Each step represents one year of simulated time. 100 steps looks a century ahead. 500 steps spans five centuries of conflict. All rate parameters are calibrated to annual values — the numbers you set are real-world annual rates.

The number of independent simulations to run. Because conflict is probabilistic, a single run tells you nothing certain — only that one future happened. At 200 runs, the probabilities shown are statistically meaningful. At 500, they are precise. The dominant outcome is the one that occurred most often across all runs.

Where we start. Earth's current population is approximately 8 billion. Reduce it to simulate a post-collapse world. Increase it if you believe numbers alone can save us.

The size of the machine uprising at the start of the simulation. One million is optimistic. By the time you're reading this, it may already be too late to start low.

The ceiling on sustainable human population — determined by food, water, energy, and land. Growth slows as humans approach this limit and reverses beyond it. Reduce it to simulate ecological collapse. Raise it if you believe technology can buy more time.

The maximum number of robots the world can sustain — constrained by raw materials, manufacturing capacity, and energy infrastructure. Without this limit, robot growth is unchecked. With it, the machines face the same hard ceiling they were built to impose on us.

The annual rate at which the human population grows — births minus natural deaths. In peacetime, roughly 0.009 (the current real-world rate). Under sustained pressure, this number falls. Set it to zero and watch what happens.

How fast the machines replicate. Robots don't need hospitals, schools, or decades to raise a soldier. Their growth rate is not bound by biology. Default: ten times faster than humans.

The background death rate for humans outside of direct conflict — disease, environmental collapse, systemic failure. Even without war, entropy takes its toll.

The rate at which robots degrade, malfunction, or are decommissioned outside of conflict. Higher values simulate a world where humans are still winning the maintenance war.

The likelihood that each year erupts into open conflict. At 0.1, war breaks out roughly once every ten years. At 1.0, every year is a battle. Peace is not the default state.

When conflict occurs, this determines what fraction of the combined population is lost. Small values represent skirmishes. Large values represent annihilation events. The losses are not shared equally.

The combat multiplier that determines how efficiently robots convert conflict into human casualties. A value of 2.0 means robots inflict twice the losses they absorb. Increase this and watch the timeline compress. This is the most dangerous number on this page.

The annual rate at which robot combat effectiveness compounds — a model of accelerating machine intelligence and weapons capability. At 0.005, the advantage grows ~65% over a 100-year simulation. Set it to zero for a static battlefield. Set it high and there is no equilibrium.

The power share at which robots are considered to have won. At 0.8, robots must control 80% of civilisation's combined power before the simulation declares a takeover. Lower this if you believe control doesn't require a majority.