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.
United Kingdom: Brexit, Cummings and the "Dark Matter" of Non-Voters¶
Chapter: 09 — Cases: Other Countries File: 09_055d · v1 · 17 September 2026 (analysis of the film "Brexit: The Uncivil War" + verification against primary sources) Source: the film by James Graham / Toby Haynes (Channel 4, 7 January 2019; HBO, 19 January 2019); D. Cummings's blog "On the referendum #20, #21" (2016–2017); the letter of UK Statistics Authority chair A. Dilnot to Cummings of 10 May 2016; the Electoral Commission decision of 17 July 2018; the Lord Ashcroft Polls referendum-day survey (12,369 respondents); NatCen / UK in a Changing Europe on non-voters; Clarke–Goodwin–Whiteley (CUP 2017), Evans–Menon (2017), Hobolt (JEPP 2016), Shipman "All Out War" (2016). The text sent to the architect by an external model was checked line by line — discrepancies are in §6. The repository has no transcript of the film; the analysis rests on the verified plot.
How to read this chapter (note of 02.10.2026). The text contains wording that is easy to misread: sums and formulas with coefficients are worked examples: the size of the payment is set as a percentage of the median income approved by referendum; a "reversible" choice means "at the next election": within one election exit is final. The exact-answers sheet 1d and the charter 048m are in force.
1. What the film shows¶
The frame is a fictional 2020 parliamentary inquiry at which Cummings explains that politicians failed to understand how technology rewrote society. Then 2015–2016: Cummings agrees to lead Vote Leave only on condition of full control; refuses to merge with Farage's and Banks's Leave.EU because the data show Farage repels the undecided; hires the Canadian Zack Massingham (AggregateIQ) to build a database of people not on the electoral register but inclined to leave; takes MP Carswell to Jaywick — a destitute settlement in his own constituency he knew nothing about; formulates "Take Back Control" and insists on the word back — "we hate losing things, especially control". Craig Oliver (Remain) works by the classics: television, economic charts, focus groups — and one focus group explodes in a woman's monologue about feeling like "nothing". Gove and Johnson accept the £350 million and Turkey lines despite doubts. After the murder of Jo Cox, Cummings and Oliver drink in a pub: Cummings compares the campaign to a train that can no longer be stopped, Oliver warns of consequences nobody controls. The closing caption: about one billion targeted adverts through AggregateIQ.
Three scenes that matter more to the protocol than the plot:
- Jaywick. A representative of the people did not know how his constituency lived. This is the "elite's perception error" (048e.2) in its British version: not a dictator but an MP of a democratic parliament.
- Oliver's focus group. "I am nothing, I have nothing, I know nothing" is not a position on the EU, it is grievance culture (033c.1). Cummings sold that grievance an address; the protocol offers it a cheque and an exit.
- The pub. A train that cannot be stopped is the catharsis spiral without reversibility (015b): a referendum is a one-way door.
2. What the primary sources confirm¶
| Claim | Status | Primary source |
|---|---|---|
| ~1 billion targeted adverts, mostly Facebook, in the last 10 days | ✅ | Cummings's blog #20; the film's caption |
| ~98 % of the budget went to digital, not traditional media | ✅ | Cummings; Wikipedia "Vote Leave" |
| AIQ adverts viewed 169 million times by a narrow segment of a few million people | ✅ | ICO investigation (2019) |
| Vote Leave exceeded the £7 million limit by ~£500 thousand via BeLeave (£675 thousand to AIQ); fined £61 thousand; referred to the police | ✅ | Electoral Commission, 17.07.2018 |
| AIQ broke privacy laws | ✅ | ICO and the Canadian regulator, November 2019 |
| £350 million/week is the gross figure before the rebate (~£85 million/week); "potentially misleading" | ✅ | Dilnot's letter, 10.05.2016 |
| Turnout 72.2 % against 66.2 % at the 2015 general election | ✅ | Electoral Commission |
| Almost all of the turnout increase came from people with "not very much" or no interest in politics; among 2015 non-voters Leave led by 16 points | ✅ | NatCen (Swales), UK in a Changing Europe |
| Main Leave motive — "decisions about the UK should be taken in the UK" 49 %; immigration 33 % | ✅ | Ashcroft, referendum day |
| Long-term attitudes outweighed campaign events | ✅ | Clarke–Goodwin–Whiteley; Evans–Menon |
| Higher youth turnout would not have overturned the result | ✅ | LSE Brexit blog (2019) |
3. "Dark matter" — Brexit's main empirical gift to the protocol¶
This is the only major referendum in an old democracy where it has been measured what happens when the apathetic mass is successfully raised: turnout +6 points, almost all of the increase from people uninterested in politics, and those people voted for change with a 16-point margin. The protocol's thesis "the passive majority is not noise but energy that any populist can direct at demolition" (§13, 019b, 040d) is here a fact, not a hypothesis.
