Continuous, saturating stress dynamics on a node graph, coupled to a depleting buffer (the drawable inventory above the allocation floor). The loop gain is now the actual Jacobian spectral radius of the dynamics — and it rises as the cushion thins, so tipping is reachable and bounded. A reasoning instrument: it propagates assumptions, it does not validate them.
Scales each node's exogenous drive by its war-sensitivity. The structural edges stay fixed; only the spark rate changes.
The stabilising loop, now physical: high system stress destroys consumption, cutting the rate at which the buffer is drawn down. Slide to 0 to remove the brake and watch the cushion empty; raise it to find where the buffer holds above its floor.
⚠ This is the new bridge assumption B+D adds: how many million barrels/day of net buffer draw a fully-stressed system implies. It is a judgement, not a measurement — it is exposed here precisely so it can't hide.
The reservoir, in billion barrels. Allocation (the endpoint) fires when it hits zero. Default 0.8 = the amount realistically drawable before operational stress; slide up toward ~1.85 to include strategic/emergency stocks (see methods). Buffer feedback strength: 1.0× — how hard scarcity feeds back into the stress network (the doom-loop coupling).
Median trajectory (solid) with the P10–P90 fan across — stochastic trials. The red line is the allocation floor; where the fan touches it is where a major economy crosses from price rationing into physical allocation.
Each input is swept ±50% of its current value (clamped to its range), holding the rest fixed; bars rank inputs by how many percentage points they move P(allocation). The long bars are the assumptions your conclusion hinges on — the ones worth defending; the short ones you can stop arguing about.
| Node | Spark % | Draw wt | Equilibrium stress |
|---|