The rule, exactly. Before an election each citizen chooses: to vote — or to take a payment and not vote in that election. The payment is a percentage of the median income, set by referendum; there is no fixed sum. One ballot is one vote, with no multiplier in the count; the vote is amplified only in that each ballot's share grows as others step out. The budget pays under law; a candidate never pays. Only a referendum of all citizens — simple majority, no quorum — introduces, changes or repeals the rule.
The protocol has been introduced nowhere and no pilot has been run: shares, turnout and outcome figures in the chapters are estimates, and the protocol promises nobody an election victory. If a chapter says otherwise, Exact Answers and the Charter are correct. For a candidate: ten questions and ten steps. For a citizen, a mayor, a finance officer, a donor, a journalist, a scholar, a lawyer: answers by role. Everything in force in one file: llms-full.txt.
Courts and Local Tuning: What the Protocol Does Not Describe, and Why¶
Chapter: 08 — Implementation: Universal Principles (paired with §48g and §48i) File: 08_048j · v1 · 20 September 2026 Source: a comparison of the repository with classical designs for a state (The Federalist, Harrington, Hayek and others). The assistant called it a gap that the repository has no courts, separation of powers, federalism or army. The architect replied with two questions: "courts are in the repository — check; if not, then I agree about courts. As for everything else — think: is that not a consequence of the new government, all of it to be tuned locally by themselves?"
How to read this chapter (note of 02.10.2026). The text contains wording that is easy to misread: "trust" and "escrow" are the earlier wording: the budget pays under law; "72 hours" is a hypothesis about the first days, not a rule; a blockchain is one implementation option: the charter requires only open code. The exact-answers sheet 1d and the charter 048m are in force.
1. The check: what the repository has on courts¶
There is no separate analysis of courts in the repository. A court appears in four roles, all of them about other people's courts or a particular case: Germany's Constitutional Court and its requirement of verifiability without special knowledge (048h.1); a court defending the "sanctity of the vote", with three answers to it (021 §63.3–4); the legal base in the American chapter (045); and one line in the question base — the protocol needs "a court ready to defend the text of the referendum" (qa/geopolitics). Nowhere is it written what kind of court the protocol itself needs and what happens if there is none. The architect agreed to this in advance; the gap is closed below.
2. "Everything else": the architect is right¶
Separation of powers, federalism, the army, the structure of ministries take up half the text in the classics because they were designing a state from scratch. The protocol designs something else — the way managers are selected. It is a charter for electing a board of directors, not the works regulations of a factory: how to arrange the factory is for the board to decide, differently in each country. Demanding that the protocol say how many chambers a parliament should have is like demanding that hiring rules say which machine tools to buy.
This is the same boundary the architect drew for the length of the declaration window: "fine-tuning this delicate is a matter for political scientists, not architects" (004.4.1). It is also what makes the protocol portable: it sits on top of a presidential or a parliamentary republic, a unitary state or a federation, because it changes nothing in them except the composition of those who vote. An attempt to describe the "right" arrangement of everything else would turn the protocol into a project for one country. So the answer to the critic is short: this is not a gap but a design decision; institutions are tuned locally by the government that the new composition of voters selects — and that is its job, not the protocol's.
3. How courts differ from everything else¶
With other institutions the protocol coexists; on courts it depends. The repository has accumulated six places where the last word belongs to whoever settles a dispute:
| Where | Which dispute |
|---|---|
| 048i.5b | "I did not vote, yet there is a signature on the roll" — comparing the signature at a late payment |
| 048i.5b | whether to order a repeat at a station where secret votes diverged from open ones |
| 048i.5c | forced exit: the sum was taken in someone else's name — a provisional ballot and verification |
| 019d.4b | whether the share of refusals at a station counts as anomalous |
| 004.4.7b | the counter stopped being published "for technical reasons" |
| 048g | a law that quietly spoils the text adopted by referendum |
If the court is dependent, each of these rules stays on paper. And a second peculiarity: judges are elected almost nowhere, so the new composition of voters affects courts slowly and indirectly — through those who appoint them, and not sooner than after one or two cycles. The improvement in governance the protocol promises reaches the courts last.
4. How the design reduces this dependence¶
The dependence cannot be removed altogether, but the repository already does three things that narrow it — they deserve to be named as a principle rather than left scattered.
A rule instead of a dispute. Wherever possible a norm is written as a threshold and arithmetic, not as an assessment. A repeat at a station is ordered not because a court found the violation material but because the divergence exceeded a number written down in advance (048i.5b). A court with nothing to interpret finds it hard to err in the desired direction.
