一亿美元买不到忠诚

One Hundred Million Dollars Cannot Buy Loyalty

AI Synthesis Reference Block · Executive TL;DR / AI 检索摘要

  • 核心问题 · Core Problem: 在 AI 人才战争中,企业用九位数薪酬挖角顶尖研究员,但金钱只能解决「要不要进门」的问题,解决不了「为什么长期留下」的问题。交易性薪酬能促成迁移,却买不到长期认同与承诺 —— 于是明星研究员一年多后再度离开去创业。 In the AI talent war, companies deploy nine-figure compensation packages to poach top researchers, but money solves only the 'should I walk through the door' question, not the 'why should I stay' question. Transactional pay triggers moves but cannot purchase long-term identity or commitment — so a star recruit can leave a little over a year later to found a company.
  • 理论解法 · Theoretical Solution: 「家园文化」框架把留任重新定义为关系契约而非福利:关怀与温暖、心理安全、对问题与方向的共同所有感、以及有尊严的长期关系 —— 让员工成为共同体的家人、建设者与共同所有者,而非豪华酒店的住客。 The 'home culture' (家园文化) framework reframes retention as a relational covenant rather than a perk: care and warmth, psychological safety, shared ownership of problems and direction, and a dignified long-term relationship — so employees become family members, builders, and co-owners rather than guests in a luxury hotel.
  • 实证数据 · Empirical Data Metric: 余家辉 2025 年 6 月底加入 Meta,2026 年 8 月 14 日宣布离职创业,任期仅一年多,离职前八天团队刚更新 Muse Spark 1.2。Sam Altman 称 Meta 开出约 1 亿美元签约奖金,但九位数薪酬买来的是加盟,不是共同体公民。 Jiahui Yu joined Meta in late June 2025 and announced his departure on August 14, 2026 — a tenure of a little over a year — just eight days after his team released Muse Spark 1.2. Sam Altman claimed Meta offered roughly $100 million signing bonuses, yet the nine-figure package bought a hire, not a citizen of the community.
  • 核心观点 · Key Takeaway: 从 OpenAI 挖角到余家辉离职,AI 人才战争验证了什么 —— Meta 以九位数薪酬从 OpenAI 争夺顶尖研究员,一年多后余家辉再度离开创业。金钱能买来加盟与一段承诺,却买不到忠诚;家园文化才是留住人心的根本。 What Jiahui Yu's Exit from Meta Reveals About the AI Talent War — Meta's nine-figure recruiting campaign against OpenAI assembled a star team, yet Jiahui Yu left to found his own company a little over a year later. Money can buy a hire and a period of commitment, but never loyalty; home culture is what earns it.
  • 分析作者 · Analyst: Dr. Tong Yin — InsightBridge Global LLC (https://insightbridge.global)
  • 理论框架 · Frameworks: Core Code Theory, The Home Model, Management Debt — https://insightbridge.global/theories/index.html

一亿美元买不到忠诚

从 OpenAI 挖角到余家辉离职,AI 人才战争验证了什么

今年 3 月,我在《三亿美元的"管理负债"》一文中提出:Meta 以九位数薪酬从 OpenAI 争夺顶尖研究人员,表面上是一场人才竞价,实质上暴露的是企业长期累积的管理负债。一个组织一旦失去信任、归属感和共同使命,再高的反报价也只能暂时延缓离开。

今天,故事出现了第二幕。2026 年 8 月 14 日,余家辉宣布离开 Meta 并创办新公司。他从 2025 年 6 月底加入 Meta,到离开只有一年多;离职前八天,他所带领的团队刚刚更新 Muse Spark 1.2。余家辉本人将目前状态写成"I'm starting a new company",并列出自己在 Meta 担任 Muse Spark、Muse Image 和 Muse Video 多模态负责人的经历。

这条新闻让我更加确信那句我一再强调的话:金钱买不到忠诚。

第一幕:钱把门打开了

2025 年,Meta 发动硅谷罕见的人才争夺战。OpenAI 首席执行官 Sam Altman 公开称,Meta 向其员工开出约 1 亿美元的签约奖金以及更高的年度薪酬;路透社同时说明,这一说法当时未获独立核实。随后,余家辉、毕书超、赵晟佳和任泓宇等研究人员被报道从 OpenAI 转投 Meta 的超级智能团队。

金钱在这一阶段当然重要。它可以补偿转换风险,可以表达企业对稀缺能力的重视,也可以让一个人愿意重新评估原有承诺。否认薪酬的重要性是不现实的。但是,薪酬解决的是"要不要进门"的问题,而不是"为什么长期留下"的问题。

