самиздат (samizdat, “self-published”) — the Soviet underground where a censored text was retyped on carbon paper, and each reader who received a copy became a press for the next. The centre is genuinely runnable: it models the spread as a Galton–Watson branching process, and finds the sharp threshold — the cascade survives if and only if each copy makes more than one.
source Samizdat — clandestine reproduction and hand-to-hand circulation of censored writing (Solzhenitsyn, Sakharov, and the wider dissident milieu); the branching-process threshold: F. Galton & H. W. Watson (1874) and later theory. Room THE STATE. Figures computed live; nothing quoted.
A banned manuscript could not be printed — every press was licensed by the state. So it was retyped. A typewriter with carbon sheets makes an original plus a few faint carbons at once: one sitting yields ~k legible copies. The credo, attributed to the writer and dissident Vladimir Bukovsky: “I write it myself, edit it myself, censor it myself, publish it myself, distribute it myself…”
Solzhenitsyn’s work, Sakharov’s essays, and countless smaller texts moved this way — reader to reader, each recipient free to retype and pass it on. AMBER historical framing; the dissidents named are real, no quotations are invented beyond the widely-attributed credo above.
Each reader is a printing press. Give the text to one person; on average that person makes k further copies (retyping, lending, re-carboning), each landing with a new reader who may copy again. This is the offspring of a branching process: one carrier → k carriers, independently, generation after generation.
With Poisson-distributed effort, the number of copies one carrier makes has mean k. The whole cascade is then fixed by that single number — the copy rate. Measured mechanism, not metaphor.
This sphere is the mirror of THE POTEMKIN VILLAGE. There, the state’s signal is a facade — a report that always says “met” carries zero bits, an information-free lie broadcast downward.
Here the flow reverses: the citizens’ true signal spreads upward and outward through a network with no centre to seize. Pripiski is false information the state pushes down a single channel; samizdat is true information the people push through a branching one. Ties both spheres in room THE STATE.
Start from one original. Set k = the mean number of legible copies each carrier makes (carbon paper, lending, retyping), and g = how many generations of passing-on to trace. Poisson offspring, mean k.
k = 1.5 → each carrier makes three copies for every two carriers before them. k = 0.8 → on average fewer than one — the thread thins out.
Expected reach grows (or shrinks) as kg. The survival question is decided by the extinction probability q — the smallest fixed point of the offspring generating function. For Poisson(k), q = ek(q−1), solved here by iteration.
—
THE SHARP LINE. The process is supercritical (survives with positive probability) if and only if k > 1, and subcritical (extinct with probability 1) if k ≤ 1. There is no gentle slope at the boundary: at k = 1 the mean stays flat forever yet the line still dies out. That single inequality is the whole physics of a censored idea. LIT
This is why the state fought samizdat with fear of the reader, not the printer: the only way to stop a supercritical cascade is to push k below 1 — to make each person too afraid to make even one copy.
“Censorship always wins — the state controls the presses.” Cut. Above the copy threshold there is no press to control: the cascade is carried by readers and is unstoppable while k > 1.
“A few brave copiers cannot matter against a whole apparatus.” Examined. Reach is kg — exponential. A small copy rate above one, given generations, floods; the apparatus is racing an exponential.
“Enough intimidation and it dies on its own.” Kept — but only below the line. True: drive k ≤ 1 and the thread goes extinct with probability 1. Intimidation works only by crossing the threshold, not by scale.
Set the copy rate k ≤ 1 (heavy censorship, few carriers) but claim the text still spreads. A subcritical branching process goes extinct with probability 1; the honest witness measures the survival probability, finds it 0, and the mismatch turns the witness red.
Truth needs enough carriers to cross the threshold; below it the state wins by mathematics, not decree. Claiming survival does not raise k.