全球智能危机的发展进程及其后果

at if our AI bullishness continues to be right...and what if that’s actually bearish? 果我们的对 AI 的乐观持续正确……⽽这反⽽成了利空，会怎样？ at follows is a scenario, not a prediction.This isn’t bear porn or AI doomer fan-fiction.e sole intent of this piece is modeling a scenario that’s been relatively underexplored. Ournd Alap Shah posed the question, and together we brainstormed the answer. We wrote thist, and he’s written two others you canfind here. 下情景为设想，⽽⾮预测。这并⾮唱空轰动或⼈⼯智能末⽇主义同⼈⽂。本⽂唯⼀的，是对⼀个相对较少被探讨的情景进⾏建模。我们的朋友 Alap Shah 提出了问，我们共同头脑风暴给出答案。这⼀部分是我们写的，他还写了另外两部分，可在处找到。 pefully, reading this leaves you more prepared for potential left tail risks as AI makes thenomy increasingly weird. 望阅读后能让你在 AI 使经济变得愈发怪异的过程中，对潜在的左尾风险有更充分的备。 is is the CitriniResearch Macro Memo from June 2028, detailing the progression and fallouthe Global Intelligence Crisis. ⽂为 CitriniResearch 2028 年 6⽉的宏观备忘，详述全球智能危机的发展进程及其后。 TheConsequencesofAbundantIntelligence充裕智能的后果 两年。仅仅两年，便将“可控的”“局部⾏业性的”状况推进到⼀个已不再像我们任何⼈所成长的经济。这⼀期宏观备忘录试图重构这⼀序列——对危机前经济的⼀次事后检讨。 狂热情绪明显。到 2026 年 10⽉，标普 500 曾触及 8000 点，纳斯达克突破 3 万点。由于⼈类被判定为过时⽽引发的⾸波裁员始于 2026 年初，且正如裁员通常会带来的效果⼀样发⽣了：利润率扩⼤、盈利超预期、股市上涨。创纪录的企业利润被直接投⼊回AI 算⼒。 The headline numbers were still great. Nominal GDP repeatedly printed mid-to-high single-igit annualized growth. Productivity was booming. Real output per hour rose at rates noteen since the 1950s, driven by AI agents that don’t sleep, take sick days or require health HowItStarted起因 直到 ServiceNow 发布 2026 财年第三季度报告，反⾝性机制才变得更清晰。 SERVICENOW NET NEW ACV GROWTH DECELERATES TO 14% FROM 23%;ANNOUNCES 15% WORKFORCE REDUCTION AND ‘STRUCTURAL EFFICIENCYPROGRAM’; SHARES FALL 18% | Bloomberg, October 2026 SERVICENOW 净新增 ACV 增长从 23% 放缓⾄14%；宣布裁员 15% 及“结构性效率计划”；股价下跌 18% | Bloomberg，2026 年 10⽉ aaS wasn’t “dead”. There was still a cost-benefit-analysis to running and supporting in-houseuilds. But in-housewas an option, and that factored into pricing negotiations. Perhaps moremportantly, the competitive landscape had changed. AI had made it easier to develop andhip new features, so differentiation collapsed. Incumbents were in a race to the bottom onricing - a knife-fight with both each other and with the new crop of upstart challengers thatopped up. Emboldened by the leap in agentic coding capabilities and with no legacy costtructure to protect, these aggressively took share. SaaS 并未“死亡”。运⾏和维护内部构建仍需进⾏成本效益分析。但内部部署是⼀个选项，这在定价谈判中被纳⼊考虑。也许更重要的是，竞争格局发⽣了变化。AI 使得开发和发布新功能变得更容易，因此差异化消失。现有⼚商在价格上陷⼊了下⾏竞赛——与彼此以及新兴挑战者之间展开⼑光剑影的厮杀。得益于代理式编码能⼒的飞跃且⽆需保护任何遗留成本结构，这些新兴企业积极抢占了市场份额。 The interconnected nature of these systems weren’t fully appreciated until this print, either.erviceNow sold seats. When Fortune 500 clients cut 15% of their workforce, they cancelled5% of their licenses. The same AI-driven headcount reductions that were boosting marginst their customers were mechanically destroying their own revenue base. 在这次刊印之前，⼈们也没有完全认识到这些系统的相互关联性。ServiceNow 卖的是席位许可。当《财富》500 强客户裁减 15% 的劳动⼒时，他们就取消 15% 的许可证。那些为客户提升利润率的、由 AI 驱动的裁员，正在机械性地摧毁它们⾃⼰的收⼊基础。 The company that sold workflow automation was being disrupted by better workflowutomation, and its response was to cut headcount and use the savings to fund the veryechnology disrupting it. 销售⼯作流⾃动化的公司正被更优的⼯作流⾃动化所颠覆，他们的应对⽅式是裁员，并把节省下来的资⾦⽤来资助正在颠覆它们的技术。 What else were they supposed to do?Sit still and die slower?The companies most threatenedy AI became AI’s most aggressive adopters. 他们还能做什么？坐着不动等着慢性死亡吗？那些最受⼈⼯智能威胁的公司，反⽽成了⼈⼯智能最积极的采⽤者。 事后看来这似乎显⽽易见，但在当时并⾮如此（⾄少对我来说不是）。历史上的破坏模型表明， incumbents 会抵制新技术，失去市场份额给灵活的后起者，然后慢慢死亡。柯达、百视达、⿊莓就是如此。然⽽2026 年发⽣的不同；那些既有企业并没有抵制，因为他们⽆法承受不这么做的代价。 在股价下跌 40%-60%、董事会要求给出答案的压⼒下，⾯临 AI 威胁的公司别⽆选择。裁员，将节省下来的费⽤重新投⼊AI⼯具，利⽤这些⼯具在更低成本下维持产出。 Each company’s individual response was rational. The collective result was catastrophic.Every dollar saved on headcountflowed into AI capability that made the next round of jobuts possible. 每家公司单独的反应都是理性的。但集体结果却是灾难性的。每⼀美元的裁员节省都流⼊了能够促成下⼀轮裁员的 AI 能⼒建设。 WhenFrictionWenttoZero 当摩擦降为零 By early 2027, LLM usage had become default. People were using AI agents who didn’t evennow what an AI agent was, in the same way people who never learned what “cloud omputing” was used streaming services. They thought of it the same way they thought ofutocomplete or spell-check - a thing their phone just did now. Qwen’s open-source agentic shopper was the catalyst for AI handling consumer decisions.Within weeks, every major AI assistant had integrated some agentic commerce feature.Distilled models meant these agents could run on phones and laptops, not just cloudnstances, reducing the marginal cost of inference significantly. Qwen 的开源 agentic 购物助⼿成为 AI 处理消费者决策的催化剂。数周内，每个主要的 AI 助⼿都集成了某种 agentic 商务功能。精简模型意味着这些代理可以在⼿机和笔记本上运⾏，⽽不仅仅是在云实例上，从⽽显著降低了推理的边际成本。 The part that should have unsettled investors more than it did was that these agents didn’twait to be asked. They ran in the background according to the user’s preferences. Commercetopped being a series of discrete human decisions and became a continuous optimizationrocess, running 24/7 on be