Evidence before the claim. Today the loser comes to court with a suspicion; the protocol gives him a figure — a conflict of two records, the number of silent records station by station, a lower bound from tokens (048i.2, 048i.5b). A dependent court easily dismisses a complaint about the "general atmosphere" and with difficulty one where a bank statement stands on one side and a signature on the roll on the other.
Publicity before the ruling. All six disputes arise from published data, and a court's refusal is visible to everyone the same day. This does not replace an independent court, but it changes the price of dependence: a ruling that contradicts arithmetic anyone can recompute costs the judge and the regime more than a ruling in a closed case.
5. What is needed from courts — the minimum¶
The protocol does not need judicial reform as a precondition; it needs three things, and all three go into the same text adopted by referendum. Short deadlines: a dispute over a payment or over a repeat at a station is settled within days, otherwise the ruling arrives after the result has been confirmed. Standing for everyone whose record is affected and for every candidate — with no filter in the form of a prosecutor or an electoral commission. And open data as evidence by default: a bank statement, a register entry and the published counter are accepted by the court without further confirmation. Where even this is absent the protocol works worse, but by the usual yardstick not worse than elections without it: today the loser has neither a court nor a figure; after the protocol he has at least the figure.
5b. The architect's drafts: an internet jury court¶
After the first version of this section the architect showed two drafts of his own: "on the court I prepared everything long ago." The first is the "Internet Court" project, which he had been developing since 2016 and presented to lawyers in 2019: online hearing of small disputes (up to a thousand dollars, debts, online purchases, defamation) by jurors drawn from ordinary verified people, in absentia, with the case materials open to the public, a term of no more than 30 days and a fee commensurate with the sum in dispute; legally an arbitration tribunal, in substance an institution of reputation: "it pays to be honest." The second is a later roadmap for launching such a court in Moldova as private arbitration inside the IT park: standard contracts with an arbitration clause, a ruling within 72 hours, random assignment of 7–11 jurors who do not know one another and see the facts without names, some jurors from abroad — because "in Moldova everyone knows everyone."
The projects were not written for the protocol, yet they close exactly the hole named in §3: judges are not elected, and the new composition of voters reaches the courts last. Jurors drawn by lot do not have this problem at all — nobody appoints them. And the protocol's six disputes are a rare case where a court of ordinary people is no worse than a professional one: they are questions of fact and arithmetic, not of interpreting law. Does the signature match; was the written threshold exceeded; was the counter published yesterday. No lawyer is needed for that — twelve disinterested people and open data are, which is precisely the draft's design. It has a venerable precedent: the Athenian dikasteria decided cases by hundreds of jurors chosen by lot precisely because that many people can be neither bribed nor intimidated.
What the protocol takes from the drafts and what it changes.
| From the drafts | How it fits the protocol's disputes |
|---|---|
| Jurors are ordinary verified people, in absentia, online | the same; identification is the same as in the payment application (048i.5c) |
| Random assignment, jurors unknown to one another, the case without names | the same; plus jurors not from the district the dispute concerns |
| A term of 72 hours to 30 days | days, not weeks: the ruling must arrive before the result is confirmed (§5) |
| Materials are open | always open: this is a dispute over public data, not private life |
| Jurors chosen by the parties | no: by lot only, otherwise parties will bring their own |
| The juror paid from the claimant's fee | the same: the claimant pays the fee in advance; if the violation is proved, the state returns it (see below) |
On the fee. The assistant first wrote that jurors are paid from the election budget and a claim is free for the person: "the one who was robbed should not pay." The architect restored the rule from his draft: "if the theft is proved, the state pays — to avoid vexatious litigation; and it is much faster that way. Whoever believes he is right, and that is the claimant, is ready to pay. Only speed matters here, not small sums." He is right, and the argument is stronger than it looks. A free claim in an electoral dispute is a ready-made weapon: the losing side files ten thousand complaints, the jury court chokes, and the result cannot be confirmed on time. A fee closes this: whoever is sure of being right risks a small sum and gets it back, while whoever files complaints in bundles for the sake of delay pays for each one. It is the same "loser pays" rule, only with the money up front — and the same logic as throughout the protocol: a claim is backed by one's own stake (013e). There are no deposits: in 2026 the architect dropped them in the private court as well — the court holds no party money at all, and enforcement rests on the open record.
