Introduction

Artificial intelligence is here to stay, and it is rewriting the rules of software development as we knew them. Knowing the fine technical details of a specific framework or language used to be the whole game. Increasingly, it isn't. What matters now is knowing how to orchestrate AI components so they do the heavy lifting for you. If the end result works, does it really matter that a human didn't write every line of it?

This shift isn't theoretical. It's already changing how we interview candidates, how we spend our working hours, and how junior developers are expected to prove their worth. Below are four ideas I've been sitting with—not because I have all the answers, but because I think we need to start asking the questions out loud.

When the Result Matters More Than the Author

For decades, a developer's value was tied to their ability to personally produce working code. Interviews tested algorithmic recall. Performance reviews rewarded lines shipped. But if an AI can generate a correct, maintainable solution faster than most of us, the old yardstick stops making sense.

What starts to matter instead is the ability to:

  • Frame the right problem before asking a model to solve it
  • Evaluate output critically, catching subtle bugs or bad assumptions an AI won't flag itself
  • Orchestrate multiple tools and components into a coherent system
  • Take responsibility for what ships, regardless of who—or what—wrote it

Job interviews built around "write this function on a whiteboard" are measuring a skill that's rapidly losing relevance. The day-to-day of a developer is changing too: less time typing boilerplate, more time reviewing, directing, and deciding. The uncomfortable question is whether our hiring processes and career ladders will adapt fast enough, or whether we'll keep rewarding a skill set that AI has already surpassed.

Who Really Wrote This?

Here's a confession: I didn't write this article. I gave an AI the premises I wanted to discuss—the ideas, the concerns, the angle—and it produced the text you're reading. I'm Mario, and my contribution here was direction, not prose.

That raises a question worth sitting with: what actually holds more value—writing the article yourself, or deciding what it should say?

There's no clean answer. A carefully directed AI output can carry more insight than a mediocre article written from scratch by a human with nothing original to say. At the same time, something is lost when the friction of writing—the process that often generates new ideas as you struggle to articulate them—disappears. Original thought might become harder to find, not because people have stopped thinking, but because the visible artifact (the article, the code, the report) no longer proves that thinking happened at all.

We may need new signals for what counts as "original" or "valuable" content, because the old signal—a human typed this—no longer tells us much.

Who Owns the Machine?

Right now, most of the AI capability we rely on comes from a handful of external companies—Anthropic, OpenAI, and a few others. The possibilities of generative AI are close to limitless. What's limited is the money we have available, whether as individuals or as companies, to keep paying for access to it.

That dependency raises hard strategic questions:

  • Do we build our entire operation on top of third-party APIs we don't control, accepting the pricing, rate limits, and policy changes that come with them?
  • Do we invest in running our own models, open-source or otherwise, on infrastructure we own?
  • Or do we, at least partially, retreat to the world we knew before—where humans did the thinking and the executing, simply because it's more predictable and self-contained?

There's no universally right answer. A startup optimizing for speed might happily rent intelligence by the token. A company handling sensitive data, or one that's been burned by a sudden pricing change, might see self-hosted models as a matter of survival rather than preference. What's certain is that "which AI strategy do we bet on" is now a business decision as consequential as "which cloud provider do we use"—maybe more so.

The Learning Gap Nobody Is Pricing In

This is the part that worries me the most. If an AI can already outperform the average junior—or even a mid-level or senior developer—on raw productivity, what happens to the incentive to learn?

Learning used to be a professional necessity: you studied because your job depended on it. If that pressure disappears, learning risks becoming a personal hobby rather than a career requirement—something only the curious will keep doing, while everyone else leans entirely on the tool.

The people who will feel this first are the ones just starting out:

  • Junior developers may struggle to justify their hiring when an AI is more productive on day one, even if that junior would have grown into a strong senior with time and mentorship.
  • The traditional ladder—junior to mid to senior, built on years of hands-on repetition—may simply stop working the way it used to, because the repetition itself is being automated away.
  • Mentorship and codebases as learning grounds may shrink, since fewer humans are writing the code that used to teach the next generation how things work.

If we don't deliberately protect space for people to struggle, fail, and learn, we risk ending up with a generation of developers who are excellent at directing AI but never developed the deeper intuition that comes from having done the hard thing themselves at least once.

Conclusion

None of this means AI is a threat to be resisted—it's a shift to be understood. The developers, teams, and companies that thrive won't be the ones who cling to the old definition of technical skill, but the ones who figure out, deliberately, what to keep human: judgment, direction, responsibility, and the willingness to keep learning even when it's no longer strictly necessary.

Appendix: The Prompt Behind This Article

In the spirit of the second section above, here is the actual prompt I gave the AI, translated into English, along with the model that wrote the article: Claude Sonnet 5.

Let's create a new article in English with these premises:

  • Artificial intelligence is here to stay and to revolutionize the software development world as we knew it. Knowing specific technical details of certain technologies is no longer as important as knowing how to orchestrate and use AI components to do the work for you. The technical result shouldn't matter just because a human wasn't the one who produced it. Job interviews and the day-to-day of workers must change to adapt to this new world.
  • Original content can be hard to find. This very article was written as the result of an AI prompt. I (Mario) didn't write this, I only gave the premises I wanted to talk about. What holds more value: that I developed the article myself, or that I directed what it should be about?
  • The tremendous dependency on external companies (Anthropic or OpenAI). The possibilities of generative AI are limitless; what's limited is the money we have available as individuals and as companies. Do we base our operations on using these tools? Do we prepare our own or open-source execution models on machines we control? Do we go back to the world we knew, where the human being was the one who thought and executed?
  • I'm concerned that our learning process is becoming almost a personal exercise rather than a professional necessity. Also, those starting out in the workforce may have it much harder when an AI is far more productive than any average junior, mid-level, or even senior talent.