Monopoly: Exemption from Questioning
垄断的免问
Once a position has formed, is the road by which it is questioned still there? If the road is there, does questioning have effect? This paper writes two forms of "exemption from questioning": the road of questioning is closed; or the road remains, and questioning does not move it. For the first, it looks at how a firm makes rules, changes rules and adjudicates appeals on its own, keeps its rules opaque, substitutes a compliant text for being questioned, or writes its rules into programs. For the rules one firm or a few make for many people, the scope is drawn by the institution, not fixed and changed by the firm on its own; changing rules is the same as making rules; the channel of questioning is functionally independent of it. Where rules are written into a program that executes automatically without a corresponding stage of fresh judgment, every run carries out the corresponding rule once, and the entry points for questioning are set accordingly (LD4). For the second, it looks at too big to fail, deep integration that makes separation difficult, the threat of exit, the public support that rescues it, and the reviewer's capability dependence on the party reviewed. That the channel of questioning is there does not prove that questioning has effect; actual efficacy can fall on reasons, actions, rules or the holding of office; not every questioning must reverse a decision, still less must there be a breakup before questioning counts as having effect. Too big to fail is structural influence, not an exemption from it. Capability dependence is set up as a separate category: if what the reviewing side looks at, how it measures and how it interprets can be supplied only by the party reviewed, questioning may also have no effect on it; that dependence is present does not mean capture has already occurred (LD5). The constructed examples use a model company's agent interface, a city's allocation of water in a drought year, a body that reviews models reading filings with the reviewed party's tools, and the three days in which that model company went down. In reality, in Epic v. Apple the district court found that Apple's new commission arrangement violated an injunction, and the court of appeals affirmed; in the two Google cases, search and advertising technology, the courts weighed the risks of integration, migration and taking over when choosing remedies, and these show only the costs and constraints of changing a structure after the fact; several decisions of the European Commission are one form that checks by several states acting together take in reality. Whether a company's monopoly can grow larger than a state this paper writes as a possible form, not as an existing fact.
Writing Declaration: This paper was co-drafted with Claude (Anthropic). All intellectual decisions, framework design, and final editorial judgments were made by the author.
Introduction
In the winter of Year 36, Beacon Assistant went down for three days. An update went wrong, and during the rollback it ran into a failure on the platform of its cloud partner, Stone Granary.
In those three days, Apple City's building permit approvals stopped: from Year 34 the initial review and scheduling of permits had been done by Beacon Assistant. The initial review of the State Labor Agency's work-injury claims stopped, with a backlog of about four thousand cases. The four clinics' queue numbers went back to the telephone, and the lines were jammed. Crossways Ask-the-Way stopped, and Crossways Market's orders in those three days fell by about forty percent. Neither Apple City nor the State Labor Agency had kept a usable manual process: the section of Apple City that did initial review had been cut from twelve people to three in Year 35.
Afterward, Beacon made two requests to the state government: support for "backup and resilience"; and no liability for breach for "outages caused by force majeure" within a year. Beacon's explanation was: "If we go down, half the state stops."
The Economic Committee of the State Legislature held a hearing. One member asked: "If it can't be allowed to fail, do we have to hold it up?" By the end of Year 36, the requests had not been put to a vote.
Before the outage, the average time for the initial review of permits had shrunk from about six weeks to about one; the backlog of initial claims reviews had been cleared once in Year 35. These are true as well.
Ten years earlier, settlement at Crossways Pay had been delayed for three days, and the state gave it an emergency loan; the state government's spokesperson said: "If it falls, what falls is forty percent of the small merchants."
The two matters are ten years apart and of the same shape. The roads by which it is questioned were there all along: the State Legislature can hold hearings, the state government can refuse, the courts can take up cases. The roads are there; does questioning have effect?
This paper asks about two things. The road of questioning closed: a firm makes rules, changes rules and adjudicates appeals on its own, or writes its rules into programs (§§1, 2). The road remains, and questioning does not move it: the reviewer relying on the capacity of the party reviewed, too big to fail, deep integration, the threat of exit (§§3–6).
§1 The Road of Questioning Does Not Belong to the Party Questioned
In Year 32, Beacon released an "agent interface": only when other services connect through this interface can Beacon Assistant call on them for users, to book rooms, get queue numbers or pay bills. Connecting requires Beacon's approval; the standards are written in Beacon's Access Code, which Beacon can amend unilaterally.
In Year 33, a room-booking service in Willow Ferry, the queue systems of four clinics in Apple City, and Apple City's municipal payment system all connected. Other firms' assistants could also use this interface to call on the services connected, but with a daily cap on the number of calls, and queued behind Beacon Assistant. In Year 34, when another model company's assistant called on the Willow Ferry room-booking service, at peak times it had to wait about ten times as long.
That same year, Beacon amended the Access Code: a connected service that also connected to another firm's interface had to "ensure a consistent experience." That same year, the room-booking service stopped another firm's interface; how the two matters relate, the example does not record.
In the spring of Year 34, Beacon retired an old version of the assistant, with ninety days' notice. Apple City's payment system, connected to the old version, had to be modified; the modifications took about fifteen percent of the city's information budget for that year, and for three weeks during the modifications bills could not be paid in the assistant.
