A Conversation with 孙宇晨: How Young People Can Seize the Opportunities of the AI Era?
Summary
For the version of yourself starting over at 20 with no network or resources, 孙宇晨’s answer is not to bet on a sector first, but to turn yourself into an AI-native individual. Even with less than $1,000 in monthly disposable income and unable to afford Claude at roughly $200 a month, you can use free or pay-per-use services such as 智谱, Kimi, Z code, and B点AI. Start from your phone, and turn health, learning, work, schedules, and decisions into portable local MD files. “智谱 is free today, Kimi tomorrow”; the key is not loyalty to any model, but completing the AI transformation of your personal system first.
In the AI era, career advantage starts with revealing the map, not building more units earlier or grinding harder. 孙宇晨 recommends spending 2–4 months accumulating personal data, then having multiple models analyze it and produce best-, base-, and worst-case options. Before choosing a major, industry, home, or car, study 100 examples—like sending scouts across a real-time strategy game to identify resources, enemy positions, and the current meta before committing. His core commercial lens is “discover opportunities globally”: China’s 400 million young people should not compete only with one another, but serve 8 billion people and the 2 billion AI entities he imagines will exist in the future.
What he is most bullish on is not a single isolated hot sector, but every field where data is being generated at scale and remains underprocessed. Mathematics, astronomy, biology, gene sequencing, pharmaceuticals, proactive health management, insurance, crypto payments, LLMs, data cleaning, and labeling all make the list. The common mechanism is that AI lowers the barriers to mentorship, language, analysis, and experimental search. The most investable intersections are AI infrastructure and data services, biotech and health finance, and global payments and agent settlement—not an unresearched fashionable label.
The asset-allocation message he repeats most often is to maintain the “carbon-based server” first, then amplify it with silicon-based leverage. 孙宇晨 is giving Claude Code, Oura Ring, photos, and training videos access to his 172 cm, 75 kg body at roughly 29% body fat, targeting 140–150 grams of protein a day and a calorie deficit capped at 600 calories, while using progressive overload through squats and deadlifts to reshape himself within 3–6 months. He calls muscle “cash on your body” and “the body’s savings,” while acknowledging that he once equated muscle with chasing women, steroid use, and poor health—a misconception that required a complete reversal.
His foundational definition of crypto is that it translates money into a language machines and the internet can directly understand, invoke, and settle. Banks and PayPal merely digitize traditional ledgers; he sees decentralized cryptographic ledgers as machine-native, programmable money that can support tiny denominations, instant settlement, stablecoin payments, and transactions between AI agents. By his account on the show, Tron carries tens of billions of dollars in daily transactions and several trillion dollars in annual transfer volume; the host summarizes it as infrastructure that “builds highways and collects tolls.”
“Freedom first, wealth second” is the least changed and most investment-relevant principle from 孙宇晨’s theory a decade ago. Homes, cars, marriage, and a single geographic location can all lock up future opportunities. He advocates “following the pasture,” keeping location, sector, and assets highly mobile, while treating income as feedback on the value created. Freedom improves the ability to capture opportunities, and wealth then expands freedom, creating a positive flywheel; fixed costs and missed opportunities create the opposite downward spiral.
What determines long-term compounding is not existing wealth, but whether you can maintain a small ego and keep installing new versions. He compares life to “Earth online”: real estate may be nerfed by 50% while AI productivity is buffed by 100%; patches may arrive with a lag, but they eventually take effect. Winners of the previous version often refuse to update because of vested interests and ego. His overarching principle is that cognition determines destiny, but he openly admits he still has no answer on marriage or life’s ultimate meaning. After achieving financial freedom, he would rather create more and expand his experiences than consume forever or lie flat—and he remains willing to let new information overturn today’s conclusions.
Deep dive
1. Fitness Is Only the Entry Point; 孙宇晨 Is Really Building a Personal AI Operating System
孙宇晨’s first-stage target is 3–6 months. He is currently 172 cm tall, weighs roughly 75 kg, and has about 29% body fat. He has just started training and intends to observe the changes through measurable data rather than judge success by feel.
He photographs each meal before and after eating and sends the images to Claude Code, which estimates his actual intake and recommends the eating order. His current supplement plan is 25 grams of protein powder and 5 grams of creatine, with a daily protein target of roughly 140–150 grams.
He alternates training and rest days, with exercises including squats, deadlifts, and stretching. Oura Ring supplies sleep and activity data; after subtracting food calories, he keeps his daily deficit to roughly 600 calories in an attempt to gain muscle and lose fat simultaneously.
2. Portable Data Assets Matter More Than Any Single AI Model
Every exercise is recorded by his coach and handed to Claude Code to log sets, reps, and weights. His first day included several movements performed for 3 sets of 12 reps, with estimated expenditure of 275 calories; that data then informs his progressive overload plan.
AI combines his photos, training, sleep, and calorie data into a daily report. Everything is stored in local folders, MD files, and image archives. He is pursuing a system that is visual, data-driven, and easy to summarize, turning improvement into a continuous experiment.
The architecture is deliberately vendor-agnostic. If Claude Code is unavailable, Codex can read the README and MD files and take over directly. What compounds is personal context and structured history—not any individual chat log.
3. Even With Less Than RMB1,000 in Monthly Disposable Income, You Can AI-Enable Yourself From a Phone
Asked how a 20-year-old with no network or resources could start over, 孙宇晨 set the baseline at his own time at Peking University: no more than RMB1,000 in monthly disposable income. He therefore does not assume the user can afford Claude’s roughly $200 monthly fee.
His alternatives are free or low-cost tools such as 智谱, Kimi, Z code, and B点AI. Use whichever one is running a promotion—“智谱 is free today, Kimi tomorrow”—and switch APIs when necessary rather than abandon the method because of budget constraints.
A computer is not a prerequisite: “You may not have a computer, but you probably have a phone.” Create folders, save materials, and let AI build long-term context from the phone. Tool reliability can wait until income rises enough to justify paying for it.
