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.

48n. If You Are an Underdog: A Candidate's Ten Questions¶

Chapter: 08 — Implementation File: 08_048n · v1 · 1 October 2026 (session 01.10.26) Source: the architect's experiment of 01.10.2026 — with an AI adviser he walked the path of a candidate with ten per cent and asked nine questions in the order in which they come, and then a tenth — a request to re-check the figures (the analysis of the experiment is in 059e §5b). The questions are given in his words; the answers are assembled from the repository's current norms, not from the model's replies.

Why this chapter¶

Objections in the repository are collected from the critic's side. Here they are collected from the side of a person who wants to apply this and has doubts. The answers are short; behind each is a link into the depth. There is no promise of victory here: the protocol gives a chance a standard campaign does not have, and nothing more.

Three things to know before the first question.

The protocol is a rule, not a campaign technique. It is introduced by a referendum, not by a candidate (048m, Articles 3 and 14).
There is no sum — there is a percentage of the median income, approved by referendum. "$780", "£465", "30,000 roubles" in other chapters are worked examples for particular countries, not a norm (048k §3).
"The vote weighs triple" is not a norm but arithmetic. Every ballot counts as one; if half have exited, a remaining vote weighs twice as much, if two-thirds — three times (001b, 033c §9b).

Ten questions¶

1. "I am an underdog: ten per cent at most, clearly losing to the first two, and I badly want to win. How will the protocol help me, and can I win?"¶

The protocol gives you a question none of your rivals has: "the vote is already paid for — to parties, to contractors, before elections; why not to the voter himself?" That question reaches those no programme reaches — the non-voters. And it leaves the favourites without a good move: to support it is to adopt your theme; to object is to defend the right to take the vote for nothing (001c; 056f §13j).

Can you win — honestly: nobody knows. The calculation "66 % against 33 %" in 059e is an estimate made in a role-play, not data. What is firmly known is something else: a standard campaign from ten per cent gives no chance, because the favourite's machine determines who turns up. A referendum on introducing the protocol changes who turns up: those who usually stay away come too, because they are voting on their own money (048g §6b).

2. "Are there other simple ideas that would work as well? And is it realistic to beat the party bosses in two months?"¶

The comparison with ten known ideas is in 07/036: ranked-choice voting, basic income, citizens' assemblies and others. Each either requires the parties' consent or gives the voter nothing personally; the protocol avoids the first and gives the second.

In two months — no. In two months the question can be made known: one blogger's address in April 2026 gathered tens of millions of views within a day and forced the power to answer (056f §13f). But introducing the rule means a referendum, signature collection and the courts, and the timing is set by your country's law. The speed of the question is days; the speed of the law is months and years.

3. "Why has nobody done this before?"¶

Three reasons. For two and a half thousand years the question was "whom to admit to the vote", and nobody asked "how much to pay for a voluntary exit" (059e §5; 006). The vote was never a transaction, it had no second side, and the count belonged to those being elected (048f §4e). And those with the resources for such a campaign have won under the present rules; they have no reason to change them. The closest designs and what each lacked are in 006b.

4. "In my city sixty per cent never vote, and it is a mayoral election. Does this fit me?"¶

It fits: the more non-voters, the more people to whom the payment is addressed. Each level of government pays at its own elections, from its own budget and on its own median (048m, Article 7).

Two caveats. A mayor cannot introduce the protocol by decree: a local referendum or initiative is needed where the law allows one; the legal route is specific to each country (048m, annex). And the sixty per cent are not one group but three states: the apathetic will take the money, the burned will come back to vote, the meek will gain a lawful ground not to take part (011b).

5. "Will activists who like the idea help me, or will people be against?"¶

Both. Those to whom the protocol gives something will help: the burned, to whom it returns the meaning of a vote; the meek, to whom it gives an exit; those who want their vote to weigh more. Against, at the referendum on introduction — by the repository's estimates, from 9 to 16 %: the machines and their core, principled opponents, those who lose status (08/045 §45.6; 001b). For repeal after the first cycle — 7–12 % (048g §4).

Expect an attack not on the content but on you personally: a sum is hard to refute, a person is easy (056f §13g).