In Year 35, Beacon delisted a small connected clinic on grounds of "non-compliance with the Code." The appeal was handled by Beacon's connection review team, which gave no reasons; three months later the clinic was restored.
The other face is to be recorded too: after connecting, the clinics' telephone queues shrank, and orders at the Willow Ferry room-booking service doubled.
How an interface serves as a gate is examined in Paper 5. What this paper reads is the road of questioning: who makes the Code, who changes it, and who adjudicates appeals against it. All three sit with Beacon.
The Political Economy specialist paper has fixed this: a position others cannot get around makes rules for many people, and how ranking, access, settlement and appeals are handled falls on their way to sustaining their lives or unfolding their own paths; "The scope of these rules is drawn by the institution, not fixed and changed by the rule-making party on its own"; the rules made can be questioned, and the channel of questioning is functionally independent of it (A Way Out, a Way Forward, E41a). The Theory of the Market applied this sentence to infrastructure that cannot be got around (Theory of the Market Paper 3, SC6). That the scope is drawn by the institution does not mean that every rule must be submitted for approval; what is required is that on what kinds of matters it may make rules, and to what limit, is not for it alone to say; and that once it has made them, there is a road it cannot close by which they can be questioned.
Independence is functional. Institution Theory has fixed three conditions for the channel of questioning: "it is independent of the party questioned, internal subjects can actually initiate it, and initiating it brings no retaliation. Independence is functional and does not require being outside the institution; what it guards against is the party under review sealing off review of itself on its own" (Institution Theory Paper 6, Z34). That appeals go to Beacon's own review team, and that the team sits inside Beacon, does not in itself mean that the road of questioning has been closed; what is to be looked at is whether the small clinic can ask through a channel that is functionally independent and that Beacon cannot close, whether asking has effect, and whether asking brings retaliation. In the example, the review team gave no reasons and the clinic was restored three months later; where else besides the review team the clinic could have gone to ask, the example does not record.
Changing rules is the same as making rules. Retiring an old version looks like a technical matter; but it falls on those connected to the old version, just as making a new rule does. Apple City spent fifteen percent of its information budget on it, and for three weeks bills could not be paid in the assistant. Adding the line "ensure a consistent experience" to the Code is also changing rules. The room-booking service stopped another firm's interface in the same year; how the two matters relate, the example does not record, and this paper does not read cause from sequence. Changing a rule and making a rule are questioned in the same way: who draws the scope, and to whom the road of questioning belongs. Whether ninety days' notice is enough lies at the rule layer; that changing rules is to be questioned lies at the normative layer.
In reality, a court's finding shows how changing rules can close again a road just opened. In Epic v. Apple, the United States District Court for the Northern District of California issued an injunction in 2021 requiring Apple to allow app developers to steer users within apps to purchases outside the store. In April 2025 the same court found that Apple had willfully violated the injunction: Apple changed its rules to allow outside links, while charging a 27% commission on purchases made outside through those links; the court found that this arrangement was meant to make outside purchases economically unworkable, and held Apple in civil contempt. On 11 December 2025 the Ninth Circuit affirmed the contempt finding, and at the same time reversed the part of the district court's order that banned any commission altogether, holding that Apple may charge a commission on linked-out purchases corresponding to costs that are "genuinely and reasonably necessary," and remanded for that to be set. On 30 June 2026 the United States Supreme Court granted certiorari on the question of civil contempt, with argument set for 11 January 2027; at the time of writing it had not been decided; if the judgment is reversed, this text will follow the new judgment.
This series writes it as the two levels of court have found, and gives no verdict of its own on Apple's character. It uses it to show one thing: after a road has been opened by a court, the side holding the chokepoint can, with a new rule, make it unworkable again while allowing it. Changing rules and making rules can have the same effect.
Opacity and compliant texts. Beacon's Code is public, but how it was applied in review, and why that clinic was delisted, no reasons are given. A rule written on paper is not a rule that has been questioned; substituting a compliant text for being questioned answers "we acted under the Code," and does not answer "why the Code is written this way, and why it was applied this way this time." When the party questioned produces a text it wrote itself, the road of questioning is still in its own hands.
§2 Written into Programs
This section uses an imagined far-future scenario. After machines have taken over work such as farming and drinking water, Apple City's drinking water is supplied by the city-owned Water Department, and from Year 42 the operation and maintenance of the treatment plants and mains have been handed to an operations contractor, the operator: Hundredcraft's machines do the maintenance, and Beacon's model does the dispatching. What is written in this passage is imagined, not a prediction of the real future.
In the summer of Year 53 there was a drought, and the reservoirs held about forty percent of their usual water. Apple City Council set the goals for allocating water: "Households first; hospitals and schools are not cut off; everything else is reduced in proportion." Which district is cut first each time, and by how many hours, is decided by the operator's dispatching model; the model recalculates every six hours.
