AI has evolved dramatically over the past few months. What started as a nice-to-have tab completion feature has become an essential part of many developers’ workflows. But this rapid adoption isn’t without risks—and those risks look very different depending on where you are in your career.

The exoskeleton metaphor#

I came up with this image back in early 2024, and it still holds: AI is like an exoskeleton for developers.

If you already know how to walk, run, and jump, an exoskeleton will help you do it longer, faster, and higher. But if you don’t have those fundamentals down, chances are high you’ll hurt yourself trying.

This is precisely the dilemma facing junior developers. They don’t yet have much to augment. The AI isn’t amplifying their skills—it’s substituting for skills they haven’t built yet. Worse, it often robs them of the opportunity to search, dig deep, and ultimately learn.

The junior developer’s trap#

To be clear, AI isn’t inherently bad for juniors. But it can create a dangerous illusion of progress.

It’s easier to hit Enter, Enter, Enter and accept the AI’s suggestions than to actually understand them. The sheer volume of code produced gives the impression of being “at the right level”. But quantity isn’t competence.

That said, AI can also be a powerful pedagogical tool when used intentionally. It can demonstrate patterns, explain constructs, and point to references. The key word is references—you still need to verify everything yourself. AI hallucinates. A lot. Treating its output as a starting point for learning rather than an answer is what separates growth from stagnation.

A different challenge for seniors#

For experienced developers, the calculus is different.

Seniors typically know where they’re strong and where they need to grow. AI should be deployed strategically—specifically for tasks that offer little or no learning opportunity.

Writing unit tests? Let the AI handle the bulk of it. Moving code from one place to another? Simple refactoring? These are perfect candidates for AI assistance (though always verify it hasn’t “forgotten” a piece of code along the way).

The danger for seniors is the opposite extreme: dogmatism. Dismissing AI entirely because it has weaknesses is a path to becoming someone who produces less—or produces worse. The tool has flaws, but so does refusing to use it.

The muscles underneath#

For junior developers: Don’t let AI steal your learning opportunities. Before accepting a suggestion, ask yourself honestly: Do I actually know how to do this? And no, watching someone else do it on YouTube doesn’t count.

For senior developers: Stop doing work that teaches you nothing, especially when it involves writing lots of boilerplate code. Let AI handle that. But also treat it as a pair programmer who can challenge your thinking—and as an assistant who can take low-value tasks off your plate.

And for everyone: don’t forget books. In an age of instant AI answers, sitting down with a well-written technical book remains one of the best ways to build deep, lasting knowledge. Books offer structure, depth, and context that no chat interface can replicate. They’re how you build the muscles in the first place.

The exoskeleton only works if you’ve already built the muscles underneath.