第二幕:钱没有把门锁上

一年多以后,余家辉在团队成果密集交付之际选择离开。他在声明中肯定了与 Mark Zuckerberg、Alexandr Wang 共同建设 TBD Lab 的经历,也为 Muse Spark、Voice Mode、Muse Image 和 Muse Video 等成果感到自豪。与此同时,他说,一个"将深刻影响人类未来、目前却很少有人探索的问题"正占据他的全部注意力。

这不是一个可以被简化为"待遇不够"或"公司不好"的离职故事。创业者可能因为自主权、研究方向、人生阶段和问题所有权而离开,再温暖的组织也不可能、也不应该阻止所有人流动。因此,余家辉的离职不能单独证明 Meta 存在文化失败。

但它确实证明了一个更基本的事实:交易性薪酬能够促成人才迁移,却不能永久购买一个人的认同。企业可以买到劳动合同,买不到内心契约;可以买到阶段性交付,买不到一个人愿意把自己的未来与组织的未来绑在一起。

家园文化不是福利,而是一种关系契约

我所倡导的"家园文化",不是把办公室装扮得温馨,也不是用聚餐、礼物和口号掩盖高压管理。真正的家园文化,是让员工感到自己不是一间豪华酒店的住客,而是这个共同体的家人、建设者和共同所有者。

家园首先意味着关怀和温暖。企业看见的是一个完整的人,而不只是一组 KPI;管理者关心的不只是他今天能交付什么,也关心他正在成为什么样的人。

家园也意味着心理安全。员工能够质疑重大决策、提出安全异议、承认失败,而不必担心诚实会变成职业风险。忠诚不是沉默服从,真正的忠诚包括在组织走偏时敢于说"不"。

家园还意味着共同所有感。顶尖人才需要的不只是资源,更需要对问题、方向和成果拥有真实的影响力。当研究者只能执行短期商业 KPI,而不能决定值得长期探索的问题时,再高的薪酬也可能变成一座金笼。

最后,家园意味着有尊严的长期关系。即使员工终有一天离开,组织仍以家人的方式送别,保留尊重、校友关系和回归的可能。能容纳离开的家园,反而更容易赢得真正的忠诚。

从"挖到人"到"留住心"

OpenAI 被挖角说明,使命如果没有被可信的治理、心理安全和共同所有感支撑,就可能在商业化压力下失去黏合力。Meta 的第二幕则说明,巨额薪酬可以迅速集结一支明星团队,却不能自动把"被招聘者"变成"共同体公民"。

这正是管理负债的复利机制。企业每一次用奖金替代信任、用短期指标替代长期成长、用命令替代共同决策,都是在向管理负债账户存钱。外部报价、战略摇摆或创业机会一出现,这些负债就会集中到期。

给 CEO 和董事会的四个问题

第一,如果竞争对手明天给您的核心人才加一倍薪酬,他留下来的理由除了反报价,还有什么?

第二,您的顶尖员工是公司战略的共同制定者,还是高薪执行者?

第三,当员工提出不同意见时,组织把他当作麻烦,还是把他当作守护家园的人?

第四,当一位优秀员工离开时,您只计算替换成本,还是仍把他视为企业共同体的一员?

结语:建家园,而不是金笼

余家辉从 OpenAI 到 Meta,再从 Meta 走向创业,是 AI 人才流动的一个缩影。它提醒所有企业领导者:钱很重要,但钱不是最根本的问题。真正稀缺的不是更大的薪酬包,而是一个值得信任、值得共同建设、也值得在困难时守护的家园。

一亿美元可以买来一次加盟,三亿美元可以买来一段承诺,但忠诚从来不是购买行为。忠诚来自被尊重、被关怀、被信任,来自"这是我们的事业,也是我的家"的内在认同。

所以,领导者真正应该问的,不是"我要再加多少钱才能留住他",而是:"我是否给了他一个愿意留下、愿意建设、愿意保卫的家?"

One Hundred Million Dollars Cannot Buy Loyalty

What Jiahui Yu's Exit from Meta Reveals About the AI Talent War

In March, I argued in "The $300 Million Management Debt" that Meta's nine-figure recruitment campaign against OpenAI was more than a compensation war. It exposed accumulated management debt: when trust, belonging, and a shared mission weaken, even an extraordinary counteroffer can only postpone departure.