From July to September, the hours of reduced pressure and water cut-off in the East Market in the old city were about twice those in the newer northern districts. The reason the operator gave was: "The East Market's mains are old and lose a lot of pressure; cutting the East Market saves more water." The full calculation is produced by the model, and the City Council cannot read it.
Residents' complaints are answered by the operator's assistant. That year there were about twelve thousand complaints, and fewer than one percent were passed to a person.
The City Council asked the operator to submit its dispatching rules. The operator submitted a forty-page explanation, saying that the model's weights "change with every recalculation," and that it could not give fixed rules. In the autumn, a council member proposed that the rules for allocating water must be approved by the City Council. By the end of the year the proposal had not been put to a vote.
Both faces are recorded: in that dry season no hospital or school lost water, and every household in the city had water for at least four hours a day.
The goals of allocation were set by the City Council. Whose water is cut each time was decided by the model. Between the two, where are the rules?
Reading what a standard looks like once it is written into a program, the Theory of Disciplines says: "In arrangements that execute automatically, without a corresponding stage of fresh judgment, every run of the program carries out that line of ought once; arrangements that embed review, appeal and human judgment are checked separately." (Theory of Disciplines Paper 7, S36) The original speaks of the automatic execution of disciplinary premises; here it is extended to the automatic execution of chokepoint rules (own undertaking). The dispatching model recalculates every six hours, and each time it carries out a rule of "where to cut first"; no run idles without carrying out a rule. So the entry points for questioning must be set accordingly: not only to question the goal the City Council set, but also the rule each recalculation carries out.
That the weights "change with every recalculation" does not mean there are no rules. What it says is that the rules change; if the rules change, the road of questioning must fall on how they change: who decided to let them change this way, whether the range of change has limits, and where, once they have changed, the people of the East Market can go to ask. Saying "it cannot give fixed rules" answers the form of the rules, not the very matter being asked: each time, where is cut first, by what is that decided, and who is responsible.
The Theory of the Company specialist paper said something of the same form when reading platforms' dispatching, pricing, rating and deactivation: algorithmic management is management, these are decisions that fall at positions, and "the reasons can be given, they can be questioned, and questioning has actual efficacy; a decision does not lose its place of explanation and of being questioned because a system output it" (Platforms and Gig Work, PT2). The Theory of Government says that when judgment is handed to a machine, the bearing is not handed over with it; the bearing falls on the positions that hand the judgment out, set how it is to be used, and act on its results (Theory of Government Paper 1, G8). In the matter of allocation, the bearing falls on the City Council and the Water Department, which set the goals of allocation and handed over the dispatching; the operator holds its own accountability toward the construct that engaged it, and does not answer, in that construct's place, to those the decision acts on (ibid.). The model is not an address.
Arrangements that embed review, appeal and human judgment are examined separately. The East Market's complaints have a road: the assistant receives them, records them, and passes fewer than one percent to a person. That few are passed to a person does not by itself show that complaints have no effect; perhaps most complaints are answered completely at the assistant. What is to be examined is whether that road can carry a matter to a position that can change the outcome, and whether the position that can change it is independent of the operator. What became of the fewer than one percent once they were passed to a person, the example does not record.
That the East Market was cut more does not in itself show that a line was crossed: old mains and large losses are a reason that can be examined, and the example has not examined it. What this section reads is not whether the East Market should be cut, but where the road of questioning falls once the rule of "whom to cut" has been written into a program.
LD4 The road of questioning does not belong to the party questioned; changing rules is the same as making rules. For the rules one firm or a few make for many people (ranking, access, delisting, account bans, interface terms), the scope is drawn by the institution, not fixed and changed by the firm on its own, and the channel of questioning is functionally independent of it (per A Way Out, a Way Forward, E41a; Institution Theory Z34; Theory of the Market SC6). Where rules are written into a program that executes automatically without a corresponding stage of fresh judgment, every run of the program carries out the corresponding rule once, and the entry points for questioning are set accordingly; arrangements that embed review, appeal and human judgment are examined separately (per Theory of Disciplines S36: the original speaks of the automatic execution of disciplinary premises; here it is extended to the automatic execution of chokepoint rules; own undertaking; the same form as Platforms and Gig Work, PT2, "Algorithmic management is management").
§3 Those Who Review It Also Use It
In the autumn of Year 33, the Kingdom State set up the Kingdom State Office of Model Affairs, to handle the approval and filing of models. At its founding it had twenty staff, five of them engineers.
Approval and routine supervision rely on each firm's self-test reports. The Office has no testing equipment of its own; to carry out spot inspections, it has to ask the party inspected, or a body that party designates, to do them. From Year 34 the Office used the institutional version of Beacon Assistant to read the filings each firm submitted and write initial review summaries; Beacon's own filings were read by it too.
Of the five engineers, two came from Beacon and one from Stone Granary. In Year 35 one of them returned to a post at Beacon.
That same year, a report by the Kingdom State Audit Office said that in the Office's spot inspections "the test methods used were all provided by the party inspected," and that for the initial review summaries "no independent review was found." The Office replied: "We have no other way; Beacon's tools are the best."
