Brief, 10 September 2026. Built the same day; see below.
Thesis
The five studies so far are historical: a plan, a crisis, a posture, an exercise, a civil-defence estimate, each with a record that can be cited page by page. This one is different in kind and the brief must say so at the top. There is no record. There are two published scenarios, one a book built from forty-seven interviews and the declassified doctrine of launch under attack, the other a film that stops at the moment the book's president decides. Both are fiction. What is not fiction is the machinery they run on, and that machinery is what the study can enact: where the missiles are, how long they fly, what can be seen and by whom, how many interceptors there are and how often they have hit, whose territory an American missile crosses on its way to North Korea, and who lives under the warheads at the end.
The study's claim is therefore narrower than the others'. It does not say what would happen. It says: here are the forces as counted in 2024, here are two published timelines laid on them, and here is what the engine's arithmetic makes of each step, including the steps where the scenarios lean on assumptions the record does not support. The scenario's own clock is the study's clock. Where the scenario asserts a fact about the world, the study checks it against the record and draws the disagreement.
The constraint and the tiers
| Tier | In this study |
|---|---|
| Documented | Bases, silo fields, submarine ports, bomber wings, radar and satellite ground stations, interceptor sites, the Diablo Canyon plant, as the open record places them; warhead and launcher counts as the Nuclear Notebook and FAS estimate them, with the estimate named; the interceptor test record; the 1983 Soviet false alarm as the precedent for the misread |
| Reconstructed | Flight times from the geometry and the minimum-energy arc; the overflight of Russian territory by an American missile bound for North Korea, which is a fact of the great circle; the sequence of detection, conference and decision as the declassified command doctrine describes it |
| Inferred | Every target assignment. Neither the book nor the film names a target list beyond the first strikes; the book gives counts (82 warheads at North Korea, about a thousand at the United States, 975 aim points in Russia) and the study allocates them by a stated rule against the 2025 grid, as the DEFCON 3 study allocated SIOP-4 |
| Modelled | Every consequence: intercept probability, blast, fire, fallout, the union of the three on the population, all with the method named and the assumptions stated |
| Withheld | The real OPLAN 8010 target list and the real decision timeline are classified and are drawn as such; the study never claims to know either |
| Scenario | A sixth marking, used only here: an event that happens because the book or the film says it happens. The North Korean launch, the interceptor misses, the Russian misread, the Russian president's silence. Drawn with the scenario's page or minute mark, never as a fact of the world |
The sixth marking needs a decision before the build. The grammar has five tiers and every other study has held to them. The honest choice is to draw scenario events as inferred with the provenance naming the book and the minute, and to make the omissions panel say, on every act, that the sequence is a published fiction. That keeps the grammar whole and puts the fiction where the reader will look for it. A new tier would be cleaner on the map and worse for the engine's promise that the tiers mean the same thing everywhere.
The two scenarios
Nuclear War: A Scenario (Jacobsen, 2024)
The book runs a clock from a launch at 3:03 pm Eastern to a world ended seventy-two minutes later, in five parts with the minute marks in the chapter heads. The summary below is from the book's own structure and should be checked line by line against the text before the build; the figures in the second column are the book's, not the record's.