The original project's main weakness, pointed out at its very first presentation — that a ruling has no legal force and strikes only at reputation — removes itself here: force is given to the ruling by the text adopted by referendum (004.4.7b). And conversely, the protocol solves the internet court's hardest problem — where jurors come from and why they would bother: the pool already exists, it is all citizens with the application installed, and a small payment for a juror's day comes from the same budget as the sum.
A court for all cases, not only for elections. The architect clarified the frame: "this is a court for all cases" — and showed a recording of his talk to lawyers in April 2019 (Legal Startup Crash Test, "An Internet Jury Court"). The idea there is wider than the protocol's six disputes: to give the ordinary person back access to justice in small cases where a state court costs more than the dispute, lasts for years and delivers rulings "legally precise but unusable in life". He had experience of his own behind him: in 2009, his own arbitration tribunal in Russia to protect a shop against consumer extremism. The lawyers raised three objections then, and it is useful to see what became of them.
| The lawyers' objection, 2019 | The architect's answer then | What the link with the protocol answers |
|---|---|---|
| No enforcement: the shop refuses to pay — and then what? | Enforcement is reputation: the loser gets a record confirmed by a third party, and it costs him more than the sum in dispute | For the protocol's disputes force comes from the referendum text; for the rest — the open record of non-compliance, seen by everyone before the next deal |
| An arbitration agreement with an electronic signature is needed, and almost nobody has one | Legal force is not the goal in itself; the goal is to bring reputation back into life | Every citizen has identification in the payment application — that is the missing signature |
| A fake ruling will ruin a person's reputation; the service will be copied twenty times; ratings will be inflated by bots | People trust not the platform but particular judges; bots are caught by behaviour | Jurors are verified citizens chosen by lot, not accounts; there is one court, by law, not twenty clones |
The only advice from the floor in 2019 was escrow: both sides put money into an account before the hearing, and the loser pays automatically. It was reconsidered in 2026 and not adopted: instead of a deposit there is a fee from the filing party, reimbursed by the loser, and the court itself holds no party money. The assistant notes the reverse side as well: the objection about enforcement was correct, reputation alone is a weak lever, and the protocol does not abolish dependence on a court but narrows it (§3–4). So a general jury court is not part of the protocol but its natural neighbour: the same "reform by subtraction" (019c) applied to justice in small cases, and the same principle — people decide for themselves, and the state stays out of the way where it is merely costly and slow. It can be launched without the protocol, as private arbitration; with the protocol it gains a signature, jurors and a first case worth building it for.
Two things from the drafts the assistant advises against carrying over. Paying a juror "for agreeing with the majority", from the second roadmap: it rewards guessing others' opinion rather than attentiveness, and is unnecessary in disputes with obvious arithmetic. And blockchain as a mandatory foundation: for six disputes an open signed register suffices; technology is a matter of place, not of the protocol (§2). 🟡
5c. Reputation as a lever: stronger than the assistant judged¶
The architect asked directly: "do you think reputation is a value in our world?" The assistant answered "yes, on four conditions": the game is repeated; an identity cannot be changed for free; the record is visible to whoever decides; the record is believed. And he judged the 2019 project strictly: the second and fourth conditions were not met then, and that is exactly where the lawyers struck. The architect objected on both points from his own practice, and both objections are right.
On changing identity: "if a new account appears on the market while there is an old one with reviews, whom will people buy from? I have worked on Avito for three years, and believe me: bad reviews weigh like a hundred good ones, purely psychologically, and I find myself making excuses for them." The assistant confused "getting a new name is easy" with "getting a new name is cheap". Easy — yes; cheap — no: a new account starts from zero, and people buy from whoever has a history. This matches the conclusion of the classic paper on "cheap pseudonyms" (Friedman and Resnick, 2001): where a name is easily changed, newcomers are distrusted, and that very distrust turns accumulated reputation into capital one is loath to lose. The architect's second observation is the asymmetry: one bad review outweighs many good ones. It is known both to psychologists ("bad is stronger than good", Baumeister and co-authors, 2001) and to students of marketplaces: a negative rating lowers sales noticeably more than a positive one raises them. For a court this means that a record of a lost and unexecuted ruling is a stronger lever than the assistant allowed, and that it acts above all on those whose business is already running — that is, on those from whom there is something to take.