4. “孙学, Fighting, Yellow Hair, AI” Form a Personal Modernization Framework
The first layer, “孙学,” means reinstalling the cognitive operating system and rethinking money, financial freedom, and how society works. The second layer, “fighting,” refers here mainly to muscle gain, strength training, and proactive health management, preventing the body from becoming a veto point.
The third layer, “Yellow Hair Theory,” deals with social resistance. Once an unconventional approach begins producing advantages in muscle, wealth, or cognition, it may attract mockery or jealousy. The response is to execute validated judgments fully rather than roll them back for the sake of fitting in.
孙宇晨 places AI in the captain’s seat, even saying he “trusts AI 100%.” He estimates that China has roughly 200 million men aged 20–40 and about 400 million people in that age cohort overall, all of whom could install this cognitive, physical, and psychological upgrade at zero cost.
The framework is highly personalized and carries obvious slogan-like elements. The host’s more restrained summary is that it addresses cognition, health, resilience, and personalized decision-making, helping young people keep acting even on nontraditional paths.
5. Accumulate Data for 2–4 Months, Then Let Multiple Models Cross-Diagnose You
孙宇晨 does not recommend changing your life the moment you install AI. He suggests spending 2–4 months collecting data on health, work, interests, income, and behavior, allowing models to understand the real starting point before discussing direction or ways to make money.
For decisions, feed the same local materials to 3 AIs and ask each to produce best-, base-, and worst-case strategies, then “take what is useful from them.” He mentions Claude, OpenAI’s “Aster,” and “G4”—the exact model names used in the conversation. The point is to cross-check Chinese and US models, not to worship any single answer.
B点AI serves as a multi-model API gateway in his account. Users pay per query and can switch between models. It attempts to turn expensive subscriptions into pay-as-needed purchases, though the episode does not discuss specific pricing or performance comparisons.
6. Finding Direction Is Not a One-Time Forecast but Continuous Tuning
Beyond having AI understand you, young people must continuously track technological change. 孙宇晨 follows X and Elon Musk, and also has agents scan developments automatically, but he does not expect everyone to cover every industry.
A more workable approach is to choose one area that genuinely interests you and could become a specialty, then go deep. AI expands the map; humans still need to build local expertise. Otherwise, “looking at everything” becomes information consumption without execution.
Every decision is only an initial set of weights. You then tune it continuously as new data arrives, like training a model. It is not a one-shot transaction: sleep, performance, market feedback, and changes in interest all require adjustments.
7. He Says He Trusts AI 100%, Yet Still Keeps the Final Human Tuning Right
孙宇晨 uses flying a plane to describe the division of labor: let AI fly 90% of the time and use the remaining 10% to make adjustments. He may disagree with a model, but first asks it to explain the underlying principles before deciding whether to update an old belief.
Protein powder was a concrete reversal for him. He once suspected it was “technology” or close to taking drugs; after AI explained that it was simply a convenient way to supplement protein, he accepted taking 25 grams a day because getting 140–150 grams from regular meals alone is difficult.
The conversation preserves an important tension: he first says, “I execute whatever Claude says,” then later adds, “Of course I also apply my own judgment.” The actionable version is not abandoning judgment, but having AI handle search and rebuttal while humans set the goals and risk boundaries.
8. From “Talk Less on WeChat” to “Use APIs to Talk to Global AI,” the Upgrade Is Context Ownership
孙宇晨 upgrades his earlier line—“don’t chat with WeChat, chat with 豆包”—once again. A chat interface can still trap materials inside a platform; the better approach is to save files locally and call global models through APIs.
In his view, an API is not a programmer-only barrier. It requires only “a tiny bit more channel change” than using 豆包. If you do not know how to create an MD file or connect an API, ask AI how to do it and let the tool teach you how to use the tool.
邵艾伦 acknowledges that this would change his own content workflow. The problem is not that he never uses AI, but that he has not yet connected AI to every layer of his materials, decisions, and execution.
9. The Scarcest Asset in the AI Era Is Not Knowledge but Remaining Life
孙宇晨’s optimism about young people comes from an extreme but clear judgment: AI will make most goods, knowledge, and services steadily cheaper. The only thing that cannot currently be replenished is lifespan, so “youth is wealth.”
AI remains far from solving immortality. Older people may not catch the next rounds of technological iteration, while a 20-year-old has more version cycles to live through and can wait for biology, computing, and commercial infrastructure to mature.
He therefore does not limit opportunity to one sector: “Every sector is full of opportunities.” Young people’s structural advantage is having time to learn new tools, make mistakes, change direction, and wait for compounding to appear.
10. Mathematics’ Old Barrier Was Mentors Owning the Way of Thinking; AI Is Flattening That Gap
Explaining why Chinese students perform strongly in competitions while relatively few win top mathematics prizes, 孙宇晨 attributes the bottleneck to “the approach.” If a mentor does not know a route to a proof, even a brilliant student may spend years without access to the entry point of a frontier problem.
AI can absorb methods published globally and show the routes simultaneously to a Peking University graduate student, a high-school student, and a doctoral student at a top university. If a student can understand the explanation, they may access research context that previously required years of network-based accumulation.
He cites 陶哲轩 and Einstein as examples of young mathematical achievement, arguing that breakthroughs never inherently required waiting until one’s 30s or 40s. AI is not changing the age of genius; it is giving more young people a “PhD-level map” earlier.
11. The “Solve the Navier–Stokes Equations in 2 Days” Story Is About Combinatorial Speed, Not AI Creating From Nothing
As 孙宇晨 recounts it, a mathematician at New York University shared a new idea for researching the Navier–Stokes equations with OpenAI. An agent inside the system recognized its potential, then marshaled roughly $7 million of compute and thousands of agents to complete the route in 2 days.
He specifically adds a limitation: “AI itself does not create anything either; it cannot build a skyscraper on flat ground.” If humans do not have a viable route to controlled nuclear fusion, AI cannot produce one from nothing. Its strength is aggregating existing directions, comparing paths, and exploring rapidly.