6. "And if I have no money — can I raise it? Will money be given for the protocol?"¶

Unknown: there is no precedent. Three things can be said firmly. The payment to voters is made by the jurisdiction's budget under law, not by the candidate; you need money only for the campaign and the lawyers (048m, Article 7 and the annex). The cheapest test is a pilot in a courtyard or village, of the order of 30–35 thousand euros (048l). And the donor's motive is named in 059e §1: to be the first to pay for the campaign for a rule that others will then adopt. A donor pays for signature collection, lawyers and explanation; he never pays the payments to voters.

Raise and spend strictly under campaign-finance law, with a lawyer.

A large donor. The architect, 02.10.2026: "any billionaire would readily finance this; they give money to far less realistic things, and here one can change the world". The sums are indeed within what has been seen: about 60 million dollars for a study of unconditional payments, about 350 million for aid to election offices in 2020, a million a day to voters in 2024 (023f). What matters is where the money goes.

Where	Is it allowed
A pilot: a courtyard or village, both rounds	yes — it is not an election (048l)
The campaign for a referendum, lawyers, calculations	yes, under campaign-finance law
The payments to voters at an election	no: the budget pays, not a private person (048m, Articles 1 and 8)

The third row is not a formality: after private money in the 2020 elections at least twenty-four US states banned private funding of elections altogether. A suggestion of the assistant's, not a decision of the architect's: a donor may give not the payments but a guarantee on the first city's loan (Article 8, paragraph 5) — the city pays from its budget, and the risk of too high a percentage does not fall on the residents. Whether that is permissible is for the country's lawyer.

7. "Is this a trick, a catch — or is it logical and fair?"¶

The test for a trick is simple: what is taken from you? Nothing. The right to vote remains and returns automatically every cycle; the right not to vote remains and is worth something for the first time; the charter imposes no duties (048m, Articles 5 and 16).

The argument from fairness: the vote is already paid for, only not to the voter (029.4b; 056f §13c). And selection is not by purse but by how much the outcome matters to the person: a mother whose school is being closed will forgo the sum; the well-off and indifferent will take it (001b §3; 006). To those for whom the vote is a value in itself it will still look like bribery; the analysis is in 029, 06/021 and 040j.

8. "What will my voters say if I propose this? Might they turn against me?"¶

Some will leave — those for whom the vote is sacred. Those who never came before will come. Speak in one sentence and in ordinary words, without the author's language (10/051, rule 2).

And one rule without exceptions: you promise not money for your victory but a referendum on the rule. A candidate who promises to pay voters if he is elected is bribing them, and the protocol has nothing to do with that.

9. "You said the protocol gives a chance of winning. Read it again — is that true?"¶

True to the extent it is written down. What is written down formally is not "who will win" but "who stays to vote and who takes the payment": the one who values his vote above the payment stays (013f). It is a statement with sketched proofs and named assumptions; the full proofs are for an economist to write, and one of the assumptions says outright that where money is worth different amounts to poor and rich, selection runs partly by income too. Tested on data: nothing so far — there has been no pilot (048l). Everything else is estimates, and at the end of each chapter there is a section on the weak point saying which ones.

A chance, not a guarantee.

10. "Check the logic and the figures again. Perhaps there is an error: not one of my political scientists or consultants gave me such results, and here there is some kind of magic"¶

The suspicion is correct. To this question the AI adviser replied with a calculation containing a gross error — and itself gave an assurance that there was none.

The error. The model introduced a rule the protocol does not have: "whoever refuses the payment receives a vote multiplier of 1.2", and applied it only to the underdog's voters, counting the mayor's votes at one. The result was "22,800 against 22,000". In fact every ballot counts as one (048m, Article 5). "Weight" is a share: if half have exited, each remaining vote matters twice as much, but equally for all who remain, the mayor's voters included — they did not take the money either. By the model's own headcount the result is 19,000 against 22,000, a defeat. And "eight thousand non-voters took the money and left" changes nothing: they were not voting anyway.