But the film and the text sent to the architect skip an uncomfortable symmetry. Cummings raised the "dark matter" through a promise of money (the bus: £350 million for the NHS) and a sense of loss (back). The campaign for an AB-EXIT referendum in France (055b.7) forecasts 85–90 % turnout — from the same mass, raised by the same lever: "your lawful income is at stake". The difference is not in the physics of mobilisation — it is identical — but in what happens the next day: Cummings's voter went home without money and without a lever; the protocol's voter receives a payment and chooses again next cycle. The honest formulation: Brexit proved the protocol can gather turnout for its referendum; it also showed how such turnout ends without a thermostat.
4. Orders of magnitude: the bus and the dividend¶
£350 million a week is ~£18 billion a year, ~£270 per resident per year, ~£1,350 per five-year cycle. The protocol's dividend for Britain by the formula D = M × 1.5 × K, at a median annual wage of ~£31 thousand, is about £465 per cycle. Cummings's fake figure was of the same order as a real dividend, and even larger. The conclusion is not "the protocol is more generous" but the reverse: the sum is not the main argument — delivery is. The promised £350 million never arrived in any account; out of that grew the "£200 million hole" and the bus's reputation as a symbol of deceit. The protocol's dividend arrives by a formula any student can recompute (048f.5), and its absence is a bank SMS, not a dispute between experts (048f.3.1). Cummings himself explained: the row about the figure matters more than the figure — if Remain argues with £350 million, everyone remembers £350 million. The protocol inverts the trick: D cannot be argued with, because it arrives.
5. Five correlations with the film — and where they break¶
- Data against persuasion. Oliver persuaded the centre with charts; Cummings persuaded nobody — he found those already inclined and brought them to the urn. The protocol follows the same logic: "stop persuading; take them out of the loop" (033c.6). But Cummings included the apathetic in the decision, the protocol excludes them for money. That is not a shade — it is the opposite sign.
- A separating equilibrium inside the camp. Refusing to merge with Farage was a data decision: a toxic ally costs more than the votes he brings. The analogue of 056e.2 at the level of a campaign, not a voter.
- The word back. Loss aversion as the campaign's engine is exactly what 040b.3 describes as "forgone gain, not confiscation": the protocol takes nothing away so as not to start that engine against itself.
- Lies as fuel. The bus and "76 million Turks" — Cummings exploited the bug "the vote is free, so believing a lie costs nothing" (Caplan, 059d.4). The protocol closes the bug with a price: to vote for Turkey-in-the-EU you must give up D.
- The 2020 scene. Cummings leaves the inquiry in disgust: the system did not change, only the result of one vote did. This is the sent text's thesis — "won the campaign, lost reality" — but it is the film's position, not the historians' (§6).
Where the correlation seemed to break. The first version of this section claimed: 49 % of Leave voted for the principle of sovereignty, that is a high-intensity value, so the Leave core would refuse the dividend and come with x3. Checking against the same Ashcroft poll refuted this — see §8: the reason people name and the stake they perceive are different things, and by stake the Leave camp turned out to be cheap.
6. Verification of the sent text¶
| Thesis of the external model | Verdict |
|---|---|
| Cummings found "3 million who had never voted" | ⚠️ A figure from the film; not found in Cummings's blog. What is confirmed: the target was a few million, and the turnout increase came from the uninterested |
| "Billions of personalised adverts" | ❌ About one billion impressions; AIQ adverts — 169 million views |
| The bus was "a deliberate lie, Cummings knew" | ⚠️ More precisely: a gross figure presented as net; the UKSA — "potentially misleading"; Cummings defended the figure as gross and deliberately provoked the row |
| "Take Back Control — a masterpiece of emptiness" (Shipman) | ✅ in substance; Shipman describes the slogan's power and the word back; the phrasing is the model's |
| "Historians reached a consensus: won the campaign, lost reality" | ❌ The academic consensus is different: long-term factors decided the result (decades of detachment, UKIP, identity); the campaign worked at the 51.9/48.1 margin |
| "The apathetic crowd voted and went back to the sofa" | ⚠️ Turnout at the 2017 general election — 68.8 %, 2019 — 67.3 %: higher than 2015, not lower. "Went back" is not supported |
| "Cohort A had to clean up for four years" | ⚠️ Roughly fair as an assessment, but "cohort A" here is the protocol's term imposed in hindsight |
| Cambridge Analytica / Leave.EU / Banks | The sent text is silent, as is the film; The Conversation and Cadwalladr consider this the film's main omission |
| The comparison "Cummings = an architect who found a zero-day" | Rhetorically apt, factually incomplete: Cummings then spent two years as the prime minister's chief adviser (2019–2020) and left — "could not build a new one" applies to his own attempt as well |
What is worth taking from the sent text into a pitch: the table "mobilisation versus buyout" (§5 of the external text) is correct and useful as a contrast. What is not: "historians agree" and "billions of adverts" — errors an opponent finds in a minute.