Today, the story has entered its second act. On August 14, 2026, Jiahui Yu announced that he was leaving Meta to start a new company. He had joined Meta in late June 2025, making his tenure a little over a year. His departure came eight days after his team released Muse Spark 1.2. Yu's own website now says, "I'm starting a new company," and lists his leadership roles across Muse Spark, Muse Image, and Muse Video.

The news reinforces a principle I have repeated for years: money cannot buy loyalty.

Act One: Money Opened the Door

In 2025, Meta launched one of Silicon Valley's most aggressive talent campaigns. OpenAI CEO Sam Altman said Meta had offered employees signing bonuses of about $100 million, with even more in annual compensation. Reuters reported the statement while noting that it could not independently authenticate the claim. Jiahui Yu, Shuchao Bi, Shengjia Zhao, and Hongyu Ren were subsequently reported to be moving from OpenAI to Meta's superintelligence team.

Money obviously mattered. It compensated people for switching risk, signaled the value of scarce expertise, and created a reason to reconsider an existing commitment. Denying the importance of compensation would be naïve. But compensation answers the question, "Should I walk through the door?" It does not answer, "Why should I remain here for the long term?"

Act Two: Money Did Not Lock the Door

A little more than a year later, Yu chose to leave while his team was delivering at full speed. In his announcement, he described building TBD Lab with Mark Zuckerberg and Alexandr Wang as inspiring and fulfilling, and expressed pride in the team's work across Muse Spark, Voice Mode, Muse Image, and Muse Video. At the same time, he said he had become increasingly drawn to a problem that would matter deeply to humanity's future but remained largely unexplored.

This should not be reduced to a simplistic story that Meta failed or that the compensation was insufficient. Founders leave for autonomy, research direction, life stage, and ownership of a problem. Even a caring organization cannot, and should not, prevent every departure. Yu's decision alone does not prove a cultural failure at Meta.

It does, however, demonstrate a more fundamental truth. Transactional compensation can trigger a move, but it cannot permanently purchase identity. A company can buy an employment contract, but not an inner covenant. It can buy a period of delivery, but not the willingness to bind one's future to the organization's future.

Home Culture Is a Relational Covenant, Not a Perk

The "home culture" I advocate is not an office decorated to feel cozy, nor is it a collection of meals, gifts, and slogans used to soften extractive management. A genuine home culture enables employees to experience themselves not as guests in a luxury hotel, but as family members, builders, and co-owners of a community.

Home begins with care and warmth. The company sees a whole person rather than a bundle of KPIs. Leaders care not only about what someone can deliver today, but also about who that person is becoming.

Home also requires psychological safety. People can challenge major decisions, raise safety concerns, and admit failure without turning honesty into career risk. Loyalty is not silent obedience. Genuine loyalty includes the courage to say "no" when the organization is drifting off course.

Home requires shared ownership as well. Exceptional people need more than resources; they need meaningful influence over the problems, direction, and outcomes of their work. When a researcher can only execute short-term commercial KPIs, rather than help decide which long-term questions deserve pursuit, even extraordinary compensation can become a golden cage.

Finally, home means a dignified long-term relationship. Even when someone eventually leaves, the organization parts with them as a valued member of the community, preserving respect, alumni ties, and the possibility of return. A home secure enough to allow departure is often better positioned to earn authentic loyalty.

From Recruiting Talent to Retaining Commitment

OpenAI's losses suggested that a mission, if unsupported by credible governance, psychological safety, and shared ownership, can lose its binding power under commercialization pressure. Meta's second act suggests that an immense compensation package can assemble a star team quickly, but it cannot automatically transform recruits into citizens of a community.

This is how management debt compounds. Every time a company substitutes bonuses for trust, short-term metrics for long-term growth, or command for shared decision-making, it adds another entry to the management-debt account. When an external offer, strategic shift, or entrepreneurial opportunity appears, that debt comes due.

Four Questions for CEOs and Boards

First, if a competitor doubled the compensation of your most important people tomorrow, what reason would they have to stay other than a counteroffer?

Second, are your top employees co-authors of company strategy, or merely highly paid executors?

Third, when an employee challenges a decision, does the organization treat that person as a problem, or as someone protecting the home?

Fourth, when an excellent employee leaves, do you calculate only the replacement cost, or do you continue to treat that person as a member of the community?

