Why System Design matters more in the age of AI
After years in tech, I've never seen a shift like this. We're no longer writing code line by line, we're designing systems and letting AI handle the rest.
For the past 12 years I’ve been working in tech, and this last year has felt completely different. AI didn’t just improve workflows, it changed the nature of software development. Coding has shifted dramatically, and in my opinion, it’s good… but also maybe (sort of) bad.
1. System design is as relevant as ever
Not long ago (literally less than 2 years), we wrote every line of code. We worried about syntax errors, stack traces, condition logic, structure. Hours went into what now feels like menial work in programming and software engineering.
We’ve moved from “how to write” to “what should exist.”
We build. We direct. We describe systems, and AI tools generate code under our guidance (and hopefully supervision). This shift is redefining how developers use AI to build software.
2. The shift from Developer to Builder
“Programmer” feels too narrow now.
You’re no longer constrained by the stack you know, any language, any framework is within reach. With current AI tools, developers can build MVPs faster than ever. What used to take days or weeks in traditional software development now takes hours.
Iteration is insanely fast, and the role of a developer is evolving into something closer to a software builder or system designer.
3. Real implications
Are junior developers learning less or just differently?
Are senior developers becoming obsolete or more valuable?
What are we even testing in technical interviews today?
I talk about this often with colleagues. Junior developers seem to be struggling more, not because they’re worse, but because AI in software development can replace a lot of entry-level output.
Where you needed 3–4 developers to build an MVP, now maybe you need 1–2.
And interviews? Take-home projects feel strange now. You can’t evaluate them the same way when anyone can generate a fully working application, with solid architecture, and deploy it in hours using AI tools.
Technical tests are losing meaning. System design is king
4. Vibing the code
“Vibe coding” went from being almost an insult to something everyone does, and openly. The rise of AI-assisted coding and AI code generation has made this approach mainstream.
Late night vibe coding is honestly fun. You can go from idea -> MVP in a day. One-shot prompts are getting scary good with modern large language models.
But I do wonder: is this hurting programming fundamentals?
Probably yes and no. It depends on how much you let AI think for you. If it replaces your decisions entirely… that’s where things might start to break in the long term.
cf