Assisted Superficiality

A long time ago—what feels like an eternity today—I made the decision to take web development very seriously. As a true child of the 1980s, my concept of preparation involved a routine of repetition, set to inspirational rock music. For instance, singing that line: "Try to be best, 'Cause you're only a man, And a man's gotta learn to take it." Putting the time in fast forward, I figured that after a few months of writing a functional website from scratch every day, I would internalize every step and every element. Eventually, making a website would be as simple as walking.

I downloaded every related publication, from Net to Web Designer, searching for anything that gave me ideas to practice CSS and JavaScript. The routine was simple: I’d take an idea I had read about the day before, write a page from start to finish, set up its entire folder structure, and complete the exercise in under an hour, only to delete it afterwards. The goal was practice and reinforcement. Of course, I wasn't thinking about building a project portfolio to look for a job, and that's definitely something that, in hindsight, I would recommend to anyone starting out: do not delete your practice exercises.

But, without a doubt, that practice was incredibly useful to me. And not because I’ve needed to write many websites from scratch since then, but because the most valuable outcome was gaining a deep understanding of how everything worked. Much like a soldier who disassembles and reassembles their weapon repeatedly to know every piece, so they can react when it jams in the field, deep learning enables developers to confront situations that require a quick and precise reaction.

Today, I am very aware of my limitations and my weak points. With an enormous resource library like the internet at my disposal and a good LLM to process it, I can tackle challenges in my daily work that fall outside my core experience. I’m talking about, for example, contributing to projects with languages I don't know, or making cloud infrastructure changes. If time is not pressing, I can tackle them without an issue and with more than acceptable quality.

But that way of working is superficial; it doesn't give me lasting experience. As a casual user of cloud infrastructure, making a couple of fully assisted changes per year, if I have to face a complex failure within the AWS ecosystem, I am completely defenseless. That is why it is important that we make responsible use of AI tools: in addition to streamlining and boosting our work, we must continue practicing and doing deep work. If we let ourselves be carried away by the illusion of "knowing" just because we are capable of completing complex tasks with help, we run the risk of being left unarmed in the middle of a battle.

exit(0);