Read the explanation
The local default assumes one hundred twenty decorative agents, velocity eighty five, guardrails thirty and coordination forty. Risk adds velocity times point fifty two, agents times point twenty eight, uncoordinated share times point eighteen and three, then subtracts guardrails times point forty four. Positive contributions total ninety one point six and the guardrail subtraction is thirteen point two, leaving seventy eight point four. At three pixels per point the positive bar is two hundred seventy four point eight, reduction thirty nine point six and total two hundred thirty five point two. These fitted teaching coefficients are not measured systemic risk probabilities or evidence for a real development freeze. Containment adds guardrails times point seventy six plus coordination times point thirty five plus eighteen point nine. It subtracts velocity times point fourteen and agents times point zero eight. Default positive contributions fifty five point seven minus twenty one point five give thirty four point two. At four pixels per point those bars are two hundred twenty two point eight, eighty six and one hundred thirty six point eight. The vulnerability formula separately returns forty two point eight assigned operations per minute. Canvas rogue nodes and moving targets are a visual simulation; these counts do not establish attacks, operational breaches or actual containment effectiveness. The slowdown preset assigns fifty agents, velocity thirty, guardrails eighty five and coordination seventy five. Its raw risk is negative point three but the displayed floor raises it to five. Raw containment one hundred one point five five is capped at ninety eight point five; raw vulnerability negative three point eight becomes point five. Compare default risk seventy eight point four with five, a reduction seventy three point four. At three pixels per risk point the bars are two hundred thirty five point two, fifteen and two hundred twenty point two. This clamp enforces a preset result rather than verifying security. Native controls, JSON briefing and actual drawn two dimensional canvas operate locally, with random particle positions. No real agents, policy enactment or model backend runs.