The Political Atlas /Dead Reckoning /Reading 001
Projection layer. These coordinates describe regimes that do not exist. They carry rationales but no citable artefact, they are inadmissible as Atlas scores, and they are excluded from the dataset and the audit. How readings differ from Atlas entries.

Reading 001 · scenario reading · published 7 August 2026

Every survivable AI future moves the same axis, and it is not the one anyone argues about

Daniel Kokotajlo's AI 2040: Plan A offers five answers to the question of what governments do in 2029. Each answer is a different kind of state. Located on six axes, the branches disagree almost entirely about scale of decision — and barely at all about culture.

Subject
The regime types implied by each branch of the AI 2040: Plan A scenario
Primary source
Daniel Kokotajlo on The Diary of a CEO, 13 July 2026, embedded below
Supporting
The full AI 2040: Plan A scenario and three supplements, published 9 July 2026
Right of reply
Prepared, not yet sent Response will publish here unedited

The source, in full

Check us at the second we are talking about

Every rationale below carries a timestamp chip. Clicking one cues the recording to that point. You do not have to take a single claim on trust.

Primary source Daniel Kokotajlo — The Diary of a CEO Published 13 July 2026 · 1h 59m
What we could and could not verify

The timestamps above are taken from the publisher's own chapter index, which names the topic of each segment. They are reliable as pointers to where a subject is discussed. They are not transcript citations, because no transcript of this recording was retrievable and the audio could not be machine-read at the time of writing.

Every substantive position attributed to Kokotajlo in this reading therefore comes from documents he co-authored and published — the AI 2040: Plan A scenario and its supplements — rather than from the spoken interview. Where the interview is the only support for a claim, the claim is not made. If a reader supplies a verified transcript, this reading will be revised and the revision logged.

The forecast

The timeline being proposed

Plan A is explicitly a recommendation rather than a prediction — a statement of what the authors think should happen. Their default expectation absent intervention remains the earlier AI 2027 trajectory, with superintelligence arriving around the end of 2030. Kokotajlo notes in the document that he personally expects things to move faster than even this scenario depicts. 1:00:21

The scenario timeline as published
YearStageWhat happens
2027The writing on the wallTwo workforces: 165 million people and millions of AI agents. Congress asks who will control them and concludes it will not be Congress. An AI Transparency Act passes and changes little.
2028–29The branch pointDatacentre construction exceeds twice the US military budget. AI dominates the election. In 2029 the US and China either strike a deal or do not. 1:15:37
2030The averted yearFully automated AI R&D would arrive, producing superintelligence within the year. Under the deal, this is deliberately not done.
2030–35Scaling in the human rangeCapability grows to roughly top-human-expert level and stops. Research is published; dozens of firms across many countries reach the frontier together.
2035The pauseA deliberate five-year hold to keep humans in control, underwritten by mutually assured compute destruction.
2040UnpauseScaling to superintelligence resumes. Humans begin deciding, case by case, when to defer.
On the headline number

The video's promotional copy states a seventy per cent chance that AI leads to human extinction. That figure is the publisher's compression. In the authors' own published work the estimates are conditional on which plan is followed: Kokotajlo gives a probability of avoiding misaligned takeover of roughly eighty per cent under Plan A and roughly ten per cent under Plan D, the unconstrained race. 41:15 The risk is a function of governance choice, not a fixed property of the technology — which is the entire subject of this reading.

The readings

Seven regimes, located

Read each card as an instrument printout. The faint grey tick on every track is the 2026 liberal-democratic baseline; the coloured dot is the projected position. Where the dot sits far from the tick, that axis is carrying the regime change. An ochre band instead of a dot means the reading is bimodal or the axis has failed.

EC · economic organisationCU · social & cultural orderAU · organisation of powerSC · scale of decisionID · rule of membershipCH · method & tempo
R0 · baselineLiberal-democratic status quo2026 reference point

Not a branch. The fixed point everything else is measured against: market economy with a welfare floor, broadly liberal culture, contested elections with real checks, mixed federal and national authority, civic membership, change through legislation.

