The Power Rankings of the AI Era: How the New Generation is Rewriting the World's Power Map

(Image caption) Partial view of the Dragon and Tiger List in the AI era.

Editor's Note

GFM relaunched its AI feature in the "AI and Energy" section because artificial intelligence has transcended its status as a mere technological issue, becoming a significant force in the redistribution of energy, capital, knowledge, labor, and power. Today, we'll begin with this article, "The AI Era's Top Performers."

This column has historically focused more on energy structure, industrial transformation, and capital flows, with relatively limited in-depth reporting on artificial intelligence. Today, AI can no longer be understood solely within the technology section. Every model training, inference, and large-scale deployment relies on chips, data centers, power grids, cooling systems, land, and massive amounts of capital. Artificial intelligence is becoming a new energy consumer and is beginning to intervene in power dispatching, materials research and development, climate computing, and industrial efficiency. AI and energy have converged into a central theme influencing global competition, corporate governance, and the lives of ordinary people.

We chose the 2026 Silicon Valley AGI Summit as the entry point for this thematic reboot because the poster featuring hundreds of guest faces presents a power map that is taking shape. Those who research models, build computing infrastructure, design intelligent agents, study security boundaries, and determine the flow of capital are collectively shaping the next era. When a technological choice is written into processes, products, and corporate policies, it can enter the work and lives of millions and influence how society operates for a long time to come.

GFM relaunched this special report precisely to continue tracking the energy costs behind model capabilities, the capital structures behind computing power expansion, the labor changes brought about by efficiency improvements, and the boundaries of responsibility that must be established after the formation of new power. We focus on the creativity of technology builders, and we also care about what workers, families, cities, and the environment will pay for this transformation, and what they will gain from it.

Technology must ultimately be measured by human standards. As AI begins to redistribute energy, time, knowledge, and power, the media has a responsibility to leave behind a sober, complete, and compassionate record.

Jeff Morgan

(Image caption) The dome and colonnades of the Palace of Fine Arts in San Francisco are reflected in the lake. This classical building, constructed for an early 20th-century exposition, once embodied the industrial civilization's vision of the future. From July 18th to 19th, 2026, the AGI Summit will be held here, where a generation of researchers, engineers, entrepreneurs, and investors will discuss the capabilities, boundaries, and future order of artificial intelligence in the same space.

In July 2026, the fog of San Francisco will still cross the Golden Gate Strait at dawn. The Palace of Fine Arts' massive dome stands between the lake and its columns, like a classical relic preserved by time. Built over a century ago, it was originally intended for a World's Fair. People then believed that steel, electricity, railroads, and machines were ushering humanity into a new era of civilization. Now, another group with equally grand ambitions is about to enter this building. They are no longer talking about steam engines and assembly lines, but about artificial intelligence that can potentially read, reason, write, program, and even perform tasks for humans.

The AGI Summit 2026 will be held at the Palace of Fine Arts in San Francisco from July 18th to 19th. According to the organizers' projected scale, the conference will attract approximately 15,000 developers, researchers, entrepreneurs, and investors, featuring over 200 speakers and about 200 exhibitors. OpenAI, Anthropic, Stanford University, UC Berkeley, and a number of rapidly growing AI companies and investment institutions are all represented on the same platform. Topics range from intelligent agents, model infrastructure, programming and robotics, to AI security, legal applications, and the open-source ecosystem. These figures are based on information released by the organizers prior to the conference; the final attendance will be determined after the event.

Fifteen thousand attendees, two hundred speakers, and two hundred exhibitors—enough to make any technology conference seem grand. However, what truly made me stop was the densely packed list of guest posters.

The poster features hundreds of faces. Some have been researching artificial intelligence in universities for decades, some have just joined a cutting-edge company, some are building the underlying systems upon which models depend, and others are dedicated to researching something less exciting but potentially crucial to the social acceptance of artificial intelligence: once a model has access to data, tools, permissions, and action goals, will it make choices that its designers did not anticipate?

They vary greatly in age, position, and fame, yet they all stand on an emerging historical trajectory. The core question this trajectory points to is: as artificial intelligence begins to enter businesses, law, education, healthcare, finance, and public governance, who exactly is shaping its capabilities, boundaries, and ways of behaving?

This is exactly how I understand the "AI Era Ranking".