In Year 36, the state funded a testing laboratory at Oakgrove College: eight engineers, testing equipment of its own; its reports go directly to the Office and to the State Legislature; its funding and the term of its head are set by the State Legislature. That same year the laboratory retested two self-test results of one company: one matched, one did not; for the one that did not, the company tested changed its statement.
There is one more related matter, read in Paper 3: in Year 34 the state government issued rules on model safety under which a model whose parameter scale is above a set level must be approved before release; that year only two bodies in the state could do the tests approval required, one being Beacon's testing department and the other a subsidiary of Stone Granary.
The road of questioning is there: the Office was set up, the Audit Office inspected, the State Legislature can hear. Yet what the reviewing side looks at, how it measures and how it interprets all rely on what the party reviewed provides: its self-test reports, the tests it designates, its tools, the engineers who came out of it. The road of questioning is there, but the eyes of the questioning side are borrowed from the party questioned.
This is capability dependence: the reviewer's dependence on the reviewed party's own information, technology and personnel. This series sets it up as a separate category (own undertaking).
It has the same form as a sentence in the Theory of the Company specialist paper. That paper says that a judgment entrusted to a position requires that the position be able to judge: "the information needed can be obtained and understood, there are corresponding competence, training, time and support" (The Company in the Age of AI, ZN1). The original speaks of positions of judgment within a company; here it is extended to the reviewer's position (own undertaking). If a reviewer cannot obtain independent information, cannot understand what the party reviewed hands in, and has no capacity to test for itself, its review is review in form, and in efficacy may not move the party reviewed.
Kept apart from another kind of dependence. The Political Economy specialist paper has already set out one kind of dependence: the reviewer does not depend in substance on the party reviewed for its income, its term, its promotion or its later prospects (A Way Out, a Way Forward, E43). That kind asks whether the reviewer relies on the party reviewed to live; this kind asks whether the reviewer relies on the party reviewed to see. The two meet on the item of personnel: engineers come from the party reviewed and go back to it. Under E43, what is to be looked at is whether their income, term and later prospects rely on the party reviewed; under capability dependence, what is to be looked at is whether, without them, the Office can still understand and still measure. For the same matter, the two are examined each on its own, and neither stands in for the other.
That dependence is present does not mean capture has already occurred. This sentence from the specialist paper (A Way Out, a Way Forward, E43) holds here too: the Office used Beacon's tools to read Beacon's filings, and the example does not record that its reading was biased; nor does where the engineers came from, by itself, prove a lack of independence. Capability dependence is not a charge but a structural reading: part of the efficacy of questioning is pledged into the hands of the party questioned.
The testing laboratory is one answer. It has its own people and its own equipment, its reports go out directly, and its funding and term are not in the Office's hands, still less in Beacon's. Once set up, it retested and found one result that did not match, and the company tested changed its statement: questioning moved things a little. This does not show that the dependence has been removed, since the laboratory has only eight people; nor does it show that self-tests are generally untrustworthy, since one of the two matched. What it shows is that capability dependence can be reduced, and the way to reduce it is for the reviewer to acquire the capacity to see and to test for itself.
One question is left for Paper 6: what difference it would have made had the laboratory been built three years earlier.
"Too complex for outsiders to regulate" is what capability dependence looks like on the side of what is said. The Office's line, "We have no other way; Beacon's tools are the best," is its account of its own situation: it really had no equipment of its own; whether there truly was no alternative, and whether Beacon's tools were the best, the example has not examined. If the party reviewed said the same, what it would be answering with is a kind of name. How such statements are examined is in Paper 3; this section examines the structure.
§4 Questioning Must Have Effect
Back to the three days in which Beacon went down.
The road of questioning was there. The State Legislature held a hearing, and a member asked; the state government could give, and could also refuse. Yet this time, in what position did the questioning side stand? Permit approvals, the initial review of claims, the clinics' queue numbers, the market's orders, all pressed on Beacon at one place; the manual processes had been withdrawn and could not catch them. Those who questioned it could not themselves leave it.
When Political Economy asks about control within the line, it asks three things: "who can activate it; what actual efficacy it has on reasons, actions, rules, or the holding of office; how the party questioned cannot cancel it on its own." (Political Economy Paper 3, E16) The first question can be answered here: the Legislature can activate it. The second is the business of this section: once questioned, is there actual efficacy?
Being open to questioning does not cancel monopoly. Where dependence leaves questioning without actual efficacy, the structural influence remains.
Reading the exit of the strong, Political Economy said that only when structural dependence is present does exit show a power-face; "What can leave is the construct; what cannot are those who depend on it" (Political Economy Paper 3, E20). The original speaks of a construct tying its relocation to a piece of legislation. Political Economy Paper 4 has already extended structural dependence to a position's failure, and what it says is that the position obtains public support (Political Economy Paper 4, E25); here it is extended one step further, to the actual efficacy of questioning: if it goes down, those who cannot afford its failure are those who depend on it, and those who question it are among them (own undertaking). The line "If we go down, half the state stops" answers exactly this: the reading of dependence. It does not answer the very matter being asked: on what ground support should be given, on what ground liability should be waived.