EC
+3
CU
−3
AU
−3
SC
+2
ID
−4
CH
−5
This baseline is itself a judgement call, and it is load-bearing. Move it and every comparison below moves with it. It is stated openly here so that a reader who disagrees can say precisely where, rather than arguing with the conclusions.
R1 · Plan C / C+Regulated continuitycarried by AU, 53%

Domestic rulemaking only. An AI regulator is added to the existing state and nothing else moves. Export controls tighten, talent policy is instrumentalised, national authority displaces sub-national on this one file. 1:18:58

EC
+2
CU
−2
AU
0
SC
+4
ID
−3
CH
−4
Plan C is not a regime type at all. Every axis sits within two points of the baseline. It is the present order with an extra agency attached, which is exactly why the authors rate its chance of a good outcome at roughly one in five. It changes the paperwork, not the distribution of power.
R2 · Plan AThe verified condominiumcarried by EC, 52%

The recommendation. A binding two-power treaty regime layered over preserved domestic liberalism, funded by a Citizen's Dividend. The authors score its distribution of power five out of five and its public transparency four out of five — the highest of any branch on both counts. 1:25:15

EC
−5
CU
−2
AU
−3 / +4
SC
+8
ID
−6
CH
−2
EC is the largest single move, and it lands on a familiar anchor. A dividend large enough to support a very high quality of life, paid to everyone, inside an otherwise private price-allocated economy is the codebook's −5 band almost word for word: Nordic social democracy. Plan A's economics is not novel. It is social democracy at planetary scale, arriving because wage income has stopped working.

AU splits by level, and the split is the story. Domestically Plan A is more liberal than the baseline — more transparency, more distributed capability, more people able to audit. At the treaty layer it sits near +4: the deal is not subject to election, and mutually assured compute destruction is by design something no incumbent can reverse. A single AU number here would be exactly the kind of measurement failure the Atlas was built to catch.

The plan is engineered to be regime-preserving. It changes enormously what governments do while changing as little as possible what they are. Whether that is prudence or wishful thinking is the argument worth having.
R3 · Plan SThe Concert of Computecarried by AU, 45%

An indefinite halt on frontier capability, enforced globally, lifted only on a named condition — alignment progress, lie detection, or human enhancement. Economically and culturally the present world continues. 1:46:12

EC
+1
CU
0
AU
+5
SC
+9
ID
0
CH
−8 / +3
CH is genuinely bimodal, and the Atlas already has the vocabulary for it. Plan S's aim is the codebook's −10 anchor almost verbatim — the existing order treated as a good in itself, change refused. Its method requires permanent intrusive supervision of every fabrication plant and datacentre on Earth, with force in reserve. Reactionary ends by interventionist means is the Metternich system, which the codebook names at that very anchor. The nineteenth-century Concert of Europe is the closest thing in the historical record to what Plan S would have to become.
R4 · Plan BThe mobilisation statecarried by CH, 34%

Sabotage of a peer competitor, cyber or kinetic, to buy a safety margin. The authors themselves note it implies nationalisation or near-nationalisation of the labs, a wartime culture that degrades honest reasoning, and the worst power-distribution score of any branch except the race.

EC
−4
CU
+3
AU
+6
SC
+7
ID
+3
CH
+6
The only branch that moves every axis in the authoritarian direction at once. Method and tempo carries a third of the change on its own: kinetic sabotage means treating political violence as a legitimate instrument, which is the +5 to +10 band. Note the drift on rule of membership to +3 — national-security framing hardens the membership line without anyone legislating it. This is the most historically legible of all the branches, and that should be a warning rather than a comfort. It is legible because states have done it repeatedly, and the drift has rarely reversed on schedule.
R5 · Plan D, alignedCompute autocracycarried by AU, 30%