The episode does not independently verify this example. Its value lies in the competitive picture 孙宇晨 draws from it: future leaders may be the people who provide a key new idea, combined with those who can rapidly mobilize large-scale agents to validate it.
12. Astronomy, Biology, and Genetics Share the Same Opportunity: More Data Than Humans Can Process
Telescopes such as Hubble have already returned enormous volumes of observational data, but manual analysis is tedious and expensive. 孙宇晨 believes that putting AI into the loop for continuous screening could produce “fundamental progress” in astronomy.
Biology, pharmaceuticals, and gene sequencing follow the same pattern: abundant experimental and sequencing data, with the possibility that an effective correlation could flow directly into drugs, health prediction, and commercialization—“when the cannon fires, gold pours in.”
He extends the opportunity to proactive health management and insurance. If risk identification, lifespan prediction, and personal health data are recalculated, the financial algorithms built on top of them will also need to be updated.
13. Regional API Price Gaps Are His Simplest Example of Global Arbitrage
孙宇晨 says the same AI service may cost less on a monthly basis in Argentina, Turkey, or Nigeria than in the United States. If high-cost-market users can legally access lower-cost service, the intermediary creates measurable value.
The host quickly turns this into the idea of a “token relay station.” 孙宇晨 agrees it is a good concept, but emphasizes that the underlying business is not exploiting loopholes; it is identifying global frictions in prices, payments, and service delivery, then charging to remove them.
The core capability is not local distribution, but continuously scanning supply, demand, regulation, and pricing across countries. AI can reduce language barriers through translation, but compliance boundaries remain subject to local rules.
14. In Business, Ask What Value You Created Before Asking How Much You Can Make
For 孙宇晨, a price gap only matters if users actually receive a cheaper or more convenient service as a result. “Can I make my value visible to the customer?” is the first question in every commercial judgment.
China is the world’s factory, and connecting its capacity and products to global markets offers vast room for expansion. He does not encourage young people to stare only at the small shop around the corner, but to imagine Turkish, Argentine, and Nigerian customers as potential users.
AI lowers the cost of translation, customer service, market research, and content localization. Young people do not need to abandon Argentina because they do not speak Spanish. Basic language learning helps, but language should not narrow the market in advance.
15. If 400 Million Young People Serve the World, Competition Shifts From Involution to Finding Gaps
孙宇晨 imagines the show’s audience as roughly 400 million Chinese young people. Equipped with AI and “孙学, fighting, and Yellow Hair,” they should not continue fighting among themselves for the same demand, but serve 8 billion people.
He adds a future assumption: the world may soon contain another 2 billion AI entities, making the potential customer base “10 billion people and AI.” The figure is aspirational, but the commercial principle is clear: expanding the total market matters more than optimizing competition within a small local pool.
Once the demand pool is large enough, failure becomes more reversible. If one industry has no room, you can switch countries, customers, or products rather than interpret one local failure as proof that you lack ability.
16. Before a Major Decision, Study 100 Samples; Information Is Often Scarcer Than Courage
Although 孙宇晨 advocates not buying a home, car, or getting married before 30, if you must do so, his advice is not to search for one universal answer. First study 100 homes, 100 cars, or a large number of potential options.
With enough examples, location, pricing, negotiation tactics, and real differences naturally form a benchmark. “Read a book 100 times and its meaning reveals itself”; do not commit immediately because the first option looks good.
The same method applies to careers. Have AI research the next 10 years of employment and industry change for 100 majors, then assess interest and odds, with particular attention to niche paths hidden by elite-university prestige.
17. His “10-Home Rule” Places Exploration and Action in the Same Framework
Among 10 irreversible options, 孙宇晨 says never buy the first 3; use them to establish a benchmark. If a home among options 4–6 is better than every home in the first 3, choose it immediately.
He links the sixth option to 0.618 and the intuition behind the optimal stopping problem, though the episode does not provide a rigorous mathematical derivation. The point is to avoid both extremes: deciding impulsively on the first option or still being unable to act when the final option arrives.
For young people, the first 3 research projects, internships, or ventures can serve as a calibration period. But exploration cannot continue forever; once a new option clearly crosses the threshold already established, execute.
18. “Reveal the Map” Is the Episode’s Most Actionable Career Metaphor
孙宇晨 compares many people’s problems to “building units behind closed doors” in a real-time strategy game: they study, work overtime, and accumulate resources without knowing where the resources are or what units the opponent is building, only to discover that the map has already been claimed.
The smarter move is to send out fast, lightly armed scouts to identify gold, resources, and the enemy’s composition. If the opponent is mass-producing war elephants, build spearmen in response rather than continue producing ineffective units.
The host adds the analogy of placing vision wards, controlling the jungle, and taking the dragon in 王者荣耀. Without visibility, the faster you execute, the more likely you are to pour resources in the wrong direction. AI’s first value is expanding visibility for free or at low cost.
19. Asked for “Five Sectors,” He Offers a Set of Data-Intensive Cross-Sector Opportunities
邵艾伦 is not satisfied with an abstract method and asks 孙宇晨 to name 5 trends directly. 孙宇晨 does not mechanically list 5 isolated industries, but gives clusters including mathematics and astronomy, biology and genetics, pharmaceuticals and health management, insurance and finance, crypto payments, and AI infrastructure.
The screening criteria are consistent: Has the field accumulated large amounts of theoretical, experimental, or observational data while lacking enough people to process it? The more AI can reduce search costs, the more likely young people are to bypass old credential barriers.
He specifically warns against dismissing data cleaning and labeling, citing the claim that Scale AI “sold more than 10 billion” to Facebook. That figure is his verbal description; the real point is that basic data services can also capture value from model growth.
For investors, this is not a list to buy mechanically. It is a filter: look for nodes where data density, automation potential, a commercial loop, and global settlement capability coexist.