Too big to fail is structural influence, not an exemption. Political Economy says that structural dependence can bring an economic position public support; obtaining support does not issue legitimacy to its existing acts and its continued existence; the public position that gives the support must state its grounds, authority, object and cost, and must answer the very matter it is questioned about; "public support obtained by scale itself needs a ground of its own" (Political Economy Paper 4, E25). If Beacon's two requests are granted, their ground must be set somewhere other than its being unable to fail: which supply is to be secured, to what degree, who takes over, and on whom the cost falls. Political Economy has already read Crossways Pay's emergency loan in this way: the interest rate was below the market rate, and the condition was that no dividends be paid to the parent company for three years; a loan with conditions attached is not the setting up of a new general guarantee. Beacon's case is the same shape, coming round again ten years later.
Whose act it was to withdraw the manual processes. Apple City cut its initial-review section from twelve people to three, and the State Labor Agency kept no manual road. These are their own acts, not Beacon's. The Prequel said that the construct that adopts one firm's product as a gate answers for its adoption; Apple City and the State Labor Agency adopted Beacon Assistant, and also withdrew the roads that could catch things. Part of the reading of being unable to fail is booked to them. This paper does not read Beacon as using the outage to exert pressure; its requests are requests, and what is recorded is what it said.
In the far future, the same shape runs deeper. An imagined scenario: from Year 40, Hundredcraft's farm machines are driven by Beacon's model, and Fieldworks, the agricultural operating company the two set up together, operates about seventy percent of the Kingdom State's farmland by Year 50. At the autumn harvest of Year 47, an update to Beacon's model went wrong, the combines in two counties stopped for nine days, and about twenty percent of the harvest was lost. The Kingdom State Agriculture Department had no farm machines of its own, nor any system that could schedule the farming season; in those nine days no one in the state could be found with the people and machines to take over. After the outage, the state gave Fieldworks a special "food security" fund, to be used for backup systems. In the same imagined scenario, Apple City's Water Department handed over its operation and maintenance, and the department's engineers fell from forty to four, all four managing the contract rather than the operations; in Year 49 another operator gave an estimate for taking over, saying the systems would first have to be retrofitted, with water cut district by district across the city for about two months; at the renewal in Year 50 the operating fee rose by about forty percent, the City Council voted it through, and a council member who voted in favor said: "Without renewal, there is no water."
These two matters are imagined, not predicted; Papers 5 and 7 each read them in their own way. Here they are only kept apart from the reviewer's capability dependence. The matters of Fieldworks and the water are structural dependence in supply and taking over: in the nine days Fieldworks was down, no one in the state could be found with the people and machines to take over; for the water, another operator did give an estimate for taking over, but the systems would first have to be retrofitted, with water cut district by district, and the choices at renewal were narrowed by this. The Office's matter is the reviewer's capability dependence: it reviews, and may not understand what it reviews. Both kinds of dependence can weaken the efficacy of questioning, weaken it in different places, and call for different remedies; whether they have already left a particular questioning without efficacy must be examined case by case.
Guarding against overreach. That questioning did not reverse a decision does not mean that questioning has no effect. Actual efficacy can fall on reasons: the party questioned must give its reasons, and if it cannot, that itself is a reading; on actions: pausing, correcting, adding a manual road; on rules: amending a code; or on the holding of office. Not every questioning must reverse a decision, still less must there be a breakup before questioning counts as having effect. That Beacon's requests had not been put to a vote by the end of the year does not show that the committee was persuaded, nor that it was held by the party making the requests. The explanation recorded, "If we go down, half the state stops," still did not answer the very matter the member asked.
§5 The Cost of Changing a Structure After the Fact
In reality, two courts' opinions on remedies show what changing a position's structure after the fact costs.
The Google search case. After finding in August 2024 that Google had unlawfully maintained a monopoly in search, the United States District Court for the District of Columbia issued its remedies opinion on 2 September 2025. The plaintiffs asked for Google's Chrome browser to be divested, and the court refused. The court's reasons were several: structural remedies require a "significant causal connection" between the unlawful conduct and the monopoly power, which the plaintiffs had not shown; Chrome relies heavily on Google's technical systems, infrastructure and personnel, and the court said that divesting it would be "incredibly messy and highly risky." What the court ordered were behavioral remedies: exclusive distribution contracts are prohibited, and payment for default placement is permitted but for terms of no more than one year; parts of the search index and user-interaction data are to be shared with qualified competitors; search results are to be syndicated to them; a technical committee is to assist with enforcement; the judgment runs for six years. The final judgment was entered on 5 December 2025; both sides have appealed, and at the time of writing the appeal had not been decided.
