The PreCrime system works because three precogs always see the same murder.

This is the film's foundational lie — and it's the only reason the entire apparatus functions.

Nobody in 2002 noticed the architecture was broken because the film never let us see it fail gracefully. Only catastrophically, through John Anderton's exception. But the real catastrophe was baked into the premise: the system demands consensus from three independent neural networks processing identical inputs and somehow reaching identical outputs. In reality, they wouldn't. Agatha, Arthur, and Dashiell would generate three slightly different murder scenarios — different angles, different motivations, different outcomes — and the PreCrime machine would have to choose which vision to trust.

Which precog gets overruled. Which dataset counts as truth. The film stages this exactly once: when Agatha's vision contradicts the other two, and the system automatically privileges the majority. Spielberg treats this as a character flaw — a procedural loophole — rather than a fundamental vulnerability. By 2024, this had become the actual problem crushing every significant AI deployment: facial recognition systems trained on different datasets give conflicting results; medical AI trained on different patient populations produce different diagnoses; recommendation algorithms optimized for different metrics contradict each other.

The solution isn't better precogs. It's accepting that multiple truths coexist, and no amount of processing power reconciles them into a single vision. Minority Report understood the mathematics of plurality would break prediction. It just thought that was a plot twist instead of the whole game.