Every technological revolution produces a curious psychological phenomenon: those who have reached a certain point through great effort tend to turn their own story into a universal law.

It is happening with AI too. Every discussion about programming ends the same way. If AI writes the code, how will junior programmers become seniors? Who will correct the mistakes? Who will acquire the experience that, after ten or fifteen years, distinguishes a true professional from a beginner?

It sounds convincing because, after all, almost everyone leading development teams today has followed the same path: thousands of hours spent writing code, reading documentation, chasing bugs and rewriting functions that a model now produces in a few seconds. It is natural to think that this long apprenticeship was precisely what built their expertise.

No one is surprised that an architect born today does not know how to use a parallel ruler. No photographer insists that a young person should understand silver emulsions before taking a photograph or using Lightroom. Yet when it comes to programming, the reasoning changes. The assumption is that the new generation must retrace the previous generation's steps, even when that path no longer exists.

It is an odd assumption, because it presupposes that AI is a temporary interlude destined to be swept away.

“Enough of this. Back to programming without AI, developing plates with emulsions, starting engines with a crank and getting tangled up in the cords of a parallel ruler.”

It has never worked like that. Those thousands of hours are no longer needed. What is needed is an understanding of how to talk to a model to get it to program what you want. End of story. Just turn the key or press a button to start the engine.

When an invention profoundly changes the way we work, it almost never disappears. It becomes the environment in which we work. It happened with the electric motor, the operating system, the browser and search engines. No one proposes uninstalling Windows to force young people to learn DOS. No one pines for Telnet because the new system administrators have never used it. The fax disappeared without trauma. Changing the little roll of thermal paper was a genuine skill; today it is simply useless. Someone born in the age of email is not an incomplete professional because they cannot do something that progress has made irrelevant.

A young person with an average level of training and enough intelligence can do a senior's work if they have the right subscriptions and the software's functional specifications. 

Artificial intelligence is the new technological context in which the new juniors, who are seniors in all but name, will train and work. Asking whether they should learn to program without AI is equivalent to asking whether an architect should forgo CAD.

Behind this reasoning lies a kind of nostalgia for toil. Every generation tends to attach an almost moral value to the difficulties it has faced. If learning was difficult or laborious, then that difficulty must have had some value. I understand that some people may think this, but it is not true. What once had value no longer does, suddenly, as in the crash of ’29. From 100 to zero in less time than it takes a sports car.

The meaning of the word junior has changed: it is merely administrative, because today a new graduate produces results that would once have required a programmer with years of experience and time to spare. This is because they work with a tool that enormously amplifies their abilities. Comparing this young person with their counterpart in 2015 means comparing two different professions.

From now on, the difference between programmers will probably emerge from the quality of their decisions, their ability to design systems, to recognise an implicit requirement, to understand when a solution proposed by artificial intelligence is elegant or merely suitable. The keyboard will play an ever smaller part. Judgement, by contrast, will gain in value.