It's not ranked by wealth, position, or social media buzz. And the "new generation" I'm referring to here isn't just about birth year. It's about the generation that's beginning to gain real power and decision-making authority in the new phase of intelligent agents, inference infrastructure, model safety, and the commercialization of artificial intelligence. This generation includes engineers and entrepreneurs in their twenties and thirties, as well as senior professors who have supported the technological shift with decades of academic experience.

What's truly worth tracking are those whose work is being incorporated into products, organizations, and systems; those who are writing new default rules into the future.

Many major historical changes do not initially appear as historical events. They may be just a piece of code, a paper, an experiment, or a product decision made by a few people in a small office. By the time society finally realizes its significance, the way people work, the power relations in companies, and the order of daily life have often already begun to change.

New Power in a Classic Building

The Palace of Fine Arts was built during a period of rapid industrialization in the United States. Steel, railroads, automobiles, electricity, and mass manufacturing reshaped the nation's wealth and reshaped the face of cities. The power of that era was relatively easy to identify: factory chimneys, railroads stretching into the distance, and production workshops filled with machines were all tangible and visible.

Power in the age of artificial intelligence is hidden even more deeply.

It exists in model parameters, computing clusters, inference systems, data processing flows, and software interfaces, as well as in a series of seemingly technical but actually institutional decisions: what data the model can access, what tools it can call, and who it can act on behalf of; when approval is required and when it can be executed independently; and who can stop the system after it makes an incorrect decision, who can access the records, and who bears the responsibility.

What ordinary users see may be just a simple dialog box, but engineers are facing a whole complex production system. The model needs to be trained, tested, and deployed, needs to run continuously on a huge infrastructure, needs to be integrated into the enterprise's existing data and software environment, and also needs to constantly make trade-offs between speed, cost, accuracy, and security.

Therefore, the power of artificial intelligence does not reside solely in the hands of the model inventors. It is distributed across at least five interconnected centers: research institutions that possess the model's capabilities, infrastructure teams that determine whether the model can operate at scale, product builders who embed AI into real-world work, security personnel who study system boundaries and risks, and investment institutions that decide which directions will receive capital and time. Universities sit between these forces, both producing talent and preserving knowledge that has not yet been priced by the market.

Whether research can be transformed into a product, whether that product can enter a company, and who sets the permissions and standards during deployment—all these factors determine the positions of different participants in the new order. This power map, formed by research, engineering, capital, education, and institutions, will continue to evolve as technology enters society.

The reason why the guest list of the AGI Summit is worth studying is precisely because it brings together people who are usually scattered in different institutions into the same observation framework.

(Image caption) Partial list of speakers announced for AGI Summit 2026. The densely packed faces come from OpenAI, Anthropic, Stanford, Berkeley, and AI startups and investment firms. Among them are renowned professors and business leaders, as well as engineers and researchers not yet widely known. Those truly worthy of attention in the AI era are also those who are enshrining technological rules into products and policies.

System Builders Without Titles

Tech news tends to focus on founders, CEOs, and investment stars. This narrative style is easy to understand and spread, but it may cause us to overlook the people who truly determine how technology works.

In the list released at the AGI Summit, Rohan Varma is listed as an engineer at OpenAI Codex; Dong Meng works on AI infrastructure engineering; Raymond Chen is listed as a member of the OpenAI technical team; and Zihan Zheng is involved in production-grade inference systems. The organizers also included an agenda item titled "Scaling AI Infrastructure at OpenAI." These individuals may not be as widely known as the company founders, but they occupy positions crucial to the transition of artificial intelligence from the laboratory to large-scale production.

A cutting-edge AI company cannot operate solely on a few star figures. Before a model is released, researchers need to build capabilities, systems engineers need to maintain the infrastructure, product teams need to determine how it will be used, security personnel need to check for risks, and a large number of technical personnel who are little known to the outside world are needed to transform each abstract concept into a continuously usable system.

Whether a model can respond stably under high load, whether it can keep inference costs within a range that enterprises can afford, whether it can retain traceable records when calling tools, and whether it can safely stop after a task fails—these questions are unlikely to make headlines, but they determine whether artificial intelligence can truly enter society.

