One brief for each model and each study: what it computes, which document or dataset every figure comes from, what it deliberately does not model, and where its output has been set against a published one.
These are not the same thing as the research dossiers, and the difference is worth stating. A dossier is about the literature: what has been published on a subject, who disagrees with whom, and which widely repeated figures do not survive checking. A brief is about this instrument: what it computes, how, and what it leaves out.
They are rendered from the working files the project keeps, so what is here is the working record itself — including the passages recording that something was got wrong and corrected.
The planning rules the solver applies, and the documents each of them comes from.
The four phases of interception, the arithmetic of each, and why the endgame is the easy part.
How a strike on any point on Earth is planned, sized and flown, step by step.
Every figure the engine computes, set against a published one: what agrees, what does not, and by how much.
The chain from the fires to the famine in five coupled models, and what each was fitted to.
The optimiser: what it is asked to maximise, what it may spend, and what it cannot trade away.
One missile, one decision, and the seventy-two minutes between them.
The exercise of November 1983, and the question of how close it came.
Square Leg and the Home Office estimates, against the engine’s own.
October 1962, and the forces actually in place on each day of it.
The alert of October 1973: what moved, when, and on whose order.
A Soviet strike on the Chinese nuclear complex in 1969 — the surgery nobody has evidence was planned.
The Single Integrated Operational Plan of 1962, from the declassified record.
Tactical and theatre doctrine in Europe, and the five scenarios drawn from it.
The British bomber force, its profiles and its timings.
The window of vulnerability, as the forces of 1983 actually stood in it.