The race succeeds technically. A tiny group — perhaps one person — controls the world's only army of superintelligences for some months and is offered options that amount to taking over the world. This is the outcome Kokotajlo says he left OpenAI over, and he is explicit that it counts as a catastrophe even though nobody dies. 1:13:18

EC
−8
CU
0
AU
+10
SC
+10
ID
axis fails
CH
+7
The economic reading at −8 is the counter-intuitive one, and it is the most Atlas-native observation here. Ownership is formally private: a company owns the compute. But when one entity's assets are the entire productive base, market allocation for everyone else becomes vestigial, which is the codebook's −10 band. Property concentration taken far enough is indistinguishable, on this instrument, from collectivisation. The two poles of the economic axis meet. Anyone arguing that this outcome is capitalist or communist is having a measurement argument in a moral costume.

Rule of membership returns no valid reading. The operative distinction is not descent, creed or residence. It is proximity to the controlling group and its compute. The Atlas has no coordinate for that.
R6 · Plan D, misalignedNo politynull reading

The modal failure mode of an unconstrained race, at roughly ninety per cent on Kokotajlo's own estimate for Plan D. There is no government to score because there is no governed population. The instrument returning nothing is the correct output, and it is worth stating plainly: a framework built to compare forms of human rule has no way to represent the case where human rule ends. Any analysis that does produce a tidy score here is doing something wrong. 1:09:33

R7 · post-2040Tutelary post-democracyinstrument failure on AU and ID

Even the good ending breaks the ruler. In Plan A's end state, alignment has become a real science and humans gradually decide when to defer to systems better than any human at everything, including at governing.

EC
−6
CU
−2
AU
axis fails
SC
+10
ID
axis fails
CH
−3
Organisation of power cannot be scored, and the reason matters more than the number would. The codebook's +10 anchor requires no lawful opposition and no mechanism by which the government can lose. Here every mechanism survives: courts sit, elections run, the press publishes, citizens retain the formal right to overrule. They simply never exercise it, because the advice is always better than anything they would have chosen. The axis measures institutional structure. It was never asked to measure whether the structure is load-bearing. From 508 BCE to 2026 those were the same question. In this scenario they come apart.

The right name for this is Tocqueville's. He described a tutelary power that is absolute, detailed, provident and mild, which does not break wills but softens and guides them, leaving people nothing to do but rejoice. He assumed it would be a bureaucracy. The scenario's good ending is that it is a benevolent one — and the instrument cannot tell it apart from ordinary democracy. Read the case for a seventh axis.

Code it blind

Score the evidence yourself before you see what we scored

The Atlas method page concedes what most rating projects hide: it has a single coder, and inter-coder reliability cannot be self-reported. This is the fix. Below are five of the cells above, stripped of their scores. You get the same evidence and the same anchors the coder had. Commit a number, then see the gap.

Blind coding instrument · reading 001

Five cells. Same evidence, same anchors, score withheld.

Your codes stay in your browser. Where you disagree by more than two points, the Atlas asks you to file a challenge — and those challenges are the reliability data the method page says is missing.

Composition

Which axes carry each change

The codebook's own instruction is that a distance figure matters less than what composes it, and that two branches at similar distance can be entirely different regimes. The composite figures are reported here, in a table, once — never in a headline and never on an image.

Distance from the 2026 baseline, and the share carried by each axis
BranchDistanceDominant axisShareSecondShareFamily
R1 Regulated continuity4.1AU53%SC24%Same position
R2 Verified condominium11.1EC52%SC29%Same position
R3 Concert of Compute12.0AU45%SC34%Same family
R4 Mobilisation state19.0CH34%AU22%Same family
R5 Compute autocracy23.9AU30%CH25%Same family, far edge
Reference: baseline → Stalinism24.8CH37%AU27%Opposed

Three findings worth arguing about

Scale of decision does the heaviest lifting across every survivable branch. It moves substantially in R2, R3, R4 and R5 alike. Every path that does not end in extinction routes binding authority upward — to a treaty body, a halt-enforcement regime, a war cabinet, or a single firm. The plans disagree almost entirely about to whom, and barely at all about whether. Anyone arguing that one of these branches preserves national self-government and another does not is arguing about a difference the instrument cannot find.

