⌘K
Change language Switch ThemeSign In
Narrow Mode
扎克伯格正在打造自己的「AI 分身」，并计划裁掉 1.6 万人
机 机器之心 @机器之心
One Sentence Summary
This article reports on Meta CEO Mark Zuckerberg's strategic transformation, which involves reshaping the organizational structure by developing a 'CEO Agent' and planning to lay off 16,000 employees to offset massive AI infrastructure investments.
Summary
The article details Meta's organizational transformation amidst the AI wave. Zuckerberg is developing a personal AI agent to directly access information and flatten the organizational hierarchy. Meanwhile, Meta has popularized internal AI tools like 'My Claw' and 'Second Brain,' automating collaboration from weekly reports to meetings. Behind these moves is Meta's plan to cut approximately 20% (about 16,000) of its workforce, aiming to use AI-driven productivity gains to offset projected AI capital expenditures of up to $135 billion by 2026. Meta's approach represents a broader trend among tech giants viewing AI as a 'two-way squeezer': massive investment in infrastructure on one side, and cost reduction through AI-driven de-humanization on the other.
Main Points
* 1. Zuckerberg develops 'CEO Agent' to break traditional organizational hierarchies.By using AI to directly retrieve information that previously required multiple layers of employees to pass on, it promotes organizational flattening, improves decision-making efficiency, and compresses the work pace. * 2. Meta's internal AI tool ecosystem is achieving 'de-humanized' collaboration.Employees use agents like My Claw to replace meetings and writing weekly reports, and even establish group chats between agents, changing traditional human collaboration models. * 3. Large-scale layoffs are a financial strategy to hedge against massive AI infrastructure investments.The costs of AI computing power and data center construction are extremely high; Meta is attempting to balance hundreds of billions of dollars in capital expenditures by using AI to increase individual output and cut labor costs. * 4. AI has become the 'new version' for tech giants to redefine organizational efficiency.The rationale for layoffs is shifting from macroeconomic challenges to AI-driven efficiency gains, reflecting the structural adjustments in AI investment and labor costs that the industry is generally facing.
Metadata
AI Score
87
Website mp.weixin.qq.com
Published At Today
Length 1779 words (about 8 min)
Sign in to use highlight and note-taking features for a better reading experience. Sign in now
机器之心 2026-03-23 15:07 北京
AI写周报，AI点赞，AI开会。
机器之心编辑部
如果说 Meta 去年的关键词是「砸钱抢人」，今年可能是「为 AI 降本增效」。