The Google advertising technology case. In April 2025 the United States District Court for the Eastern District of Virginia found that Google held monopolies in two markets, publisher ad servers and ad exchanges, at the core of which was the tie between the two. On 2 September 2026 the court issued its memorandum opinion on remedies (unsealed on 16 September), refusing to order the sale of the ad exchange and saying that structural remedies are "neither realistic nor needed." Its reasons: the plaintiffs had not met the standard of significant causal connection, nor shown that behavioral remedies would be ineffective without divestiture; the plaintiffs' divestiture plan did not identify any prospective buyer; a sale would take years, and the market might change first. The opinion recorded Google's side's statement that these two businesses had for seventeen years been tightly bound up with Google's core infrastructure; it also recorded an internal analysis by Google itself in 2023 to 2024 which considered divesting the ad exchange as a business technically feasible, and noted that this analysis differed in scope from the plaintiffs' plan. What the court decided to impose were behavioral remedies: tying terms between the two businesses are prohibited; they are to interoperate with an open-source bidding framework; the ad exchange's real-time bids are to be offered to other publisher ad servers; publishers' data can be exported; a monitor and a technical committee are to be appointed. An order the same day required the parties to submit jointly, within thirty days, a proposed final judgment; on 2 October the parties jointly submitted a proposed judgment, with the disputed terms set side by side for decision. As of the quarterly report Alphabet filed on 29 October 2026, the final judgment had not been entered; the company said in the report that, once the final judgment is entered, it plans to appeal the parts of the April 2025 decision adverse to it. Later developments were not examined at the time of writing.
This series writes these as the two courts have found, and gives no verdict of its own on Google's character. What these two matters are used to show is the cost and constraints of changing a structure after the fact: in choosing remedies, the courts put the risks of integration, migration, taking over and product degradation into the balance. In these two matters, the courts recorded the depth to which things had grown together as a risk of separating them; how large the cost is, and what is uncertain, must be examined by the particular structure and the particular plan of separation.
These two matters are not used to show that questioning has no effect. The courts found violations, ordered behavioral remedies, and appointed a technical committee and a monitor; questioning moved things. What actual efficacy these arrangements have remains to be seen later, and is examined separately. Nor are they used to show that things cannot be separated: what the courts said is that under the standards and evidence of these cases, breakup was not needed and not realistic, not that no kind of separation can be done; the opinion even records an internal analysis that considered it technically feasible.
What they point to is something else: where integration makes separation after the fact difficult, preserving conditions for taking over in advance can prepare for later separation and for someone else taking over: separate interfaces, separate accounts, arrangements for taking over made in advance. Doing it in advance also has costs that must be stated as they stand; for the costs of separating after the fact, these two opinions on remedies give the shape. The two cannot be directly offset: which conditions are preserved, how much is paid in advance, and which costs are later reduced must be examined separately for the same arrangement. That preparation for separability comes before the fact is established in Paper 5; letting it be able to fail, in Paper 7. That it cannot be separated does not exempt it from checks.
§6 Larger Than a State
Questioning without effect has one larger shape.
A state holds the power to approve and to prohibit: whether a company can operate within its territory is in the end for it to decide. But when many of a state's affairs press on one company, the cost of using this power can grow so large that the power cannot be used: prohibit it, and what stops is the state itself. Then the power is there, and questioning does not move it.
Read it by the sentence with which Political Economy read the exit of the strong: what can leave is the construct; what cannot are those who depend on it (Political Economy Paper 3, E20). When a construct says "if this happens, we leave," under structural dependence this sentence presses on the side that makes the rules. In the Kingdom State example, Crossways declared in Year 24 that if the draft Online Marketplaces Act passed together with its breakup clause, it would move its headquarters and settlement center out of the state; Political Economy and the Theory of the Market have both read this matter, and it is not redone here.
Whether a company's monopoly can grow larger than a state this series does not rule out; it writes it as a possible form, not as an existing fact. A state's power has many sources, and companies usually have no army, courts or taxes; that it is costly for a state to prohibit a company does not mean it cannot be done. Only one sentence is recorded here: when things reach that point, a state may itself be unable to question with effect, and several states have to question together.
In reality, there is a form of checks by several states acting together that can be looked at as an example: the European Union. The European Commission is a body set up by treaty among a number of states, and its decisions fall on companies operating in each member state. On 23 April 2025 the Commission adopted its first two non-compliance decisions under the Digital Markets Act: it found that Apple had breached its anti-steering obligation, with a fine of 500 million euros; and it found that Meta's "consent or pay" model of the time did not give users an equivalent option that uses less personal data, with a fine of 200 million euros. Apple and Meta have both brought actions before the EU General Court (the Meta case is T-435/25, brought on 4 July 2025). On 5 September 2025 the Commission found that Google had favored itself in advertising technology, fined it 2.95 billion euros, and ordered it to end the self-preferencing and to remove the conflicts of interest inherent in the advertising technology chain; at the same time it set out a preliminary view that only the sale of part of the business could resolve this conflict, while first hearing Google's proposal, with the particular remedy to be decided later. Google submitted a remedy proposal in November 2025, which the Commission is assessing; Google and Alphabet brought an action before the General Court on 20 November 2025.
This series writes these as the Commission has found, with their status marked, and gives no verdict of its own. These decisions serve here as examples, not as proof: they show that checks by several states acting together have one form in reality; they do not show that it is the only form, nor that it necessarily has effect. How several states check, supervise and question together is handed to the Theory of Inter-State Cooperation; the functions between a state and a multinational company, who makes the rules, whom they bind, how they are carried out and where the accounts fall, have already been written in Theory of the Market Paper 7. This paper only leaves this interface.
