AI Leaders Are Refineries, Not Winners of Native Apps
AI enters the agent era: first-phase leaders are just refineries, not winners of native apps; new platforms don't come from old platforms; organizations shift from roles to tasks; ultimately, humans compete on creativity that creates something from nothing.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Building agents today is like building websites in 1992
Zeng Ming divides the industrialization of general-purpose technology into three phases: first, it becomes social infrastructure; then, applications explode; finally, native applications emerge. In 2026, token as a unit of measurement reaching consensus signals the maturity of the first phase; the agent events around Chinese New Year (commonly called "lobster") mark AI officially entering the agent era. He compares it to the website-building movement of 1992: back then people built websites to share information; today they build agents to share capabilities. The browser hasn't appeared yet; those who can establish agent standards and capture user entry points may be the biggest opportunity in the next two years.
— Zeng MingA ten-trillion-dollar company will emerge, but not these two
Zeng Ming states bluntly that OpenAI and Anthropic are remarkable technical and commercial achievements, but business history shows that the first wave of companies to run out are likely not the big winners of the native application phase. Yahoo was worth $120 billion five years after founding, and AOL reached a valuation of $220 billion, yet neither became the Google of the next era. Model companies are more like AI cloud companies, refineries that extract base oil; the real application value belongs to chemical plants and car companies—and cars were not built by oil companies. He believes a ten-trillion-dollar company will definitely emerge, but it may not be the current two leaders. Model companies will move toward a mature business of oligopoly plus strong government regulation.
— Zeng MingThe unit of AI-native organizations is tasks, not roles
Companies correspond to the industrial-era bureaucracy; AI disrupts the industrial era, so the corporate system will decay. The basic unit of an AI-native organization is not a role but a task: people are not organized by reporting lines but flow around "what problem needs solving," with humans and AI collaborating around tasks. Therefore, well-designed positions and hierarchies will disappear, and traditional middle management will die out. Future founders must first define the organization's task network. Silicon Valley's NewLab is a prototype of the new organizational form; OpenAI itself resembles a Lab more than a traditional company. The culture of new organizations is more like a sports team, emphasizing transparency, sharing, and co-creation, not command and assessment.
— Zeng MingExcellence is judged only in hindsight; being good is visible on the spot
Zeng Ming believes "excellent ≠ extraordinary." Excellence is growth sustained by positive feedback, while extraordinary requires overcoming repeated negative feedback, persisting when everyone disapproves, and ultimately proving oneself a prophet of the era through enormous success, so "extraordinary is judged in hindsight." Those who are merely excellent easily fall into a cycle of self-validation, leading to conservative strategies. Extraordinary people have a small ego, tend to attribute success to the era and partners, and are mission-driven rather than externally valuation-driven. When identifying founders, he asks, "What kind of self would you be satisfied with in ten years?" If the answer is "a ten-billion-dollar company," that is mainly ego-driven, not an inner mission. Truly extraordinary founders are often altruistic and empathetic.
— Zeng MingStrategy in the AI era cannot be planned; it must be generated
In the AI era, strategy is not planning but generation. Zeng Ming proposes a "strategy generation system": organizations must build an environment and network that allows strategic insights to emerge naturally. Because decision frequency is higher and quality requirements are higher, CEOs cannot rely on linear extrapolation for planning, nor can they outsource strategy to McKinsey—McKinsey's rise was a product of the slow maturation of the industrial era. His methodology is "look ten years, think three years, do one year," with the core being "think three years": you must see at least two or three milestones ahead, creating tension between short-term actions and the medium-term picture. Organizational building must ensure sufficient context rather than control, allowing the right people to make the right decisions at the right time.
— Zeng MingThe correct analogy for robots is appliances, not cars
Zeng Ming believes robots are still in the strategic exploration phase, far from convergence. Currently, two paths—building a generalized brain first or closing the loop in specific scenarios first—are both logically valid, and it's impossible to judge which is faster in advance. Historically, the automobile industry had thousands of companies from 1900 to 1920, and it wasn't until Ford built the assembly line in 1913 that scaling began; robots haven't yet had their Model T moment, and whoever truly sells ten thousand robots first is already remarkable. But a better analogy for robots is appliances rather than cars: after electricity was invented, refrigerators, washing machines, air conditioners, and countless other categories emerged. Robots interacting with the physical world will have N major scenarios, and home, companionship, and industrial uses may each produce their own native giants.