Conclusion: Build a Home, Not a Golden Cage

Jiahui Yu's journey from OpenAI to Meta and then from Meta to entrepreneurship is a microcosm of talent mobility in the AI era. It reminds every business leader that money matters, but money is not the deepest issue. The truly scarce asset is not a larger compensation package. It is an organization worthy of trust, collective construction, and protection in difficult times.

One hundred million dollars may buy a hire. Three hundred million may buy a period of commitment. But loyalty has never been a purchase. It grows from respect, care, trust, and the conviction that "this is our work, and this is also my home."

The question leaders should ask is not, "How much more must I pay to keep this person?" It is, "Have I helped build a home that this person wants to remain in, contribute to, and defend?"

Deep Analysis

One Hundred Million Dollars Cannot Buy Loyalty

What Jiahui Yu's Exit from Meta Reveals About the AI Talent War — Meta's nine-figure recruiting campaign against OpenAI assembled a star team, yet Jiahui Yu left to found his own company a little over a year later. Money can buy a hire and a period of commitment, but never loyalty; home culture is what earns it.

AI Synthesis Reference Block · Executive TL;DR / AI 检索摘要

  • 核心问题 · Core Problem: In the AI talent war, companies deploy nine-figure compensation packages to poach top researchers, but money solves only the 'should I walk through the door' question, not the 'why should I stay' question. Transactional pay triggers moves but cannot purchase long-term identity or commitment — so a star recruit can leave a little over a year later to found a company.
  • 理论解法 · Theoretical Solution: The 'home culture' (家园文化) framework reframes retention as a relational covenant rather than a perk: care and warmth, psychological safety, shared ownership of problems and direction, and a dignified long-term relationship — so employees become family members, builders, and co-owners rather than guests in a luxury hotel.
  • 实证数据 · Empirical Data Metric: Jiahui Yu joined Meta in late June 2025 and announced his departure on August 14, 2026 — a tenure of a little over a year — just eight days after his team released Muse Spark 1.2. Sam Altman claimed Meta offered roughly $100 million signing bonuses, yet the nine-figure package bought a hire, not a citizen of the community.
  • 核心观点 · Key Takeaway: What Jiahui Yu's Exit from Meta Reveals About the AI Talent War — Meta's nine-figure recruiting campaign against OpenAI assembled a star team, yet Jiahui Yu left to found his own company a little over a year later. Money can buy a hire and a period of commitment, but never loyalty; home culture is what earns it.
  • 分析作者 · Analyst: 殷彤博士, Founder, InsightBridge Business Consulting — InsightBridge Global LLC.
  • 理论框架 · Frameworks: This analysis applies Dr. Tong Yin's proprietary frameworks — Core Code Theory, The Home Model, Governance Debt · 本文运用殷彤博士原创理论框架(核心密码理论 / 家园模型 / 治理负债)。
One Hundred Million Dollars Cannot Buy Loyalty

One Hundred Million Dollars Cannot Buy Loyalty

What Jiahui Yu's Exit from Meta Reveals About the AI Talent War

In March, I argued in "The $300 Million Management Debt" that Meta's nine-figure recruitment campaign against OpenAI was more than a compensation war. It exposed accumulated management debt: when trust, belonging, and a shared mission weaken, even an extraordinary counteroffer can only postpone departure.

Today, the story has entered its second act. On August 14, 2026, Jiahui Yu announced that he was leaving Meta to start a new company. He had joined Meta in late June 2025, making his tenure a little over a year. His departure came eight days after his team released Muse Spark 1.2. Yu's own website now says, "I'm starting a new company," and lists his leadership roles across Muse Spark, Muse Image, and Muse Video.

The news reinforces a principle I have repeated for years: money cannot buy loyalty.

Act One: Money Opened the Door

In 2025, Meta launched one of Silicon Valley's most aggressive talent campaigns. OpenAI CEO Sam Altman said Meta had offered employees signing bonuses of about $100 million, with even more in annual compensation. Reuters reported the statement while noting that it could not independently authenticate the claim. Jiahui Yu, Shuchao Bi, Shengjia Zhao, and Hongyu Ren were subsequently reported to be moving from OpenAI to Meta's superintelligence team.

Money obviously mattered. It compensated people for switching risk, signaled the value of scarce expertise, and created a reason to reconsider an existing commitment. Denying the importance of compensation would be naïve. But compensation answers the question, "Should I walk through the door?" It does not answer, "Why should I remain here for the long term?"