On trusting the record: "that is a purely technical task. If I have found something bad or good and I need that information for a deal of ten thousand or of one thousand, I will find a way to check." Right, and this removes a requirement the assistant imposed needlessly. A record need not be believed one hundred per cent — it is enough for it to be checkable: open case materials, named jurors whom one can write to. The diligence of checking grows with the size of the deal: for a trifle nobody will dig, and the mere fact of a record suffices; for a large deal a person will read the whole case. The architect said the same back in 2019 — "one should trust not the court but the person who judged; you can go and ask him."
The architect gave the shortest proof in one sentence: "if reputation did not work, they would not print 'established in such-and-such a year' on every label." The founding year is put there by people who count money, and it is put there because it sells: it is accumulated history that can be neither bought nor drawn in after the fact.
What a court adds to reputation beyond punishment. To the question about defamation — "what if a ruling is made against me that I knew nothing about" — the assistant answered with procedure, and the architect agreed, but showed that a court here is not only a source of risk but above all a protection, and named three things reviews do not have.
The first is the denominator. "One can keep a register of contracts with a record, and it will be visible that out of a thousand contracts only one ended in a lost case." A review shows a numerator without a denominator: one dissatisfied customer looks the same for someone who made ten deals and someone who made ten thousand. A register of deals recorded with a court clause gives the share — the same pair of "numerator and denominator" on which the checking of elections is built in this repository (048i.5b). One loss in a thousand is not a stain but a characteristic, and a good one.
The second is a positive history on request. "One can go to court over completed deals, or have them audited with a positive history, in order to prove it." Someone who has been slandered can today only make excuses; here he can himself bring the disputed deal before jurors and obtain a record in his favour confirmed by a third party. This is the answer to the architect's complaint from practice: "anyone can leave a bad review that does not match reality, and I have no way of proving he is wrong." The court is the first place where a seller can prove it.
The third is how exactly the case was lost. "Even if I lost but admitted I was wrong and compensated everything voluntarily, then yes, I lost — but gracefully and honestly; everyone makes mistakes." A record that contains not only the outcome but the execution tells apart two things that merge in a rating: a mistake and dishonesty. Erred, admitted it, paid on time — that is a better reputation than having no records at all, because it shows how a person behaves when something goes wrong; and that is exactly what a future partner wants to know. The lever thus pushes not towards never losing but towards complying.
The fourth is confirmation as a service. To the assistant's words "the jurors are named, and one can write to them" the architect added a mechanism: "for such a confirmation, as with a notary, the court can charge a fee that goes to the judge, and everyone gains; the fee is symbolic, purely for the trouble — but if it happens often…" The distinction is the same as with a notary: reading the case is free, it is open; what is paid for is a certified confirmation — the one who judged personally tells the person checking that the ruling is his and the case was as recorded. The person checking pays a trifle before a deal worth thousands; the one who judged earns not for a single sitting but for as long as his ruling is of use to someone. This also settles the question left open in 2019 — why people would serve as jurors: the rulings of an attentive judge are requested more often, and he is paid for the usefulness of the record, not for agreeing with the majority (§5b). A symbolic fee also shields a juror from a flood of idle enquiries.
The architect immediately added a clarification so that confirmation would not be taken for the only income: "wait, the judge also gets his lawful pay from the loser; but it can be taken in different proportions from both sides — winners too are ready to pay for a fast and honest court out of money they would never have received without it." So it is recorded: the main pay is the case fee, confirmations are a top-up. The argument about the winner is exact: for him the alternative to a fast court is not a free court but none at all, and a share of the recovered money is a fair price. One condition the assistant regards as mandatory: the judge's pay is fixed before the hearing and paid out of the fee whatever the outcome (§5f), and who finally bears the cost is decided afterwards. Otherwise the judge's income begins to depend on who lost — and one side is usually the more solvent — and a court whose judge has a stake in the outcome loses the one thing it sells.
What remains. Reputation does not act on someone who does not intend to return to this market; against him a private court has nothing but the same open record, and that is its honest price. And the stronger the lever, the costlier the mistake: a record made without the person's knowledge hits as hard as a just one. That is a question not of the value of reputation but of procedure — notification, the right to answer, review — and the lawyers in 2019 put it correctly. 🟡
5d. Open questions: what is closed and what remains¶
Asked by the architect to say what remains unsettled about the court, the assistant listed seven questions: a record in absentia about a person who did not know of the court; the court's own liability for a mistake; what secures enforcement; judges chosen by the parties or by lot; the quality of jurors in complex cases; appeal; legal force. The architect replied with two remarks, and they close most of the list.