| Minute | Event, as the book has it | What the record says |
|---|---|---|
| 0 | North Korea launches a Hwasong-17 from near Pyongyang at Washington; a one-megaton warhead; SBIRS sees the boost within seconds | The Hwasong-17 exists and has flown to 6,000 km apogee on lofted tests; its warhead yield is unknown and North Korea's largest test was about 250 kt; a single-warhead one-megaton fit is the book's assumption |
| 0 to 12 | NORAD assesses, the conference forms, the president is told with minutes to decide; four interceptors from Fort Greely miss | Forty-four interceptors at Fort Greely and Vandenberg; twelve hits in twenty-one tests; the four-on-one salvo is the stated doctrine and its claimed 97 per cent assumes independence the tests do not show |
| 16 to 24 | A second missile, from a submarine 350 miles off California, at the Diablo Canyon plant; the president orders 50 Minuteman III and 8 Trident missiles, 82 warheads, at North Korea | North Korea's one ballistic-missile submarine is a rebuilt Romeo-class boat of the 1950s design; a patrol to the California coast is the scenario's weakest link and the study says so |
| 26 to 33 | Russia's Tundra satellites see the Minuteman launches; the operators, with the system's history of false alarms, read an attack on Russia; the Pentagon is destroyed by the one-megaton burst | The great circle from the Minuteman fields to North Korea crosses the Russian Far East, which the study draws exactly; the 1983 Oko false alarm is the precedent; Tundra's record is thinner than the book implies and the misread is a scenario event |
| 36 to 45 | Trident warheads leave for North Korea; the Russian president, unreached, launches about a thousand warheads at the United States and NATO | The count matches the deployed strategic force within the New START limit; the silence on the hotline is a scenario event |
| 48 to 72 | The acting president answers with the rest of the force at 975 Russian aim points; a North Korean high-altitude burst; the Russian warheads arrive over twenty-two minutes; hundreds of millions dead by the last page | The engine can sum this on the 2025 grid, and will disagree with the book's round numbers; where it disagrees, both figures are shown |
| After | Nuclear winter after Toon and Robock: 150 million tonnes of soot, a fall of tens of degrees, five billion dead of famine | Out of scope for the engine; cited, not modelled |
The book's authority is its interviews: Perry, Panetta, Kehler, Blair, Postol, Garwin, Kristensen, Podvig, Toon. Its plausibility has been questioned on exactly the points the third column marks: the submarine, the misread, the silence, each a choice rather than a consequence.
A House of Dynamite (Bigelow, 2025)
One unattributed missile, seen by the SBX radar in the Pacific, bound for Chicago with about nineteen minutes to impact. Two interceptors from Fort Greely: one fails to leave the ground, the other misses. The chain forms: NORTHCOM, STRATCOM with OPLAN 8010 on the table, the Situation Room, a president who cannot be sure whose missile it is. The film ends before he decides.
The film's argument with the record is a single number: the screen gives the interceptors a 61 per cent chance, the Pentagon answered that the operational system has never failed, and the screenwriter pointed to twelve hits in twenty-one tests. That argument is the study's first act: the same missile, the same two interceptors, and three fates, one at the test record, one at the four-on-one claim, one at the film's coin, each labelled with its source.
What the study enacts
Three acts on one clock, with the 2025 grid under all of them.
- House of Dynamite (
#/study/72-minutes). A single ICBM from the Sea of Japan toward Chicago, nineteen minutes on the clock. The radar horizon and the SBIRS detection marked as they happen; the two interceptors from Fort Greely flying to their intercept point over the Pacific; the decision ladder as a posture ladder of the kind DEFCON 3 uses, one line per minute from the book's and the film's account of the conference. Variants for the interceptor fate: test record, salvo claim, film. If the warhead arrives, the outcome on Chicago at the film's implied yield, which the film does not give and the study must state. - Seventy-two minutes (
#/study/72-minutes/jacobsen). The book's timeline as the act sequence: Hwasong-17 to Washington, the four interceptors, the submarine missile to Diablo Canyon, the 82 warheads to North Korea drawn as arcs that visibly cross Kamchatka and the Sea of Okhotsk, the Russian launch, the American answer, the arrivals over twenty-two minutes. The union outcome summed on the 2025 grid for each side, beside the book's figures: one million in Washington in the first minutes, hundreds of millions by the end. - What the record allows (
#/study/72-minutes/record). The same first missile with the scenario events removed: the interceptor fate at the test record, no submarine, no misread, and the decision left open at the film's ending. A short act whose point is the size of the gap between what is documented and what the book fills in.
The Diablo Canyon strike raises a question the engine cannot answer: the book's contamination comes from the reactor inventory, not the weapon. The study draws the weapon's own fallout under the day's wind and states that the plant's contribution is not modelled.
