I remember the hours on Stack Overflow.

Early in my career, when documentation did not explain how something worked and I had not seen enough systems to reason through how it was probably built, I would dig through forums and threads, trying to piece together an answer. The process was slow. Often frustrating. Sometimes I would be stuck for days.

Then one of the senior engineers would walk over. "Yeah, I ran into that a few years ago. It is not documented."

The flag itself was not the revelation. Everything I learned while searching for it; that was the revelation. The rabbit holes taught me how systems fit together. The wrong turns taught me what to eliminate next time. The mentor showing up at my desk taught me which instincts to trust.

The journey was the point.

Now there is AI. You ask a question, watch a blinking cursor, see a few dots animate, and there is your answer. No journey. No struggle. No development of judgment.

The current narrative says this is progress. AI makes junior engineers unnecessary. Why train someone when the machine can produce the code? Companies are cutting entry-level roles and celebrating the efficiency.

They are making a generational mistake.

Senior engineers, the ones who can debug production systems, understand why things break, design something novel did not arrive knowing these things. They were juniors once. They struggled through exactly the process that AI now bypasses.

AI amplifies capability. It does not create capability from nothing.

If you have decades of experience, AI makes you faster. You know what questions to ask. You can evaluate whether the answer is right. You have the judgment to recognize when the machine is confidently wrong.

If you are starting out, you get the answer without understanding the answer. You skip the struggle that builds mental models. You never develop the instincts that come from failing a hundred times.

Here is the problem no one is talking about. The experienced engineers will not be around forever. They will retire. They will move on. And if we spend this decade not hiring juniors, not training them, not maintaining the pipeline—who replaces them?

This is not hypothetical. We have watched it happen already.

Destin Sandlin from Smarter Every Day spent four years trying to manufacture a BBQ grill scrubber entirely in the United States. Everyone he talked to told him it could not be done. They were right.

I think we're screwed as a nation if we can't do the intelligent work of tool and die—making the tools that make the things.

The turning point came when he needed an injection mold made. He found Chris Robson, a local toolmaker in Alabama who could build it. But Chris was retiring. If he had retired a year earlier, the mold could not have been made locally at all.

Destin put it bluntly: "I think we're screwed as a nation if we can't do the intelligent work of tool and die-making the tools that make the things."

It used to be that America did the smart work—designed the machines, built the molds—and sent them elsewhere to be operated. That has flipped. Now the smart work happens overseas and we push the buttons on the machines they send us.

Tim Cook confirmed this at the 2017 Fortune Global Forum. The reason Apple manufactures in China is not cheap labor. It is skill. "In the US, you could have a meeting of tooling engineers, and I'm not sure we could fill the room. In China, you could fill multiple football fields."

The abstract became concrete when Destin tried to source chainmail for his scrubber. He found a US supplier, but they could not produce commercial quantities. So he turned to an Indian supplier to avoid China. When the shipment arrived, the packaging had Chinese characters on it. The Indian company was drop-shipping from Shanghai.

He could not even avoid China when he was actively trying to. The capability simply does not exist here anymore.

We went up the technology stack software, services, finance—without maintaining the foundation. Now we have brilliant designs that can only be manufactured by someone else.

Software is heading down the same path. We are cutting junior roles, celebrating the efficiency, congratulating ourselves on running leaner. And we are dismantling the system that produces the next generation of capable engineers.

I will allow that AI handles routine work effectively. I will allow that it is cheaper for certain tasks. But there is no clear path to AI understanding what is actually happening when systems behave unexpectedly. It has not shown it can debug novel problems or create something genuinely new.

The best way to learn a skill is by working alongside capable people who coach you through real problems. Not in a classroom. On the job. I benefited from this directly. Mentors who took the time. Engineers who walked over and explained what the documentation left out.

That transfer does not happen when there are no juniors to transfer it to.

So stop treating junior roles as inefficiency to eliminate. Start asking who will operate your systems when the current experts are gone.

Better tools do not reduce the need for capable operators. They increase it. A more powerful instrument requires more skill to play, not less.

We are optimizing for today while breaking the system that produces tomorrow's capability.

The journey was never the inefficiency. The journey was the entire point.