SA-001 made five dated, falsifiable calls on RAND's Infinite Potential scenario, then promised to track the distance between the scenario's world and the measured one. Two sweeps later, here is the motion. Five calls still hold, no falsifier has tripped, and the sixth signal opened in June moved again in July. Substrate has now strengthened at both sweeps, and July supplied the cleanest evidence yet from a direction the May read did not anticipate: the binding constraint showed up as a memory producer's capacity decision and a rate environment pricing physical plant, not as a capability result.
A wargame cannot show you a month of motion. A radar can. Each call carried a stated falsifier on 28 May. Here is what moved, what the falsifier required, and whether it tripped. Nothing was graded after the fact.
SA-001 ran three desks and had no East–West axis. Adding Hé does not add a fourth opinion. It surfaces that the US–China compute fight, read past the "race" frame, is organizing around the same binding constraint Manticus already named: the physical substrate.
Washington shifted H200 and MI325X chips to case-by-case licensing and cleared roughly ten Chinese firms to buy. Beijing declined — telling firms not to buy, prioritizing a domestically controllable stack built on Huawei's rack-scale Ascend systems. The US tried to re-bind China through dependence; China refused the binding to own its own substrate. Both systems are now racing for substrate sovereignty. The seam confirms the Substrate call from a second direction.
Hé's falsifier: full indigenous substitution or full embargo — either marks the seam crossing out irreversibly. Status: inside the band. · Sources: Brookings, CFR, Geopolitical Monitor · Mar–Jun 2026.
The wargame imagines the day after.
The Radar tells you how close it is — and, after a month, which way it moved.
"Two months on, nothing in the read broke, and the market spent July trading the substrate call: a fifth off the chip complex while every operating number underneath it improved."