When Switching Stops Being a Choice为什么越来越难换一家為什麼越來越難換一家 · 02
From Opening a Route to Owning the Passage
当年替你开路,后来等你经过
當年替你開路,後來等你經過
Networks, records, defaults and acquisitions can turn a useful service into an unavoidable passage. Its history does not answer for what it does now.
网络、记录、默认位置与收购,可以使一件好用的服务逐渐难以绕开。要认的不是它曾经是什么,而是它如今怎样作用。
網路、記錄、默認位置與收購,可以使一件好用的服務逐漸難以繞開。要認的不是它曾經是什麼,而是它如今怎樣作用。
The workshop owner sends his first confident letter to an overseas buyer with Beacon's help. Product descriptions once passed through a translator; now he can revise them himself. After a year, quotations, customer exchanges, and accounts live inside the assistant too. At the beginning this is not an additional gate. It is a route he could not previously use.
We remain in the imagined Kingdom State. Four researchers founded Beacon, offering a free basic assistant and additional paid services. Distant orders increased for workshops using it. These are stipulated details, not findings about the AI industry. They let us examine a change: when does reliance on something genuinely useful begin to restrict the routes available later?
Reliance alone cannot be the accusation. We rely on electricity, language, colleagues, and acquired skills. Doing better with a tool does not establish that the tool has taken freedom away. The issue becomes sharper when someone wants to refuse, switch, or reorganize a life: can the capability gained through it still be carried forward on terms they can use?
Beacon adds a memory service for years of correspondence and preferences. When our owner considers another assistant, it cannot read the format. He could rebuild three years of records, but stays. We know that rebuilding is costly. We do not know that every alternative is unusable, so we cannot yet declare withdrawal completely blocked. Neither should we pretend the difficulty is imaginary.
Here is one path of formation: a tool understands more, receives more, and becomes more expensive to replace. Another provider might build a workable bridge and change that path. A data feedback loop tells us where to look. It does not predict that concentration must end in monopoly, or that every personalized service already exercises wrongful control.
Crossways' marketplace grows along a different path. More buyers attract sellers; more sellers attract buyers. Meeting in one place is a real network benefit. But a seller leaving alone does not take the buyers along, and a buyer leaving alone may leave the desired goods behind. Coordinating a departure is harder than closing an individual account.
Network effects are not guilty simply for doing what makes a network valuable. The further questions concern available connections and restrictions added to them. People coming together differs from someone preventing every alternative means of continuing those relationships. That difference is obscured when network effects are treated either as a complete defense or as a prediction of inevitable monopoly.
Another path sits inside a settings menu many users never open: the default. An alternative is not prohibited, yet the route is selected before an active choice occurs. Defaults can save effort. They can also put competitors where many people never encounter them. Convenience is real without answering for the particular effects of the arrangement through which it is delivered.
In the Google search case, the federal district court in Washington, D.C., found in 2024 that exclusive distribution arrangements unlawfully maintained monopolies in general search services and general search text advertising. The district court’s December 2025 final judgment imposed prohibitions and remedies. Both Google and the plaintiffs subsequently appealed. That is a judicial finding within particular markets, evidence, and law, not a declaration that every default is illegal.
The distinction it illuminates is ordinary enough: the ability to change a setting does not establish equivalent practical access for an alternative. We use the court's findings rather than pronounce independently on the company's entire conduct. The legal assessment and this account's inquiry into usable routes meet at a structural point without becoming the same framework or answering every question for each other.
Acquisition is another possible route. Crossways buys Willow Basket, a young marketplace that subsequently closes. Years later it offers to buy Ferry Crossing, a small cold-chain distributor. Those are different events. The first has an outcome; the second is an offer. We cannot fill in the later story by borrowing the earlier ending, however suggestive the sequence appears.
A seller's agreement matters, but it answers for the seller's undertaking, not everyone relying on the route. If an established position can repeatedly buy alternatives before they grow, examine what can still develop around it. An offer alone proves neither illegality nor a wish to suppress competition. Looking at effects does not authorize us to invent intentions missing from the account.