7. What this gives the protocol¶
- A negative reference for investors (as in the sent text) — works, with a caveat: the film shows not a "hack of democracy" but its normal mode. Cummings did not break the rules of mobilisation — he executed them better than Oliver. The only violations — the overspend via BeLeave and AIQ's privacy breaches — have nothing to do with the mechanics of mobilisation.
- A test for the protocol's own campaign. The protocol's referendum will pass through the same industry: targeting, slogan, bus. The defence is not "we are better" but a public formula that cannot be turned into £350 million: the five statute parameters (059e.3, iteration 3).
- An argument against "we need special filters" (Milov, 055c.3): Brexit is the case where the liberal establishment, after losing, demanded censorship of targeting. The protocol proposes not filtering adverts but making the vote cost something.
8. A "money" button on the ballot: who would have pressed it¶
The architect's question: if on 23 June 2016 there had been a button "take the dividend and don't vote" next to Leave/Remain — would Brexit have happened? The answer depends on one thing: which camp is cheaper to buy out. On four measurable indicators — Leave.
| Indicator | Leave | Remain | What it means for choosing B |
|---|---|---|---|
| Perceived stake (Ashcroft, referendum day) | 69 % — "might make us a bit better or worse off, but probably isn't much in it" | 77 % — "could have disastrous consequences" | Low stake → cheap vote → B |
| Social grade | C2DE — 64 % Leave | AB — 57 % Remain | D weighs more for the poor → B |
| Interest in politics (NatCen) | Almost all of the turnout increase came from the uninterested; among 2015 non-voters Leave +16 pts | Increase among the interested ≈ 0 | Uninterested → B |
| Expected result | 54 % of Leave expected Remain to win | 70 % of everyone expected Remain | A "just in case" vote → B |
| Timing of decision | 43 % decided more than a year before | 43 % — the same | Does not separate the camps |
A control figure with no dividend at all: the final Ipsos poll with the filter "certain to vote + the result is personally important" gave 52 to 48 for Remain. Among those who cared, Remain already led; Leave won on those for whom "there isn't much in it". Those are exactly whom the protocol buys out.
Orders of magnitude. Suppose 60 % of Leave and 35 % of Remain choose B (the shares illustrate the direction set by the rows of the table): at the urns remain 7.0 million Leave against 10.5 million Remain — 40 to 60. With a milder asymmetry of 50 % against 40 %: 8.7 against 9.7 million — 47 to 53. Leave's margin was 3.8 points; any asymmetry in the direction of the data flips it. With a "money" button, Brexit most likely would not have happened. The stated reason "sovereignty" (49 %) does not contradict this: it is what a person says; "there isn't much in it" is how much he is willing to pay for it. The protocol measures the second (033c.8: revealed preference).
The protocol's caveat: the mechanism is designed for electing managers, not for a one-off constitutional question; a button on the EU referendum is a thought experiment, not a statute norm (see 013b.3: a referendum is a one-shot game, the protocol turns it into a repeated one).
9. Two cycles before 2016: the causes, not the ballot¶
The architect's second question: isn't the point that an incompetent government busy with its ratings created the popular anger, and wouldn't one or two cycles of the protocol before the referendum have removed it. Three decisions from which Brexit grew were taken not by populists but by the mainstream — and all three were ratings decisions, not engineering ones:
- The referendum was called for ratings. Cameron promised it in 2013 to stop the leak of voters to UKIP, and expected to win. This is a pure case of the "fatally incompetent" decision from 059: betting the country on party tactics. Under the protocol there is no UKIP threat — its voter, by the rows of the table in §8, sits on the dividend — and the promise is unnecessary. Brexit was not lost in 2016; it was scheduled in 2013.
- The 2010 cuts. Fetzer (American Economic Review, 2019): districts hit harder by the welfare reform produced a rise in UKIP and in direct support for Leave; by his counterfactual, without that policy Remain could have won the referendum.
- The 2004 migration forecast. A forecast of 5–13 thousand a year when the labour market was opened to the new EU states missed reality by an order of magnitude — an execution error that became Leave's main story.
