When Removing Redundancy Becomes Negligence Again

Michael Rejent sits in a darkened room in Jacksonville watching planes that don't exist—the radar screen in front of him shows aircraft that his eyes, through the window behind him, cannot see.

The reverse also happens. A controller spots movement on the tarmac that the system hasn't logged yet, or a pilot calls in a position that contradicts what the automation recorded.

These gaps are not malfunctions but the normal operating state of American air traffic control, and they are getting wider. After a Lockheed Constellation and a Douglas DC-7 collided over the Grand Canyon on July 10, 1956, killing 128 people, the federal government realized that a single source of truth was not enough—the planes had followed their assigned altitudes perfectly, yet neither aircraft could see the other. No human controller could see them either.

The solution was to add humans into the loop not as backup but as a second independent detection system. Controllers became eyes the radar could not be. For seventy years this worked because everyone understood the trade-off. Automation detects what humans miss in the noise, while humans catch what automation cannot perceive, namely the subtle, the contextual. The physical detail that changes meaning.

Then the system got better, and automation improved so much that regulators began removing the redundant layer. Controllers could be consolidated, facilities could be automated. The human perception that had been insurance against single-point failure looked now like inefficiency. What has changed is not the planes or the sky but the confidence that the gap between what the system detects and what a human can actually perceive in real time has finally closed. Every organization that removes oversight layers tells itself the same thing. The people you trained for decades to catch what you missed are now optional. That confidence is where the gaps that killed 128 people in 1956 stop being redundancy and start being negligence again.