AI-Era Engineers as Airline Pilots

Modern airline pilots don’t actually fly the plane the entire time.

During cruise, things are heavily automated. The job is mostly monitoring instruments.

Humans really matter at three moments:

• Takeoff
• Landing
• Anomalies

Recent AI development is starting to feel a lot like this.

AI implements, humans monitor

These days, AI handles a lot — and fast:

• Code generation
• Test creation
• Research
• Refactoring

So humans spend more time:

• Checking that the design isn’t drifting
• Reviewing security
• Watching for AI doing something weird

We’re moving from “I write everything myself” to “I pilot and monitor the AI.”

But only humans can stop things in an emergency

Even airline pilots aren’t fully hands-off.

When something goes wrong, the pilot’s experience is what saves the day.

AI development is the same.

• Hallucinations
• Dangerous designs
• Production incidents
• Broken logic

When these happen, the human is the last line of defense.

And what matters most there is gut feeling — the sense that something’s off.

Pilots also start with manual flying

Airline pilots don’t start in big jets.

They start with small propeller planes and learn manual flying inside out.

In other words, they’re drilled in the fundamentals that existed before automation.

The same is probably true for AI-era engineers

Implementation will keep getting more automated.

But the value of fundamentals will probably go up, not down:

• OS
• Databases
• Networks
• Algorithms

Most of the time, AI flies the plane.

But when things really go wrong, only someone who understands the fundamentals can take over.

The AI-era engineer might be less of a “person who writes code” and more of a “pilot monitoring a massive AI system.”