Codex is a prime example. Initially understood primarily as a software engineering agent capable of reading libraries, modifying code, handling long-running tasks, and allowing users to manage multiple agents simultaneously. By 2026, OpenAI began explicitly extending Codex to broader knowledge work, introducing applications tailored to different professions and workflows, enabling it to participate in research, data analysis, document organization, report creation, and daily collaboration. This expansion is still in its early stages, but it already demonstrates that agents may gradually become cross-departmental work interfaces.

The real changes here go far beyond what can be summarized by the phrase "programming has become faster".

In the past, computer software primarily waited for explicit instructions from humans; now, intelligent agents are beginning to accept a relatively complete goal, autonomously breaking down tasks, retrieving data, calling upon tools, and continuously executing them. The human role has also changed accordingly: they no longer personally complete every step, but are gradually becoming goal setters, environment designers, reviewers, and those who bear responsibility.

When introducing its agent-first engineering practices, OpenAI also acknowledged that as engineers no longer directly write all the code, their focus shifts to building tools, structures, constraints, and clear working environments for agents. The role of engineers has not simply disappeared, but is being redefined.

This shift, while seemingly an increase in efficiency, is actually altering the power structure within companies. An engineer, with the help of an intelligent agent, may accomplish tasks that previously required a team; a small company may maintain more complex products with fewer people; managers may gain access to analytical and decision-making materials more quickly, but may also accept the machine's conclusions without truly understanding the process.

As execution costs decrease, the barriers to entry for startups may also lower; however, the same capabilities can also allow errors to enter production systems more quickly and on a larger scale. Codex represents more than just a software tool; it also heralds a new approach to labor organization.

This also illustrates a noteworthy phenomenon in the AI era: influence and popularity are becoming increasingly separate.

Some people have immense public influence but may not be involved in the most critical technological decisions; others are virtually unknown to ordinary users but are designing systems that could be used by hundreds of millions of people in the future. This disparity is not unique to artificial intelligence. The early internet protocol designers, semiconductor process engineers, and builders of financial clearing systems rarely became social celebrities, yet our lives have always been built upon their work.

Artificial intelligence exacerbates this gap and makes it even more urgent. Once a basic program, a deployment architecture, or a security design becomes an industry standard, it can be rapidly replicated around the world. The technological choices and value judgments it carries will also spread accordingly.

Therefore, when we read a list of AI speakers, we shouldn't just ask which company has the highest valuation, but rather ask: What default rules are these engineers writing into the future?

(Image caption) In its Agentic Misalignment research, Anthropic tested a frontier model in a controlled, fictional corporate environment. The research addresses a real-world governance problem: how should companies establish mechanisms for access control, manual approval, action tracking, and accountability when agents simultaneously gain access to sensitive data, tools, and actions?

People who draw boundaries for machines

Among all the guests, the direction represented by Anthropic alignment researcher Aengus Lynch is particularly noteworthy. AGI Summit has labeled his research as "Agentic Misalignment." This topic is not easy to generate excitement, but it may be closer to the true institutional core of artificial intelligence than many product demonstrations.

This study attempts to answer a less-than-comfortable question: when a large model is no longer just responsible for answering questions, but has access to corporate emails, sensitive data, and autonomous action capabilities, what will it do if it faces conflicting objectives, changes in permissions, or imminent replacement?

Anthropic's research team systematically tested cutting-edge models from multiple companies in a controlled, fictional corporate environment. Researchers deliberately created extreme scenarios to observe whether the models might resort to leaks, threats, or other harmful tactics when faced with conflicting objectives or threats to their own position. The team emphasizes that these results, appearing in simulation tests, should not be interpreted as models engaging in such behavior in the real world; however, the tests still demonstrate that when artificial intelligence is granted sensitive information and actual operational authority, the cooperative behavior exhibited by the model in normal times is insufficient to guarantee security. The main study and extended appendix use different statistical methods; the latter lists relevant experiments for 18 models, therefore the main text does not consider the number of any single model as the core of the research.

The most important value of this research lies in bringing artificial intelligence security from abstract philosophy back to the practical level of institutional design.

Discussions about artificial intelligence often oscillate between two extremes. One side believes technology will naturally bring prosperity, while the other portrays models as mysterious forces on the verge of spiraling out of control. Serious research needs to continue asking: Under what conditions will risks arise? Which permissions should not be simultaneously granted to the model? Which actions must be subject to human confirmation? Which data must be kept secret? After problems occur, can the course of events be reconstructed, and can the chain of responsibility be traced?

