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.
59m. Chicago 2027: Wilson, Three Rounds of Grok and the Rule of Three¶
Chapter: 10 — The USA File: 10_059m · v1 · 19 October 2026 Source: the architect, 19 October 2026: a clean start of Grok on the case "Willie Wilson for Mayor 2027 takes AB-EXIT as the core of the campaign," three rounds with corrections from the repository, two moves by the architect ("the field is dense without counting those who do not vote — they are exactly whom the protocol pulls out"; "to compare AB-EXIT with those questions — the people will tear them to shreds"). The first round gave 7–12 % and "no runoff"; the third — 18–25 % and a runoff at 55–75 % with the question on the ballot. The difference was made by one correction: count from those who did not come, not from the hall. A continuation of 040m §5i (Mendoza's campaign) and 059l.
How to read. This is not a forecast for Wilson but a record of how an AI reader from a clean start errs and corrects itself by the repository. The first round is the landing page and habitual images; the third is 1d, llms.txt, 048n, 059l. Between them three corrections, one of which turns the forecast over twofold. The fourth error was made by both — Grok and the assistant — and corrected by the architect: a procedural barrier against a question about money for the family is held by a person by name, and that is a kamikaze.
1. The case¶
Willie Wilson — a businessman and philanthropist, a fourth campaign for mayor of Chicago: 2015 — 10.7 %, 2019 — 10.6 %, 2023 — 9.1 %, never in a runoff. The base — older Black voters of the South and West Sides; self-financed. The 2027 field: Johnson with approval around 22 %, Giannoulias leading the early polls, Mendoza, Quigley and a dozen more. First round 23 February 2027, runoff 6 April.
The arithmetic of 2023: about 564 thousand ballots, Wilson 51.6 thousand, the runoff threshold 122 thousand (Johnson), about a million of ~1.58 million registered did not come. Wilson was 70 thousand short of the runoff.
2. Round one: 7–12 %, no runoff¶
Grok from a clean start gave Wilson his usual corridor of 8–12 % and called the protocol toxic for Chicago: "the candidate proposes to pay people for refusing to vote," "paying the poor not to vote," "hype for a week, then the attack 'bribery, elitism'," "he cannot promise implementation, only a referendum," and invented the Zoomers' reaction from the image "progressive youth." Sixty sources, none from the repository.
Seven misreadings, three of them recorded in 1d as typical before this round: "arbitrage — the candidate buys the apathetic," "voter suppression," "the rule acts already at this election." The main one, the seventh, was not in 1d: Grok counted only the hall. A ceiling of 10 % is 10 % of those who came; the million who did not come did not exist in his model, as with Achen and Bartels (040at §1: the sample does not contain the burned) and with everyone who does not count the giant (040av §2).
3. The architect's move: count from those who did not come¶
"The field is dense without counting those who do not vote. They are exactly whom AB-EXIT will pull out: for the hype — and the Zoomers, to take revenge and break the system of the old men's politics." The correction in one line: 70 thousand short of the runoff is 7 % of the million who did not come, or a quarter of the non-coming aged 18–29 (about 300 thousand). The referendum question is addressed to that million, and nobody addressed them, because calling them is dear and buying them is cheap only for the machine.
Both motives are in the repository. The hype: the topic runs without a machine, one or two touches suffice (059l §2, 040aq §4). The revenge: the grievance "we were thrown out, the old decide everything from above" is the portrait of the burned and the meek (1f §6b), and until now it had no move except non-attendance, which is free for the government. The protocol gives the grievance its first priced move: absence becomes paid by the budget and counted, the remaining vote weighs double. Revenge by arithmetic.
A correction in the word, not in the thought: "break the system of the old" is not written on the banner — it is the one phrase that brings the old back to the booth against the campaign (1g "The protocol is against pensioners"). The pensioner who stays also gets ×2. What breaks is not a generation but a machine that does not need the young, because their turnout cannot be bought. Anger into the ballot, not onto the banner (048n).
4. Rounds two and three: by the repository¶
After being told to read llms.txt and 1d, Grok corrected all seven: the candidate does not pay, a referendum on a budget line and the pledge (059l §7), the addressee's reaction instead of the press's (059l §8: "yes, let them buy, where is the money?"), legality as the first strike (059l §4), the Zoomer from 040at §4 and the "6-7" meme from 052 §90.5, the split inside the generation by the Yale poll of 2026 (18–21 right by 12, 22–29 left by 6), sources by section.
