Concise Summary简洁概述
Software evolved from a one-way artifact (disks, no feedback channel) into an always-connected system that can measure, experiment on, and nudge its users — each step individually defensible, the sum totalizing.
Once usage became measurable, the goal quietly shifted from 'is this good software' to 'does this raise engagement,' and Goodhart's law took over: metrics that become targets stop reflecting reality.
软件从单向的物理商品（磁盘、无反馈渠道）演变为始终联网、能够衡量用户、对用户做实验并推动其行为的系统——每一步单独看都说得通，累加起来却面目全非。
一旦使用行为可被衡量，目标便从「这是不是好软件」悄然转向「这能不能提升参与度」，古德哈特定律随之生效：一旦指标成为目标，它就不再真实反映现实。
Infographic信息图
From artifact to channel
从商品到渠道
Disk-era software was a one-way artifact you bought, installed, and used — it never spoke back. Always-on internet enabled updates and crash reports, opening a return channel; once data could flow back, the question inevitably became 'what else can we learn?'
磁盘时代的软件是单向商品：买下、安装、使用，它从不回话。始终在线的互联网让更新和崩溃报告成为可能，也顺带打开了回传通道——一旦数据能传回总部，下一个问题必然是「还能了解些什么」。
Goodhart's trap in product metrics
指标里的古德哈特陷阱
Swanson's Microsoft example — a feature buried six menus deep, then killed because 'nobody uses it' — shows metrics measure what's currently easy to observe, not what matters. Once a metric becomes a target, people get promoted for moving it, no malice required.
斯旺森举的微软案例——某功能被逐层深埋进六级菜单后，又因「没人用」被下架——说明指标衡量的是当下容易观察到的东西，而非真正重要的东西。一旦指标变成考核目标，推动它上升就能带来晋升，全程无需任何恶意。
A/B testing turns products into labs
A/B 测试把产品变成实验室
Experimentation looks like engineering rigor, but it quietly displaces product vision: judgment gives way to iteration because charts are politically safer than opinions. Once experimentation is the default decision tool, teams optimize whatever moves fastest — even the wrong direction.
实验化听起来像工程严谨性，实则悄悄取代了产品愿景：判断力让位于迭代，因为支持图表远比支持个人观点更「安全」。一旦实验成为默认决策方式，团队就会优化那个进展最快的指标——哪怕方向本身是错的。
Prescription: a real quiet mode
药方:真正的静音模式
Swanson's fixes are concrete, not abstract: separate health telemetry from growth telemetry, treat opt-out as permanent, use analytics as a flashlight rather than a steering wheel, and ship a mode with zero prompts, tours, surveys or push — a real test of whether the product stands on its own value.
斯旺森给出的方案很具体：把健康遥测和增长遥测彻底分离；退出选择必须永久生效；把数据分析当手电筒而非方向盘；并推出一个真正零弹窗、零导览、零调查、零推送的静音模式——这也是检验产品是否真有独立价值的试金石。
Detailed Summary详细解读
Swanson opens with an absurd thought experiment — a car that pulls over to ask for feedback — to make visible a pattern we've normalized in software but would never accept elsewhere. His core historical claim is that no single decision caused this: disk-era software was inert, then always-on internet enabled updates and crash reporting, both genuine improvements. The essay's method is genealogical, tracing how each incremental, well-intentioned step set up the conditions for the next.
The pivot point is the 'back channel': once software could report crashes home, the next question was inevitable — 'what else can we learn?' This led to analytics becoming mainstream (Google's 2005 Urchin acquisition is his marker), and Swanson is careful to credit the genuine upside: analytics replaced guesswork about user behavior with real signal, saving teams from listening only to their loudest customers.
His sharpest analytical move is invoking Goodhart's law with a concrete case: a Microsoft feature buried under six layers of menus, then cut because usage data showed 'nobody uses it.' This demonstrates that metrics measure what's currently easy to observe, not what's actually valuable — and once a metric becomes a promotion criterion, people will optimize it regardless of whether that helps users.
He then traces A/B testing as the next escalation: experimentation looks like engineering rigor but quietly replaces product vision, because backing a chart is politically safer than backing an opinion. This is Swanson's most original observation — that experimentation culture doesn't just change tactics, it erodes the incentive to hold a coherent product judgment at all.