Similar terms across several firms do not automatically establish coordination either. Each might independently recognize the same advantage and choose the same practice. Coordination needs evidence. A lack of evidence for coordination does not make the resulting practical restrictions disappear. The origins of conduct and the situation it produces are distinct inquiries, and each requires an answer at its own level.
Standards and interfaces can connect these paths. One firm writes the format through which others enter. A record exports but cannot be read elsewhere. A partner's tasks receive priority while other customers cannot examine the queue. Technical explanations may be valid. They still need to identify what is restricted, why that restriction is needed, and where the explanation can be examined.
Usually several mechanisms arrive together: buyers and sellers in one place, logistics tied to it, house products prominently displayed, records retained in accounts. Looking at one mechanism alone may miss why leaving is difficult. Collapsing the whole into the word large conceals where a change might help. We need the combined picture and the separate examination, not a choice between them.
The depth of influence can also change. We begin by asking how much commission sellers pay or income workers lose. Later we must ask which information reaches people, which possibilities appear first, and which never become visible. These are not a person's judgment itself. They are materials through which judgment is exercised, and controlling their entrance can therefore reach unusually far.
Crossways moves the product list lower on its page and replaces the entrance with a shopping assistant. A buyer states a need; the assistant presents one or two products and places the order. The list remains two clicks away. Buyers have not thereby lost judgment. The service has nevertheless moved from displaying many goods to selecting a very small set before the buyer sees them.
Wanting to make fewer decisions can itself be a choice. Nobody needs to compare every product by hand to qualify as independent. What matters is the scope and rules of delegation, who can change them, and how usable the route back to a wider view remains. Lower prices cannot alone answer what happens when the materials of choice pass through fewer entrances.
Time complicates diagnosis further. Crossways once used accumulated resources to waive fees for newcomers. Later it ends that program and replaces seller-elected appeal seats with management appointments. The same accumulated capacity is exercised in both periods, without necessarily working in the same direction. Earlier usefulness does not provide the grounds for every later arrangement or exempt a changed position from examination.
Reassessment must also work the other way. In November 2025, the district court hearing the FTC's case against Meta found that the agency had not established present monopoly power; the FTC filed an appeal in January 2026. The court required proof about the present. That is a particular legal judgment and a subsequent procedural step, not a certificate resolving every possible structural question.
Take the temporal discipline, not an invented extra verdict. The judgment does not establish that Meta definitely held monopoly power in the past; it says past power would not alone settle the required showing now. Nor does a finding about a legally defined market automatically settle another framework's questions about departure, alternatives, and effective challenge. Similarity at one point does not collapse their different reach.
If Crossways' position genuinely changes, earlier remedies should be reconsidered too. They may remain necessary or may be narrowed. A justification must identify what problem it still addresses, rather than deny a company the possibility of being examined afresh. This is not a promise to relax constraints whenever a market-share number falls; actual routes and actual effects remain the relevant evidence.
Reconsideration does not annul history. Why Willow Basket was acquired and what followed remain questions about that time. Crossways' present operation requires a present account. Erasing old conduct differs entirely from refusing to let an old conclusion perform every subsequent assessment. The two can be kept apart without either permanent condemnation or an amnesty written by changes in market conditions.
Beware two apparently opposite statements: it helped us once, so no further questions; it controlled us once, so no further examination. Both surrender the present to a completed story. They spare us the work of looking at the people, arrangements, and routes before us now. This account asks for that work even when either familiar story would be more satisfying.
Beacon helped the owner write his first overseas reply. That capability deserves acknowledgment. Years later, the portability of his records deserves an answer too. We do not have to deny a tool's usefulness before asking what it holds. Its usefulness is precisely why the later problem of leaving acquires a concrete shape rather than remaining an abstract preference for more competition.
Formation is not the hunt for a single year when a company became bad forever. It is an account of how a position grows, connects to other layers, and changes available paths. The next essay follows the remaining routes of challenge. When the hearing room, complaint form, and regulator still exist, what can make the arrangements inside them so difficult to move?