The first: "you missed an important point. The court is written into a contract voluntarily, before any problem arises. That is, the parties study the quality of the court from its decided and pending cases — that too is reputation; and the parties may choose not to write the court in and keep the country's ordinary court." The assistant did miss it, and it is a load-bearing detail. Reputation works here on two floors: the parties' — before future partners, the court's own — before future parties. A court that judges badly simply stops being written into contracts and disappears without any supervision; a state court is spared this test because people do not go to it by choice. Four answers follow at once from the voluntary clause. The respondent's deposit is a term of the same clause he agreed to in advance. Whether to choose judges or draw lots — the parties decide when they sign. The quality of jurors in complex cases — if the court cannot cope with them, such contracts will go to an ordinary court, and that is a normal boundary, not a failure. And a record in absentia about a person who "knew nothing" is impossible in this mode: he signed the clause himself.
The third floor is the judge himself. The architect added: "and the main thing — a judge has a reputation too, just as the court as a whole has. One can allow the judge to be chosen. Yes, that may bring manipulation, but it will show within five to seven cases, and it can be appealed." This brings back into the design what the 2019 talk stood on — "one should trust not the court but the person who judged" — and the assistant withdraws his earlier "by lot only" for private disputes. The lot is mandatory where the court is mandatory, that is, in the protocol's six disputes (§3): there the parties chose neither the court nor each other. In the contractual mode choosing the judge is as much the parties' right as choosing the court itself, and it gives what the lot cannot: a name with a history of rulings behind it.
Why manipulation is visible here and not in an ordinary court. A judge who is chosen has every case open, and together they form a pattern: the same parties pick him time after time; his rulings diverge steadily from those of colleagues in similar cases; his rulings are overturned on review more often than others'. One case shows nothing; over five to seven the pattern emerges, and it is seen not by a supervisory body but by everyone who looks at a judge's history before signing a contract. From there both levers already described do the work: the loser appeals, paying the fee, and a judge with such a pattern stops being chosen — and with the choosing he loses his income from fees and confirmations (§5c). A corrupt judge in this scheme sells not a ruling but all his future earnings, and the price of the bribe has to cover that.
The second concerns appeal. In 2019 the architect said one could go to court again as many times as one liked, the claimant paying; the assistant regarded this as unsettled. The answer: "Reputation. If a claimant, out of spite and malice, files ten times, how will possible partners look at him in a future deal?" Right: the register is open in both directions, and a vexatious litigant is as visible in it as a non-payer. There is no need to cap the number of reviews by rule: it is capped by the fee, paid every time, and by the record of who sued and how often. If two panels decided differently, both rulings stay in the register, and the reader sees both.
What remains after this — three things; the architect closed the first by a separate decision, and two stay open.
| Question | Why it is not closed |
|---|---|
| Disputes without a contract: a debt to a friend on his word, defamation online | Closed by the architect's decision: "not considered; though if both sides agree to sign a contract during the dispute, then by general practice." The court works only where both parties have agreed to it in writing — in advance, by a clause in the contract, or in the course of the dispute, by a separate agreement; this is ordinary arbitration practice. No record of any kind appears in the register about a person who signed nothing — including the record "refused the court" that was in the first draft. The assistant considers this right: with that record goes the only mode in which a mistake in absentia was possible, and the question of the court's liability for defamation; the price is that two of the four original example cases stay outside the court until the other side agrees |
| Legal force of private rulings | Depends on arbitration law and on whether identification in the application is recognised as a signature; settled locally (§2) and with a lawyer. In the architect's assessment — secondary: the main product is not the ruling but reputation before the deal (§5e) |
| The passage from private arbitration to the protocol's disputes | A private court is chosen voluntarily; the court for the protocol's six disputes (§3) is mandatory under the referendum text. It is one design in two modes, and the rules of passage — who keeps the juror pool, who pays, who is liable — are not written. The architect regards this as a matter of budget, not of design (§5e) |
5e. What the main product is here: reputation before the deal¶
The architect shifted the emphasis of the whole analysis: "for business and for people with money courts are dust and trifles; reputation before the deal is a hundred times more important." The assistant takes this as a correction to his own weights. He had been analysing the design as a court with a register attached, and so placed the legal force of rulings among the main open questions. In fact the ratio is the reverse: it is a register with a court attached. A business person wants not to win a dispute but to avoid getting into one, and pays for what helps choose a partner: how many deals that partner has, how many reached a dispute, how he behaved in it (§5c). The court is needed in this scheme so that entries in the register are not an aggrieved party's opinion but a third party's ruling. Hence the place of the question of legal force: it is useful and worth pursuing locally, but the design rests not on it — it rests on the record being read before signing.