Push notifications (iOS, 2009) mark the point where interruption became essentially free — software no longer needed the user to open it to demand attention. Swanson pairs this with nudge theory / choice architecture and B.J. Fogg's behavior model to show that the same mechanisms helping users discover features are structurally identical to those used to redirect them toward metrics.
The essay closes with five concrete prescriptions rather than vague calls for restraint: make opt-out permanent, separate health telemetry from growth telemetry, treat analytics as a flashlight not a steering wheel, optimize for trust over return visits, and ship a genuine quiet mode. This practical turn is what elevates the piece above a nostalgia essay — it names a testable criterion (does quiet mode improve retention?) for whether a product actually has value.
斯旺森以一个荒谬的思想实验开篇——一辆会中途靠边索要反馈的汽车——让读者看清一种我们在软件领域已习以为常、却在其他任何领域都绝不会容忍的模式。他的核心历史论断是：这不是任何单一决策造成的，磁盘时代的软件本是惰性的，而始终在线的互联网带来了更新和崩溃报告，两者都是真实的进步。全文采用谱系式写法，逐一追溯每个善意的渐进步骤如何为下一步埋下伏笔。
转折点在于「后门通道」：软件一旦能把崩溃信息传回总部，下一个问题就不可避免——「还能了解些什么？」这催生了数据分析的主流化（他以谷歌 2005 年收购 Urchin 为标志），而斯旺森也公允地承认了其真实价值：数据分析用真实信号取代了对用户行为的猜测，让团队不必只听最会喊的客户意见。
文中最锐利的分析动作是用一个具体案例来援引古德哈特定律：某功能被逐层埋进六级菜单后，又因数据显示「没人用」而被砍掉。这说明指标衡量的是当下容易观测的东西，而非真正有价值的东西——一旦某项指标成为晋升考核标准，人们就会去优化它，无论这是否真的对用户有益。
接着他把 A/B 测试视为下一次升级：实验化看似工程严谨，实则悄悄取代了产品愿景，因为支持一张图表远比支持一个个人观点更「政治安全」。这是全文最具原创性的观察——实验文化改变的不只是战术，更是消解了团队坚持连贯产品判断的动力本身。
推送通知（iOS，2009 年）是打扰几乎零成本的分水岭——软件不再需要用户主动打开就能索取注意力。斯旺森将其与「助推理论/选择架构」以及 B.J. Fogg 的行为模型并置，说明帮助用户发现功能的机制，与将用户引向指标的机制在结构上是同一套东西。
文末不是泛泛呼吁「克制」，而是给出五条具体处方：退出选择必须永久生效、把健康遥测与增长遥测分开、把数据分析当手电筒而非方向盘、优化信任而非回访率、推出真正的静音模式。这一实操转向让本文超越了一篇怀旧散文——它给出了一个可检验的标准（静音模式能否提升留存）来判断产品是否真有价值。
FAQ常见问答
Is Swanson arguing we should go back to disk-era software with no updates?斯旺森是在主张回到没有更新的磁盘时代吗？
No — he explicitly says he loves fast updates, security patches, and crash reports. His target is the growth-driven interruption layer, not connectivity itself.
不是——他明确表示自己喜欢快速更新、安全补丁和崩溃报告。他批判的是以增长为驱动的打扰层，而不是联网能力本身。
What's the actual mechanism that turns helpful telemetry into manipulation?有用的遥测究竟是怎样一步步变成操纵的？
Once a data pipeline exists for one purpose (crash reports), it's cheap to reuse for another (engagement tracking), and incentives favor growth metrics because they're easier to show as wins.
一旦为某个目的（崩溃报告）搭建了数据管道，将其复用于另一目的（参与度追踪）就变得极其廉价，而激励机制又天然偏向增长指标，因为它更容易被包装成「成果」。
Does the essay provide evidence beyond anecdotes like the Microsoft menu example?除了微软菜单这类轶事，文章还有其他证据支撑吗？
Mostly a documented timeline (PointCast 1996, Office 97's Clippy, App Store 2008, push notifications 2009) plus references to Goodhart's law and Fogg's behavior model — strong on narrative, thin on quantitative data.
