The Brave New AI-First Development World

The Old vs. The New

In the good ol' days you'd have an idea, spend hours buried in documentation, searching Stack Overflow, and probably end up in a YouTube rabbit hole watching videos completely unrelated to your project (but hey, now you know how to make sourdough bread). Or maybe it was the other way around—you saw an interesting new tool, got inspired, and decided to build something with it. Either way, it was a journey.

Now? I just chat with an AI. Boom—code appears. It's like magic... except when it's not. Sometimes the LLM gives you code that's simply not as useful. Soon enough you'll find yourself into prompt fatigue fighting the AI to do what you want. It's like pair programming with a really smart but occasionally confused partner.

The Developer Skillset In The AI Era

We're heading towards a world where the most productive developers won't necessarily be the ones who know every framework or memorize syntax like it's trivia night at the local pub. Instead, they'll be the ones who can dance with AI—knowing when to lead and when to follow.

But what does this dance look like? Here's my take:

1. Clear Communication is King

   The ability to verbalize your ideas clearly is going to be the superpower. As Andrej Karpathy said, "English is becoming the hottest new programming language". If you can't explain what you want in plain terms, good luck getting your AI buddy to deliver anything useful. For example, if your app is suddenly crawling because of a user spike and you just tell your AI tool "Make it faster", don't expect miracles. It's not going to magically shard your database or optimize your queries unless you know what needs fixing and guide it accordingly.

2. Debugging: The Eternal Skill

   While AI tools are great at generating code, they're not perfect at inferring your specific context. Sure, hallucinations (AI making stuff up) might become less of an issue as these tools improve, but logic errors? Misunderstood requirements? Those aren't going anywhere. Just like how you occasionally write buggy code yourself (don't lie; we all do), you'll sometimes instruct the AI to create something that doesn't quite work as intended. Debugging those issues will still fall squarely on your shoulders.

3. Adaptability Over Memorization

   The tools we use today are incredible productivity boosters, but they're not omniscient problem-solvers. They won't replace engineers because engineering isn't just about writing code; it's about solving problems creatively and adapting when things inevitably go sideways. Think about it: AI can help you write boilerplate code or even suggest optimizations, but can it understand why your product needs a certain feature? Or predict how users will misuse it? Nope—that's still on you.

   And let's face it—users will always find ways to use your product in ways you never anticipated. That's just reality. Your initial design will inevitably miss edge cases or rely on abstractions that turn out to be wrong later on. Complexity creeps in. Entropy wins. And suddenly, you're rewriting code because the problem statement has shifted so much that your old decisions are now dead ends. Someone will need to take decisions on how to adapt to this new requirements and it won't be your AI, at least not for now.

It’s Humans All the Way Down

Everybody thinks everyone else's job is easy, thats why AI "is going to replace so many people". The current scenarios however is that we've all been handed a super-smart sidekick who can handle tedious tasks and speed things up—but only if we know how to work together without stepping on each other's toes (or circuits). Even with all the automatization we'll still need a human in the loop.

Ultimately, being a great engineer in this new world isn't about knowing every tool or writing perfect code on the first try—it's about problem-solving:

• Can you identify what needs fixing when things break?
• Can you design systems that scale and evolve as requirements change?
• Can you guide your AI co-pilot effectively?

So yeah—AI is changing how we build software, we won't be writing as much code as before, but at its core? It's still all about solving problems creatively and adapting as we go. That part hasn't changed—and probably never will.