20. “A Nobel Prize Per Person” Is Hyperbole for a Revaluation of Researcher Productivity
孙宇晨 believes people aged 20–40 are at a creative peak and have arrived at the moment of “humanity’s greatest invention.” At 36, he even feels he has only 4 years left in the traditional prime window.
He describes the productivity leap as “a college student with AI surpassing a Nobel laureate in minutes,” and predicts that future output could reach “Nobel-level results per person.” This is not a forecast for the number of prizes, but an extreme description of lower barriers to achievement.
Compared with when he entered university at 17, when the internet and campus networks were the main connectivity tools, AI provides parallel labor directly across research, writing, analysis, and execution. That is why he expects multiple industries to flower simultaneously.
21. Strength Training Was Chosen as the AI Experiment With the Shortest Feedback Cycle
Choosing an industry, reviewing 100 homes, or building a global business all require long validation periods. Strength training can show changes in physique, weight, and health metrics in roughly 3 months, making it his first test of the data–advice–execution–review loop.
He believes total muscle mass peaks around age 30–40, when the return on training investment is highest. Starting at 70 is more about preserving mobility, with longer recovery cycles.
That is why his message to people aged 20–40 is not that they “can train,” but that they “must train.” The certainty comes from his recent cognitive reversal and is notably stronger than his tone on most commercial sectors.
22. “At 80, You Either Walk Into the Restaurant or Get Pushed In” Reframes the Fitness Comparison
Responding to the fact that Buffett, Trump, and Elon Musk are not known for fitness, 孙宇晨 relays the question AI posed to him: when you have lunch with Buffett, does he walk into the restaurant himself or get pushed in in a wheelchair?
The question shifts from whether successful people exercise to how you want to enter a restaurant at 80. In his accounting, roughly 40 minutes a day buys mobility from 40 to 80; it is an investment in quality of life, not appearance.
He therefore concludes that the 3 business icons are “wrong” on health. Only if Neuralink, mind uploading, or an Iron Man-like body becomes real can biological constraints be bypassed; otherwise, wealth cannot repeal natural law.
23. The More Agents You Have, the More Important the Human Body’s Veto Right Becomes
孙宇晨 says he has more than a dozen agents monitoring and working simultaneously each day, requiring only occasional tuning. In theory, one person could eventually direct hundreds or thousands of agents, giving individual time far more leverage than before.
But the brain remains “forced to operate inside carbon-based flesh.” No matter how powerful the silicon server, if the carbon-based server falls ill, the entire command system goes offline. He defines training as maintaining the hardware that carries the brain.
His daily priorities therefore compress into 2 tasks: maintain the carbon-based body and direct silicon-based AI. The body does not serve the work; the body determines how long all the AI leverage can operate.
24. “25 Points for Cardio, 25 Points for Muscle” Becomes His Proactive Health Scorecard
孙宇晨 says that when he asked AI about the factors determining lifespan, it gave him a 100-point framework: cardio accounts for 25 points and muscle for 25, together making up half the total. The rest covers smoking and alcohol, genetics, cancer, eyes and teeth, and organ checks.
He believes resistance training builds muscle while also training the cardiovascular system through movement and load, allowing one program to capture the first 50 points. Pure cardio could cause muscle loss and create a trade-off.
This is a personal health model formed by 孙宇晨 and AI, not a medical scoring standard offered by the program. Its importance lies in changing his resource allocation toward variables that are both highly controllable and high impact.
25. He Reprices Muscle as Cash, Wages, and Strategic Reserves
The line AI gave him was “electrifying”: “You may be materially rich, but physically you are poor.” Training is paying the body a wage and increasing future deployable physiological capital.
In his interpretation, illness, inadequate calories, and excessive long-duration cardio consume muscle first. Muscle supports movement, recovery, and metabolism, so it should be “saved” through sufficient protein and training.
The host calls muscle a strategic reserve asset. 孙宇晨 goes further: “Muscle is basically cash on your body.” If dollars disappear, the body can still help you earn; if the body’s reserves disappear, money may not be able to buy the full function back.
26. He Reverses 180 Degrees From “Muscle Is Bad for Health”
孙宇晨 admits that when he was younger, he understood building muscle as a way to attract women, associated muscle mass with steroid use and sudden death, and deliberately did less resistance training and more cardio in the name of health.
He even treated muscle and health as opposites and used busyness as a reason not to train. Protein powder was also misunderstood as “technology.” This was not a minor adjustment, but what he calls a shift from “closing the country” to “reform and opening up.”
His new conclusion is that muscle itself is beneficial to health, while protein powder is simply a supplement when it is difficult to obtain enough protein from food. Appearance and dating advantages are side effects, not reasons to surrender the physical benefits because of concerns about how others interpret your motives.
27. Online Abuse Was Not Treated as Truth, but Became Anomaly Detection for Blind Spots
After an online controversy, some people claimed he was only 163 cm tall, with a large head, small limbs, and obesity. He clarified that he is 172 cm and said he would not change because of the criticism itself, but was willing to check whether the attacks contained any real problems.
Comparing photos made him realize that his leg muscles were underdeveloped. He immediately consulted AI and moved squats up the priority list. “If there is a problem, correct it; if not, take it as encouragement” here means extracting actionable signals rather than accepting humiliation.
Yellow Hair Theory provides the psychological firewall, while a small ego allows real feedback into the system. The former prevents group emotion from carrying you away; the latter prevents self-esteem from filtering out all negative information.
28. Proactive Health Management Means Testing Before You Are Sick, Not Paying Only in the Emergency Room
Before starting training, AI advised 孙宇晨 to check blood sugar, blood lipids, HDL, LDL, and other markers. Without obvious symptoms, he went to the hospital proactively and listed the tests he wanted.
The doctor saw the list and asked whether he had used AI, saying that most patients previously arrived only after symptoms appeared. 孙宇晨 concluded that AI would shift medicine from passive treatment toward everyday monitoring.
