Vai al contenuto
vati.it

← All reflections

Zero lines of code

Valentino Pietrobon · 12/11/2026

All my programs were built without my writing a single line of code. Not because programming no longer matters: quite the opposite.

Osserva, Prenota, Progetta, Gestisci, Pianifica, the interactive tools I make for my classes, this very website; and the ones still in the works, like Allena and Jarvis: all my projects have one thing in common. The number of lines of code I have written myself is zero. Every line was written by an artificial intelligence, in a dialogue that is now called vibe coding: you describe what you want, the AI writes the program, you test it and ask for corrections.

Put like that, it sounds as if programming were no longer needed. My experience says the opposite, and I'll try to explain why.

If you don't try it, you'll never know what you can do

The first lesson is the simplest: you have to try. As long as you only talk about it, vibe coding stays a vague idea, somewhere between enthusiasm and suspicion. When you use it on a real problem (booking the labs, recording observations in class, publishing materials without depending on anyone else) you discover that things you would have put off for years become possible in a few days. And you also discover where it gets stuck. Neither can be learned by reading.

Design and testing become the main job

AI gets better month after month, and writing code costs less and less. What does not shrink, and in fact grows, is the weight of two activities: design and testing.

Designing means deciding what the program must do, for whom, and under which constraints. For this website, for example, the most important choice was not a technical one: I wanted it to keep working even if one day I no longer manage the school's accounts, so nothing in it depends on services that others could take away. No AI could have known that for me.

Testing means actually trying every feature, the way a student or a colleague will use it, rather than trusting a "done, it works". More than once the program was "ready" and, when I opened it, nothing had changed. That is where you find out whether a project really stands up.

Knowing how to program changes the dialogue

I don't write code, but I know how a program works: what a database is, a form, a session, a file that needs protecting. That knowledge matters at two moments.

  • It helps you imagine: if you know programming, you can see what could be built and sense how.
  • It helps you talk to the AI: the more you know, the more precise your requests, the better you understand the answers, and the sooner you notice when something doesn't add up.

The dialogue with an AI is only as productive as your understanding of what you are talking about. Those who know nothing get something that looks like it works; those who know get what they actually need.

The AI doesn't always understand, and doesn't always suggest the best

The AI is an excellent executor and often a good adviser, but it is not infallible. Sometimes it misreads the request; sometimes it proposes a solution that works but is not the best one. For the students' area of this website, for example, a complete system with an account for every student had been built, with registration and a privacy notice. Thinking it over, I suggested another way: one code per class. Simpler for the students and, above all, no personal data to collect. The best solution came from the discussion, not from the first proposal.

Programs are no longer set in stone

There is a change that is easy to underestimate: a program is no longer a finished object. Software used to be bought as it was, and if a feature was missing you waited for the next version, or made do. Today, if I need something new, adding it is almost always immediate. While I was preparing the materials for this website, I realised I wanted to choose the order in which they appear and to publish students' group projects as web pages: I asked in the morning, and by the afternoon it worked. The program grows along with the way you use it, and that is a different world.

Software springing up like mushrooms

The same phenomenon has another side: free programs are appearing everywhere, written in very little time, that work just like the originals that inspired them. What once required a company and months of work can now be born in a school, in an office, at the home of someone with a specific need.

It is a great opportunity, but it moves the value elsewhere. If code is cheap, what matters is the idea, the care and the trust: who keeps the program alive, who updates it, who looks after the security and the data of the people using it. Once again, the real work is not writing the code, but designing it well and taking responsibility for it.

What it means for school

If the AI writes the lines of code, what is left to teach? Everything else, which is the most important part: analysing a problem, designing a solution, breaking it down, testing it, spotting the mistake. In a word, computational thinking, and the knowledge of programming that makes it concrete.

It is another way of saying something I believe in: AI doesn't serve the school; the school serves AI. We need people who know what to ask, how to check it, and when to say "no, that's not right". Zero lines of code written does not mean zero programming: it means programming has moved from the keyboard to the mind.

More reflections