to attribute success to luck, not mainly to appear humble, but because I have become too accustomed to failure. After all, if the goal is innovation, most effort must fail. So you can begin with the attitude: “Even if failure is highly likely, I will put my whole self into it; failure is fine, and success would be wonderful.” Try more things and you will always succeed eventually. This habit of “low expectations” makes me more alert to danger, more likely to feel happy, and somewhat better at doing things. Its drawback is that it makes me relatively conservative.
Resources can be exchanged for one another: talent density, culture, morale, brand awareness, cash reserves, vision, and the Founder's physical energy. As long as a weakness in one area is not beyond an acceptable level, it can be made up for by strengths elsewhere.
Begin with the end in mind. The most important thing I learned from Professor Zeng Ming is this: do not think about what you can do in the future based on the cards currently in your hand; instead, first think about what you want to accomplish in the future, then see which cards you can still add. The latter way of thinking can accomplish things ten times bigger than the former, and does not necessarily make them harder. In other words, asking realistically “Why can't I do this today?” is a basic skill, but a better question is “What can I do today that might make it possible to do this in the future?” If you do not try, how do you know you cannot do it?
The most important lesson my time studying at Tsinghua and MIT gave me is: “The smartest people around you are often many times more hardworking than you.” The market rewards only the people with the best results, so in most cases a smart but unhardworking person cannot compete with someone who is both smart and hardworking. Competition is this ruthless. Compared with people who have not experienced such an environment, my greatest advantage may be two firm beliefs: “Nobody can succeed casually,” and “Anything gained by luck will eventually be lost through ability in the long run.” Once you think this way, many things become simpler: just do every task solidly and well.
The second-greatest lesson from those two places was the chance to observe closely how the “geniuses” in other people's eyes work. Many people in these two groups became outstanding in their industries. From what I observed, the most important trait was that they deeply loved what they did. Someone who is merely “hardworking + smart” will certainly lose to someone who is “hardworking + smart + passionate.” For the former, “hard work” is depletion; for the latter, “hard work” means “doing what I like”: the former thinks, “Good heavens, that's enough, I cannot do any more,” while the latter thinks, “I can make it a little better, because I have not had enough fun yet.” They also all handle every detail solidly: experiments must be rigorous, and every punctuation mark in a paper must be correct. Their patience for things ordinary people find boring comes from passion.
Pursue excellence rather than muddling through. Polish without limit; it can always be made a little better. I used to think, “Spending so much effort on one punctuation mark—isn't that being overly picky and controlling?” Later I found that the top performers in every field are like this, including Hayao Miyazaki making animated films, Michael Jackson rehearsing dance, Shigeru Miyamoto designing Mario, Lei Jun making phones, Elon Musk making Tesla, Steve Jobs designing the iPhone, Steve Lin teaching me English writing, Professor Jiang Chenfan guiding me to write a SIGGRAPH paper, and Zhang Yimou directing the 2008 Olympics. People who are “so-so” basically never achieve greatness. There is one prerequisite: they clearly understand priorities and put “important things first.” If they all follow this priority principle, then “nitpicking” becomes an astonishing ability and asset: they have already solved all the basic problems others have not, leaving only the pursuit of excellence, while others are not even qualified to pursue those details because they have not reached that stage. On the other hand, they understand exactly what they want and what they can give up, and know clearly that setting high requirements the team cannot understand may create great pressure today and even get them called an idiot, but will produce good results later.
Because the world is a giant amateur production, becoming top 1% is no problem if you simply do every basic thing solidly. The difficulty is that “doing the basic things well” is not easy either. What counts as basic? For example: “Think clearly about what you really want,” “do what you say,” “go to sleep on time,” “think independently,” “make sure readers can understand what you write,” “make sure you ask only high-quality questions,” “use your own product every day,” “give people who contribute more a larger share,” “put facts above your own pride,” and “do not speak correct nonsense.” All of these require strong confidence, restraint, execution, independent thinking, and the ability to help others succeed. There is a reason that “top 1%” is only 1%.
Since doing the basics well is already difficult, let us not do too many things at the same time. As the company grows, “focus” and choosing what not to do become more important. “What not to do” does not mean “losing”; it means “gaining”: opening fewer projects gives more time to think about the company's strategic center; attending fewer events gives more time with family; meeting fewer new friends makes it possible to build deep friendships with truly like-minded partners; hiring fewer people who do not fit the team gives higher talent density and cultural density.
Deep thinking is fine even if someone else takes it away. Such thinking is the final product of considering market demand, the founder's abilities, the team's abilities, the times, the company's genes, culture, and more together. Therefore, only you can implement this deep thinking at 100 points. If someone takes it and implements it at 80 points in their own company, they still cannot compete with you, because the market chooses only the best. If something cannot be done best, it is better not to do it.
When I was very young, my father once took me to dinner with a student of his who had left to start a company. The two of them said: “A scientist is someone who takes science as their family; an entrepreneur is someone who takes the enterprise as their family.” More than twenty years have passed, and I still remember those words. Having a life's work worth striving for and enduring its ups and downs every day may look like a great burden to some people, but for the goal of becoming an interesting “scientist-entrepreneur” who brings some value to society, I find it endlessly enjoyable. So what do science and running a company have in common? My answer is: find a sufficiently important and interesting problem, then do the most basic things solidly and well.