He rejects outsourcing all health decisions to doctors. Doctors cannot eat, sleep, train, or track long-term behavior on a patient’s behalf. By the time someone rushes into an emergency room and says “do whatever it takes,” many choices are already irreversible.
In his imagined future, muscle will become a basic configuration for successful people, like blood pressure and blood lipids. Without proactive management, the problem is not merely appearance; a piece of the cognitive system is missing.
29. The Bryan Johnson–Musk Debate Is, in His View, Health Version Path Dependence
孙宇晨 mentions Bryan Johnson’s Don’t Die practice. Johnson posted before-and-after photos, and Elon Musk commented, “You look better before.” 孙宇晨 sees that as using aesthetics to reject a health intervention.
He explains it as “the butt determines the head.” Musk has relatively more torso fat and thinner limbs; if he endorsed strength training, he would be admitting that his previous public position might have been wrong, making it easier to defend the status quo.
He also relays AI’s view that muscle may decline by roughly 1%–2% per year after middle age, while recovery intervals after training lengthen with age. Delaying does not merely cost one year of muscle; it raises the future cost of catching up.
The host broadly agrees, but the entire judgment is an inference based on the two men’s public images, not a clinical assessment of Musk’s health.
30. The Post-30 Strategy Is Summed Up as “Following the Pasture”
邵艾伦 asks how someone who has largely achieved financial freedom should plan the next stage. 孙宇晨 answers with the nomadic principle of following the pasture: wherever the grass and water are, that is where the herd goes; life need not be permanently anchored to one place.
The modern pasture is the location and industry where you can have the greatest impact, create the most value, and earn the highest return. Geography, careers, investments, and business models should all retain the ability to move.
Money is not the only objective. It is feedback data: high income often means customers are willing to pay for the value, and may indicate that the industry ranks highly in society’s allocation of resources.
31. The Main Cost of a Home, Car, or Early Marriage Is Pledging Future Optionality to the Present
孙宇晨 opposes buying a home or car and getting married too early—not because any of them are always wrong, but because each adds fixed-location, cash-flow, and family-coordination costs, making it harder to “pick up and go.”
His hypothetical is that if OpenAI invited you to San Francisco tomorrow morning, someone who had already bought a home in Beijing and built a fixed life might first ask whether the opportunity could be moved to Beijing, leaving the scarce opportunity to someone more mobile.
He has also seen someone whose monthly salary could rise from RMB20,000 to RMB40,000 reject a big-city offer because a spouse did not want to give up a stable RMB5,000 local job. The example does not settle the value of marriage; it shows how household decision rights change career optionality.
32. In His Framework, High Returns Are Signals of Social Value, Not Moral Medals
Comparing restaurants with AI companies, 孙宇晨 argues that if the same person can create more output in AI, the market will generally assign a larger premium. He contrasts the difficulties of restaurant businesses with expectations for DeepSeek’s high valuation.
This does not mean all high income is more virtuous. Money can be treated as a measuring instrument: how much customers, capital, and society are willing to invest reflects the size of the demand connected to an activity at that moment.
The important point is to keep watching the measuring instrument change. AI may have high value density today, but that does not guarantee the same sector will dominate in 10 years. Following the pasture means remaining open to new pastures rather than turning “do AI” into a new lifelong doctrine.
33. The $6 Million Banana Reflects a Preference for Mobility, Not Merely Conspicuous Consumption
Responding to criticism over spending $6 million on banana concept art instead of gold or diamonds, 孙宇晨 emphasizes that the work is a network-distributable liquid IP. With his authorization, any museum in the world can re-stage it according to the concept.
Gold and jewelry are physical, location-bound, costly to store, and potentially stealable. The banana work is tied to his name; the core asset is narrative and distribution rights, not the physical banana that will rot.
He places this preference for mobility in the same framework as crypto, personal IP, and muscle: digital assets move through networks, knowledge moves through local files, and physical assets move with you on your body.
“Niche proof that something is a trend” is his further judgment. Mass incomprehension does not automatically prove that an idea is correct, but if the logic is sound, it may mean dissemination and pricing are not yet complete.
34. Tron’s Job Is to Translate Money Into a Language Machines Understand Natively
For people who know nothing about crypto, 孙宇晨 defines it not as trading coins but as translating thousands of years of human money into a form the internet, machines, and AI can read, invoke, and verify.
China Merchants Bank and PayPal digitize balances, but their ledgers remain controlled by centralized servers. Truly machine-native money, he argues, should run on a decentralized ledger according to cryptographic and mathematical rules.
This structure exposes APIs that programs can compose directly. Tron, Bitcoin, and other crypto networks therefore do more than “display numbers”; they let machines understand ownership, transfers, and settlement conditions.
35. Mathematical Formalization and Crypto Payments Are the Same Translation Project in His Eyes
孙宇晨 praises the effort to translate mathematical proofs into Lean formal proofs so that every deduction can be checked by machines. The subtitles render Lean as “Lin,” but the context clearly refers to a machine-readable proof system.
Fermat’s Last Theorem and the Poincaré conjecture may already have been proven by humans, but if the proof still depends on natural language and a small circle of experts, AI cannot fully absorb, reuse, and extend it.
Mathematics therefore needs to be translated for machines, and money needs to be translated for machines. Humanity is moving from a carbon-based knowledge system to a silicon-readable one; the translators in between will become key infrastructure builders.
36. A Civilization That Never Enters the Internet or AI Corpus May Eventually Be Equivalent to One That Never Existed
孙宇晨 makes an aggressive claim: if history, books, poetry, and personal thoughts are not digitized, uploaded, and included in AI corpora, later machines will have no way to know they existed.
He therefore supports digitizing Chinese history, culture, and library collections and putting as much as possible online. This is not merely preservation; it is a bid to enter the memory and reasoning systems of the next civilization.
The podcast itself is an example. A private flash of insight disappears quickly; once recorded, published, and transcribed, it can be searched, cited, and developed further by other people and models.