一位小作坊主第一次用灯塔助手,终于敢给海外买家回信了。原先要托人译的商品页,自己改几遍就能发出去。用了一年,他把报价、客户往来和账也放进去。它不是先给他添了一道门,而是替他开了一条路。
这里继续用虚构的王国州。灯塔由四位研究者办起来,个人可以免费用基础档,作坊付费得到更多服务。镇上的订单里,来自远方的部分增加了。数字是设例,不是 AI 行业的调查;我们借它看一种可能的变化:越有用,越依赖,什么时候依赖开始限制后来想走的路?
先别急着把依赖都说成受制。我们依赖供电、语言、同事,也依赖学会的一门手艺。靠一个工具做得更好,本身不是工具偷走了自由。需要查的是,当人要换、要拒绝、要重新安排时,那份先前得到的能力还能不能由他继续使用。
灯塔后来推出记事,保存多年的来往与偏好。作坊主想换一家时,发现新助手读不了旧格式。他仍能走,只是要重建三年的记录。他留下了。我们知道这份成本,不知道其他替代是否真的不可用,因此还不能直接宣布退出已经封死。
这是一条形成的路:一个工具越懂你,交给它的东西越多;交得越多,换走越费力。也可能有别家做出足够好的接续,改变这条路。数据回路说明要往哪里看,不是一条最后一定垄断的预言。
通衢的市集走过另一条路。买家多,卖家愿意来;卖家多,买家也更愿意来。相遇容易,确实是网络带来的好处。可是卖家一个人搬出去,原来的买家不跟着走;买家搬出去,想买的货也许仍在这里。两边很难约好同时搬家。
不能由此说网络效应本身有罪。让人互相找得到,是许多服务存在的理由。要继续查的是,有没有可以跨过来往的办法,平台后来又加上了什么限制。关系聚在一处与别人把所有接续封在一处,不是同一句话。
再有一条路,藏在你从来没有打开过的设置里:默认。不是不许选别家,而是在你尚未作选择时,路已经替你排好了。默认可能替人省事,也可能把竞争者放到许多人始终不会看见的地方。省事是真的,作用仍须核。
美国谷歌搜索案给过一个具体的法律认定。哥伦比亚特区联邦地区法院在2024年认定,谷歌通过排他的分发安排,非法维持通用搜索服务与通用搜索文字广告市场的垄断;2025年12月的最终判决列出相应禁令与救济,谷歌和原告方随后均提起上诉。法院处理的是特定市场、证据与法律,不是在宣布所有默认设置都违法。
这一案照亮的区别很日常:设置里可以改,不等于替代者实际得到了同样的接近机会。这里用法院已经作出的认定,不替法院另判一家公司的全部行为。法律上的排他效果与我们正在追问的可用路线相接,却不因此变成同一个判断框架。
形成还可能发生在收购里。通衢买下柳筐市集,后者随后停了。几年后它又出价收购做冷链的小公司渡口。这是两笔不同的事:柳筐已经关掉,渡口只有出价,不能把前一笔的结局预先填到后一笔。
卖方愿意成交,当然重要。可卖方处理自己的事业,不替所有依赖那条路的人回答。若一个位置有能力在替代者长大之前不断买下它们,就要查周围还有什么路能够长起来。出价本身不证明违法,也不证明买方心里只想消灭竞争。
同样,几家的条款相像,不自动说明商量过。它们也可能各自看见同一种收益,各自选择相似做法。是否协调需要证据;没有证据证明协调,也不使共同出现的实际限制忽然不存在。行为的来处与形成的处境,要分别说明。
标准与接口也会接在这些路上。一家写了入口的格式,别人只有照它的规格才能接进来;记录导得出,别家读不了;伙伴的任务排在前面,其他人不知道怎样等。每一件都可能有技术上的理由,技术理由仍要说到实际限制哪里为止。
