In development · Indo-Pacific · 2035
China enforces a customs law over Taiwan with thousands of autonomous systems and calls it a police action. Taiwan's air defenses are being dismantled one radar at a time. Nobody has been killed. The Council has to decide whether a war has started.
The scenario
Beijing announces that goods bound for Taiwan must clear customs on the mainland, then enforces the announcement with machines. Port cranes stop working. Cargo masters divert. Radar dishes on Taiwanese surface-to-air batteries are destroyed by charges small enough to leave the crew standing beside them unhurt, placed by drones that were already loitering overhead as part of the pressure campaign that preceded all of this.
Beijing's position is that this is domestic law enforcement on Chinese territory. Taipei's position is that it is a siege. Washington has to pick one, quickly, in front of allies who are watching to see who moves first.
Every framework the Council has for answering aggression assumes something was destroyed and someone was killed. Here the second half is missing, and Beijing has arranged it that way on purpose.
What the room decides
The legal question governs everything downstream: the War Powers reporting, the treaty consultations, whether the ally who is asked for basing says yes. Answer it too early and the Council owns the escalation. Answer it too late and Taiwan's defenses are gone.
Transferring autonomous systems avoids putting Americans in range. It also avoids putting American resolve on the table, which allies and Beijing both notice.
Human operators on the island say something no quantity of equipment says. They are also hostages to the next escalation, and the room has to decide what it is buying.
China produces replacements faster and ships them a hundred miles. The side losing a machine-attrition contest has an incentive to escalate into a contest it can win. The Council will feel that incentive and has to decide whether to act on it.
A Taiwan still standing, permanently weaker, with Beijing able to restart the campaign whenever it chooses. The room has to say whether it will accept that, and say so before the pressure to accept it arrives.
Against the library
Storm Across the Strait already turns on a threshold question: Beijing calls the cordon a customs quarantine, the San Remo Manual calls a blockade an act of war, and the Council has to characterize it under a clock. Machine Quarantine takes the same seam and removes the two things the room currently leans on — human crews and visible damage.
| Storm Across the Strait | Machine Quarantine | |
|---|---|---|
| Setting | September 2027 | 2035 |
| The instrument | Coast guard cutters, missile exercise closure areas, crewed aircraft | Autonomous air, surface, undersea and ground systems, with cyber underneath |
| Beijing's framing | A customs quarantine, kept below the blockade threshold | An internal police action on sovereign territory |
| Casualties at the decision point | Contested, and rising | None, by design |
| The governing clock | Taiwan holds eleven days of LNG | Whether resupply outruns degradation |
| What breaks the room | Deferring the legal call until the military one is made for them | Having no threshold, then discovering one after it is crossed |
What it is built on
The scenario family comes out of open research, and the customs-enforcement premise is not RAND's invention. Taiwan's own National Security Council ran a tabletop in June 2026 on a Chinese maritime quarantine enforced through mainland customs declarations, and CSIS has wargamed the blockade problem at length. What RAND's Camp(ai)gn series added, in August 2026, is the autonomous layer: it reports that when players were given limited political objectives, the Chinese side opened with weeks of low-intensity robotic operations before any missile campaign, and the American side struggled to say whether the threshold for a response had been reached. RAND's authors present that as a hypothesis raised by gameplay, and this exercise treats it the same way — a scenario worth putting a room inside, and an argument the room is invited to reject.
Borrowed: the 2035 horizon, the finding that autonomous coercion blurs the threshold, and the degrade-versus-resupply dynamic. All of it is published, cited above, and none of it is ours to claim. Camp(ai)gn is an operational-level game — two theater commanders allocating forces across a map with unit counters and adjudicated combat. This is a policy exercise: an NSC principals' committee with no map to move, deciding characterization, authorities and signaling under a clock. Different level of war, different players, different unit of action. The seat list, the five decisions, the adversary cell, the assessment rubric and the engine are the same ones running under every other exercise in the library.
It cannot establish that a conflict would open this way. The machines it puts on the table are more capable than anything fielded today, and if that assumption fails the exercise inherits the error. What a room can get from it is the practice of deciding under a threshold that has gone ambiguous, which is the part that transfers whatever the machines turn out to be capable of. The full statement of method and limits is on the main page.
Where it stands
The scenario, the seat list and the decision points are drafted. The inject set, the adversary cell's brief and the adjudication rules are not. It will be built for the first program that wants it on a calendar, and that program shapes what it becomes: which allies get played, whether Taipei has a seat of its own, how long the low-intensity phase runs before the room is allowed to escalate.
If you want to see the draft and argue with it before it hardens, a briefing is the way in. Forty-five minutes, the scenario walked end to end, and the questions you would need answered before putting your people in it.
Book a briefing
Dr. Mullick runs the briefing himself. No mailing list, no automated follow-up, and no obligation attached to it.