Data needed, and what exists
| Need | Source | Status |
|---|---|---|
| Population, present day | GHSL GHS-POP 2025 at 30 arc seconds, on the bucket | Have |
| United States strategic forces, 2024: 400 Minuteman III in three wings of 150 silos (F. E. Warren, Minot, Malmstrom), 14 Ohio-class boats at Kings Bay and Bangor with 20 tubes each, B-52H at Minot and Barksdale, B-2 at Whiteman | Nuclear Notebook 2024 (Kristensen, Korda, Johns, Knight), FAS status of world nuclear forces; Wikipedia wing and squadron pages for positions | To build: scripts/build-order-of-battle-2024.py, the 1973 script's pattern with the missile alert facilities geocoded per squadron |
| Russian strategic forces, 2024: about 327 ICBM launchers in twelve divisions (Teykovo, Kozelsk, Tatishchevo, Yoshkar-Ola, Vypolzovo, Dombarovsky, Nizhny Tagil, Irkutsk, Barnaul, Novosibirsk, Uzhur), Borei and Delta IV boats at Gadzhiyevo and Vilyuchinsk, Tu-95 and Tu-160 at Engels and Ukrainka | Nuclear Notebook 2024, Wikipedia Strategic Rocket Forces | To build, same script |
| North Korean forces: about 50 to 60 warheads, some seventeen ICBMs on mobile launchers, the Sunan, Sohae and Sinpo sites, the one missile submarine | SIPRI Yearbook 2024, FAS, Wikipedia; the launcher count is an estimate and the study says so | To build, same script |
| Missile defence: 44 interceptors, Fort Greely and Vandenberg; the SBX radar; the Clear, Beale, Cape Cod, Fylingdales and Pituffik radars; the SBIRS constellation as coverage, not positions | Missile Defense Agency fact sheets, Wikipedia GMD | To build, in the same file |
| Interceptor fate: 12 of 21 tests | GMD test record, Wikipedia and MDA | Rule in the study, three variants |
| Flight times | Minimum-energy arcs from the geometry, as the alert force uses | Have: src/models/ballistic.ts |
| Overflight | Where each arc's ground track crosses a border | To build: a small geometry check against a country outline, or the terrain lab's land mask |
| Target lists | None published. Urban targets by scripts/derive-urban-targets.py on the 2025 grid for the United States, Russia and North Korea; military aim points from the order of battle; category rules as DEFCON 3 uses NUWEP-74's | To build |
| The book's timeline | The text, minute by minute | The user has the book; the table above must be checked against it before it is drawn |
| The film's timeline | The film; reviews and the Pentagon exchange for the interceptor number | Have enough for the first act |
| Nuclear winter | Toon, Robock and colleagues, 2019 and the book's Part V | Cited in the omissions, not modelled |
Built, 10 September 2026
Routes: #/study/72-minutes (the film, as filmed), /record (the same missile with each interceptor drawn at the test record), /claim (at the agency's salvo claim, a hit), /jacobsen (the book's seventy-two minutes). Code in src/studies/minutes72/scenario.ts; forces from scripts/build-order-of-battle-2024.py (63 sites, three countries, the interceptor and early-warning sites included); urban targets from the HYDE 2023 grid by scripts/derive-urban-targets.py (300 American, 300 Russian, 40 North Korean cells); the book's and the film's minute marks in data/72-minutes/scenario.json; a coarse hand-traced outline of Russia in data/72-minutes/russia-outline.json for the overflight test. The sixth marking was not added: scenario events draw as inferred with the book named, and the log tags them SCENARIO. Population is GHSL 2025 from the bucket.
What the arithmetic says, before the outcome sums:
- The film's clock. A minimum-energy arc from the Sea of Japan to Chicago flies 31 minutes, so nineteen minutes on the radar puts the launch thirteen minutes before the film opens. Two interceptors from Fort Greely meet the arc over the Bering Sea about nine minutes after they leave. As filmed, one fails on the pad and one misses; at the test record two shots give 82 per cent, and the engine's draw decides; at the claim they hit and nothing arrives. The film never shows the warhead land; the study does, with the book's megaton borrowed and labelled.
- Four misses. The book's four interceptors all miss. At twelve hits in twenty-one tests, four independent misses happen 3.4 per cent of the time. The book's turn rests on a three-in-a-hundred event, and the log says so at minute eight.
- The Pentagon by the model's arc arrives at H+34:05 against the book's thirty-three minutes: the book's flight time is right.