这些东西往往不是单独来的。市集聚着买卖双方,物流接着市集,自营商品排在首页,记录留在账户里。只盯其中一个环节,有时看不出人为什么走不动;把它们全揉成一个大字,又看不出该改哪里。需要合起来看,再一项项拆开核。
而且,它握住的未必只是价格。原先我们问,卖家多交了多少佣金,工人少了多少收入。后来还得问,他取得信息要经过哪里,哪些东西先到眼前,哪些东西根本不会出现。这些不是人的判断本身,却是他拿来判断的材料。
在设例里,通衢把搜索列表移到下面,首页换成问路助手。买家说要什么,助手只给一两件,替他下单。列表还在,多点两下也能看;不能因此说买家已经失去判断,但服务已经从让人看见许多货,走到先替人挑出极少的货。
愿意少操心,可以是买家自己的选择。问题不是每个人都该亲手比较所有商品,而是这份委托按什么规则执行,谁能改,想看更多时有多难。若给材料的一端集中起来,需要核的深度也变了,不能只用价格便宜没便宜答完。
时间又给判断添了一层。通衢早年拿积累替新店免佣金,后来取消扶持,把卖家互选的申诉席位也改成管理层任命。两回使用的是同一家公司积下的能力,作用却不必相同。早年好用,不是后来每一项安排的根据。
不过,照现在看也不能只向一个方向用。FTC诉 Meta 案里,地区法院2025年判委员会没有证明 Meta 现在仍有相应的垄断力,强调需要证明现在,而非仅讲过去。FTC在2026年提出上诉。这是一个具体法律争议中的判决与后续步骤,不是所有结构问题都已消失的证书。
我们从中取的是提问的时间:今天的力量要有今天的根据。不能反过来补成法院已经认定过去确有垄断;判词没有这样替过去定案。也不能拿特定法律市场中的结论,省掉另一个框架下对去留、选择和申诉实际效力的核对。
假如有一天通衢份额降了,替代者真的接得上,原有的限制也应重看。曾经需要的救济可能仍然需要,也可能可以收窄。要说明现在还处理什么问题,而不是让一家公司因一份旧诊断永远失去重新被判断的机会。
这一重看不撤销已发生的行为。柳筐当年为何被买下、买下以后发生什么,仍按当年的材料问;今天的通衢有什么作用,按今天问。把旧账抹掉与拒绝拿旧结论包办现在,完全是两件事。
因此,最该小心的也许不是一句大公司不好,而是两句听来相反、做法相同的话:它从前帮过我,所以现在不用问;它从前控制过,所以现在无需再看。两句都把眼前的人和路,交给了一个已经说完的故事。
灯塔让那位作坊主能够写出第一封外文回信,这份能力值得承认。等到他想换一家时,记录怎样带走,同样值得认真答。并不是先否认工具的用处,才有资格问它握住了什么;恰恰因为那些用处真实,后来离开的事才如此具体。
看形成,不是抓住某一年宣布从此变坏。是把一个位置怎样长出来、怎样连到别的层上、怎样改变人的去处读清楚。下一篇再把目光移到已经留下的问路:如果所有窗口都还开着,为什么有时谁也推动不了里面的事?
一位小作坊主第一次用燈塔助手,終於敢給海外買家回信了。原先要托人譯的商品頁,自己改幾遍就能發出去。用了一年,他把報價、客戶往來和帳也放進去。它不是先給他添了一道門,而是替他開了一條路。
這裡繼續用虛構的王國州。燈塔由四位研究者辦起來,個人可以免費用基礎檔,作坊付費得到更多服務。鎮上的訂單裡,來自遠方的部分增加了。數字是設例,不是 AI 行業的調查;我們借它看一種可能的變化:越有用,越依賴,什麼時候依賴開始限制後來想走的路?
