🔮 Exponential View #595 公开部分：AI 采用的决策陷阱与 Leopold 基金的崩解｜英文原文 + 中文翻译 (2026)

原文信息

阅读 Exponential View 官方原文：Leopold & exponential markets; transformative GLP-1s; runaway AI & the future of safety++ #595

副标题：Plus: Citizen agents, protein & longevity, nervous Europe
作者：Azeem Azhar
发布时间：2026 年 8 月 2 日 11:19（北京时间）1
公开范围：官方页面在「Too brittle to go exponential」最后一句之后显示付费订阅门槛。以下保留该门槛之前公开可读的英文原文与中文翻译；付费墙后的 GLP-1、AI 安全等内容不补写。1

English original

"It is a lot to cope with the rollercoaster of the last decade and deep uncertainty of what’s coming next. Over many years, the quality and depth of the newsletter has ensured I am better informed and inspired." — Hugh K., a paying member 1

Get 30% off for 1 year 1

AI adopter’s decision trap

We modeled three types of companies adopting AI. They have the same starting economics, the same 5% hit rate, but different learning practices. Two years into their investment, all three are losing similar amounts of money. In year five, the eventual loser looks best. It takes eight years to see which approach leads to outsized ROI. 1

A line chart showing three AI-adoption project-budget curves and a shaded decision-trap region — *Exponential View 的 EV company J-curve model，图表标注为示意图。*1

Most CEOs are facing the decision trap right now – you likely don’t know if your firm’s spending is learning that will compound to ROI, or waste. The FT calls Zuck "the king of the side quest" as he works away on a portfolio of bets: 1

An English text excerpt about Meta developing coding agents, task-executing assistants, data-centre rental, and customer-service bots — *The official post reproduces an excerpt about Meta’s products and Zuckerberg saying that exciting things would happen "soon".*1

An English text excerpt about the stakes of Meta’s side projects and its possible five-year investment — *The excerpt says Meta may invest nearly $750bn over the next five years, attributing the estimate to analysts polled by LSEG; the screenshot cuts off mid-sentence.*1

Mark doesn’t need each project to succeed, as long as the experiments deepen Meta’s infrastructure and inform the next move. In our model, system builders that consistently compound their learnings over time have the winning formula. See our framework and the interactive model: 1

Too brittle to go exponential

A $45 billion fund at its peak, started by an AI researcher with no hedge fund experience, betting on the AGI capex build-out at roughly four times leverage, was forced to liquidate this week. Leopold Aschenbrenner started Situational Awareness LP in late 2024 on a thesis that the path to superintelligence would put trillions into compute, chips and power. 1

A bar chart showing Situational Awareness fund assets rising to $45bn and falling to $10bn in 2026 — *Exponential View 图表：Situational Awareness 基金资产，2026 年 3 月至 7 月；图注说明数值来自引用匿名消息人士的新闻报道，均为近似值。*1

He did great. And then, the trade turned. The Philadelphia Semiconductor Index fell 28.6% from its June peak, and software gained. 1

Situational Awareness’ unraveling is not proof that Leopold’s thesis is wrong. 1

This post is for paid subscribers

Already a paid subscriber? Sign in 1

Public-access boundary: The official page shows the paid-subscriber gate after "Situational Awareness’ unraveling is not proof that Leopold’s thesis is wrong." The remainder of the article is not publicly readable here, so it is not translated or reconstructed. 1

中文翻译

「要应对过去十年的过山车，以及接下来会发生什么带来的深度不确定性，需要消化很多东西。多年来，这份 newsletter 的质量和深度让我获得了更充分的信息，也不断受到启发。」——Hugh K.，付费会员 1

一年订阅可享 7 折 1

AI 采用者的决策陷阱

我们对三类采用 AI 的公司做了建模。它们的起始经济条件相同，命中率同为 5%，但学习方式不同。投入两年后，三者都在亏损，而且亏损额相近。到了第五年，最终会失败的那一类反而看起来最好。要等到第八年，才看得出哪种做法能带来超额投资回报（ROI）。 1

上图是 Exponential View 的 EV company J 曲线模型，图表明确标注为示意图。它把「决策陷阱」画在早期阶段：不同的 AI 采用路径，在很长一段时间里可能呈现相似的亏损曲线。1

现在，大多数 CEO 都在面对这个决策陷阱：你很可能还不知道，公司在 AI 上的支出究竟是在积累会复利到投资回报的学习，还是在浪费钱。金融时报把扎克伯格称为「side quest 之王」——他正同时押注一组项目。1

原文配图中的摘录写道：

Meta 还在开发许多偏离自身舒适区的产品，包括编程代理、执行任务的助手、数据中心租赁和客服机器人。这些创新中，有相当一部分的收入和时间表细节仍很少。在公司周三的财报电话会上，扎克伯格至少十次说，各种令人兴奋的事情会「很快」发生。1

另一张原文配图写道，Meta 的「副业」风险更高；扎克伯格需要这些项目做成，才能证明他庞大的支出计划合理。配图可见的下一句是：LSEG 访问的分析师估计，未来五年公司可能投资近 7500 亿美元。截图文字在此处截断。1

马克不需要每个项目都成功，只要这些实验能不断夯实 Meta 的基础设施，并为下一步行动提供信息。在我们的模型中，长期持续复利累积所学的系统构建者拥有制胜公式。你可以查看完整框架和互动模型。1

脆弱到无法指数增长

一个高峰期规模达到 450 亿美元的基金，由一位没有对冲基金经验的 AI 研究者创办，押注 AGI 的资本支出建设，并使用大约四倍杠杆；本周被迫清算。Leopold Aschenbrenner 于 2024 年末创办 Situational Awareness LP，依据这样的判断：通往超级智能的道路会把数万亿美元投入计算、芯片和电力。1

上图是 Exponential View 绘制的 Situational Awareness 基金资产变化图，覆盖 2026 年 3 月至 7 月。图注说明，这些数字来自引用匿名消息人士的新闻报道，均为近似值。1

他的交易一度做得很好。然后，交易风向变了。费城半导体指数从 6 月高点下跌 28.6%，软件股上涨。1

Situational Awareness 的瓦解，并不能证明 Leopold 的论点是错的。1

付费订阅边界

已经是付费订阅者？登录 1

公开内容边界： 官方页面在「Situational Awareness 的瓦解，并不能证明 Leopold 的论点是错的」之后显示付费订阅门槛。其后的文章内容在当前公开页面中不可读，因此这里不翻译，也不根据上下文重建。 1

References

1Exponential View #595