How to count correctly. A city: 100,000 voters; the mayor 25,000, the underdog 10,000, 60,000 do not vote. After the protocol is adopted the outcome is set by two numbers, both unknown in advance:

a — what share of the mayor's votes are not convinced but brought in: the dependent, the bussed, those voting for a trifle. They take the payment and leave (015c §7b; 011b).
r — how many non-voters come back and vote for the underdog. These are the burned: they return for the vote and do not take the payment. The architect holds that everyone underestimates this number: the burned have been humiliated for years and will not miss the first real chance (011b §11b.2c — "the compressed spring" and three cases in which it was released).

The underdog wins if 10,000 + r > 25,000 × (1 − a).

Share of the mayor's vote that is brought in	The mayor keeps	The underdog must bring back
20 %	20,000	more than 10,000 — one in six of the non-voters
40 %	15,000	more than 5,000 — one in twelve
60 %	10,000	it is enough to keep his own

For simplicity the calculation leaves out the other candidates (the remaining 5,000 votes) and the fact that some of the underdog's own voters will take the payment too.

What follows. There is no magic: everything turns on the share brought in. That number can be estimated before any campaign, and the one who knows it best is precisely the political consultant — this is the question to take to him. And a second caveat the model missed: this arithmetic belongs to elections held after the protocol is adopted. At the coming election nobody is paid anything, and all the candidate has from the protocol is the question (answer 1).

The consultants did not give "such results" because there are no such results.

Magic above arbitrage¶

The AI adviser said twice: "this is not magic, this is arbitrage". The architect, 02.10.2026: "this is magic above arbitrage; it is pure mathematics, not deception".

Arbitrage is a middleman's profit on the difference between two prices. Here there is no middleman: nobody buys some and sells to others, and nobody earns on a difference. What looks like magic is something else: both sides get what they wanted, and nothing is taken from anyone. The one who does not need the vote gets the payment; the one who needs it gets a larger share in the decision. That happens not in trading within the rules but when the rule itself is changed.

Why it is not deception. Deception needs a hidden part, and here there is none: everything the protocol rests on can be recalculated by anyone.

The share of one vote equals one divided by the number who voted.
The total of payments equals the number who took one multiplied by the size of the payment, and it is visible in the treasury.
The number of ballots at a precinct cannot exceed the roll less those who took the payment.
The one who stays to vote is the one who values his vote above the payment.

The limit. The mathematics here is the arithmetic of the rule. The outcome of an election does not follow from it: who wins depends on people, and all the numbers on that in the repository remain estimates (answers 1 and 9).

A candidate's ten steps: what actually follows from the protocol¶

The architect asked the AI adviser one more question: "give me an exact answer on how I can win with this protocol, and ten steps of what I need to do, and what the result will be". The model replied with a plan "from ten per cent to victory in eight weeks", with a budget and percentages. This question will always be answered, so here is the answer that follows from the protocol itself.

The exact answer. One cannot win "by means of the protocol" at the coming election: it is not in force there. Two things can be done. The first is to campaign with a question the rivals do not have (answer 1). The second is to obtain a referendum on the rule, after which the next election is held under it. The first takes weeks, the second months and years.