— Zeng MingGiants that feel safe are the most unsafe
Zeng Ming says that in the face of a giant wave, no one is safe; feeling safe is the most unsafe. In business history, no company has ever smoothly transitioned from one era to the next. IBM and Microsoft are the only two possible exceptions, and both are full of uncertainty. AI is a productivity revolution that disrupts the industrial revolution, not a continuation innovation like mobile internet, so the technological reserves and organizational culture of old giants are liabilities. Google can survive as an AI cloud company, but it may not capture the consumer entry point; ByteDance has an AI cloud opportunity, but Doubao is not the future application; Tencent's social relationships will be restructured; Alibaba is not naturally safe either. New platforms are likely to be created by new companies.
— Zeng MingHuman value lies in creating from nothing, not in writing songs or painting
Borrowing Drucker's framework, Zeng Ming defines the AI era as the era of creativity: all knowledge work that can be structured will be taken over by AI. Humans are both forced and finally liberated to develop new potential. This potential is creativity—not writing songs or painting, but the ability to originally define complex problems and create from nothing. AI has already taken all existing knowledge; human value lies in creating things that do not yet exist. Therefore, he calls his mission "to be a happy researcher," and believes the next step of civilization will reorganize around creativity. Even the education system will be forced to shift from "knowledge indoctrination" to "open exploration," because young people already know the old path leads nowhere.
— Zeng MingIn their own words · checked verbatim
The most successful companies must be contrarian, but being contrarian does not guarantee success.
最成功的企业一定是反共识的,但反共识不一定成功。
Zeng Ming7:19
It may not survive to the end. It will likely do well later, but it is probably not the big player, the big winner of the native application phase.
它不一定能活到最后。它很可能会往后活得不错,但它大概率不是原生应用阶段的大玩家,大赢家。
Zeng Ming27:49
Companies in the first phase find it hard to survive into the second phase, and companies in the second phase find it hard to survive into the third.
第一阶段的企业很难活到第二阶段,第二阶段的企业很难活到第三阶段。
Zeng Ming29:50
So companies will disappear; companies will die out. There's nothing to regret about that. In the future, there will be more fun things. How many people go to work at a company happily now? They're not happy.
所以公司会消失,公司会消亡。这没有什么可以遗憾的呀。将来有更好玩的,你说多少人现在去公司上班是开心的?不开心。
Zeng Ming1:15:33
My technology is so great, I've created so much value, why might my company not be worth much? Because of the basic laws of economics: as long as there is homogeneous supply, you cannot capture high profits. You are just a company that isn't worth much.
我的技术这么牛,我创造了这么大的价值,为什么我公司可能不值钱?因为经济学的基本规律,只要有同质化的供给,你就不可能获取高额利润,你就是一家不怎么值钱的公司。
Zeng Ming1:28:01
Excellence is judged in hindsight, because you overcome negative feedback again and again. When everyone disapproves of you, you finally prove with enormous success that you were the earliest prophet of the era.
卓越是事后论定的,是因为你克服了一次一次的负反馈,让所有人都不认可你的时候,你最后会以巨大的成功证明,你是时代最早的那个先知
Zeng Ming1:31:32
I have always believed that in the face of a giant wave, no one is safe. Whoever feels safe is the most unsafe.
我从来觉得在巨浪面前没有人是安全的,谁觉得安全,谁就是最不安全的
Zeng Ming2:02:39
Figures
| K3 model parameter scale | 2.8T | 35:58 |
| Years since OpenAI's founding (as of recording) | 11 years | 2:17:47 |
Glossary
- Token factory
- Treating AI compute as infrastructure that produces tokens, symbolizing intelligence entering a standardized pricing stage.
- Agentic OS
- An operating system that can directly dispatch all agents to complete tasks based on intent; Zeng Ming believes it belongs to the third phase and cannot be implemented yet.
- New Lab
- A new organizational form dissatisfied with the corporate system, primarily composed of researchers; OpenAI is seen as the earliest example.
- High agency
- A Silicon Valley buzzword meaning taking control of one's own destiny, actively seeking tasks and growth, replacing hierarchical command.
- Context not control
- A management philosophy: provide sufficient shared context and information so members can autonomously make correct decisions.
How to listen
CEOs making AI strategy, executives at traditional giants seeking a second curve, AI investors, and tech founders torn between building models or applications.
The rapid-fire Q&A and book list at the end can be skipped (after 2:29).