主要依靠一条有据可查的时间线（1996 年 PointCast、1997 年 Office 的 Clippy、2008 年 App Store、2009 年推送通知），并援引古德哈特定律与 Fogg 行为模型——叙事有力，但量化数据较薄弱。
How is 'backseat software' different from the more common term 'enshittification'?「后座软件」和更常见的「垃圾化(enshittification)」概念有何不同？
Enshittification describes platform decay from monopoly power over time; backseat software isolates one specific mechanism within that decay — turning user attention into a KPI via telemetry, testing, and nudges.
「垃圾化」描述的是平台凭借垄断地位随时间衰败的整体过程；「后座软件」则聚焦其中一个具体机制——通过遥测、实验和助推，把用户注意力转化为一项 KPI。
Would the proposed 'quiet mode' actually work as a business incentive?文末建议的「静音模式」在商业上真的可行吗？
Swanson argues it's self-selecting: if quiet mode hurts retention, that reveals the product depends on nagging rather than real value — a useful diagnostic, though he doesn't address how teams measured on growth would adopt it.
斯旺森认为这具有自我筛选效应：如果静音模式会拉低留存，恰恰说明产品依赖的是骚扰而非真实价值——这是个有用的诊断工具，但他没有说明以增长为考核标准的团队要如何被说服去采纳它。
In-depth Analysis · Pros & Cons深入解读 · 优缺点
Mike Swanson traces how software drifted from a tool you operate into a channel that operates on you, through a chain of individually reasonable engineering decisions. He names the mechanism—measurement turning into optimization turning into manipulation—and proposes concrete fixes to reverse it.
迈克·斯旺森梳理了软件如何从「你操作的工具」一步步滑向「操控你的渠道」——这一切并非源于恶意，而是一连串看似合理的工程决策的累积。他点出了核心机制：衡量演变为优化、优化演变为操纵，并给出了具体的逆转方案。
- Clear causal chain因果链条清晰Rather than blaming greed, Swanson builds a step-by-step genealogy where each move is individually defensible, making the critique harder to dismiss as cynicism.斯旺森没有简单归咎于贪婪，而是构建了一条逐步展开的谱系，每一步单独看都合情合理，这让整篇批判更难被斥为愤世嫉俗。
- Concrete, testable fixes给出具体可验证的方案The five prescriptions (permanent opt-out, separated telemetry, real quiet mode) are actionable engineering decisions, not vague calls for 'better ethics.'五条方案（永久退出、遥测分离、真正的静音模式）都是可落地的工程决策，而非「更讲道德」式的空泛呼吁。
- Well-chosen concrete example举例具体有说服力The buried Microsoft feature illustrates Goodhart's law far more memorably than an abstract definition would, grounding the theory in lived product experience.被深埋的微软功能案例，比抽象定义更能让古德哈特定律深入人心，把理论落到了真实的产品经验上。
- Honest about the good uses坦诚承认合理用途Swanson doesn't caricature telemetry as purely evil — he explicitly defends crash reports, security patches, and time-sensitive alerts for messaging/banking apps.斯旺森没有把遥测一律妖魔化——他明确为崩溃报告、安全补丁以及即时通讯/银行类应用的时效性提醒做了辩护。
- Anecdote-driven evidence证据以轶事为主The central claims rest on one Microsoft anecdote and a historical timeline rather than systematic data on how widespread or severe the pattern is across the industry.核心论点主要靠一个微软轶事和一条历史时间线支撑，缺乏关于该模式在行业内普遍程度和严重程度的系统性数据。
- No org-incentive solution未解决组织激励问题The fixes assume a team with the authority to separate health from growth telemetry, but doesn't address what happens when growth teams are measured and paid on the very metrics being critiqued.方案假设团队有权将健康遥测与增长遥测分开，却没有说明当增长团队正是靠被批判的这些指标来考核和发薪时该如何推进。
- US consumer-app bias偏重美国消费级应用视角Examples (Apple, Microsoft, App Store policy) center on Western consumer software; B2B, enterprise, and non-US regulatory contexts (e.g. GDPR-driven consent flows) are largely absent.举例（苹果、微软、App Store 政策）集中在西方消费级软件，B2B、企业软件以及非美国监管语境（如 GDPR 驱动的同意流程）基本未涉及。
- Quiet mode's business case is asserted, not tested静音模式的商业逻辑只是断言，未经检验The claim that quiet mode 'should' improve retention if the product has real value is a hypothesis, not a validated finding — no product cited actually ships this feature.「若产品真有价值，静音模式应能提升留存」这一说法只是假设，并非经过验证的结论——文中也没有举出真正上线过该功能的产品案例。
Worth reading for anyone in product, growth, or UX roles who wants a clear vocabulary for a pattern they've likely felt but not named — the Goodhart's-law framing and the five-point fix list are the most useful takeaways. Treat the historical narrative as a compelling frame rather than a rigorously evidenced study, since it leans on select anecdotes over systematic data.
