Essay · 长文Talent & Organization · 人才与组织

The Next Biggest
Moat in AI

AI 时代最大的护城河
2026 · 05 · 09
When products can be copied and categories renamed in months, the shape of the company itself becomes the moat.
当产品可被复制、品类可被重命名,真正难以复制的,是公司本身的"形状"。
JAYA GUPTA · @JayaGup10 · 3.67M VIEWS
The Premise · 前提01 / Convergence
Everything in AI is converging
AI 领域的一切都在收敛
The application layer is collapsing into infrastructure; infrastructure is moving up into workflows; nearly every startup rebrands as a "transformation company."
应用层正坍缩进基础设施,基础设施又向上侵入工作流,几乎每家创业公司都把自己重新包装成某种"转型公司"。
The words change every few months — context graph, system of action, organizational world model. A category gets named, every site absorbs it, and "inevitable platforms" multiply.
名词每隔几个月就换一批——上下文图谱、行动系统、组织世界模型。一个新品类被命名,所有网站都吸收它,"必然平台"随即泛滥。
When models improve fast and product velocity gets cheap, the visible parts of company-building become easy to imitate.
当模型快速迭代、产品速度变得廉价,公司建设中"看得见"的部分就变得容易模仿。
The harder thing to copy is the institution underneath: how a company attracts exceptional people, organizes ambition, concentrates judgment, and compounds work into a system no rival can reproduce.
真正难以复制的,是底层的"制度":一家公司如何吸引顶尖人才、组织他们的野心、汇聚判断力,并把工作沉淀成对手无法复刻的复利系统。
02 — 15
PREMISE
The Thesis · 核心论点People are the company
“
People are not an input to the company. They are the company.
人不是公司的"投入项",人就是公司本身。因此,公司的"形状"正在成为护城河。
— The shape of the company itself is becoming the moat
03 — 15
THESIS
I
Part One · 第一部分

Great companies are organizational inventions

伟大的公司,本质上是一种组织发明
They build a new institution around a new kind of work — and make a new kind of person possible.
它们围绕一种新工作建立起新制度,并由此让一种新人成为可能。
04 — 15
PART I
Case Studies · 案例New institutions, new protagonists
A structure lets a certain talent finally exist
一种结构,让某类人才终于得以存在
01

OpenAI

前沿模型训练 · Frontier model training
Not academia, not a corporate lab. Safety, policy, product and infrastructure all orbit one center — enabling the researcher who works at the edge of science, product, geopolitics and civilizational risk at once.
既非学术界,也非企业实验室。安全、政策、产品、基础设施都围绕同一个核心运转,催生出能同时游走于科学、产品、地缘政治与文明风险之间的研究者。
02

Palantir

前线部署 · Forward deployment
A new operating institution for broken systems. It took work that was low-status elsewhere — absorbing institutional mess, translating politics into product — and made it central.
为失灵系统打造的新型运营制度。它把别处地位低下的工作——消化制度乱局、把政治翻译成产品——放到了舞台中央。
05 — 15
PART I
II
Part Two · 第二部分

Shape determines who can exist there

"形状"决定了谁能在这里存在
The best companies don't only compete on market or pay. They compete on identity.
最好的公司不只在市场或薪酬上竞争,它们在"身份认同"上竞争。
06 — 15
PART II
The Emotional Promise · 情感承诺What ambitious people are starving for
What do people actually want to feel?
人们真正想要的,是哪种"感觉"?
01Special独一无二Rare, seen, not interchangeable — "only you could build it here."稀有、被看见、不可替代——"只有你才能在这里造出它"。
02Destined命中注定Life bending toward something inevitable; credible only inside a shape positioned to matter.人生正朝某种必然弯曲;只有当公司确实身处关键位置时,这种感觉才可信。
03Not missing out不愿错过Inside the room where the compounding happens. Talent density is itself a shape decision.置身于复利正在发生的房间里。人才密度本身就是一种结构选择。
04Something to prove想要证明The credentialed performer who suspects none of it proves anything — or optionality (McKinsey perfected it).那个被光环包裹、却怀疑这一切毫无意义的人;或是对"可选性"的渴望(麦肯锡将其做到极致)。
05A mission使命召唤A cult around something believed viscerally. The strongest missions make some people refuse to work there.围绕一种发自本能的信念形成的"教派"。最强的使命,会让一部分人拒绝加入。
07 — 15
PART II
The Test · 检验Every emotional promise is structural
“
If it says ownership matters but decision rights are centralized — the promise is fake.
如果它说"主人翁很重要",决策权却高度集中——这个承诺就是假的。每一个情感承诺,都是一个结构承诺。
— Each emotional promise is also a structural promise
08 — 15
TEST
III
Part Three · 第三部分

