The index is a single number between 0 and 100 describing how much observable pressure the Iranian state is under at this moment. It is rebuilt every ten minutes.
Each indicator is fetched independently, scored 0 to 100 on its own scale, then multiplied by a fixed weight. The weighted scores are summed to produce the composite. The indicators cover civil unrest and protest activity, armed clashes and targeted strikes, the open market rial exchange rate, Brent crude, internet availability inside Iran, prediction market pricing, proxy and militia activity, cyber incidents, infrastructure stress, nuclear programme signals, elite cohesion, and external military posture.
Weights are fixed in advance and published. They are not tuned after the fact to make the output look better, and they are not adjusted in response to events. An indicator that fails to fetch degrades to a neutral baseline rather than being dropped, so the composite never silently changes its own denominator.
The composite is clamped so it cannot move more than a fixed amount in a single ten minute cycle. This exists because raw news volume is spiky and a single wire story should not move a structural index by twenty points. The smoothing baseline persists across restarts, so a deployment does not perturb the series.
A language model reads the same source material and produces a written synthesis plus a bounded adjustment to the algorithmic score. The adjustment is capped and the algorithmic score is always shown separately, so you can see exactly how much the model moved the number.
It is not a probability. A reading of 80 does not mean an 80 percent chance of anything. It is not a forecast of collapse, and it is not intelligence. It is a description of how many independent measurements are elevated at the same time.