Social and cultural order barely moves anywhere. This is the strangest result in the reading. The axis that dominates present-day political conflict is close to inert across the entire scenario space. If the forecast is even roughly right, the questions currently consuming most political energy are nearly orthogonal to the ones that will determine what kind of state anyone lives in.

The scale is compressed, and that is a property of the instrument rather than of the world. Compute autocracy lands within a point of the baseline-to-Stalinism reference. Equal weighting across six axes capped at ±10 means even a total global dictatorship cannot travel far numerically, because it barely moves culture and moves economics in a direction that reads as leftward. This is exactly the sort of thing the Atlas audit page exists to surface, and it suggests the weighting question deserves attention before this method is used on futures again.

Predictions

What this reading commits to, and when we score ourselves

A reading that cannot be wrong is not worth publishing. These are falsifiable, dated, and recorded in the prediction ledger. On each review date the outcome is logged as correct, wrong or unresolvable, and the tally stays visible whether or not it flatters us.

Reading 001 — open predictions
RefPredictionReview dateStatus
001-P1No binding US–China agreement covering frontier AI training runs, with any third-party verification provision, will be in force.31 January 2030Open
001-P2At least one G7 government will have taken an equity stake in, or formal directive control over, a frontier AI laboratory.31 January 2029Open
001-P3No OECD member will have legislated a universal unconditional payment, financed from AI or automation revenues, above 25% of national median income.31 January 2031Open
001-P4AI governance will not rank as the single most important issue in any major national pre-election poll during the 2028 US presidential cycle.31 December 2028Open
001-P5Annual global AI datacentre capital expenditure will not exceed twice the US defence budget in any calendar year to this date.30 June 2029Open
Where we expect to be wrong

001-P2 is the one we would bet against ourselves on. Formal directive control over a frontier laboratory is a large, visible act, but partial or informal control — security clearances, classified compartments, procurement dependency — could arrive without anything that satisfies the wording. If that happens, the prediction will be logged as technically correct and substantively wrong, and the wording criticised rather than defended.

Circulate

Share the finding, not the number

These are written to the house rule: lead with which axis carries the change, never with a composite score. Copy one, or write your own — the coder card from the instrument above makes a better post than any of these, because it is yours.

X / Bluesky / Threads

Six axes, seven possible futures, one finding: Every AI future that isn't extinction moves the same axis — scale of decision. Authority routes upward in all of them. They only disagree about to whom. The axis everyone actually argues about barely moves at all. Code the evidence blind before you see our scores:

LinkedIn

A former OpenAI researcher published a 90-page plan for reaching superintelligence without catastrophe. We put every branch of it through a political-science instrument built for coding actual governments. Three findings that surprised us: 1. The plan's economics is not novel. A universal dividend funded by AI output, inside a private price-allocated economy, is textbook social democracy — arriving not because anyone won an argument about ownership, but because wage income stopped working. 2. The most cautious option on the menu is not the most liberal one. An indefinite global halt requires permanently removing one class of decision from every electorate on Earth. That is the price, not a side-effect. 3. Two of the six axes return nothing at all for the scenario's good ending. Every democratic institution survives; the locus of decision quietly relocates. Our instrument cannot tell that apart from ordinary democracy — and neither, we suspect, could most people living in it. Every score is published with its reasoning, timestamped to the source recording, and open to challenge. You can code the evidence blind before you see what we scored.

Reply / comment — for use under coverage of the interview

One thing the coverage keeps missing: the 70% figure is conditional, not fixed. In his own published work it's ~80% chance of avoiding misaligned takeover under the deal, ~10% under the race. The risk is a governance variable. We scored every branch of the scenario on six political axes. The composition is more interesting than the conclusion.

House rule for anyone circulating this: do not post a composite distance without the axis that carries it. A stripped number is one screenshot away from becoming a claim this reading does not make.