LD5 Being open to questioning does not cancel monopoly; questioning must have effect. That the channel of questioning is there does not prove that questioning has effect. Where dependence leaves questioning without actual efficacy, the structural influence remains. Actual efficacy can fall on reasons, actions, rules or the holding of office; not every questioning must reverse a decision, still less must there be a breakup before questioning counts as having effect (per E16, "actual efficacy"; per E20 and E25: E25 has already extended structural dependence to a position's failure, and what it says is that the position obtains public support; here it is extended to the actual efficacy of questioning; own undertaking). Too big to fail is structural influence, not an exemption from it; public support that rescues it does not issue legitimacy to its continued existence (per E25). Where integration makes substitution, separation or taking over difficult, it is examined by actual dependence; that it cannot be separated does not exempt it from checks (under LD11). Capability dependence: if what the reviewing side looks at, how it measures and how it interprets can be supplied only by the party reviewed (its data, its engineers, its self-tests), questioning may also have no effect on it. The way of examining is whether the reviewer has the capacity to obtain information independently and to test independently (the same form as The Company in the Age of AI, ZN1: the original says of a position of judgment that "the information needed can be obtained and understood," and that there is corresponding competence; here it is extended to the reviewer's position; own undertaking). The boundary with E43: on the item of personnel, capability dependence looks at whether, without the reviewed party's people, the reviewer can still understand and still measure; E43 looks at whether the reviewer's income, term and prospects depend on it; the same matter is examined under each of the two separately. That dependence is present does not mean capture has already occurred (set up as a separate category; own undertaking).
Where This Paper Stops
This paper has established two propositions: the road of questioning does not belong to the party questioned, changing rules is the same as making rules, rules written into programs are carried out once with every run, and the entry points for questioning are set accordingly (LD4); being open to questioning does not cancel monopoly, questioning must have effect, too big to fail is not an exemption, and capability dependence is set up as a separate category (LD5).
What it has not shown: that any regulator has already been captured; in which case structural remedies should be used; that capability dependence is capture; that a court's not ordering breakup proves that questioning has no effect. The Office of Model Affairs has not been judged to be held by the party it reviews; whether Beacon's requests should be granted this paper has not judged; whether the East Market should have been cut more it has not judged.
Handed to Paper 3: how the party questioned answers. How statements such as "too complex for outsiders to regulate" and "If we go down, half the state stops" are examined.
Handed to Paper 7: where questioning has no effect, the gap is filled by going back to checks before the fact; how the testing laboratory, manual processes and preparations for taking over serve as remedies.
The remainders it leaves: how several states cooperate is handed to the Theory of Inter-State Cooperation; the functions between a state and multinational companies follow Theory of the Market Paper 7; the design of cross-border mechanisms is still a remainder registered by Theory of the Market Paper 7, and this series does not take it up.
The Two Propositions of This Paper
LD4 The road of questioning does not belong to the party questioned; changing rules is the same as making rules. For the rules one firm or a few make for many people (ranking, access, delisting, account bans, interface terms), the scope is drawn by the institution, not fixed and changed by the firm on its own, and the channel of questioning is functionally independent of it (per A Way Out, a Way Forward, E41a; Institution Theory Z34; Theory of the Market SC6). Where rules are written into a program that executes automatically without a corresponding stage of fresh judgment, every run of the program carries out the corresponding rule once, and the entry points for questioning are set accordingly; arrangements that embed review, appeal and human judgment are examined separately (per Theory of Disciplines S36: the original speaks of the automatic execution of disciplinary premises; here it is extended to the automatic execution of chokepoint rules; own undertaking; the same form as Platforms and Gig Work, PT2, "Algorithmic management is management").
LD5 Being open to questioning does not cancel monopoly; questioning must have effect. That the channel of questioning is there does not prove that questioning has effect. Where dependence leaves questioning without actual efficacy, the structural influence remains. Actual efficacy can fall on reasons, actions, rules or the holding of office; not every questioning must reverse a decision, still less must there be a breakup before questioning counts as having effect (per E16, "actual efficacy"; per E20 and E25: E25 has already extended structural dependence to a position's failure, and what it says is that the position obtains public support; here it is extended to the actual efficacy of questioning; own undertaking). Too big to fail is structural influence, not an exemption from it; public support that rescues it does not issue legitimacy to its continued existence (per E25). Where integration makes substitution, separation or taking over difficult, it is examined by actual dependence; that it cannot be separated does not exempt it from checks (under LD11). Capability dependence: if what the reviewing side looks at, how it measures and how it interprets can be supplied only by the party reviewed (its data, its engineers, its self-tests), questioning may also have no effect on it. The way of examining is whether the reviewer has the capacity to obtain information independently and to test independently (the same form as The Company in the Age of AI, ZN1: the original says of a position of judgment that "the information needed can be obtained and understood," and that there is corresponding competence; here it is extended to the reviewer's position; own undertaking). The boundary with E43: on the item of personnel, capability dependence looks at whether, without the reviewed party's people, the reviewer can still understand and still measure; E43 looks at whether the reviewer's income, term and prospects depend on it; the same matter is examined under each of the two separately. That dependence is present does not mean capture has already occurred (set up as a separate category; own undertaking).