A lawyer. Find out what route to a referendum exists in your jurisdiction and how long it takes: a local initiative, signature collection, a council decision to call a referendum (048m, annex).
Two numbers for your city. How many people do not vote, and what share of the incumbent's vote is brought in rather than convinced. Whether you have a chance at all depends on this (answer 10).
The text. Take the charter (048m) and, with an economist, fill in the square brackets: which median, what percentage is desirable to 40–50 % of residents, what it costs the budget of your level.
One sentence. You promise not money but a referendum: "I will obtain a vote on paying for your vote to you, not to the parties". Addition of 05.10.2026 after the Grok test (040m §5i). Lawfulness: advocating a referendum or a ballot measure is the core of political speech; vote-buying statutes (in Illinois 10 ILCS 5/29-1, federally 18 U.S.C. § 597) prohibit paying or promising money for a vote, not promising to work for a rule. The architect's formula: "I support and will promote a referendum by every lawful means". The trap in the US is not legal but political — the ad "wants to pay people not to vote" will be made whatever the wording — and the answer to it is political: nobody is stopped from voting; today the one who does not vote sits at home unpaid and uncounted, and under the rule he will be counted and paid; your turnout is paid for now to the mobilisation industry — billions to consultants (008) — and I propose to ask you whether it is you who should be paid. The American frame of the question is not "the parties are paid for the vote" (in the US parties are not funded from the budget) but "your turnout is paid for to consultants — why not to you". The Chicago frame, after the Gemini test (040m §5j) and the architect's correction: in a city where votes have been paid for for a hundred years — with turkeys, with jobs at City Hall, with "walking-around money", and now with cash and petrol from the candidate Wilson — the protocol does not introduce payment for the vote but takes it away from the candidate: makes it lawful, open, equal for everyone and from the budget. The architect: "it all goes on already, without anyone's consent — bribery and populism; the protocol is precisely what should de facto eliminate it and make it lawful". The answer to the ad "wants to pay": "in Chicago they pay already — not you, but their own; I propose that they pay by law to everyone or to no one".
A pilot. A courtyard or a village, a real question and private money — the only place where private money is appropriate, because it is not an election (048l). Correction of 05.10.2026 after a clean Grok test (Chicago, 040m §5i): the pilot is run by an institute or a homeowners' association, not by the candidate's campaign and not on campaign money. A pilot arranged by a campaign during the campaign reads in the US as payment for not taking part, whatever is on the courtyard's ballot; the campaign may point to a pilot, but not run it.
Signatures, if the law requires them. Collecting signatures is your field campaign.
A debate on the percentage. Open it yourself: an argument over how much to pay advertises the rule better than any advertising (056f §13k).
Attacks. Answer with the text of the rule, not with a story about yourself. You will be attacked personally: a sum is harder to refute (056f §13g).
Your own election is held under the old rules: observers and a parallel count, as in any campaign.
The referendum, and after it elections under the rule, with the counter and the reconciliation of two records.

The result. At the coming election — a question the city knows, and an outcome nobody knows. After the referendum — elections from which the votes brought in by the incumbent leave. There are no support figures or probabilities of victory here, because nobody has them.

What in the AI adviser's plan must not be done.

What it proposed	Why not
Put half a million or a million in an account as "a first contribution to the payment fund if the candidate wins"	this is bribery of voters; the protocol expressly prohibits it (048m, Article 1)
Message voters "your vote will weigh triple"	at an election where the rule has not been adopted nobody has exited and nobody's share has changed: it is a lie
Write "a mayor's decree" introducing it	the rule is introduced only by referendum (Article 14)
The formula "median × 1.5 %" and "an independent trust"	early drafts: the size is a percentage approved by referendum, the budget pays
"A 25–40 % chance", "35–50 % support", "plus 2–5 % after every attack"	the numbers were invented by the model
"A factory of compromising material"	has nothing to do with the protocol

Addition of 05.10.2026 after the second conversation with Gemini (040m §5j). Of the fourteen turns of the "campaign analyst" three things and one boundary are fit for a campaign. First, the image: not a renunciation of the auditor but her continuation — "I control the schemes and return the money to the people", Robin Hood with a ledger; "where is the money" is closed not by theory but by Chicago practice — a TIF surplus, no-bid contracts, a hiring freeze. Second, the length: the rule in two sentences, the sums left to the clever. Third, the answer to "the radicals remain": we do not need the convinced activist with a stake; we need the one who is asked without wishing it, and the burned who will return. The boundary is one word: return, not hand out. The source of the money is the audit: what is taken back from no-bid contracts is returned to those it was taken from — that is the image, and it is right; the way it is paid is the rule, equal for all — that is the charter. The Wilson ad strikes at the word "hand-out", and it must not be in the campaign. And any percentage of gain — the architect's "+20–30 %", the model's "22–25 %" — remains an invention until the measurement below.

