Influence OS treats human society as a dynamic network where every relationship has weight, direction, and propagation speed. Six instrumented modules — graph, scoring, propagation, narrative, power, strategy — let you map who matters, measure how much, and simulate what happens next.
Most digital tools surface content. Influence OS surfaces structure: who actually moves opinion, how fast a claim mutates as it spreads, where the bottlenecks of attention sit, and which seven nodes — if you removed them — would collapse a coalition. It is not a social network. It is a microscope and a wind tunnel for the social field.
Posts, likes, follower counts. A flat ranking by recency and engagement. You see the foam, never the current. The actual structure of who-influences-whom remains invisible — and that's exactly the structure that decides what you see next.
A weighted, directed graph with trust, frequency, authority on every edge. Centrality metrics on every node. A propagation engine that simulates how an idea flows from any seed across any topology. The structure is the product.
Drag the trust, frequency, and authority weights to re-render the graph topology. Edge thickness = composite influence weight; node size = aggregated outgoing influence; color = sector.
Influence is not a single number. The leaderboard ranks nodes by composite score, but each node has direct outflow (whom it speaks to), indirect cascade (whom those speak to), and topological position (eigenvector / betweenness). Toggle the metric to re-rank.
Choose a seed node, set transmission rate (k), recovery rate (λ), and mutation noise (σ). The simulator runs an SIR-with-mutation model and animates the wavefront across the network. Track reach, speed, and how the message drifts from its origin.
Opinions are not static — they are an attractor landscape on the network. The engine runs a bounded-confidence model where each agent updates toward neighbors within tolerance, while narrative entrepreneurs inject new framings. Bars show real-time share. Click a narrative to amplify it.
Run a structural sweep across the live graph: k-core decomposition for backbone density, community detection for clusters, and betweenness centrality for the bottleneck nodes that — if removed — would partition the network.
Strategy is search over interventions. Choose an objective; the simulator scans seed-set choices, message timings, and edge re-weightings, and surfaces the play that maximizes expected reach (or minimizes time-to-saturation, depending on the goal).
The AI layer doesn't generate posts. It generates insight. It watches the network state, identifies anomalies, predicts where a cascade will go next, and explains why in plain language.
Cluster C-04 (Tech sector) is forming an unusually tight subgraph — average pairwise trust is up 38% over the 7-day baseline. This is the structural precursor to coordinated messaging. Expect a unified narrative within ~48 hours.
If you seed at node N17 with the current message, expected reach is 412 of 512 nodes within 38 ticks. The bottleneck is N09 — its rejection caps your saturation at 71%. To break the cap, pre-condition N09 via N04, which holds 0.81 trust toward it.
Narratives N-02 and N-04 are converging on the same audience (~22% overlap) but propose mutually exclusive framings. Expected outcome by T+30: one collapses, the other captures both pools. Current win-probability for N-04: 0.62.
Node N42 is not in the visible top 10 by direct influence, but its eigenvector centrality is 4th-highest in the system. Translation: it influences influencers. Removing N42 fragments two coalitions that currently appear unified.
The next decade of digital power will not belong to those with the loudest channel. It will belong to those who can read the network, simulate what it does next, and intervene at the point of maximum leverage.