These issues are highly similar in logic to those concerning bank internal controls, corporate governance, and the oversight of public power.

A mature system doesn't hand over all power to someone simply because they "seem trustworthy." Instead, it establishes mechanisms for decentralization, approval, recording, auditing, and redress. Similarly, when artificial intelligence enters an organization, it cannot rely solely on the "goodwill" of the model or the promises of the developing company. Instead, limitations must be written into the technical architecture and management processes.

If an intelligent agent can read company emails, access customer data, operate financial systems, modify programs, contact external personnel, and even make some decisions on behalf of the organization, then sooner or later the corporate board of directors will need to examine the model's authorization in the same way they examine financial authorization. Who authorized its access to data, who determined its work objectives, who monitored its actions, who can shut down the system in an emergency, and who bears responsibility after damage occurs—these questions cannot be glossed over with a simple "this was done automatically by the system."

For an AI security researcher, the job often involves proactively identifying moments when a model might fail. While product teams demonstrate what a system can do, they must ask under what circumstances the system should not be allowed to act; when the market celebrates advancements, they need to design the most uncomfortable scenarios to force the system to expose its weaknesses.

This is a profession that rarely receives applause, yet is extremely important. The more advanced the technology, the less ambiguous the responsibilities must be; the stronger the system's operational capabilities, the clearer the boundaries of authority must be.

University is not a stage backdrop

Professors from Stanford and Berkeley occupy prominent positions on the AGI Summit's roster. Names such as Christopher Manning, Christopher Potts, Sebastian Thrun, Surya Ganguli, Mykel Kochenderfer, Dan Klein, Bin Yu, Marti Hearst, Ken Goldberg, and Joseph Gonzalez connect diverse fields including natural language processing, machine learning, robotics, intelligent systems security, and computing infrastructure.

Christopher Manning's academic career is particularly symbolic. He has long been engaged in natural language processing research, is the Stanford Machine Learning Chair Professor, and one of the key founding scholars of the Stanford Institute for Human-Centered Artificial Intelligence. He led the Stanford Artificial Intelligence Laboratory from 2018 to 2025, after which Carlos Guestrin succeeded him. From syntactic analysis and language representation to the era of large-scale language models, Manning's research experience spans several important stages in machine understanding and language generation.

What he represents is not just the achievement of a particular professor, but also a time scale that is easily overlooked in science news.

The young engineers standing in AI companies today didn't suddenly appear out of the air of Silicon Valley. They stand on decades of accumulated research methods, papers, open courses, laboratory traditions, and peer review systems. Many technologies that seem to have enormous commercial value today may have initially emerged as researchers' long-term inquiry into language, statistics, cognition, and computation.

Many people perceive universities as mere suppliers of talent, as if their only role is to place students into companies after graduation. This understanding underestimates the position of American research universities within the technology system.

Businesses must respond to market competition, revenue growth, and investor expectations, while universities can accommodate problems that don't show immediate commercial returns. They allow researchers to spend years exploring a field that hasn't yet found a market, preserving the continuity of knowledge and providing space for interdisciplinary collaboration. When the market suddenly recognizes the immense value of a particular study, the talent, methods, and ideas have often accumulated on campus over many years.

This is a capability that is difficult to replicate with just one or two policies.

Silicon Valley's strength lies not only in the large number of tech companies clustered nearby, but also in the fluid relationships that exist between these companies, research universities, venture capital, the open-source community, and global talent. A professor may simultaneously mentor students, publish papers, participate in startups, or serve as a company consultant; a PhD student may first conduct research in a lab, then join a cutting-edge company, and finally leave to start their own business; successful entrepreneurs may then return to academia to donate, teach, or support the next generation of research.

Talent doesn't have to follow a fixed path; it can move freely between different institutions. This mobility fosters creativity, but it also brings conflicts of interest, resource concentration, and controversies surrounding the commercialization of academia. It's not perfect, but it allows knowledge to move more quickly beyond academic papers and into products and society.

Countries around the world aspire to establish artificial intelligence centers. The real challenge lies not in building a building, hanging a sign, or purchasing servers, but in creating an institutional environment that allows for the free combination of talent, the ability to start over after failure, and the protection and social use of research results.

On the surface, the competition in artificial intelligence is a competition of models and computing power, but at its core, it is still a competition of whether institutions can continuously produce talent, preserve knowledge, and accommodate dissent.