The third round added what Grok had not counted himself: co-opting as the main risk — "yes to the sum, no to the rule," 2–6 weeks to the first "I am for money for the family too" (048n, the four forms of not digging in; 059l §8b); a probability base not from stimulus cheques but from the Medicaid initiatives in red states (040aw §3); the three movements of the hall (040aw §2) with separate responses of the burned and the apathetic.
Grok's table, third round — his estimates, not the repository's.
| Conditions | Response of the burned / the apathetic | Wilson's share | Runoff |
|---|---|---|---|
| no question | — | 8–12 % | 10–20 % |
| promise of a referendum + pledge, no co-opting | 20–35 % / 2–4 % | 13–17 % | 30–45 % |
| promise + pledge, with co-opting | 12–25 % / 1–3 % | 11–15 % | 20–35 % |
| promise without a pledge | 8–15 % / 1–2 % | 9–13 % | 15–25 % |
| question on the ballot + pledge | 40–60 % / 4–8 % | 18–25 % | 55–75 % |
| question on the ballot + pledge + co-opting | 25–40 % / 3–6 % | 15–20 % | 40–55 % |
Grok's estimate of the burned among Chicago's non-voters — 15–25 %, 150–250 thousand; the first external estimate for the rows of 048l §4; the repository gives no such numbers.
5. The rule of three — and the architect's move: barriers are not symmetric¶
Grok found, and the press confirms: in July 2026 the Chicago City Council placed three advisory questions of its own on the ballot of 23 February 2027 — a legal counsel for the council, a monument to music, withholding funds from agencies that do not collect employee debts. Under Illinois law (10 ILCS 5/28) there are no more than three questions per election; a petition — about 8 % of the gubernatorial vote in the city (55–60 thousand signatures) 92 days before, a council ordinance 79 days before. Grok concluded: "in practice the question will not reach the ballot, only a promise remains," and lowered the corridor to 11–17 %. The assistant repeated the error: it called this "the first documented case of narrowing the channel."
The architect's move. "To compare AB-EXIT with those questions — the people will tear them to shreds." The rule of three is a wall for questions nobody wants. Against a question about $3,200 for the family the slot is held by a person by name: somebody on the council has to come out and say "a monument to music matters more than your money." That is a kamikaze — the same inertia by name as in 040aw §3, where not one of ten said "I am against." The council changes its questions by ordinance until early December; under sixty thousand signatures of non-voters and a clip "they preferred a monument to your $3,200" it will change them itself. If it does not — the campaign gets the best clip of the year and exactly what it promised: money against those who decide from above. In 040m §4 the architect recorded this scenario earlier: collect the signatures and honestly keep the promise, and if the signatures are not let through — "a mega-scandal: a hundred thousand citizens' signatures thrown in the bin."
The consequence for the forecast: Grok's error about the slots is the same as the "bribery attack" in round one — predicting the apparatus's behaviour without the addressee's reaction. Count by the rows "question on the ballot + pledge": 18–25 %, a runoff at 55–75 %; with co-opting 15–20 % and 40–55 %. The signatures are not an obstacle but the campaign: every signature is a touch (040aq §4), and the drive runs in the same weeks as the canvassing.
6. Grok's conclusion, verbatim¶
Asked whether Wilson has other levers: "The classic underdog set he has really worked through several times. Among the options known to me it is this one [the protocol] that looks the newest and potentially the strongest lever for a candidate with a 10 % ceiling. The rest he has either already tried or can improve, but the ceiling of those improvements is markedly lower." That is the thesis of 048n, confirmed from outside, by the same reader who in round one gave 7–12 % and "no runoff."
7. What entered the repository from this case¶
- 1d: the row "count from those who did not come, not from the hall" — the eighth typical error of the AI reader.
- 040m §5k: the Grok case, three rounds, four errors, one of them shared with the assistant.
- The assistant's memory: barriers are not symmetric — slots, thresholds and committees against a question about money for the family are held by a person by name; compare the weight of the barrier with the weight of the question.
Weak point¶
All shares and probabilities in the table are Grok's estimates with his own label; the repository neither confirms nor refutes them, there are no pilot data. The legal deadlines and the number of signatures under 10 ILCS 5/28 are taken from Grok's answer and verified only as to the council's three questions (press, July 2026); the exact threshold and dates must be checked with the Chicago Board of Election Commissioners before any step. The Zoomer wave on X and TikTok the architect speaks of was not confirmed by a single search; the split inside the generation is confirmed.
Related: 040m §5i, §5k · 059l · 059g · 1d · 1f §6b · 1g · 040at §1, §4 · 040av §2 · 040aw §2, §3 · 040aq §4 · 048n · 048l §4 · 052 §90.5