The "signatures" step, added 05.10.2026 (040m §5j). The promise "I will seek a referendum" is kept by collecting signatures, and its keeping is visible. Chicago, by the Board of Election Commissioners' 2026 guidelines: a question by petition is advisory only; signatures — 8 % of the votes cast for governor in the city in 2022 (of the order of fifty to sixty thousand 🟡, the exact base is given by the Board); filed 92 days before the election; and the "rule of three" — no more than three citywide questions at one election, the first three initiated are certified. The Council has used this rule to crowd out other people's questions since 2012, and on 22 July 2026 it already took all three places on the February 2027 ballot with questions of its own. So the scenario is known in advance: collect, file, receive a refusal citing the rule of three — and that refusal is addressed to the signers, not to the candidate; for the next election, file first, before the Council (for the lawyer: how the "first three" are counted). None of these figures promises the candidate votes; it promises only that the promise can be kept and shown.

The lawyer ally, added 05.10.2026 (040m §5j). The architect: on the wave of hype a constitutional scholar will be found who says "women and Black people did not vote once either — and I will prove this is admissible"; he will come by himself, in his own interest, because a scandal of this size makes him famous, and that is a natural continuation of the protocol, not an accident. So it is: a campaign everyone is sinking has as allies those who profit from being against everyone, and the lawyer is the first of them. What matters is only the argument he comes out with. The losing argument is the one the model suggested — "monetising the right, freedom of contract"; that is the opponents' label spoken in one's own words, and he loses the first broadcast. The winning argument is his own analogy carried through: the history of the franchise is the history of ever more people coming to count; today two-thirds of the city do not count at all — they sit at home unpaid; the rule makes their choice visible, counted and paid, and takes nothing from anyone. Legally three things go with it, which he must have in hand: an advisory question by petition is governed by no vote-buying statute; advocacy of a ballot measure is political speech; the budget pays a person for not turning out, not anyone for a vote (045, 048m Art. 1 part 4, step 4 above). That his interest coincides with the campaign's is not a weakness but the reason he will come.

Wanting the payment and voting for the candidate are different numbers¶

The objection came from a clean chat with another AI (Grok, 02.10.2026; analysis in 040m §5c). The candidate asked: "I propose a payment to all residents, and this is my main idea; how many will support me?" The answer: wanting to take the money and voting for you are different numbers, and the repository does not count the second. This is correct, and it should be known before a campaign begins.

What was cited against. Four cases in which live money lost at the ballot: Switzerland in 2016 — 23.1 % for an unconditional income; Oregon in 2024, Measure 118, a payment to everyone from a new tax on companies — 22.5 % for; Hamburg in October 2025, a basic-income experiment — 37.3 % for; Andrew Yang with a thousand dollars a month dropped out of the primaries.

How these cases differ from the protocol.

	The four cases	The protocol
What is offered	a large regular payment to everyone	a small payment once per cycle to one who does not vote in that election
Who pays	a new tax or the whole economy; people voted against the price	the budget, in the amount of a percentage the voters themselves approve
What the one who does not take it gets	nothing, only the bill	a larger share for his vote in the decision
To whom it is addressed	in Hamburg — to two thousand participants in an experiment	to every voter

Two arguments from the architect, 02.10.2026. "There are no people there who are for the vote, and they come on top as a percentage. And that is income for everyone, every year, and large as a percentage of income, not once in four years."

The second group in favour. For an unconditional income only the one who wants money votes yes; the one who pays is given nothing but the bill, and votes no. Under the protocol the one who does not take the payment gets a larger share for his vote: "yes" has two motives and two groups, and the second is the very people who in Switzerland and Oregon were against. None of the four cases had this.

The size. Brought to one four-year cycle per person:

	Over four years
Switzerland, 2,500 francs a month	about 120,000 francs
Yang, 1,000 dollars a month	48,000 dollars
Oregon, about 1,600 dollars a year	about 6,400 dollars
The protocol, the repository's worked examples	from 700–800 dollars to a month's earnings

The difference runs from eight times to over a hundred.

The assistant's caveat. It does not follow from the difference in size that opponents are fewer by the same factor: nobody knows how the share "against" depends on the sum. And the protocol has an opponent the unconditional income did not have — the one for whom the vote is sacred and payment for not taking part looks like bribery; the repository estimates him at 9–16 %, and that too is an estimate (08/045 §45.6).