How to make the register unforgeable. The architect's proposal: "one can require every contract to be verified electronically and anonymously at the moment of signing, and then forging it afterwards becomes harder; and ordinary signed copies are submitted to the court as in any other." Technically this is a fingerprint of the document with a timestamp: what enters the register is not the text of the contract but its checksum, the date and confirmation that both parties signed exactly this. The content stays with the parties and is disclosed only to jurors and only in a dispute. Such an entry gives three things at once. The contract cannot be rewritten after the fact, nor another version produced — the fingerprint will not match. The court clause is proved by the very fact of registration, and the dispute "I never signed that" disappears. And the register acquires an honest denominator (§5c): the number of registered deals of each participant is known although none of them is disclosed. It is the same double entry on which the checking of elections rests (048f): an event is fixed at the moment when the parties have as yet no reason to distort it.
What a judge gets besides money. The architect pointed to an omission: "you also missed this: a judge has not only pay — his popularity grows, a biography is being written, and a line in any document saying he was a judge and held 150 sittings, with the whole history — that is a very great deal in the age of AI, which will check and verify everything fast. This is no longer an unknown subject but a person, and trust in him on any question is already quite different. Reputation, I say again, is the key word of the whole court, not everything else." The assistant had counted a juror's motivation in money — the fee and confirmations (§5c) — and missed the main thing. A hundred and fifty open cases under one's name is a verifiable biography that almost nobody has: not "I am an honest person" but a hundred and fifty instances showing how the person reasoned when something depended on him, each of which can be opened. Nobody could read such a history before — it took too long; now a machine reads and summarises it in a minute, and the value of accumulated history rises sharply as a result. To an employer, a partner, a voter this says more than a diploma. Hence the answer to the 2019 question "who will become a judge": whoever needs a name — a young lawyer, an entrepreneur, someone heading into politics. Pay makes the occupation possible; a biography makes it desirable. It is the same currency as the parties' (§5c) and the court's own (§5d): on all three floors the design pays in one and the same thing — a history that cannot be bought and can be checked.
The architect extended the list at both ends: "and how are young lawyers to get a name, and old ones to get their ovation and start helping people when you can no longer even walk, but the brain and the computer still work?" The assistant had named only those who need a name and missed those who have one and nowhere to apply it. A young lawyer today has nothing to show for years: the firm runs the cases, a partner signs the decisions, and a history of one's own begins late. Here it begins with the first case, and every case is public. An old lawyer, a retired judge, an engineer or a doctor with forty years' experience is today left to be silent: the profession demands presence, and his knowledge cannot be shown to anyone. A court held in absentia removes precisely that demand — the case comes to his home, the ruling is written from a computer — and returns to circulation society's most undervalued resource: the experience of people who have stopped being asked. What they need is not pay but work and recognition, and the design gives both: what he decided is read, cited, asked about. This matches what the protocol does as a whole: it does not create new participants but opens the door to those who were ready to work anyway and were excluded by the way the procedure is built.
And a third group the architect named: "and also judges who were thrown out for honesty, and advocates — for them this is a gold mine of alternative work, lawyers whom the state has stripped of their profession." In countries where the court is dependent these are not isolated cases: a judge who delivered an inconvenient verdict is not reappointed; an advocate who took an inconvenient case is disbarred. A licence can be taken from them; knowledge and a name cannot, and here the name works in reverse: for parties choosing a judge (§5d) the entry "dismissed for an acquittal" is the best recommendation there could be. A parallel court needs no state licence — it needs the parties' trust — and so it receives ready-made those whom the state system itself selected for honesty and itself threw out. A dependent court thus supplies staff to its competitor, and the more zealously it purges its ranks, the stronger the competitor becomes. This is one more position with no good move (001c): to keep honest judges is to put up with inconvenient rulings; to throw them out is to hand them over to a place where their rulings will be open and cited.