Catalogue: not "shortcomings" but decisions that would not have passed a board of directors. The architect's correction: the people's anger was created not by "difficulties" but by decisions at the level of logical absurdity, where any schoolchild would have done better. Checking against the documents confirms this — every row below rests on an official report, a department's own statistics or a decision of an international body, not on journalism.
| Year | Decision | What happened | Why this is schoolchild level | What the mass sees (040b.2) |
|---|---|---|---|---|
| 2004 | Open the labour market to eight new EU states on a forecast of 5–13 thousand a year | In three years 430 thousand registrations from Poland alone; net inflow above 50 thousand a year | The forecast itself stated it held only if all EU states opened their borders; three did — the figure was used anyway | Queues at the clinic and the school |
| 2009 | The parliamentary expenses system | A duck house for £1,645, moat cleaning; the Speaker resigned; a collapse of trust in politicians | MPs set the rules themselves and broke them themselves | "They steal in small change while we count pennies" |
| 2011–2014 | Reassessment of disabled people as "fit for work" | By the department's own statistics 2,380 people died shortly after being found fit; the UN in 2016 — "grave or systematic violations", the UK the first country under such an inquiry | Finding fit for work someone who dies within weeks is an error not of policy but of common sense | A disabled neighbour without benefits |
| 2013 | The "bedroom tax": cut housing benefit for a "spare" room | Councils reported there was nowhere to move — no one-bedroom stock; disabled people not exempt; the UN special rapporteur called for repeal | Fining people for not doing what cannot be done | A pensioner pays for the room her son left |
| 2013– | Universal Credit: a five-week wait for the first payment built into the design | In areas of full rollout, food parcels +30 % in a year; advances repaid by deductions from the same benefits; NAO: no value for money and unclear there ever will be; rollout years behind plan | Leaving a person without income for five weeks by design, not by failure | The queue at the food bank |
| 2016 | A referendum called — and planning for a loss forbidden | The Commons Foreign Affairs Committee: "gross negligence"; civil servants: made to work "in a climate where contingency planning was formally proscribed" | Betting the country and forbidding thought about losing | The morning of 24 June with no plan |
This is not "cohort A in a supporting role doing slightly worse". It is the selection architecture (059: the system rewards hypnosis, not intellect; 034b.7: negative selection) in which such decisions pass because they are taken for ratings and the consequences fall on those whose vote is worth nothing. Under the protocol every row hits two filters: a board of pragmatists with x3 does not approve an estimate whose forecast premise is unmet or where there is nowhere to move (048f.5: "14,000 free auditors"); and the mass, unable to design, sees the breakdown in its own wallet and comes back with its vote (040b.2: Check Engine). The anger of 2016 is not a diffuse mood but the sum of these rows.
Integrated forecast: the protocol adopted in 2005, two cycles before the referendum. The architect's correction to the first version: "the angry poor will refuse £465 on principle" underestimates the desire for money under a short planning horizon (040c.1.5). In 2016, £465 was six weeks of unemployment benefit or four weeks of the basic state pension, paid as a lump sum (033c.5); for a person whose horizon is "the end of the month" that is not a symbol but an event. The forecast below accepts that correction and computes by the repository's model (015b.6, 034b.8, 013b.3), with British data as anchors.