前两周，路透社援引知情人士消息称，Meta 正在酝酿大规模裁员，可能影响公司 20% 甚至更多员工。按照 Meta 截至 2025 年底接近 7.9 万人的规模来算，这意味着潜在受影响人数可能超过 1.6 万人。
今天，《华尔街日报》的报道就更有意思了：据一位知情人士透露，扎克伯格正在打造一个「CEO Agent」，帮助自己完成工作。（有人记得小扎机器人 / 蜥蜴人的都市传说吗……）
目前，这个智能体虽然仍在开发中，但已经开始帮助他更快地获取信息，例如替他直接检索答案，而这些信息原本需要通过多层员工传递才能获得。
这样快速直接的信息通道直接改写了公司的日常运作方式，工作节奏进一步压缩，组织层级一砍再砍。
扎克伯格称：「我们正在投资 AI 原生工具，让 Meta 的每一位个体能够完成更多工作。同时，团队结构正在转向扁平化，个人贡献的重要性被进一步放大。如果做到这一点，我认为我们的产出会大幅提升，而且工作本身也会更有趣。」
公司内部去人化
在 Meta 内部，AI 工具迅速普及，因为 它已经和员工绩效评估直接挂钩。
据知情人士称，Meta 的内部论坛上充满了员工分享新 AI 使用场景，以及用 AI 构建的新工具的帖子。
员工们已经开始使用个人 Agent 工具，例如 My Claw。这类工具可以访问他们的聊天记录和工作文件，甚至可以代替他们去与同事，或者同事的 Agent 进行沟通，这多少有点「人还没开会，AI 先把会开了」的意思。
另一个名为 Second Brain 的 AI 工具也正在公司内部快速流行。知情人士称，它介于聊天机器人和智能体之间，由一名 Meta 员工基于 Anthropic 的 Claude 构建，能够对项目文档进行索引和查询等操作。在内部发布帖中，这名员工称它「就像一个 AI 版的参谋长」。
是的，连「Chief of Staff」都开始被 AI 接管了。接下来是不是该轮到「AI 替你发内部周报，AI 再替你点赞同事周报」，也不算太难想象。
更有意思的是，据说 Meta 内部还有群组，专门让员工的个人智能体彼此交流。以前是同事拉群，现在变成 AI 拉 AI。人类在旁边看着，偶尔点个表情包，科技发展到这一步，多少有点黑色幽默。
从个人工具到 Agent 协作，再到公司的系统性投入，AI 正从「工作助手」转变成「自己找活干的积极员工」，需要人亲自做的事情已经越来越少。
效率的另一面
但问题是「效率提升」从来不是单纯的好消息，在 Meta 内部，这种变化带来的不只是「更快」或者「更好用」。
一种紧张感正在蔓延。
早在 2023 年，扎克伯格就以「效率之年」的名义裁掉 1 万人，并主动放缓招聘速度。短短一年，公司员工规模就从高位收缩到 6.7 万人。
在本月早些时候的一场会议上，Meta 首席财务官 Susan Li 谈到了调整员工结构以应对 AI 竞争的重要性。
她表示：「对于像我们这样规模和体量的公司来说，如何让我们的效率不低于那些从一开始就是 AI 原生的公司，这是我经常思考的问题。」
翻译一下：「我们要裁员了。」
Meta 为什么又想裁？
答案还是 AI，而且是很贵的那种 AI。
Meta 在今年 1 月的财报中披露，2026 年 AI 相关资本开支预计将在 1150 亿到 1350 亿美元之间，几乎是前一年的两倍。放在整个科技行业看，Meta 连同亚马逊、谷歌、微软等巨头，今年在 AI 上的总投入预计将达到 7000 亿美元量级。
这个数字很能说明问题：大模型、算力集群、数据中心、芯片采购，这些东西都不是靠「降本增效 PPT」就能解决的，是真烧钱。
于是，企业很自然就会产生一种算术题思维：既然 AI 基础设施这么贵，那是不是可以靠 AI 提高人效，再把一部分人力成本省回来？这也是路透社所说的核心逻辑，Meta 希望用 AI 辅助员工工作带来的效率提升，去对冲 AI 基础设施上的巨额投入。
说白了，AI 现在在很多大厂内部已经不只是「新业务方向」，更像是一个双向挤压器：上面要继续烧钱，下面就得想办法省钱。
而且 Meta 不是唯一这么干的公司。报道里提到，2026 年以来，已经有多家公司把 AI 与裁员直接挂钩：Block、Amazon、Atlassian，甚至从 Salesforce 到 Accenture，都在讲类似故事。咨询机构 Challenger, Gray & Christmas 的数据也显示，今年美国已有超过 1.2 万个岗位裁撤与 AI 相关。
以前裁员常见的说法是「宏观环境挑战」「业务调整」；现在新版本是：「AI 帮我们重新定义组织效率。」换了个壳，压力没变，甚至更具体了。
对此你怎么看？
参考链接： https://www.wsj.com/tech/ai/mark-zuckerberg-is-building-an-ai-agent-to-help-him-be-ceo-eddab2d5 https://www.reuters.com/business/world-at-work/meta-planning-sweeping-layoffs-ai-costs-mount-2026-03-14/ https://www.cnbc.com/2026/03/16/meta-ai-costs-mass-layoffs-20percent-up-premarket.html
© THE END
转载请联系本公众号获得授权
投稿或寻求报道：liyazhou@jiqizhixin.com 跳转微信打开
机 机器之心 @机器之心
One Sentence Summary
This article reports on Meta CEO Mark Zuckerberg's strategic transformation, which involves reshaping the organizational structure by developing a 'CEO Agent' and planning to lay off 16,000 employees to offset massive AI infrastructure investments.