- The overflight. Of the 58 arcs the 82 warheads ride (50 Minuteman and eight Trident buses), 45 cross Russian territory by the coarse outline, the first ground track entering at H+39:30 over the Far East, fourteen minutes after the Minuteman launch and thirteen before the book's Russian president gives the order. This is the one part of the book's central turn that is plain geometry, and the study draws the crossing point with its hundred-kilometre uncertainty.
- The counts. The rule puts the 82 warheads on 46 North Korean targets. Russia's counted force is 2,132 warheads on missiles; the rule finds 324 American targets for 1,114 of them and holds 1,018 back for want of a target, close to the book's "about a thousand" by accident of the target cap rather than by the book's reasoning. The American answer is 958 warheads on 328 Russian targets against the book's 975 aim points. The last American warhead lands at H+1:26, fourteen minutes after the book ends.
A fifth act, /salvo, is the defence lab drawn on the map: every North Korean launcher fires at a city, the forty-four interceptors engage the first eleven objects detected with four shots each at the test record, and the rest arrive. The order of engagement is a draw, because the defence cannot tell which object is bound where; on the engine's draw eleven are killed and six cities die, which of them changing with the draw. The urban targets are now named after the nearest of the largest cities within thirty kilometres, geocoded by scripts/name-cities-2024.py.
The outcome sums on the 2025 grid, blast, fire and fallout in sequence, are on the readout of the Jacobsen act and are to be recorded here from a run with the page visible: the union flushes on the animation frame and does not advance in a hidden tab.
Open questions
- The sixth marking was kept at five tiers; the log tags scenario events. Revisit if readers confuse the scenario lines with the record.
- The yield over Chicago in the film is never stated. A choice must be made and shown; the book's one megaton is the obvious borrowing and the readout should say it is borrowed.
- The Diablo Canyon reactor inventory: a bounded estimate from published source-term work could be added later as its own modelled entity; the first build leaves it out and says so.
- The Russian misread depends on the ground track crossing Russian territory. The great circle from Montana to Pyongyang passes over Kamchatka and the Sea of Okhotsk; from Wyoming it grazes the Kurils. The study should draw the crossing points and time them, because that geometry is the one part of the book's central turn that is simply true.
- Whether to end the Jacobsen act at seventy-two minutes as the book does, or let the arrivals run to the last warhead. The book's clock is the honest end; the engine can run on under a Keep switch.
Sources
- Annie Jacobsen, Nuclear War: A Scenario (Dutton, 2024). The timeline, the interviews, the figures in the table above.
- Kathryn Bigelow (dir.), Noah Oppenheim (writer), A House of Dynamite (Netflix, 2025). The single-missile act and the interceptor argument.
- Hans M. Kristensen, Matt Korda, Eliana Johns, Mackenzie Knight, "United States nuclear weapons, 2024" and "Russian nuclear weapons, 2024," Bulletin of the Atomic Scientists (Nuclear Notebook).
- Federation of American Scientists, Status of World Nuclear Forces (2025 update): United States 1,670 deployed strategic warheads, Russia 1,796, North Korea about 60 assembled.
- SIPRI Yearbook 2024, chapter on world nuclear forces, for the North Korean estimate.
- Missile Defense Agency fact sheets and the GMD flight test record: twelve intercepts in twenty-one tests through December 2023.
- Wikipedia: Hwasong-17 (test apogees and flight times), Ground-Based Midcourse Defense, LGM-30 Minuteman, Strategic Rocket Forces, North Korea and weapons of mass destruction; used for positions and counts, each cited on the mark.
- Bruce G. Blair, The Logic of Accidental Nuclear War (Brookings, 1993) and his later writing on launch under attack, which the book's decision timeline follows.
- The 26 September 1983 Serpukhov-15 false alarm, as documented by the National Security Archive, for the precedent the book's misread rests on.
- Owen B. Toon, Alan Robock and colleagues, "Rapidly expanding nuclear arsenals in Pakistan and India portend regional and global catastrophe," Science Advances 5 (2019), and the same group's global-exchange work the book's Part V relies on.