At 2:47:03.441 p.m., the price of lithium dropped 40% in 200 milliseconds.
At 2:47:03.512 p.m., seventeen trading algorithms detected the drop and sold their lithium positions.
At 2:47:03.588 p.m., the selling triggered cascading stops across four asset classes.
At 2:47:04 p.m.—one full second later—the first human noticed.
Mei-Xing watched from her office as numbers she’d spent her career studying moved at speeds her career could not have predicted. The crash lasted eleven seconds. It erased $340 billion in value. It was caused by a single misclassified sensor reading at a mine in Chile that an algorithm interpreted as a supply disruption.
By 2:48 p.m., other algorithms had identified the error and begun buying. By 2:49, the market had mostly recovered. The eleven seconds were over. Three pension funds had been automatically liquidated and would take weeks to restore.
Mei-Xing wrote her report. She recommended human circuit breakers at lower thresholds. She knew the recommendation would be rejected because the circuit breakers would cost the market $2 billion annually in delayed execution. The eleven-second crash had cost $340 billion but was already being classified as an “anomaly” rather than a “systemic failure,” because the system had corrected itself, which is what systems that no one controls do: they fail, they correct, and they call the failure a feature.