Another class of analogues: measures with a personal benefit pass. The architect, 02.10.2026: "comparable votes with serious arguments against still passed, and not with zero turnout". Checked.

Vote	Result	Who was against
Switzerland, March 2024: a thirteenth pension payment	58.2 % for on a turnout of 58.3 %	the government, parliament, business
USA, minimum-wage increases at the ballot, 1996–2025	28 of 32 passed	employers
Florida, 2020: a 15-dollar minimum wage	60.8 % for	business
USA, expansion of health coverage: Idaho 2018, Missouri 2020, Oklahoma 2020, South Dakota 2022	61 %, 53.3 %, 50.5 %, 56.2 %	those states' legislatures

The main point here is Switzerland: one country and one electorate gave 23 % to an unconditional income in 2016 and 58 % to a thirteenth pension in 2024. So what fails is not "voting for money" but a particular design: a huge regular payment to everyone. A moderate, clear addition passes against the government's will.

The limit of the comparison. None of these measures is a payment for forgoing a vote; they are additions for a particular group, and most were held together with general elections. They show that measures with a personal monetary benefit pass and draw turnout, but not how much the protocol would get.

Hungary: money moves votes in both directions. The architect, 02.10.2026: "the argument that voters need money is confirmed again by fact, in Hungary". One country gave both experiments within four years. In 2022 about 6 % of GDP was handed out before the election — and Fidesz took 54 %. By 2026 inflation had eaten what was handed out, real incomes had fallen, the frozen EU billions had not arrived — and Fidesz took 38.6 %, while the winner was the one who promised to bring the money back (039). The same people and the same ruler; what changed was what happened to their money. Here wanting the money became a vote for a candidate both times — and both times the chain was short: in 2022 the money was already in the account, in 2026 the promise was one step away. Hungary does not show whether people will take a payment for not voting; it shows that money moves votes.

What points the other way. In Alaska in 1999, in an advisory vote, 83 % refused to hand the earnings of the fund that pays the dividend over to the budget. Money that already arrives people defend; a promise of money three steps away they do not believe. That is the real content of the objection: the weak point is not the payment but the length of the chain "elect me — then a referendum — then the next election". Correction of 02.10.2026 after a second check (040m §5c): the first link — "elect me" — was added by Grok; it is not in the repository. Where the law gives citizens the initiative, the chain is one link shorter: signatures — a referendum — the next election, and electing the candidate is not part of it (steps 1 and 6 below). Election becomes a link only where the council calls the referendum. That does not make the chain short: the money still arrives a cycle later.

What follows for a candidate. Support for the rule at a referendum and votes for the one who champions it do not coincide: initiatives often pass where the candidates who back them lose. So the ten steps above lead to a referendum, not to the office, and the repository contains not a single number on support for a candidate. Numbers such as "5–15 %" from one AI and "25–40 %" from another are equally unsupported.

What nobody knows. How many people will refuse the payment for the sake of the vote. That is the one thing the pilot measures (048l), and until it is run the dispute over "whether people want money or greater weight" is not settled by argument.

And why it does not matter to the protocol. The architect, 02.10.2026: "it does not even matter to us". The exit share is not a premise the protocol rests on but a reading it produces. If many take it, those indifferent to the outcome leave and the rest decide. If few take it, people value the vote above the payment: the rule has cost little, and something nobody knew about the value of a vote has become known. Both outcomes are a result. The pilot records two limits: if fewer than a quarter take it, the payment is too small; if more than three-quarters, nobody is left to decide (048l §5). But these are limits of the sum, not of the protocol. The architect: "that is not a question for the protocol but a question for the sum; people will simply want to raise it, not to remove the protocol". A wrongly chosen percentage is corrected by referendum once per cycle (048m, Article 3); the rule itself stays. From below this works simply: few take it — it gets raised. From above it is harder, because the recipients must vote for the reduction. The answer is in 018 §34: the bill is printed on the ballot, the payment falls with the median, and the first city that votes for too much becomes a lesson for the rest, like a corporation that paid out too much in dividends. To a candidate the number still matters: his arithmetic depends on it (question 10).

