Building a JUCE Strategic Simulation Plugin

Building a JUCE Strategic Simulation Plugin -From APVTS to SIMD AVX2

A channel strip is a narrow column on a mixing console: an input, a chain of transformations—gain, filter, dynamics, send—and an output that adds to the others. Anyone who works with audio spends years looking at these channels and ends up seeing them for what they actually are: dynamic feedback systems, where a poorly set gain oscillates, a delay throws the phase, and a systematic error in the measurement silently propagates throughout the mix.

This book was born from a question that, when spoken aloud, sounds like a provocation: what happens if we treat a strategic system as a signal processing system?

The answer is more literal than you might expect. A country is a channel strip . The state it attempts to govern—the military apparatus, the economy, social cohesion—is a signal that evolves over time and that no one directly observes. The decision-making core is a controller: it reads a measurement, compares it to a reference, and produces a command. The reports that rise through the hierarchy are the measurement, and they carry with them an offset —a constant bias that goodwill does not eliminate and that the sum of many reports does not dilute, because they all lean in the same direction. The time it takes for information to rise and the order to descend is latency. Mobilization and industrial policy are actuators, and actuators saturate. Propaganda is a measurement error—and that's where the analogy begins to pay dividends, because propaganda is a measurement error of a particular kind: it doesn't increase variance. It doesn't make the signal tremble. It shifts it. And therefore, it doesn't trigger any of the alarms that are usually set up against noise. War is a feedback loop between channels, with the signal reversed: what comes out of one enters the other, subtracting. An alliance is the same loop with the opposite sign. A network of alliances and conflicts is a bank of oscillators with cross-modulation, and behaves as such: it has its own modes, it has directions in which it amplifies, and it has—as will be seen—a spectral radius that can be greater than one, in which case the world it describes diverges to infinity in little more than two seconds of simulation time.

The analogy is not merely ornamental. It is operational, and for a reason that only becomes apparent when you open the editor: the tools that real-time audio has developed over thirty years are precisely the tools that an interactive simulator needs.

An audio plugin has to solve four problems every day that a strategic simulator also has to solve, and that most scientific software solves poorly or not at all.

The first is the single source of truth . The same numerical value—let's call it Russia's informational bias—is written by the host when the user automates a session, written by the interface when someone flips a knob, and read by the numerical mechanism sixty times per second. That's three writers and three cadences, in three distinct execution threads. If there isn't a single repository of this value, the simulation becomes dependent on which of the three paths wrote last, and reproducibility—which is the condition of any scientific claim—evaporates without leaving a trace and without producing a single compilation error. JUCE's AudioProcessorValueTreeState exists precisely for this reason.

The second is the discipline of deadlines ...