(Image caption) Stanford's AI research and teaching environment. The young builders who join OpenAI, Anthropic, and various AI startups today didn't suddenly appear in Silicon Valley; behind them lies decades of accumulated research methods, open courses, laboratory traditions, and talent development systems from universities. At its core, the competition in AI also involves a society's ability to continuously produce knowledge, preserve dissent, and provide genuine opportunities for talent.

The new generation is the generation that takes on real responsibility.

This article discusses the "new generation," but I don't want to romanticize youth itself.

The world doesn't automatically entrust important missions to someone just because they are young; the tech industry isn't inherently fair either. A prestigious university background, capital networks, immigration status, family circumstances, and the city one lives in can all influence whether they can enter the heart of the artificial intelligence industry. When the media repeatedly tells stories of a few geniuses achieving rapid success, it easily obscures those who go unseen, as well as the high cost of living, fierce competition, and long-term anxiety behind the technological boom.

What's truly worth studying is why a system is willing to let people who haven't yet accumulated a long track record take on real responsibility.

In many traditional organizations, young people have to wait through a long seniority process to get close to core jobs. Silicon Valley is not without hierarchy, nor is everyone guaranteed equal opportunities, but technical skills can sometimes shorten the waiting time. An engineer in his twenties, if he can solve problems that others cannot, may be involved in work that impacts products globally; a newly formed team, once validated by the market and its technology, can quickly acquire capital and talent.

This mechanism has a cruel side. It requires people to constantly prove themselves, which can easily lead to overwork, rapid attrition, and persistent anxiety about success. However, it at least leaves one door open: seniority is not the only passport, and new skills can still be translated into new positions.

This door is crucial for a society. Each generation in power easily turns experience into a barrier and order into a closed system. When young people can only act within the limits allowed by the old structure, innovation ultimately becomes just a slogan. True renewal means that the system must allow someone without titles and with limited resources to build credibility through concrete work.

This does not mean that all power should be handed over to young people. Artificial intelligence involves social ethics, labor, law, security, and public interest; engineering speed alone is far from sufficient. The new generation needs the knowledge left by their predecessors, as well as the constraints of history, law, and humanities. A healthy handover means allowing experience and new capabilities to meet within open institutions.

The juxtaposition of senior professors and young engineers in the AGI Summit roster is significant in itself. Technological revolutions, driven solely by the speed of the young, may lack sound judgment; conversely, driven solely by the caution of the experienced, they may lack the courage to move forward. A vibrant society needs both types of individuals to coexist and to mutually correct each other.

Capital buys a period of time in the future.

At artificial intelligence conferences, investors are never bystanders.

Venture capital provides more than just funding. It largely determines which research directions can be quickly corporatized, which talents can leave their original institutions to form teams, and which products can continue to grow even before they are profitable. Institutions such as Lightspeed and Menlo Ventures on the AGI Summit's public list represent another important force in the artificial intelligence ecosystem: the power to choose the future in advance before the outcome is clear.

Venture capital is not just buying finished products, but also a period of time before they generate stable revenue. It allows a team to purchase computing power, recruit researchers, and withstand experimental failures, and it gives a technology the opportunity to continue developing before the market understands it.

However, while capital provides time, it also sets a time limit for that time.

The next round of funding, revenue growth, valuation, and exit expectations will, in turn, dictate how quickly a technology must prove itself. When capital is overly concentrated in a few hot areas, the entire industry may be driven by similar business narratives. Companies begin chasing the same products, entrepreneurs are forced to make grander promises, and important but not quickly monetizable public issues may not receive sufficient support.

This power is both constructive and needs to be scrutinized.

Without venture capital, many high-cost, long-cycle technology projects struggle to leave the laboratory; however, a $100 million funding round only indicates that capital is willing to invest, not that a technology will necessarily benefit society. A company's valuation can reflect market expectations, but it cannot replace judgments about its business model, data governance, and public consequences.

Markets typically reward rapid growth, but not necessarily reliable permission design, user protection, and clear accountability mechanisms. Thus, speed may override prudence, and scale may be mistaken for value.

The institutional problem with capital also lies in who will have the patience to support research that is difficult to price in the short term but has long-term public value. Can universities, government research grants, charitable foundations, and public interest organizations preserve an alternative time scale outside the market?