The free rider at the referendum: who will come to vote for the rule¶

A second objection from the same clean chat (Grok, 02.10.2026). The one who wants the payment will not go to the referendum on introducing the rule: the payment will be his whether he went or not, and his one ballot decides nothing. Work is arranged differently — miss the shift and there is no money; here staying away does not burn the cheque.

What is correct in the objection. The difference from work is real. And the claim "almost everyone will be in favour" (048g §6b) does not mean "almost everyone will come".

Why it is weaker than it sounds. This argument is the paradox of voting, and it proves too much: by it nobody should vote today either, because one vote never decides anything. Yet people vote. The architect: "they go and vote today for the promises of people unknown to them; that is how politics works; and here there is a simple formula and real money". A model that does not explain present turnout does not predict turnout at the referendum either; numbers built on it are unsupported.

What the protocol itself answers.

There is no quorum (048m, Article 14). One who stays home does not vote "against"; those who come decide. The free rider does not sink the referendum.
A fork for the machine. To defeat the rule the incumbent must bring its people to the polling station. But the ballot is secret, and the question is about the money of the very person brought in. If it brings them, it gets votes "for": the machine sees that it brought the person in but does not see his mark. This is an inference from the design, not an observation; a close case is the public-sector employee who clicked the wrong thing under his boss's eye, but that was an election, not a referendum on his own money (056f §13d). If it does not bring them, the burned (011b §11b.2c) and those who want their vote to count for more decide.

The fear of being left without the money brings people out — if it is said the right way. The architect, 02.10.2026: "after a little explanation the common fear — I will not come, and the others like me will not come — will be great among the apathetic, and many will come". The strongest result in field experiments on turnout is about exactly this: in Michigan in 2006 voters were sent a postcard with their own voting record and their neighbours', with a warning that an updated one would follow after the election; turnout rose by 8.1 points (Gerber, Green, Larimer, 2008). But the same researchers showed the reverse as well: the message "few will come" lowers turnout, while "many will come" raises it (Gerber, Rogers, 2009). A person who hears "people like you do not vote" does as everyone does. So the thing to say is not "nobody will come and we shall lose the money" but "everyone is going for their money — do not be the one who slept through it". Publishing turnout by precinct on referendum day does for a city what the postcard did for neighbours.

What nobody knows. How many will come. The estimate of referendum turnout in 08/045 is unsupported, as are the objector's numbers.

Measuring the question: a poll is cheaper than an ad¶

Added 05.10.2026 after a clean Grok test on the Chicago mayoral election (040m §5i). Asked what share of "yes" to expect among non-voters, the model answered "there are no data" — not for Chicago, not for the US, not on the site — named the nearest figures from elsewhere and rejected them as non-transferable (willingness to sell a vote for a candidate is not a budget payment for abstention and not a referendum; the Swiss 23 % for a basic income pays everyone and buys no turnout), and instead of a number gave a method. The method costs one line in a poll the campaign buys anyway, and works in any city:

Sample. 400–600 registered voters, weighted by turnout history: voted at the last election of this level or not.
Two cells. A — the question on the rule verbatim: "before each election one may either vote or take a payment from the city budget and not vote in that election; the right to vote is kept for ever". B — the same question plus one line on the bill: a percentage of median income and "from the city budget, that is, from taxes". The difference between A and B is the price of the question; without cell B it is not a measurement but an advertisement of the payment.
Crossing. A separate question on the intention to vote at the coming election, to separate the "yes" of voters from the "yes" of non-voters.
Channel. Non-voters are hard to reach by phone; text to the voter file is cheaper.
What to expect and what not. Among non-voters "yes" will be higher than among voters — a basis for the sample size, not a result; both shares fall as soon as the bill appears in the question. Until the table exists, any figure in a conversation with the candidate is an invention.

This is the cheap test of what the repository calls "a question the rivals do not have" (question 1): whether it moves votes nobody knows, and finding out costs one line in the tracker.