先別急著把依賴都說成受制。我們依賴供電、語言、同事,也依賴學會的一門手藝。靠一個工具做得更好,本身不是工具偷走了自由。需要查的是,當人要換、要拒絕、要重新安排時,那份先前得到的能力還能不能由他繼續使用。
燈塔後來推出記事,保存多年的來往與偏好。作坊主想換一家時,發現新助手讀不了舊格式。他仍能走,只是要重建三年的記錄。他留下了。我們知道這份成本,不知道其他替代是否真的不可用,因此還不能直接宣佈退出已經封死。
這是一條形成的路:一個工具越懂你,交給它的東西越多;交得越多,換走越費力。也可能有別家做出足夠好的接續,改變這條路。資料迴路說明要往哪裡看,不是一條最後一定壟斷的預言。
通衢的市集走過另一條路。買家多,賣家願意來;賣家多,買家也更願意來。相遇容易,確實是網路帶來的好處。可是賣家一個人搬出去,原來的買家不跟著走;買家搬出去,想買的貨也許仍在這裡。兩邊很難約好同時搬家。
不能由此說網路效應本身有罪。讓人互相找得到,是許多服務存在的理由。要繼續查的是,有沒有可以跨過來往的辦法,平台後來又加上了什麼限制。關係聚在一處與別人把所有接續封在一處,不是同一句話。
再有一條路,藏在你從來沒有打開過的設置裡:默認。不是不許選別家,而是在你尚未作選擇時,路已經替你排好了。默認可能替人省事,也可能把競爭者放到許多人始終不會看見的地方。省事是真的,作用仍須核。
美國谷歌搜尋案給過一個具體的法律認定。哥倫比亞特區聯邦地區法院在2024年認定,谷歌通過排他的分發安排,非法維持通用搜尋服務與通用搜尋文字廣告市場的壟斷;2025年12月的最終判決列出相應禁令與救濟,谷歌和原告方隨後均提起上訴。法院處理的是特定市場、證據與法律,不是在宣佈所有默認設置都違法。
這一案照亮的區別很日常:設置裡可以改,不等於替代者實際得到了同樣的接近機會。這裡用法院已經作出的認定,不替法院另判一家公司的全部行為。法律上的排他效果與我們正在追問的可用路線相接,卻不因此變成同一個判斷框架。
形成還可能發生在收購裡。通衢買下柳筐市集,後者隨後停了。幾年後它又出價收購做冷鏈的小公司渡口。這是兩筆不同的事:柳筐已經關掉,渡口只有出價,不能把前一筆的結局預先填到後一筆。
賣方願意成交,當然重要。可賣方處理自己的事業,不替所有依賴那條路的人回答。若一個位置有能力在替代者長大之前不斷買下它們,就要查周圍還有什麼路能夠長起來。出價本身不證明違法,也不證明買方心裡只想消滅競爭。
同樣,幾家的條款相像,不自動說明商量過。它們也可能各自看見同一種收益,各自選擇相似做法。是否協調需要證據;沒有證據證明協調,也不使共同出現的實際限制忽然不存在。行為的來處與形成的處境,要分別說明。
標準與接口也會接在這些路上。一家寫了入口的格式,別人只有照它的規格才能接進來;記錄導得出,別家讀不了;夥伴的任務排在前面,其他人不知道怎樣等。每一件都可能有技術上的理由,技術理由仍要說到實際限制哪裡為止。
這些東西往往不是單獨來的。市集聚著買賣雙方,物流接著市集,自營商品排在首頁,記錄留在帳戶裡。只盯其中一個環節,有時看不出人為什麼走不動;把它們全揉成一個大字,又看不出該改哪裡。需要合起來看,再一項項拆開核。
而且,它握住的未必只是價格。原先我們問,賣家多交了多少傭金,工人少了多少收入。後來還得問,他取得資訊要經過哪裡,哪些東西先到眼前,哪些東西根本不會出現。這些不是人的判斷本身,卻是他拿來判斷的材料。
在設例裡,通衢把搜尋列表移到下面,首頁換成問路助手。買家說要什麼,助手只給一兩件,替他下單。列表還在,多點兩下也能看;不能因此說買家已經失去判斷,但服務已經從讓人看見許多貨,走到先替人挑出極少的貨。
願意少操心,可以是買家自己的選擇。問題不是每個人都該親手比較所有商品,而是這份委託按什麼規則執行,誰能改,想看更多時有多難。若給材料的一端集中起來,需要核的深度也變了,不能只用價格便宜沒便宜答完。
時間又給判斷添了一層。通衢早年拿積累替新店免傭金,後來取消扶持,把賣家互選的申訴席位也改成管理層任命。兩回使用的是同一家公司積下的能力,作用卻不必相同。早年好用,不是後來每一項安排的根據。
不過,照現在看也不能只向一個方向用。FTC訴 Meta 案裡,地區法院2025年判委員會沒有證明 Meta 現在仍有相應的壟斷力,強調需要證明現在,而非僅講過去。FTC在2026年提出上訴。這是一個具體法律爭議中的判決與後續步驟,不是所有結構問題都已消失的證書。