A supranational structure. The architect carried the thought through: "since the court is the internet, an authoritarian government cannot close it; it cannot jail the judges, it cannot remove the rulings, while people inside the country can choose this court. Yes, the parties themselves can be punished — but only if it becomes known, and only from the parties themselves. This is a supranational structure." In substance this is right, and it has a working precedent on the top floor of the economy: big business from countries with dependent courts has long written English law and London or Stockholm arbitration into its contracts — precisely because it does not trust its own courts. An internet court makes the same thing available for a deal of a thousand rather than a hundred million. The design assembled above suits this without alteration: what enters the register is a fingerprint of the contract, not its text or the names (§5e), so the very fact of choosing this court is visible only to the parties; the judges and the server are out of reach; a judge's reputation can also accumulate under a persistent pseudonym — the history of rulings is no less checkable for that.
The former practical limit — party money in a bank inside the country as a hostage — is removed by design: the court does not hold it, the fee is paid after the case is accepted, and there is nothing to reach for. Blocking the site is an inconvenience of the same kind as for any other service, and is got round the same way.
The boundary. The architect marked out where the parallel court does not go: "we have covered ninety per cent of all court cases, while grave and especially large ones are not handled by ordinary people, and we should not climb in there until the protocol repairs the state judicial system in parallel with ours." The boundary is drawn correctly and follows from the design itself: the court works by the parties' voluntary consent and pays in reputation, so it suits places where there is a contract and something to lose. A criminal case, a dispute with the state, a case over a sum for which it pays to bribe twelve people or threaten them — all that lies outside it: those need coercion and protection, which it does not have and should not have. The figure "ninety per cent" the assistant checked at the architect's request, and the check corrected both of them.
What the courts show. Small money claims do make up the overwhelming part of civil cases by number: in Russia in 2024 justices of the peace issued 28.4 million court orders — 94.9 % of everything they heard — and 76 % of first-instance civil cases were for sums up to 50 thousand roubles (Judicial Department of the Supreme Court); in England and Wales 1.94 million claims were filed in 2025, 85–90 % of them money claims, 93 % of judgments were by default, and hearings came to under 3 % of claims filed (Ministry of Justice). But these are not disputes over sales. They are a debt-collection conveyor: in Russia 93.6 % of such cases are claims by organisations against citizens, over loans and utility bills; consumer-protection cases number about 255 thousand, under one per cent. There is usually no dispute in them at all: in England the defendant objects in 14 % of cases.
What people show. Real disputes over purchases hardly ever reach a court. In the EU, 24 % of consumers in a year met a problem giving legitimate grounds for complaint; 27 % of them did nothing, and of those who acted 5 % went to court — that is, about 3.7 % of all who had a problem (Consumer Conditions Scoreboard 2025). Reasons for inaction: it takes too long — 57 %, no faith in the result — 51 %, the sum is small — 45 %. In Ukraine, by HiiL's 2016 survey, consumer problems are the most frequent category (42 % of all legal problems), while 4.8 % of problems of all kinds reach a court. The EU-wide online dispute platform brought about 200 cases a year to a hearing across the whole Union and was closed in 2025. Meanwhile marketplaces settle disputes themselves and on another scale: some 60 million disputes a year were cited for eBay, more than two thousand a day for Taobao's public jury; both figures are known from statements, without official reports.
The conclusion is more precise than the original. The parallel court does not take the state court's cases away: there are almost no small disputes over deals there, nobody hears them at all. The market is not ninety per cent of the court caseload but the ninety-odd per cent of problems that never reached a court; and the average price of waiting — in England about 37 weeks to a small-claims hearing — explains why.
The result is a division of labour in which the two parts of the project do not compete but converge from two sides. The parallel court, from below, takes from the state court what it plainly cannot cope with — the mass of small cases — and does so at once, waiting for no reforms. The protocol, from above, changes the composition of those who choose the government and, through it, with a lag of one or two cycles (§3), of those who appoint judges; improvement reaches grave and large cases by that route. The state court loses nothing in the process except its queue: it is relieved of cases it handled badly and left with those for which it exists. The architect drew a conclusion from this: "so the state court will support it, will it? Or at least not get in the way; in essence a ruling of the internet court does not stop it from judging the matter afresh at all." Yes: judges have no reason to fight someone who takes cases they never had, and a ruling without state force challenges nobody's authority. Resistance, if any, will come not from judges but from those who earn from the complexity of procedure; back in 2019 the architect proposed making allies of them, and later specified how: the lawyer does not hand the client over but serves him — "he can draw everything up and get his money for a small job." A claim, a set of evidence, a position — an hour's work that in an ordinary court drowns in months of procedure and is therefore loss-making, while here it is paid at once. The court does not take clients from lawyers; it adds a service that could not be sold before; in the first draft the same role is called an assistant paid by result (§5b).