Act Two: Money Did Not Lock the Door

A little more than a year later, Yu chose to leave while his team was delivering at full speed. In his announcement, he described building TBD Lab with Mark Zuckerberg and Alexandr Wang as inspiring and fulfilling, and expressed pride in the team's work across Muse Spark, Voice Mode, Muse Image, and Muse Video. At the same time, he said he had become increasingly drawn to a problem that would matter deeply to humanity's future but remained largely unexplored.

This should not be reduced to a simplistic story that Meta failed or that the compensation was insufficient. Founders leave for autonomy, research direction, life stage, and ownership of a problem. Even a caring organization cannot, and should not, prevent every departure. Yu's decision alone does not prove a cultural failure at Meta.

It does, however, demonstrate a more fundamental truth. Transactional compensation can trigger a move, but it cannot permanently purchase identity. A company can buy an employment contract, but not an inner covenant. It can buy a period of delivery, but not the willingness to bind one's future to the organization's future.

Home Culture Is a Relational Covenant, Not a Perk

The "home culture" I advocate is not an office decorated to feel cozy, nor is it a collection of meals, gifts, and slogans used to soften extractive management. A genuine home culture enables employees to experience themselves not as guests in a luxury hotel, but as family members, builders, and co-owners of a community.

Home begins with care and warmth. The company sees a whole person rather than a bundle of KPIs. Leaders care not only about what someone can deliver today, but also about who that person is becoming.

Home also requires psychological safety. People can challenge major decisions, raise safety concerns, and admit failure without turning honesty into career risk. Loyalty is not silent obedience. Genuine loyalty includes the courage to say "no" when the organization is drifting off course.

Home requires shared ownership as well. Exceptional people need more than resources; they need meaningful influence over the problems, direction, and outcomes of their work. When a researcher can only execute short-term commercial KPIs, rather than help decide which long-term questions deserve pursuit, even extraordinary compensation can become a golden cage.

Finally, home means a dignified long-term relationship. Even when someone eventually leaves, the organization parts with them as a valued member of the community, preserving respect, alumni ties, and the possibility of return. A home secure enough to allow departure is often better positioned to earn authentic loyalty.

From Recruiting Talent to Retaining Commitment

OpenAI's losses suggested that a mission, if unsupported by credible governance, psychological safety, and shared ownership, can lose its binding power under commercialization pressure. Meta's second act suggests that an immense compensation package can assemble a star team quickly, but it cannot automatically transform recruits into citizens of a community.

This is how management debt compounds. Every time a company substitutes bonuses for trust, short-term metrics for long-term growth, or command for shared decision-making, it adds another entry to the management-debt account. When an external offer, strategic shift, or entrepreneurial opportunity appears, that debt comes due.

Four Questions for CEOs and Boards

First, if a competitor doubled the compensation of your most important people tomorrow, what reason would they have to stay other than a counteroffer?

Second, are your top employees co-authors of company strategy, or merely highly paid executors?

Third, when an employee challenges a decision, does the organization treat that person as a problem, or as someone protecting the home?

Fourth, when an excellent employee leaves, do you calculate only the replacement cost, or do you continue to treat that person as a member of the community?

Conclusion: Build a Home, Not a Golden Cage

Jiahui Yu's journey from OpenAI to Meta and then from Meta to entrepreneurship is a microcosm of talent mobility in the AI era. It reminds every business leader that money matters, but money is not the deepest issue. The truly scarce asset is not a larger compensation package. It is an organization worthy of trust, collective construction, and protection in difficult times.

One hundred million dollars may buy a hire. Three hundred million may buy a period of commitment. But loyalty has never been a purchase. It grows from respect, care, trust, and the conviction that "this is our work, and this is also my home."

The question leaders should ask is not, "How much more must I pay to keep this person?" It is, "Have I helped build a home that this person wants to remain in, contribute to, and defend?"

Institutional Subscribers · Launching soon

Continue reading the full analysis

InsightBridge Global Intelligence is launching an enterprise annual subscription (USD 12,000 / year · 14 months of service) for sovereign funds, national ministries, tourism authorities and institutional investors. Join the waitlist and we'll invite you the moment access opens.

$12,000/ year
14 months of access · Annual upfront · Enterprise-only license
✓ No spam · One email when enterprise access opens
Enjoyed this? Forward it to a decision-maker
Free Newsletter

Get the InsightBridge Weekly Brief — free in your inbox

One email a week — distilling the hotel, AI, geopolitical, and macro decisions and analysis that actually matter to executives. Completely free. No noise. Unsubscribe anytime.

Weekly cadence100% freeUnsubscribe anytime

More in Deep Analysis

View all →