| Period | What happened | What the model predicts under the protocol | Anchor |
|---|---|---|---|
| Cycle 1, 2005–2010 (phase 1) | UKIP 2010 — 3.1 %; the 2008 crisis | B 55–65 %: C2DE, the uninterested, the future UKIP electorate go onto the dividend; both parties compete for cohort A (AB/C1 — 57 % Remain); the 2008 crisis happens, it is external | §8: three markers of the Leave camp's cheapness |
| Cycle 2, 2010–2015 (start of phase 2) | The 2010 cuts; real wages −5 % over 2008–2014; UKIP 2015 — 12.6 %, 3.9 million | Fiscal consolidation — yes, but by 055b.8 first at the expense of the apparatus and "loyalty programmes", not benefits: cuts aimed at the poorest districts have no electoral payoff, and the thermostat punishes through the cycle (034b.8). Median stagnation still lowers D → B falls to 45–50 % → the returnees vote with x3 against the incumbents at the 2015 election. The anger goes into a change of government, not into UKIP | Fetzer: dissatisfaction tracks losses from the cuts; remove the losses and the dissatisfaction goes |
| 2013: the referendum promise | Given to stop the leak to UKIP | No leak: UKIP's base is on the dividend; the Conservatives' remaining base is cohort A, leaning Remain; the promise loses votes. Probability of the promise: 10–20 % instead of 100 % | Ashcroft: AB 57 % Remain |
| 2016: if a referendum is called anyway | Leave 51.9 % | With a "money" button — Remain 53–60 (§8); without a button, but after two cycles — Remain ~55–65, because the share of the angry has shrunk, not merely left the room | Ipsos with the importance filter: 52/48 Remain already in 2016 |
Result: the probability of Brexit by 2016 under two cycles of the protocol is of the order of 5 % (0.15 × 0.35) against the accomplished fact. The share of those "angry at the government" shrinks not because they were paid but because the rows of the catalogue above do not pass the board: by Fetzer's anchor, dissatisfaction tracks the losses that precisely these decisions created; by 015b.6, by the middle of phase 2 real incomes are +15–25 % instead of −5 %. The three failures of government: the first (a referendum for ratings) does not happen, because its motive disappears; the second (cuts on the poorest) changes its target, because who decides has changed (033c.6); the third (2004 migration) precedes the protocol and is not undone, but loses its electoral weight — cohort A votes for the labour market, and B is on the dividend. All of this is the model's claims; of what is measured here there are only three things: the Leave camp is cheaper by stake, income and interest; dissatisfaction tracks the cuts; among those who cared, Remain already led in 2016.
Would Brexit have lowered the median — and would a dividend recipient have felt it? The architect's question: had the protocol already been working, the one who took the money would have had a personal financial motive not to vote for a dangerous decision at a referendum. By the data: Brexit lowered the real median, not the nominal one. Median full-time weekly earnings rose from £539 in 2016 to £550 in 2017 (+2.2 %), but adjusted for inflation fell by 0.4 %; by 2019 — £585 and +0.9 % in real terms. The pound's fall after the vote raised consumer prices by 2.9 % and cost the average household about £870 a year (an estimate by the Centre for Economic Performance at the LSE); over a decade real pay is lower by about £470 per worker per year (Resolution Foundation). Now the scale of the signal: with D ≈ £465, a 0.4 % fall in the real median is minus £2 per cycle, a 3 % fall — minus £14. The loss through prices is hundreds of pounds a year, the loss through the dividend is single pounds. That is, the sensor would have moved in the right direction but too weakly for a person to notice; at K = 1 % the dividend creates no tangible personal stake in a referendum's outcome. Hence two conclusions. First: the architect's argument rests not on the median but on another link — after one or two cycles, bright and dangerous steps lose their point for the voter because the anger that fed them disappears (the catalogue above). Second, a technical one: the statute must specify the real median. With a nominal one the dividend would have risen after Brexit together with inflation, and the sensor would have shown the opposite.
The price the experts got right and wrong. The long-term forecasts came true: the OBR — minus 4 % productivity and minus 15 % trade intensity "broadly on track"; the CER — about 5 % of GDP and around £40 billion a year in lost tax. The Treasury's 2016 short-term forecast — a recession and minus 3.6 % of GDP within two years — did not come true, and precisely that error gave Gove his line about having "had enough of experts". The lesson for the protocol: cohort A beats the populists only while it does not itself trade in fear on the short horizon; a public formula protects against £350 million, but not against "minus 3.6 % in two years".
10. Weak point of the case¶
The calculation in §4 is an order of magnitude on an approximate median, not a model; it compares a promise of budget spending with a personal payment — different things for a voter. The B shares in §8 and the probabilities in §9 illustrate a direction, not an estimate; what is measured is only that Leave is lower on stake, income and interest. The weakest link of §9 is the "cycle 2" row: the claim that cohort A cuts the apparatus rather than benefits is taken from the model (055b.8, 034b.8), not from the behaviour of real British governments; had A cut benefits as in 2010, the thermostat would have brought the angry back in 2015 but would not have stopped the fall in their incomes. And the whole section is built on a drama, not a transcript: the characters' lines are Graham's script, not a document. 🟡
Related: 055b (France: referendum turnout) · 055c (Germany: Milov, "special filters") · 018 (three epochs: turnout 72.2 %) · 019b (the poor voter's price) · 040d (Brennan; the poor are not stupid) · 059d.4 (Caplan) · 048e.2 (the elite's perception error) · 048f (double entry) · 033c.1, 033c.6, 033c.8 (grievance culture; "we change who decides"; revealed preference) · 015b.2, 015b.6 (yin-yang; three-phase forecast) · 034b.8 (a father who can be fired) · 013b.3 (one-shot → repeated game) · 040b.3 (forgone gain) · 059 (avatars: decisions for ratings) · 059e.3 (five statute parameters)