He uses the story of Fermat saying the margin was too small for the proof to illustrate the cost of loss. If Fermat could have described the route to AI by voice at the time, other mathematicians might not have needed 100 years to reconstruct the idea.
37. The Value of Independent Media Is Breaking the Centralized Structure in Which a Few Editors Set the Public Agenda
孙宇晨 has worked in media and summarizes the traditional newsroom as “the editor-in-chief sets the topic, reporters execute.” A small number of people therefore determine what large audiences see and what they ignore.
Independent media allows everyone to speak, compete, and debate directly, which is why platforms such as Douyin, YouTube, and X have risen. He believes competition at least increases the probability of getting closer to the truth and weakens the influence of a single media outlet’s interests.
The host adds that many people now get news from Twitter-like feeds rather than traditional broadcasters such as BBC, because they can see different sources, rebuttals, and evidence simultaneously.
Decentralization does not guarantee that every view is correct. The mechanism that holds is lower publishing barriers and a larger set of comparable views—not the automatic equivalence of traffic and truth.
38. Tron Already Has a Killer Application: Sending Digital Dollars Across Borders
孙宇晨 says Tron currently handles tens of billions of dollars in daily transaction volume and several trillion dollars in annual transfers, a scale that has “caught up with SWIFT.” This is his account on the show; the conversation does not further break down the transaction types.
The clearest current killer application is not some distant new consumer use case, but moving money quickly around the world, especially through stablecoin payments and settlement.
The host explains the business model as a highway: Tron builds the infrastructure, vehicles move along the chain, and the network collects tolls. It can therefore be understood as a payment rail with cash flow. 孙宇晨 accepts this as a useful simplification.
39. Micropayments Between AI Agents Will Amplify the Advantages of Programmable Money
Traditional finance’s minimum denominations and settlement processes are not suited to massive volumes of tiny, high-frequency calls. 孙宇晨 believes crypto can measure amounts smaller than “one cent” and provide instant settlement.
That fits the API economy naturally. Model calls, data reads, agent delegation, and machine purchases can all settle automatically per use, without a human opening an account or approving each payment.
The host summarizes the future entry points for ordinary people as stablecoin payments, human-to-AI payments, and AI-to-AI payments. 孙宇晨 sees the greatest long-term potential in agent-to-agent payment.
He believes paper money will probably disappear in 100 years. The digitization of money is irreversible, though adoption takes time—just as AI is already widespread while many people still have not entered the system.
40. “Everything Crypto” in 2012 Was, in His Explanation, a System Upgrade Rather Than Pure Profit-Seeking
孙宇晨 says he moved all his assets into crypto after 2012, when he bought Bitcoin at roughly $1. On the show, he compares that with subsequent prices of $70,000–$80,000 and eventually $100,000. He emphasizes that his initial motivation was adopting the future system.
He compares it with the first people to buy an iPhone. Even if Apple shares or the device ecosystem later appreciated, that does not prove the original purchase was made only for profit. Early adopters first gained new capabilities; wealth was the byproduct of society increasing the weight assigned to those capabilities.
In his view at the time, homes and cars belonged to the old-era asset system, while crypto could be translated and invoked by machines. System versions “cannot roll back; they can only move forward,” and the later you upgrade, the greater the compatibility loss.
He was also early to the storage theme because AI would continue consuming storage resources. His method is to identify technological dependencies first, then wait for the market to confirm them through price rather than invent a price-increase story in advance.
41. Judge Technology by What It Does, Not by Inventing Motives for Its Advocates
Responding to people who reduce blockchain to “harvesting retail investors,” 孙宇晨 says this is equivalent to explaining fitness as a way to chase women: describe the motive as selfish or sordid, then skip the question of whether the technology solves a problem.
He uses an extreme example to show the logical error. Even if a fitness influencer has a terrible private life, that does not imply muscle is unhealthy. Likewise, even if negative rumors about 孙宇晨 himself were true, they would not automatically invalidate digital currency.
He also notes that critics of Musk and Sam Altman often say they are developing AI only for power or profit. Whether a product is useful and whether its risks are controllable still require evaluation independent of personal feelings about its creators.
The host’s summary is “Don’t worry about motives; first see whether the thing is right.” More precisely, motives can matter for governance risk, but they cannot substitute for analysis of mechanisms, evidence, and outcomes.
42. Being Bullish on an Industry Does Not Mean Your Career and Investments Must Fully Overlap
邵艾伦’s capabilities are concentrated in English, content, personal IP, and independent media, while much of his capital is allocated to crypto. He asks whether his capabilities and money should both be all-in on the same point.
孙宇晨’s answer is more permissive than the old theory of concentrating all resources. Producing content and investing in crypto are not contradictory; you can express belief in a technology by holding the asset without joining the company as an employee.
He recalls buying a large amount of Apple stock when he first saw the iPhone in 2007 and recognizing Tesla as revolutionary when he first encountered it in 2012. Neither view required him to abandon his career and join Apple or Tesla.
The core is to “upgrade your operating system.” Career, investment, consumption, and research can express the same belief separately; there is no need to sacrifice an existing advantage for a formal appearance of perfect alignment between belief and action.
43. Crypto Independent Media Has Potential, but the Larger Bet Is on Personal Media as a Whole
孙宇晨 does not endorse “crypto independent media” in isolation. He first takes a broader view on independent media itself: a personal show can now have influence approaching that of a major television network.
He mentions Joe Rogan—misrendered in the subtitles as John Logan—to show that podcasts can influence public debate and even votes. Rogan did not begin his show to control elections; power emerged from the accumulation of trust over time.
Content ability and crypto knowledge can certainly be combined, but the starting point should be sustained production of credible content, not treating a channel as a pump or traffic tool. Commercial and political spillovers increase as influence grows.
44. Bitcoin’s Competitor Is Not Every Asset, but Gold That Refuses to Upgrade
Asked about “Crypto will eat the world” and whether Bitcoin will become the ultimate store of value, 孙宇晨 agrees and narrows the comparison: Bitcoin mainly competes with gold, while gold has no capacity for version upgrades.