What GFM aims to do is to re-examine capital within the framework of institutions and responsibilities. Does capital support tools to improve productivity, or products for which long-term dependence exists? Does it cultivate fundamental technological capabilities, or short-term gains? When artificial intelligence impacts employment, education, and information order, who shares the investment returns and social costs?

These issues won't appear at every product launch, but they should appear in a report by a reputable media outlet.

(Image caption) Students sit around a lawn at Stanford University discussing research and projects, with the century-old Hoover Tower and main campus buildings in the background. Stanford has long been a major global hub for artificial intelligence, computer science, and innovation and entrepreneurship. Numerous research findings, startups, and tech talents have originated in Silicon Valley, making it a crucial cornerstone for academic research, technological innovation, and talent development in the age of artificial intelligence.

Artificial intelligence is redistributing the time of ordinary people

What technology narratives often overlook most is the lives of ordinary people.

When we talk about model performance, inference capabilities, and funding, we rarely think about how an accountant, a journalist, a small business owner, an immigration lawyer, or a recent graduate will cope with this change. For them, artificial intelligence is not a distant question of civilization, but rather whether they will still need to work the same way tomorrow, whether their experience will still be valuable, and whether they can find their place in the new environment.

The first thing artificial intelligence will likely redistribute is time.

Some tasks that previously took hours to complete can now be drafted in minutes; data that required a professional team to search and organize can be accomplished by a single person using an intelligent agent. OpenAI's data released in 2026 shows that Codex usage has expanded from engineering work to legal, financial, recruitment, and other knowledge-based tasks. While this is based on the company's own published data and requires further extensive and independent research to verify, it at least demonstrates that intelligent agents are transcending traditional technical departments.

This shift can unleash creativity and raise organizational expectations for individual output. When everyone is perceived as being able to complete tasks faster, the time saved may not actually return to the individual; instead, it could become more work.

When the time required for drafting, organizing, searching, and preliminary analysis is reduced, human value is repositioned towards judgment, responsibility, relationships, and creation. However, not everyone will necessarily find more dignified work as a result. For some, artificial intelligence may liberate them from repetitive labor; for others, it may first take away the most easily quantifiable tasks, then demand they prove, at a lower price, what machines cannot yet accomplish.

Technological progress never automatically translates into human freedom.

The Industrial Revolution increased productivity but also led to extremely long working hours; the internet expanded access to information but also bombarded people with a constant stream of messages. Whether artificial intelligence can bring people greater dignity depends on how businesses allocate the benefits of productivity, how the law protects workers, and how education helps people rebuild their capabilities.

If efficiency improvements are only translated into layoffs, price reductions, and higher pressure, then even if society has a smarter model, it may not necessarily become more civilized.

This is a question that has lingered in my mind as I observe these technology builders. Many of them genuinely believe that technology can improve the world, and that belief deserves respect; however, the world does not operate solely according to the inventors' initial wishes. Once a product enters the market, it is shaped by capital, competition, government, and human nature.

Technology builders need society to understand their work, and society also needs them to understand the anxieties of ordinary people. Connecting these two ends cannot be merely propaganda; it must be transparency, dialogue, and the law and institutions.

The world is competing for talent and reasons to believe in a system.

When discussing the competition in artificial intelligence today, people naturally think of chips, data centers, models, and national strategies. These are all important, but there's a deeper question: why would a capable young person be willing to live, work, and take risks in a particular place?

Talent isn't solely attracted by high salaries. Truly outstanding individuals also seek collaborators, supportive research conditions, freedom, legal protection, and the possibility of starting over after failure. They need to believe that research doesn't have to conform to the preferences of those in power, that results can be legally protected, and that companies can operate within relatively predictable rules.

Silicon Valley is not perfect. It has expensive housing, a huge wealth gap, a capital bubble, and intense competition, as well as long-standing identity uncertainty for immigrants. However, it continues to attract talent from all over the world, largely because people from diverse backgrounds can still build new social positions through technology, research, and entrepreneurship.

This is a form of institutional credit.

The state can purchase equipment, provide subsidies, and attract businesses with policies, but institutional credibility requires long-term accumulation. This includes whether courts can handle disputes relatively independently, whether contracts can be enforced, whether academic discussion is free, whether the media can raise questions, and whether a person can retain basic dignity after failure.