References
All SAE works below are cited by concept DOI, which always resolves to the latest version.
I · This system
Qin, H. (2026). Open to Questioning (SAE Institution Theory, Paper 6) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23271324
Qin, H. (2026). Judgment (SAE Theory of Government, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23412130
Qin, H. (2026). Checks (SAE Political Economy, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874142
Qin, H. (2026). Conversion (SAE Political Economy, Paper 4) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874146
Qin, H. (2026). A Way Out, a Way Forward (SAE Political Economy, Specialist Paper) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24021385
Qin, H. (2026). Platforms and Gig Work (SAE Theory of the Company, Specialist Paper) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24117808
Qin, H. (2026). The Company in the Age of AI (SAE Theory of the Company, Specialist Paper) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24117816
Qin, H. (2026). The Market: Chokepoints (SAE Theory of the Market, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24183372
Qin, H. (2026). The Market: Borders (SAE Theory of the Market, Paper 7) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24183384
Qin, H. (2026). Writing Standards into Programs (SAE Theory of Disciplines, Paper 7) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24177788
II · This series
Qin, H. (2026). What a Monopoly Is Not (SAE Theory of Monopoly, Prequel) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377536
Qin, H. (2026). Monopoly: Formation (SAE Theory of Monopoly, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377542
Qin, H. (2026). Monopoly: Names (SAE Theory of Monopoly, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377553
Qin, H. (2026). Technology and Monopoly (SAE Theory of Monopoly, Paper 5) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377558
Qin, H. (2026). AI and Monopoly (SAE Theory of Monopoly, Paper 6) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377562
Qin, H. (2026). Monopoly: Remedies (SAE Theory of Monopoly, Paper 7) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24377564
III · Real-world cases
Epic Games, Inc. v. Apple Inc., No. 4:20-cv-05640 (N.D. Cal.), contempt finding April 2025 (tier B, per law firms' commentary); No. 25-2935 (9th Cir. Dec. 11, 2025), contempt finding affirmed, the commission part reversed and remanded (tier A, full text of the opinion); United States Supreme Court No. 25-1311, certiorari granted 30 June 2026, argument 11 January 2027, not yet decided (tier B).
United States v. Google LLC, No. 1:20-cv-03010-APM (D.D.C.), remedies opinion 2 September 2025, final judgment 5 December 2025; both sides on appeal, not yet decided. Per the page citations of the judgment in Congressional Research Service LSB11362; tier B.
United States v. Google LLC, No. 1:23-cv-108 (E.D. Va.), liability judgment April 2025; memorandum opinion on remedies 2 September 2026 (tier A, text of the opinion); proposed final judgment jointly submitted 2 October 2026 (tier B); as of the quarterly report Alphabet filed on 29 October 2026 (Form 10-Q, quarter ended 30 September), the final judgment had not been entered, and the company states that it plans to appeal once it is (tier A, company disclosure).
European Commission, Commission finds Apple and Meta in breach of the Digital Markets Act, 2025-04-23 (tier A, Commission press release); Apple has brought an action before the General Court (tier B); Meta Platforms v Commission, T-435/25, brought 2025-07-04 (tier A, EUR-Lex).
European Commission, Commission fines Google €2.95 billion over abusive practices in online advertising technology, IP/25/1992, 2025-09-05 (press release, tier A); Google and Alphabet brought an action before the General Court on 2025-11-20 (tier B).
Interfaces
Main upstream: A Way Out, a Way Forward, E41a, E43; Institution Theory Paper 6, Z34; Theory of the Market Paper 3, SC6; Theory of Disciplines Paper 7, S36 (extension, own undertaking); Platforms and Gig Work, PT2 (the same form) and The Company in the Age of AI, ZN1 (the same form; extension, own undertaking); Political Economy Paper 3, E16, E20 (extension, own undertaking); Paper 4, E25.
Also used: Theory of Government Paper 1, G8; Political Economy Paper 3, E17 (breakup and its cost); Theory of the Market Paper 7 (the functions between a state and multinational companies).
Examples: the three days in which Beacon Assistant went down; the emergency loan to Crossways Pay (pointed to); Beacon's agent interface and Access Code; the allocation of water in a drought year (far future, imagined); the Office of Model Affairs and the Oakgrove Testing Laboratory; machines in the fields and the renewal of the water contract (far future, imagined); Crossways' relocation statement (pointed to). All are constructed examples. Real-world: Epic v. Apple; United States v. Google (search); United States v. Google (advertising technology); the European Commission's three decisions concerning Apple, Meta and Google.
This series: the Prequel (being open to questioning does not cancel it); Paper 3 ("too complex for outsiders to regulate"; review and testing in the name of safety); Paper 5 (interfaces as gates; integration and preparation for separability, LD11); Paper 6 (early checks; had the laboratory been built three years earlier); Paper 7 (all three are needed; letting it be able to fail).
Handed to Paper 3: when the party questioned answers in the name of development or of safety, whether a name buys exemption from questioning.