Bitcoin’s software, infrastructure, and ecosystem can iterate continuously; gold’s physical attributes are fixed. In his view, an upgradeable asset is more likely over time to absorb new functions and demand.
He uses “software eats the world” to explain the version framework. Being weaker today does not mean remaining weak forever; as developers, capital, and users keep updating, the boundaries of a software asset move.
This is a highly certain personal judgment, not an endpoint proved by the episode. The investment logic is to favor networks that can keep iterating and avoid static systems locked into old advantages.
45. Earth Online Updates Its Version; Reality Is Simply a Game With Slower Feedback
王者荣耀 gives young people an intuitive understanding of versions: a hero may be dominant today and nerfed in the next patch. What matters is understanding the economy, development, map, and team-composition mechanics—not mastering one hero and relying on it forever.
孙宇晨 writes reality’s update as an exaggerated patch note: “real estate nerfed by 50%, AI productivity buffed by 100%.” If an entrepreneur continues to practice only the real-estate hero, past success becomes path dependence.
Reality’s danger is delay. A patch may already be released while someone takes 3 years to go bankrupt. Health mistakes may also take 10–15 years to appear as insufficient muscle or limited mobility, but the result still arrives eventually.
“孙学, Yellow Hair, fighting” therefore do not prescribe one career. They teach players to read the rules, follow the meta, maintain the body, and tolerate a period of nonconsensus instead of mistaking an old answer for a fundamental law.
46. A Small Ego Is the Core Governance Mechanism for Surviving Across Versions
孙宇晨 cites 张一鸣’s small ego. ByteDance was able to move from Douyin to 豆包 because it remained willing to listen to younger people and learn new tools after achieving success, rather than treating the previous victory as permanent proof.
The larger the success, the smaller the ego should be, because the next version requires learning again. In reality, the opposite often happens: wealth and status make people more likely to believe, “I succeeded, so all my judgments must be correct.”
He recalls advising Buffett to invest in Tesla and Bitcoin, which Buffett declined. He also imagines that if Buffett had exchanged $1 billion for half of Tron when its valuation was $2–3 billion, he would have earned a huge return 7 years later based on Tron’s stated valuation of roughly $33–34 billion.
These counterfactuals cannot be verified, but the lesson is clear: previous-generation winners who refuse to install the new version create opportunities for the young. Starting from zero can be an advantage because there is no old system to defend.
47. In His Model, “IQ” Is the Visible Output of Information Input and Behavioral Updating
The host cites Alex Hormozi: new information is the denominator, while the behavior actually changed is the numerator. Given the same information, the larger the adjustment in action, the higher the intelligence displayed.
孙宇晨 pushes the idea further: “IQ does not exist in the world; only the amount of information exists.” He invokes Transformer and “Attention Is All You Need” to explain how model capability rises with more training data and reasoning.
He preserves the original version numbers for comparison: the apparent IQ gap between GPT 3, GPT 4.0, GPT 5.6 Aster, and the future GPT 10 is fundamentally the result of more information, pre-training, and reasoning entering the system.
The conclusion is not to passively read more, but to convert new information into behavioral patches. Knowing that strength training matters but not starting, or knowing AI matters but not changing the workflow, produces no observable increase in “IQ.”
48. The Risk Control of a Hexagonal Warrior Is Not Letting One Blind Spot Destroy All Compounding
孙宇晨 uses the Peking University mathematician 韦东奕 as an example. Even extraordinary mathematical ability can be undermined if inadequate periodontal and dental care causes irreversible damage early in life, affecting both daily life and research.
Likewise, Musk may lead in engineering and entrepreneurship while still making incorrect judgments about weight and muscle management. Expertise in one field does not automatically transfer to another.
“Sharpening the axe does not delay chopping wood” is his portfolio principle. Teeth, muscle, sleep, finances, and AI usage are all basic means of production and cannot be chronically depleted in the name of focus.
This is also why he advocates proactive health management. Wealthy people cannot wait until an emergency-room visit to mobilize resources; health requires personally owning the data, making decisions, and executing over time.
49. The Sequence of Financial Freedom Is Freedom First, With Wealth as a Byproduct
A decade later, 孙宇晨’s core understanding of financial freedom has not changed; he has only added AI and health patches: “Financial freedom is still freedom first, wealth second.”
Many people assume money creates freedom. His causal arrow runs the other way: first release your choices from cognitive constraints, fixed burdens, and social scripts, then you can move toward higher-value locations and sectors.
Freedom itself has no financial threshold. It is a subjective willingness to choose again. Wealth is feedback created after freedom by capturing opportunities and producing value.
The two form a flywheel: more freedom makes wealth easier to generate, and more wealth expands freedom. Mortgages, maintenance costs, and opportunity lock-in create the opposite downward spiral.
50. Muscle and Wealth Are Both Manifestations of a Cognitive System
孙宇晨 believes that even if obvious muscle does not appear until 6 months later, the result was already determined once he understood proactive health, protein, and progressive overload.
A migrant worker may develop muscle passively through manual labor without understanding calorie deficits, nutrition, training cycles, and recovery. Once the labor stops, the physique may quickly return to its old state.
Intentional training has a complete mechanism: load creates a stimulus, protein and recovery complete the adaptation, and the load then rises progressively. Appearance is the visible output of a system running over time.
He compares this with the wealth cognition he formed around 2012. His assets were far smaller then, but he had already “understood it.” Money and muscle both lag cognition; time eventually makes the difference visible.
51. Without a Bottom-Layer Cognitive System, a Windfall Can Turn From a Gift Into a Death Sentence
孙宇晨 suggests that people without money play a simulation in which they suddenly win RMB30 million. If the outcome is bankruptcy every time, their current cognition cannot yet manage the asset.
Money is an amplifier. Someone who could buy only one bucket of fried chicken a day can buy 10 buckets after becoming wealthy; consumption, drug use, relationship conflict, and bad investments all accelerate with greater purchasing power.