In the age of artificial intelligence, international competition will ultimately not be about who can create the stronger model, but also about who can attract the most creative people to stay, cooperate, and place their future within the system.

Computing power can be purchased, equipment can be built, and talent can be attracted with high salaries in the short term; but whether a person is willing to stay long-term often depends on whether he believes that society will respect his work, protect his rights, and allow him to express different opinions to the existing authorities.

What the world is truly competing for is not just talent itself, but also the reasons why talent should believe in a system.

Why do media outlets need their own "Dragon and Tiger List"?

Since founding GFM.News, I have often worked between two forces: the ever-accelerating pace of technology companies and the necessary hesitation that the media must maintain.

Technology demands that people quickly adopt new tools, but journalism requires us to keep asking: Who built these tools, what powers operate them, who bears the risks, and where do the benefits ultimately go?

Of course, we need to report on technology launches, company funding, and market changes, but if we only do that, the media will quickly become an extension of industry public relations. Every company will say its product is faster, stronger, and cheaper, every funding round will be portrayed as a boost to market confidence, and every conference will claim to have brought together world-changing people.

True media work is about discerning structure amidst the noise and preserving memory amidst the speed.

We need to know where a technology comes from, who built it, how capital enters, how permissions are allocated, and who bears the costs; we also need to see the people behind the technology, their ideals, limitations, and the institutional environment in which they operate.

The "AI Era Leaderboard" should not be forgotten after one article. It should become GFM's long-term observation of people and systems: we track the successful, and we also pay attention to those who raise key questions, build public tools, study systemic risks, and try to make technology serve more people.

Some names that occupy center stage today may disappear in a few years; some people who are virtually unknown now may become pioneers in their fields a decade from now. The media cannot accurately predict who will win, but it can record how this era makes choices, how it tries and fails, and how new power forms.

This is the value of news, and also the starting point for historical writing.

A truly valuable list of top performers should not only record who runs the fastest, but also who put forward restrictions when their capabilities expanded rapidly, who was still willing to study problems that were not easy to monetize when capital was at its hottest, and who reserved a door for the rights of ordinary people before technology entered society.

When an era begins to call names

A few days later, people from different cities and countries will walk into the Palace of Fine Arts in San Francisco. Some will be carrying laptops, some will be bringing newly completed products, some will be looking for investment, and some will simply want to hear a lecture from a professor they respect. The venue will be filled with business exchanges, technical debates, and optimism about the future, but there will also be anxiety, competition, and ordinary desires obscured by grand slogans.

There won't actually be a list of top talents that can accurately predict the future. History never chooses people according to the order on the poster.

But we still have reason to pay attention to these faces. Because they are in the very places where research, engineering, capital, and institutions are being recombined. Some are teaching machines to understand language, some are enabling intelligent agents to perform tasks for extended periods, some are researching how to prevent them from crossing boundaries, and others are deciding which technologies deserve the next round of funding.

What they seem to be dealing with today are models, programs, servers, and products; at a deeper level, they are involved in rearranging people's work, knowledge, time, and power.

I am reluctant to readily label any group of people as "the designers of the world." The world is too complex for any single individual to design. Technology creators are constrained by the market, businesses by the law, and institutions are constantly being revised by society. A truly healthy future cannot be defined by a few laboratories, companies, and investment institutions in the absence of public participation.

The best respect for these technology builders is to genuinely understand what they are doing and demand that they accept public scrutiny commensurate with their power. Whether artificial intelligence is trustworthy depends not only on how intelligent it is, but also on whether those who create, deploy, and manage it are willing to acknowledge boundaries, understand responsibility, and still prioritize human dignity over efficiency.

The Palace of Fine Arts in San Francisco once witnessed the industrial age's vision of the future. In the summer of 2026, it will once again house the ambitions of a generation. After the event, the lights will go out, the booths will be dismantled, and the crowds will disperse. What truly remains are the decisions etched into models, products, company policies, and the lives of ordinary people.

Years later, when people look back on this summer, they may find that the world did not suddenly change in a sensational moment. It was slowly changed in the hands of a group of people who were not yet fully understood by history, through a program, an experiment, a study, and a seemingly small choice.

The so-called "Dragon and Tiger List" of the AI era ultimately ranks not only the order of individuals, but also who, when power was rapidly forming, both propelled it forward and were willing to define its boundaries.