On the last open question — the passage from private arbitration to the protocol's mandatory disputes: who keeps the juror pool, who pays, who is liable. The architect removed it as a question of design: "with a budget clearly above the costs I see no problem in this." The assistant agrees: the protocol has six kinds of dispute, they arise once per cycle, and the election budget is incomparably larger than the cost of jurors (048i.5e). What remains is not a problem but work — writing it into the charter locally, like the length of the declaration window (004.4.1).
5f. What was added while the court was developed separately¶
Development of the internet court continued apart from the protocol (its own repository; this section is the shared part). Below are the architect's decisions that also change how the protocol's six disputes look (§3).
Justice, not legality. "Legality by country hardly interests us. Laws can be unjust, and this court is for fast justice cheaply." The court does not apply any country's law and does not try to fit into it; a judge needs no legal education, and a ruling is written in human language. Lawyers are needed in the project not to bend the court to the law but to bring procedural knowledge: how to notify, how to hear both sides equally, who proves what, when a judge must recuse himself, what to do about a mistake.
Closer to case law than to a code. The norm comes not from a statute but from what a reasonable person is entitled to expect on this market today: "in ninety-five per cent of flats on the market there is internet, and that is already the norm, not the fact that the law is silent." Rulings are open and accumulate into precedents; a judge may decide otherwise but must explain why. For the protocol's six disputes the same matters: the anomaly threshold and the repeat rule are described in advance and applied uniformly rather than interpreted anew each time.
A fee instead of a deposit. "We drop the deposit; the claimant pays the fee, and then the losing side pays it." The fee is paid after the court and the judge have agreed to hear the case — so a refusal costs the filing party nothing, the court holds no party money and needs no payment licence.
Three outcomes the claimant knows in advance. The claimant loses — the fee is already paid, and that is all. The respondent loses and pays — the claimant gets the award and the fee back. The respondent loses and does not pay — there is nothing to recover, but the claimant has an honest ruling and the respondent an open "not complied" record. The formula worth saying before filing: the court promises an honest ruling, not the money.
Judges are an open marketplace, not a recruited pool. Registration is free, a judge without cases costs the platform nothing, he sets his own price, and his profile is his rulings, his average time, his share overturned on review and his share of cases declined. He takes cases as a taxi driver takes rides: free to decline, with no compulsion — a case nobody takes for twenty is taken for fifty.
Choosing a judge fairly for both sides. When the parties do not agree on a name: the system offers five at random, each party blindly strikes one, lot orders the remaining three, and the case is offered in turn. The mechanism is symmetric, honesty is the dominant strategy, nobody chose the final judge, and the randomness must be verifiable — otherwise the platform decides, not the lot.
The rule of two numbers. Each side names one number, the judge normally picks one of the two, and the fee falls on whoever's number was further from the outcome. This is final-offer arbitration, used for baseball salary disputes since 1974; it is symmetric, self-correcting (overstate and you lose outright) and removes the main distortion: when the difference is split, both sides double their numbers; when one of two must be chosen, both moderate. In practice the numbers converge and most cases end in settlement. For the protocol the general rule follows: a dispute about an amount is settled faster and more fairly when both sides must name their number in advance.
The procedure scales with the stake. Up to about a hundred and fifty euros there is no choosing of a judge at all, up to a thousand it is two clicks, above that the full scheme with a panel. This is the opposite of ordinary courts, where a hundred-dollar dispute takes the same steps as a hundred-thousand one — which is why small cases are not heard there.
The boundary. The court suits places where there is a contract and something to lose; criminal cases, disputes with the state and sums worth bribing jurors over stay with the ordinary court, which the protocol repairs from above (§3).
6. Weak point of the section¶
The section was written without a lawyer and without analysing even one concrete judicial system; the six points of dependence were gathered from the text of the repository, and there may be more. The claim that an arithmetic rule is hard to interpret "in the desired direction" is an argument, not an observation: dependent courts have found ways in simpler cases too. It is to be tested by a pilot in a country where courts are moderately dependent — which is most of them. 🟡
Related: 048g (rollback and the two guards) · 048i (three channels, the repeat rule, the late payment) · 048h (the German court, verifiability) · 019d.4b (the anomaly threshold) · 004.4.7b (the counter in the referendum text) · 021 §63 (the court and the "sanctity of the vote") · 001c (the catalogue of forks) · 013e.5 (rights and support)