A forecast without a pilot: what an expert owes before the measurement¶

Added 05.10.2026 after the architect's remark: "there is no pilot — but experts exist precisely to give a forecast, not to say there is no experience; if there were experience I would not be asking". That is right, and "there are no data" answers the question "what is the figure", not a release from forecasting. An expert owes the nearest cases with their numbers, a range, what would move it — and the label "forecast" on all of it. Below is the assistant's forecast for the same case (Chicago, February 2027, a candidate in third place at 11 %); every range 🟡, sources named.

The nearest cases — data.

Chicago itself. At the last two mayoral elections a candidate in single digits in the polls three months out reached the runoff: Lightfoot in 2019 — from single digits to 17.5 % and first place after the Alderman Burke scandal in January; Johnson in 2023 — from fifth in the polls to 21.6 % and second place on the teachers' union's money and ground game. A jump of 12–18 points over a winter is not exotic but this city's own base rate, with no protocol at all; both had an event and a lane of their own.
Money and turnout. Hungary, April 2026: plus nine points of turnout once the voter was shown the bill (039). Alaska 1999: 83 % for keeping a payment already received. Switzerland 2016: 23 % for a basic income with a huge visible bill. Willingness to sell a vote for a candidate — 12 % for $25 and 20 % for $100 (according to Grok, polls of 2016 and 2018, not checked) — is a floor: selling a vote is stigmatised, a lawful budget payment with the right kept is not.

The forecast — the assistant's estimate 🟡.

What	Range	What it rests on
"Yes" to the rule in a poll, cell A (no bill), among 2023 non-voters	40–55 %	above vote-selling (lawful, right kept), below Alaska (not yet received)
The same among voters	30–45 %	the duty norm is stronger among those who turn up
Cell B (with the bill line)	minus 10–15 points on both	Switzerland: a bill in the question halves the "yes"
New voters from the 65 % at home — to turnout	+1–3 points	mobilising non-voters historically yields single points even with money on the ground
Their share — to the candidate	+2–5 points of the result	the question is hers; the rivals do not want it
Switch of the burned and the protest voters who vote anyway	+3–7 points	the "against the machine" lane was Vallas's in 2023; in 2027 it is free
Loss of her present moderates	−2–4 points	the "vote-buying" label hits her own pragmatic voter
Total	14–21 % against 11 %	the runoff threshold is about 20–22 %

What it means. In the median case the question brings the candidate to the runoff threshold, not over it; the top of the range needs an event of the same kind as Lightfoot's and Johnson's — and the media explosion the question itself creates is that event. So the expert's honest formula is "plus five to ten points, with a real base rate in this city" — more than "it moves not a single vote", less than "plus twenty to thirty".

What moves the range. Up: cell B above 35 % among non-voters; a lawyer or an editorial board on her side; the Council's refusal of the signatures before January. Down: the label "Jim Crow in a financial wrapper" takes hold in the Black wards of the South and West sides, where her lane is thin anyway; the rivals do not reply — silence suits the machine better than argument. The measurement in the section above replaces this table with one line in the tracker; until it exists, this is a forecast, and it is signed.

What the protocol does not propose¶

The AI adviser in the architect's experiment added four things of its own. They are not in the repository, and they should not be done.

The model's advice	Why not
Put money in an account and pay it to voters if the candidate wins	this is bribery of voters; under the protocol the budget pays by a rule adopted by referendum, and never the candidate
"Elect me — I will sign a decree"	elected bodies neither introduce nor alter the charter (048m, Article 14)
Route campaign money through a separate structure to get round the limits	a breach of campaign-finance law; the protocol has no need of it
A vote multiplier for those who refuse the payment	there is no such rule: one ballot — one vote (answer 10)

Weak point¶

No candidate has walked this path: the answers to the first, fifth and sixth questions are a forecast, not experience. The questions were asked by one person in one conversation; a candidate in another country will have more, and the first real campaign staff will add to this chapter. 🟡

Related: 059e §5b (analysis of the experiment) · 048m (the charter) · 048k §3 (decisions) · 048l (the pilot) · 001b (essence) · 001c (forks) · 011b (three states) · 013f (the formal statement) · 07/036 (comparison with ten ideas) · 08/045 (campaign economics) · 048g (rollback and who is against) · 056f §13f–13j (the question outside the count)