He compares this with receiving perfect muscle overnight. If diet, training, and health cognition remain unchanged, the body will return to its old state 6 months later. An external intervention cannot replace a sustained system.
Lottery winners losing their fortunes is therefore not merely a matter of luck. It is a mismatch between the outcome and the cognitive level. Building the system first does not mean someone is unworthy of money; it avoids amplifying old vulnerabilities.
52. The First Thing Money Cannot Solve Is Automatically Giving You Correct Cognition
Asked what he discovered money cannot solve after becoming truly wealthy, 孙宇晨 answers directly: cognition. It requires interest, learning, experience, and repeated correction; it cannot be purchased once like a service.
Buffett has enormous capital yet may have missed early Bitcoin and Tesla. 孙宇晨 has material resources but also held incorrect beliefs about health for years. Wealth expands execution power; it does not guarantee direction.
“Materially rich, physically poor” is therefore not merely a fitness slogan. It is an asset-allocation warning: financial capital cannot replace physical capital, relationship judgment, or version learning.
53. Marriage Is the Patch He Explicitly Admits He Has Not Installed
On business and investment, 孙宇晨 often gives definitive answers. On marriage and family, he repeatedly says, “I am still figuring it out,” because he has never married or had children and has only limited experience at the edges of marriage.
In online debates over whether someone would leave after being given RMB50 million, he currently leans toward yes. The host also believes that RMB500 million, RMB5 billion, or even RMB50 billion cannot rewrite a relationship with no emotional foundation.
His provisional conclusions are only 2: the relationship between love and money is smaller than people imagine, and money is more of an amplifier. A relationship with love may become easier; a relationship without love will not be fundamentally changed by the amount.
“Even an incorruptible judge cannot settle a family dispute” captures the subjectivity of relationships, but subjective does not mean patternless. He sees the roughly 200 million unmarried people in China, along with similar phenomena in Japan, South Korea, and the United States, as a global question that still deserves further study.
54. The Most Honest Answer on Life’s Meaning Is That You Can See Some Rules, but Still Do Not Know the Endgame
孙宇晨 sees himself as a player thrown into “Earth online”: play first, explore first, and gradually identify the mechanics, without claiming to understand why the game exists.
If strength training extends his life to 90 or 100, he views the first 50 years as an observation period. Like reviewing 10 homes, the first half is for accumulating samples rather than rushing to draw a final conclusion about life from limited data.
Going into space changed his childhood answer about “contributing to Earth.” Humans may leave Earth, while AI makes “only for humanity” feel too narrow. The more he knows, the less willing he is to treat the current scale as the ultimate scale.
He therefore reduces the answer to “enjoy your life” and “experience first.” To people in despair, he does not offer a grand theory of meaning; he says not to rush into an irreversible conclusion, but to look again and try more paths.
55. Business Is His Main Experience; Money Is the Scoreboard for How the Boundary Has Moved
When 孙宇晨 started his first company at 22, he used the human longevity record of roughly 123 years to set a goal of living to 122 and spending 100 years as an entrepreneur. By the time of the episode, he says he had completed roughly 14 years, or about 14% of the plan.
He compares business with Bolt continually breaking the 100-meter record and Cristiano Ronaldo chasing his 1,000th goal. After crossing 9.8 seconds, 9.7 seconds, or 900 goals, the athlete no longer represents only himself but is exploring the boundary of human performance.
Tron does not have to remain a blockchain company forever. It could move into AI or eventually space colonization. Berkshire Hathaway’s transformation from a textile mill into an investment giant shows that an organization’s name should not lock in its future mission.
He asks AI why personal wealth often seems to plateau around $200 billion. AI’s explanation is that wealth is constrained by global economic activity of roughly $40 trillion–$50 trillion, making it difficult for one person to hold more than about 1% for long. If AI or interstellar trade expands the total pie, the ceiling can move higher.
56. Musk, Buffett, and Trump Are Three North Stars for Engineering, Long-Term Capital, and Personal IP
孙宇晨 identified the 3 University of Pennsylvania alumni as benchmarks early on. Elon Musk represents the engineer, high-end industries, and engineering organizations; Warren Buffett represents finance, the snowball, early positioning, and extremely long holding periods.
Donald Trump represents personal IP, cross-sector expansion, and commercializing influence. From real estate to beauty pageants, WWE, The Apprentice, and politics, he first built “Trump the person,” then let the name empower products in return.
The 3 routes conflict. Buffett emphasizes focus and stays in Omaha to avoid the distractions of New York social life; Trump takes cross-sector expansion to the extreme; Musk mixes engineering, X, AI, and political participation.
孙宇晨 does not want to decide too early which school is correct. He chooses to “practice multiple martial arts,” then combine the experience of Jack Ma and his own life to build 孙学. He hopes that one day someone will use “becoming someone like 孙宇晨” as a coordinate, rather than copying one specific trade.
57. Financial Freedom Is a Ticket to a New Map; the Next Stage Should Expand Experience Through Creation
孙宇晨 would even pay $50 million for the chance to work at OpenAI or Anthropic. He would work to hit the targets without taking a salary. Having been a boss since 22, he lacks the first-person perspective of observing an excellent organization from the employee side.
He also uses different simulators to gain perspectives, including a homeless simulator. His friend 王纯 instead chooses to visit every country, ride different trains, and explore Antarctica and space. Different people do not need to share the same scoreboard for their side quests.
Einstein moved from patent clerk to relativity; Nobel transformed from explosives and the “merchant of death” into the name of a prize; 李煜 turned the pain of losing a kingdom into poetry; Haaland’s father’s football accumulation manifested in his son. Outcomes are often the conjugate result of ability, experience, and random collisions, not a mission written at the start.
The minimum direction is not to “make more money” or “find the one true meaning,” but to create more than you consume. Consumption restores energy; creation leaves behind something transferable. The episode ultimately returns to one line: cognition determines destiny—but cognition must keep updating and cannot become a new doctrine.