适合产品、增长或用户体验从业者阅读——它为一种大家早有体感却说不清楚的现象提供了清晰的命名工具，古德哈特定律的框架和文末五条方案是最值得带走的部分。但应把这段历史叙事当作一个有说服力的框架，而非严谨的实证研究，因为它更多依赖精选轶事而非系统数据。
Excerpt原文节选
This is a short excerpt, not the full piece — the complete essay belongs to its original author; please read it in full at the link above.
以下仅为节选，并非全文——完整文章版权归原作者所有，请点击上方链接阅读全文。
What if your car worked like
"How are you enjoying your drive so far?"
Annoyed by the interruption, and even more behind schedule, you dismiss the prompt and merge back into traffic.
A minute later it does it again.
"Did you know I have a new feature? Tap here to learn more."
It blocks your speedometer with an overlay tutorial about the turn signal. It highlights the wiper controls and refuses to go away until you demonstrate mastery.
Ridiculous, of course.
And yet, this is how a lot of modern software behaves. Not because it's broken, but because we've normalized an interruption model that would be unacceptable almost anywhere else.
I've started to think of this as backseat software : the slow shift from software as a tool you operate to software as a channel that operates on you. Once a product learns it can talk back, it's remarkably hard to keep it quiet.
[…the source continues — read the rest at the link above]
[……原文更长，完整内容请点击上方链接阅读]
As always, I love hearing from you .
如果你的车像许多应用程序那样运作会怎样？你正赶往某个重要地点……或许已经迟到了一点。行驶几分钟后，你的车突然靠边停下，问道： so many apps? You're driving somewhere important…maybe running a little bit late. A few minutes into the drive, your car pulls over to the side of the road and asks:
"到目前为止，您觉得这次驾驶体验如何？"
被这突如其来的打扰惹恼，更因行程严重延误，你关掉提示音，重新驶入车流。
一分钟后它又这样做了。
"你知道我新增了功能吗？点击此处了解更多。"
它用转向信号的覆盖式教程遮挡了你的车速表。它突出显示雨刷控制键，并坚持不消失，直到你证明自己已完全掌握操作。
荒谬，当然。
然而，这正是许多现代软件的行为模式。并非因为它们存在缺陷，而是因为我们已将一种中断模式视为常态——这种模式在其他几乎所有领域都难以被接受。
我开始将这种现象称为 后座软件 ：软件正从你操作的工具，悄然转变为作用于你的渠道。一旦产品学会了反向沟通，要让它保持沉默就变得异常困难。
这篇帖子讲述的是我们如何走到今天这一步。并非一蹴而就，而是循序渐进。一步一个脚印，稳扎稳打。
Software Came on Disks软件以磁盘形式发布
曾几何时，软件是通过物理介质发布的：软盘、CD-ROM，有时甚至附带螺旋装订的手册。
那时的软件就像一件商品。你买下它，安装它，使用它。若要升级，纯粹是出于自愿。软件不会在你不知情时不断变动，也无法通过开发者最初交付的界面之外的任何方式来索取你的关注。
那个时代确实存在诸多弊端。若发布了严重漏洞，你只能忍受它直到下个版本更新——而这可能要等上数周甚至数月。若发现安全问题，你的选择无非是"邮寄补丁"或"祝你好运"。回首往事，我们竟能挺过来，实在令人惊叹！
但还有另一件事也是真的。当你使用软件时，你与它独处。
作为软件开发者，当出现问题时，往往是用户告知你。
[…the source continues — read the rest at the link above]
[……原文更长，完整内容请点击上方链接阅读]
一如既往， 我很乐意收到您的来信 。