What becomes personal when intelligence becomes ambient?

当智能逐渐成为环境的一部分，什么会变得更加私人？

Five signals across products, workplaces and homes—and the question connecting them.

从产品、工作空间与家庭中选出五个信号，以及将它们连接起来的一个问题。

01 / The overview · 本期概览

Artificial intelligence is moving from a destination we deliberately visit toward an ambient layer inside tools and environments we already inhabit. It appears in the document before we ask, understands the room before we explain, and draws on personal context to reduce the distance between intention and action.

人工智能正在从一个需要主动前往的目的地，变成我们原本使用的工具与生活环境中的一层背景。它在我们提出请求以前出现在文档里，在我们解释以前理解空间，并通过个人语境缩短意图与行动之间的距离。

This shift makes intelligence feel less like software and more like infrastructure. Capability still matters, but the more intimate question is now: what must a system know about us in order to become genuinely useful—and who controls that knowledge?

当智能更像基础设施而不只是软件，核心问题也随之改变：为了真正有用，一个系统需要了解我们到什么程度？这些知识又由谁控制？

02 / Signals worth noticing · 值得注意的信号

Personal context becomes product architecture

Privacy is moving into the design of intelligence itself.

Apple’s current architecture combines on-device processing with Private Cloud Compute for more demanding requests. The strategic signal is larger than one company: when intelligence depends on personal context, privacy can no longer remain only a settings-page promise. It becomes part of how the product is built.

个人语境正在成为产品架构的一部分。当智能依赖越来越私人的信息，隐私不能只是一项设置或承诺，而必须进入系统本身的设计方式。

The workspace becomes a memory surface

Files, emails and the web are being connected into one context.

Google’s 2026 Workspace updates allow Gemini to work across selected files, email and web sources. The document is no longer only a page on which work is produced; it is becoming an interface to the history surrounding that work.

文档不再只是完成工作的页面，而逐渐成为连接文件、邮件、网络资料与工作历史的界面。

The home becomes interpretable

Ambient assistance depends on understanding what is happening around us.

Gemini for Home is framed as understanding what is happening in the home and what matters to the people inside it. The promise is ease; the design challenge is legibility: what did the system notice, why did it act, and how can it be corrected?

家庭智能的价值来自它对环境的理解，而信任则来自使用者能否知道系统注意到了什么、为何行动，以及如何纠正。

AI use is not one behavior

Autonomy, purpose and task complexity need to be measured separately.

Anthropic’s Economic Index distinguishes task complexity, purpose, autonomy and success rather than treating AI adoption as a single number. Learning, collaborating and delegating are different relationships even when they happen through the same interface.

“使用 AI”并不是一种单一行为。学习、协作与委托具有不同的关系和风险，即使它们发生在同一个界面中。

Adoption has a rhythm

Tools enter work through repeated tasks before they transform whole roles.

Observed use data emphasizes tasks and cadences rather than abstract occupational exposure. Transformation often arrives quietly: first through a recurring action, then through a changed workflow, and only later through a changed job description.

变化往往不是先以“职业被重塑”的形式出现，而是先进入一个重复任务，随后改变工作流，最后才改变岗位本身。

03 / One thread · 一条共同线索

The closer intelligence moves to the background, the more visible its boundaries must become.

当智能越接近无形的背景，它的边界反而越需要清晰可见。

Ambient intelligence promises less friction. Yet friction has sometimes been the moment in which consent becomes visible: opening an app, choosing a file, asking a question. When those actions disappear, good design must create new forms of awareness without returning all of the burden to the user.

无感智能承诺减少摩擦，但摩擦有时正是同意变得可见的时刻。当主动动作逐渐消失，设计需要创造新的知情方式，又不能把全部负担重新交给使用者。

04 / Primary sources · 一手来源

01Apple — Apple Intelligence, everyday experiences↗02Google — Gemini in Workspace↗03Google — Gemini for Home↗04Anthropic — Economic Index: Cadences↗05Anthropic — Economic primitives↗

Sources are read in their original context. The I Index separates reported facts from editorial interpretation.
来源均在原始语境中阅读。The I Index 明确区分事实信息与编辑判断。