The question for founders

创始人真正该问的问题
Not "how do we tell a better story?" — but what kind of person can only become themselves here?
不是"怎样讲一个更好的故事",而是——"什么样的人,只有在这里才能成为他自己?"
09 — 15
PART III
Alignment · 一致性Belief must show up in structure
The story's altitude must match the shape
故事的"高度",必须匹配公司的"形状"
If customer proximity is the moat若"贴近客户"是护城河 → customer-facing work has to be high status.面向客户的工作就必须是高地位的。
If speed is the moat若"速度"是护城河 → decision rights have to be pushed to the edge.决策权就必须下放到最前线。
If talent density is the moat若"人才密度"是护城河 → average people cannot set the operating pace.平庸者就不能定义运转的节奏。
If deployment is the moat若"部署落地"是护城河 → people closest to reality need power, not just responsibility.最贴近现实的人,需要的是权力,而不只是责任。
10 — 15
PART III
IV
Part Four · 第四部分

And for the people choosing

而对于做选择的人
You commit years to a vision and a shape — and recruiting is unusually bad at revealing either. The truth appears under pressure.
你把数年人生交付给一种愿景和一种结构,而招聘恰恰最不擅长揭示这两者。真相,要到压力之下才会显现。
11 — 15
PART IV
The Distinction · 区分Validation vs. structure
Being chosen is not being seen
"被选中",不等于"被看见"
Being chosen — emotional
被选中 · 情感层面

“We believe in you.”"我们相信你。"

  • You are special你很特别
  • Proximity instead of authority用"接近权力"替代"实际权力"
  • Reassurance instead of economics用"安抚"替代"经济回报"
  • "Trust me" instead of a mechanism用"相信我"替代书面机制
Being seen — structural
被看见 · 结构层面

Here is what's real.这些,都是真实的。

  • Here is the scope这是你的职责范围
  • Here is the authority这是你的权限
  • Here is the economic participation这是你的经济参与
  • Here is what changes if you succeed这是你成功后会改变的东西
12 — 15
PART IV
The Caution · 警示Promises denominated in time
“
The most dangerous promises are denominated in time.
最危险的承诺,是以"时间"计价的。但时间离开时,从不会向你打招呼。
— You can feel deeply valued and materially stuck at once
13 — 15
CAUTION
The New Moats · 新护城河What AI will and won't make easy
In Closing · 结语
AI makes many things easy to copy — surfaces, workflows, pitch language. It will not make it easy to build a new institution. AI 让很多东西变得易于复制——界面、工作流、话术。但它不会让"建立一种新制度"变得容易。
CONCENTRATE THE RIGHT PEOPLE · GIVE THEM THE RIGHT AUTHORITY PUT THEM CLOSE TO THE RIGHT PROBLEMS · COMPOUND THEIR JUDGMENT
14 — 15
CLOSING
— END · 完
The next market will reward companies built in shapes the old one could not produce. 下一个时代,将奖励那些以旧时代无法塑造的"形状"建立起来的公司。
SOURCE — JAYA GUPTA, "THE NEXT BIGGEST MOAT IN AI" · 2026.05.09