Christof needs people—not because 1998 lacked computing power, but because Truman is genuinely unpredictable in ways that require judgment, not just processing.

Watch the scene where Truman's about to leave Seahaven Island for the first time: the operators don't consult a decision tree, they improvise.

One technician suggests the car accident. Another argues for the accident in the parking garage instead—they negotiate, they second-guess, they override Christof's instincts and he accepts it because they've been watching Truman longer than anyone. That's institutional knowledge masquerading as control. An algorithm would have calculated Truman's escape probability and locked him down months earlier, with mathematical certainty.

The human operators are actually worse at prediction but better at adaptation—they can feel when the artificial world is about to crack and adjust the pressure valve before it shatters. Our actual systems work backward. Netflix's recommendation algorithm doesn't need human negotiation because it isn't trying to contain a person—it's trying to optimize engagement across millions of them simultaneously. It's perfectly efficient at making you stay, which is precisely why you can't feel the edges of your cage. You're not breaking against the walls—you're building them from the inside.

Christof's weakness—that he needs staff meetings and contingency arguments—was actually his system's strength. Ours has no staff meetings, no one's voting on whether you should see this story or that one, and the mechanism is too distributed, too fast, too fundamentally inhuman to negotiate with. Truman could have killed someone on that island and forced a human reckoning. What forces a reckoning in an algorithmic system?