Simulate Binary Signal Cascades Across Directed Agent Graphs
Construct directed agent topologies, configure activation thresholds, and inspect deterministic multi-round cascade dynamics with zero external dependencies.
Formulate, Calibrate, and Execute Cascade Scenarios
Use the guided stepper to adjust topology, thresholds, and simulation depth, or trigger updates directly in real-time.
Scenario Stepper
Configure parameters step-by-step or trigger execution directly.
Curated Scenarios
Quickly load specific baseline network configurations to evaluate distinct cascade phenomena.
Model Specifications
- Activation Rule Persistent Binary
- Influence Weighing Uniform Per-Link
- Execution Model Synchronous Rounds
- Data Boundary Zero Telemetry
Deterministic Spread Across Directed Graphs
Signal cascades depend directly on the structural connectivity of individual nodes. When an inactive node evaluates its inbound edges, it computes the fraction of active predecessors.
If the ratio meets or exceeds the threshold ratio, activation is triggered permanently for all subsequent discrete rounds.
Unidirectional Ring
Nodes form a closed directed loop where each agent has exactly one incoming and one outgoing edge. Guaranteed complete propagation if threshold is 100% or lower.
Hub & Spokes
A single central node is bidirectionally linked with all peripheral nodes. Demonstrates extreme asymmetric propagation depending on seed selection.
Sequential Chain
Strict forward pipeline from node 1 through node N. Useful for testing directional delay and terminal boundary conditions.
Empirical Activation Benchmarks
Single-Predecessor Invariance
In a pure directed ring, each agent receives exactly 1 inbound connection. Consequently, any threshold between 1% and 100% yields identical propagation rates of exactly 1 hop per discrete round.
Inspect, Validate, and Export Agent Graph Configurations
All computation occurs locally within your browser sandbox. Export deterministic state sequences in standardized JSON format or capture vector diagrams for documentation.
Synchronous activation on a directed graph
Read the explanation
The eight-node mesh gives each node two incoming neighbors. With one seeded source and a fifty percent threshold, one active incoming neighbor is enough. The default five rounds activate all eight nodes. Raising the threshold to one hundred percent requires both incoming neighbors; the one initial seed cannot start a cascade, so only one remains active. At thirty five pixels per finally active node the bars span two hundred eighty and thirty five. This is a synchronous discrete graph model, not live agents executing tasks. The native JSON and SVG exports record the graph result rather than external work. In an eight-node directed ring seeded at node one, each inactive node has one predecessor. At any allowed threshold one active predecessor suffices. After one round there are two active nodes; after five there are six. At forty pixels per active node the bars grow from eighty to two hundred forty. The simulator collects new activations before adding them to the active set, so a whole chain cannot activate during one loop pass. Previously active nodes remain active. These rounds model propagation steps, not elapsed wall time, measured communication latency, or successful multi-agent reasoning. The eight-node ring has eight directed edges. The star links its hub to each of seven leaves in both directions, giving fourteen. At twenty pixels per directed edge the bars span one hundred sixty and two hundred eighty. If the hub is seeded, each leaf receives its one incoming signal and activates in the first round. If a leaf is seeded instead, the hub threshold applies to its seven incoming signals, creating a different cascade. Edge count alone therefore does not determine activation. Inspect source, threshold, topology, rounds, and the exported history rather than treating a visually dense graph as proof of coordination quality.