Grey = not heard, yellow = told this round, green = has heard. Yellow lines = links that carried the news. Silver rings mark the four best-connected people.
The science
Independent-cascade model on an Erdos-Renyi random graph with mean degree k: in each round every newly informed person tells each uninformed contact once, who hears it with probability T. The reproduction number is R0 = T k. In the large-network limit the final reach S (share of people who hear it, and also the chance that one random seed starts a large cascade) solves S = 1 - e^(-R0 S): S = 0 for R0 ≤ 1 and grows quickly above it. This is bond percolation, equivalent to an SIR epidemic with a one-round infectious period.
Honest limits. Real social and professional networks are clustered and have hubs (journalists, large accounts), so the Erdos-Renyi formula over- or under-states real reach; people also differ in how likely they are to share, and sharing is not understanding. The factor 1.4 for a plain-language summary is purely illustrative. Nothing here predicts the reach of an actual paper, press release or post.
Checks. A headless run builds 300 fresh random networks of 1 500 people and compares the mean reach among large cascades and the share of runs that become large cascades with the mean-field S; close to R0 = 1 finite-size effects are noted rather than hidden. The 160-person table is one noisy run. Scene: Blender-baked room (oak table, brushed steel pedestal, tile, plaster) in Three.js.