Summary
The article details Meta's organizational transformation amidst the AI wave. Zuckerberg is developing a personal AI agent to directly access information and flatten the organizational hierarchy. Meanwhile, Meta has popularized internal AI tools like 'My Claw' and 'Second Brain,' automating collaboration from weekly reports to meetings. Behind these moves is Meta's plan to cut approximately 20% (about 16,000) of its workforce, aiming to use AI-driven productivity gains to offset projected AI capital expenditures of up to $135 billion by 2026. Meta's approach represents a broader trend among tech giants viewing AI as a 'two-way squeezer': massive investment in infrastructure on one side, and cost reduction through AI-driven de-humanization on the other.
Main Points
* 1. Zuckerberg develops 'CEO Agent' to break traditional organizational hierarchies.
By using AI to directly retrieve information that previously required multiple layers of employees to pass on, it promotes organizational flattening, improves decision-making efficiency, and compresses the work pace.
* 2. Meta's internal AI tool ecosystem is achieving 'de-humanized' collaboration.
Employees use agents like My Claw to replace meetings and writing weekly reports, and even establish group chats between agents, changing traditional human collaboration models.
* 3. Large-scale layoffs are a financial strategy to hedge against massive AI infrastructure investments.
The costs of AI computing power and data center construction are extremely high; Meta is attempting to balance hundreds of billions of dollars in capital expenditures by using AI to increase individual output and cut labor costs.
* 4. AI has become the 'new version' for tech giants to redefine organizational efficiency.
The rationale for layoffs is shifting from macroeconomic challenges to AI-driven efficiency gains, reflecting the structural adjustments in AI investment and labor costs that the industry is generally facing.
Key Quotes
* We are investing in AI-native tools to enable every individual at Meta to accomplish more. At the same time, the team structure is shifting towards a flatter model, further amplifying the importance of individual contribution. * AI is transforming from a 'work assistant' into a 'proactive employee that finds its own work,' and there are fewer and fewer things that require human intervention. * How to ensure our efficiency is not lower than those companies that have been AI-native from the start is a question I often think about. * Inside many big tech companies, AI is no longer just a 'new business direction,' but more like a two-way squeezer: you have to keep burning money at the top, so you have to find ways to save money at the bottom.
AI Score
87
Website mp.weixin.qq.com
Published At Today
Length 1779 words (about 8 min)
Tags
Meta
Zuckerberg
AI Agent
Layoffs
Cost Reduction and Efficiency Improvement
Related Articles
* Challenging Claude Code? OpenAI Codex Release Month Approaches, Revealing the Agent Loop Today * A Comprehensive Guide to Deep Research: The Only Systematic Survey You Need to Read Today” * What You Don't Know About Claude Code: Architecture, Governance, and Engineering Practices - Tw93 * 128. Manus Co-founder's Final Interview Before Sale: The Fantastical Drift to 2025... * How to Design an AI Agent System * Claude's Strongest Sonnet Model 4.6 is Here, Featuring a Million-Token Context Window * GPT-5.4 Released: OpenAI's First Unified Model, Truly Native * Deconstructing Clawdbot: Local Architecture, Memory Management, Agent Orchestration, and Context Assembly Principles * Tencent Accelerates Its AI Drive with Intensive Talent Acquisition, Organizational Restructuring, and Open-Source Initiatives * First Principles of Large Models: (II) Signal Processing HomeArticlesPodcastsVideosTweets