我們從中取的是提問的時間:今天的力量要有今天的根據。不能反過來補成法院已經認定過去確有壟斷;判詞沒有這樣替過去定案。也不能拿特定法律市場中的結論,省掉另一個框架下對去留、選擇和申訴實際效力的核對。
假如有一天通衢份額降了,替代者真的接得上,原有的限制也應重看。曾經需要的救濟可能仍然需要,也可能可以收窄。要說明現在還處理什麼問題,而不是讓一家公司因一份舊診斷永遠失去重新被判斷的機會。
這一重看不撤銷已發生的行為。柳筐當年為何被買下、買下以後發生什麼,仍按當年的材料問;今天的通衢有什麼作用,按今天問。把舊帳抹掉與拒絕拿舊結論包辦現在,完全是兩件事。
因此,最該小心的也許不是一句大公司不好,而是兩句聽來相反、做法相同的話:它從前幫過我,所以現在不用問;它從前控制過,所以現在無需再看。兩句都把眼前的人和路,交給了一個已經說完的故事。
燈塔讓那位作坊主能夠寫出第一封外文回信,這份能力值得承認。等到他想換一家時,記錄怎樣帶走,同樣值得認真答。並不是先否認工具的用處,才有資格問它握住了什麼;恰恰因為那些用處真實,後來離開的事才如此具體。
看形成,不是抓住某一年宣佈從此變壞。是把一個位置怎樣長出來、怎樣連到別的層上、怎樣改變人的去處讀清楚。下一篇再把目光移到已經留下的問路:如果所有窗口都還開著,為什麼有時誰也推動不了裡面的事?
Source & reading note原论文与阅读说明原論文與閱讀說明
Nine essays draw on the Monopoly prequel, seven papers and dialogue with earlier thinkers. Kingdom State, Crossways, Beacon and future water provision are imagined, not disguised allegations or forecasts. Linked cases distinguish investigations, allegations, jurisdiction, liability and remedies within their stated institutional and dated scope. Size, success, defaults and investment do not alone establish a violation; benefits do not answer for every exercise of control. Separation, interoperability and standby capacity are discussed as proposals with costs, not proven universal solutions. This is not legal or investment advice. English is independently written.
九篇据《垄断论》前置篇、七篇正文与传统对话篇改写。王国州、通衢、灯塔及未来供水等场景为设想,不是现实企业的化名或对未来的预测。真实案件与制度材料另列来源,区分调查、指控、管辖判断、责任判决与救济;按相应机构和材料日期说明,不以一项法律结论代答全部规范问题。规模、成功、默认设置或投资关系本身不直接证明越线;有益的一面也不替具体控制作答。关于拆分、互通与备用能力的建议不是已经验证的万能方案,不提供法律或投资建议。简繁中文同文,英文独立重写。
九篇據《壟斷論》前置篇、七篇正文與傳統對話篇改寫。王國州、通衢、燈塔及未來供水等場景為設想,不是現實企業的化名或對未來的預測。真實案件與制度材料另列來源,區分調查、指控、管轄判斷、責任判決與救濟;按相應機構和材料日期說明,不以一項法律結論代答全部規範問題。規模、成功、默認設置或投資關係本身不直接證明越線;有益的一面也不替具體控制作答。關於拆分、互通與備用能力的建議不是已經驗證的萬能方案,不提供法律或投資建議。簡繁中文同文,英文獨立重寫。
Read the source paper ↗阅读原论文 ↗閱讀原論文 ↗
FTC announcement of its January 2026 notice of appealFTC关于2026年1月上诉通知的公告FTC關於2026年1月上訴通知的公告 ↗